col-xs-12
col-sm-12
col-md-12
col-lg-12
col-xs-12
col-sm-12
col-md-12
col-lg-12
videoType
dynamic-media
videoDmUrl
https://delivery-p136806-e1377785.adobeaemcloud.com/adobe/assets/urn:aaid:aem:20128cec-af04-496d-8406-d2a1305ea4f6/play?assetname=Understanding+the+Epidemiology+of+COVID19.mp4
keyFrameImage
timestamp

00:00 Good morning everyone.

00:02 My name is Roberta Gasti.

00:03 I'm the chief economist for human development,

00:06 and it is my pleasure today to be here and

00:10 chair this important event on behalf of Art CR,

00:13 our acting chief economist,

00:15 who's very sorry not to be here to listen to what promises to be a very

00:20 interesting and informative

00:23 conversation.

00:23 Today we will be talking about

00:26 Understanding the epidemiology of coronavirus.

00:30 Um,

00:31 I think this is a very positive,

00:33 hopeful title,

00:34 and,

00:34 uh,

00:35 this is something that I think everyone since uh this virus has,

00:39 has,

00:39 uh,

00:40 emerged and wrecked,

00:41 uh,

00:42 the world's life,

00:43 uh,

00:43 would like to do.

00:46 We are fortunate to have today a

00:50 prominent epidemiologist

00:51 who will go over some of

00:53 the implications of different nonpharmaceutical interventions

00:57 on the spread of the disease,

00:59 and we also have

01:01 Our bank colleagues who will provide us with an overview of

01:07 the different tools that are available to

01:09 us as practitioners in economics and health

01:13 to understand better the spread of the disease.

01:17 Uh,

01:18 without further ado,

01:19 let me start by inviting Damien de Valke and Jed Friedman,

01:23 uh,

01:24 to present a bird's-eye view of the instruments

01:27 available to understand the epidemiology of coronavirus.

01:31 Uh,

01:32 both,

01:32 uh,

01:32 Damien and Jed are senior economists in the

01:35 Development Research Group of the World Bank.

01:37 Um,

01:38 they are both working on health.

01:40 Damien with,

01:41 uh,

01:42 an emphasis on,

01:43 um,

01:44 Uh,

01:45 the interaction between health and education and on the epidemiology

01:49 of HIV AIDS

01:51 and Jed,

01:52 uh,

01:52 on the,

01:53 um,

01:54 impact of,

01:55 um,

01:56 uh,

01:56 health on productivity,

01:58 among other things.

01:59 Uh,

01:59 why don't I turn to you,

02:01 uh,

02:01 Damien and Jed,

02:02 uh,

02:03 for your,

02:04 uh,

02:04 introductory presentation,

02:06 and then we'll pass on,

02:07 we'll move on to the next speakers.

02:09 Thank you.

02:09 Over.

02:13 Hi,

02:13 Roberto.

02:14 Thank you very much.

02:15 This is Jed.

02:16 Um,

02:17 Damien and I are just here to

02:20 give a,

02:21 uh,

02:22 a really brief orientation to the speakers to come

02:27 that we look forward to from,

02:28 uh,

02:29 we look forward to hearing.

02:30 Um,

02:31 so we will only take a few minutes to discuss,

02:35 um,

02:35 a few issues related to

02:40 of uh non-pharmaceutical interventions.

02:43 Um,

02:44 what was the motivation for this?

02:47 Um,

02:48 e seminar,

02:50 well,

02:51 as we all know,

02:52 uh,

02:52 our institution has reoriented much of staff effort

02:56 now towards,

02:57 uh,

02:57 counteracting and ameliorating the effects of the

03:01 pandemic,

03:02 and,

03:03 uh,

03:03 we all believe it would be useful to review

03:06 the underpinning epidemiological models,

03:09 um,

03:09 so that we can understand the

03:12 Anticipated

03:13 courses of illness in the countries we work in.

03:16 Um,

03:17 and so what we hope to discuss in the seminar,

03:19 as Roberta just said,

03:20 is the

03:21 epidemiology of the SARS-COV-2 virus,

03:25 uh,

03:25 as well as the effectiveness of strategies to reduce,

03:29 uh,

03:29 the spread in the population.

03:31 So,

03:32 uh,

03:33 Dommy and I will begin with a short primer.

03:37 Um,

03:38 which is largely a brief overview of key terms and some ideas.

03:41 So here's just one slide on some terminology

03:43 that I think we've all become familiar with.

03:46 We've all become chair epidemiologists in the last

03:49 month or so,

03:50 and I'm glad that we have actual epidemiologists on the line today.

03:54 Uh,

03:55 to give us a much more informed view.

03:57 Um,

03:57 but just,

03:58 uh,

03:58 to bring everyone up to speed,

04:00 um,

04:00 let's,

04:01 let's just talk about a few terms such as the reproduction or are,

04:04 are not.

04:05 Um,

04:05 we've all seen this.

04:07 It's defined as the average number of

04:08 secondary infections from an initial infected person.

04:13 Uh,

04:13 the basic R0 is the sort of,

04:15 sort of the maximum epidemic potential of a pathogen,

04:19 uh,

04:19 in the absence of any distancing or other types of interventions.

04:23 And,

04:24 uh,

04:24 estimates for the SARS-CoV-2,

04:26 um,

04:27 vary a bit,

04:28 but

04:29 they essentially in the 2 to 3.5 range,

04:32 which means that for each infected person,

04:34 we should expect,

04:35 uh,

04:36 2 to 3.5 additional infections.

04:38 Uh,

04:39 in the absence of any,

04:40 any intervention.

04:41 Um,

04:41 and then,

04:41 of course,

04:42 the effect of R0,

04:43 which,

04:44 um,

04:44 is the

04:45 empirical,

04:47 uh,

04:47 R0 of what happens in the population

04:50 as we begin,

04:51 uh,

04:51 to try to control transmission.

04:53 And of course,

04:54 this depends on susceptibility,

04:55 so the percent that,

04:57 um,

04:58 uh,

04:58 are,

04:59 have yet to be exposed to the virus as well as the effectiveness of interventions.

05:03 And I just saw a piece yesterday,

05:05 uh,

05:05 estimating that

05:06 The effective R0 for Washington DC has,

05:10 is,

05:11 has fallen from 2.5 to 1 in the last few weeks as a result of the,

05:16 the stay at home measures and other things.

05:18 So that's

05:19 a bit of good news if that's true.

05:22 Um,

05:22 the latent period,

05:24 uh,

05:25 is the time from infection to onset of infectiousness.

05:29 Uh,

05:29 this,

05:29 of course,

05:30 is important in understanding the,

05:32 the spread of the disease,

05:33 and then that,

05:34 that's followed by the infectious period.

05:37 Again,

05:37 measured usually in days of the time a person's,

05:39 uh,

05:40 is infectious.

05:42 Um,

05:42 and I,

05:43 I,

05:43 we've been told by Imperial College colleagues now that maybe the best guess of the

05:47 infectious period is that it begins roughly half a day before the onset of symptoms.

05:52 Um,

05:53 and then there's the serial interval,

05:55 which is often used in,

05:56 in modeling,

05:57 which is the time between the onset of symptoms in the

05:59 index case and then the onset of symptoms in their context.

06:03 Seasonal forcing,

06:04 of course,

06:04 uh,

06:05 this is one way to get at differences in how climate affects transmission.

06:09 Not much is known about this.

06:11 This is obviously another critical factor,

06:13 uh,

06:14 that I don't think we'll be

06:15 speaking about too much today,

06:16 but there might be questions.

06:18 And one way to model seasonal forcing is

06:20 to vary the amplitude of variation and transmission.

06:24 And of course,

06:25 there are some familiar statistics to all of us now,

06:27 such as the case fatality ratio,

06:29 which is the number of deaths divided by the total number of people diagnosed.

06:33 Uh,

06:34 for a certain period of time,

06:35 ideally it's for that period of time to represent the duration

06:38 of the course of illness in the vast majority of cases.

06:42 And one way to account for the possibility of asymptomatic infections

06:47 is to switch from the case fatality ratio to the infection fatality ratio,

06:51 where the denominator also includes estimates for

06:54 the number of asymptomatic and undiagnosed infections.

06:57 And of course,

06:58 what's very relevant uh for policy and a lot of the health teams

07:01 now working in coronavirus preparation are various features of the health system.

07:06 Hospital beds,

07:07 staffing,

07:07 ICU beds.

07:08 This all relates,

07:09 of course,

07:09 to treatment of severe cases,

07:11 the duration of treatment,

07:13 and then the presence of key equipment,

07:15 um,

07:16 such as oxygen,

07:17 ventilators,

07:18 and then the personal protective equipment that staff need,

07:21 health staff need to try to prevent,

07:23 uh,

07:23 exposure themselves.

07:25 And then a large topic of this talk will be non-pharmaceutical interventions.

07:30 Uh,

07:30 there's various types I'll discuss in a second,

07:33 and,

07:34 um,

07:35 Uh,

07:36 what's relevant though for their effectiveness is the degree of compliance or

07:39 enforcement with these types of interventions and of course their duration.

07:44 Um,

07:45 so just continuing on this,

07:46 uh,

07:46 a brief overview here in this slide

07:48 of the types of non-pharmaceutical interventions at,

07:51 at our disposal.

07:54 Um,

07:54 the first would be surveillance.

07:57 Uh,

07:57 this is important

07:58 throughout,

07:59 but it's particularly important at points in the

08:02 epidemic,

08:03 um,

08:03 when transmission

08:05 is at a relatively low level and you can hopefully

08:08 try to contain it solely through the strategy of testing,

08:11 tracing,

08:11 and isolation.

08:13 Which of course is having

08:15 widespread surveillance in the,

08:16 in the population where you're

08:18 testing either widely or,

08:20 or,

08:21 or potential cases and then tracing all their contacts and,

08:24 and isolating them.

08:26 Um,

08:27 from the international perspective,

08:29 uh,

08:29 there are trade and travel restrictions.

08:31 Many countries adopted these,

08:32 uh,

08:33 at the start of March.

08:34 They,

08:35 they range in severity

08:37 from traveler screening,

08:38 which is allowing borders to be open,

08:40 I'd say screening for,

08:42 for symptoms,

08:43 and then if found to say,

08:45 have a fever,

08:46 then forcing traveler quarantine.

08:48 Uh,

08:48 and they run all,

08:50 all the way to the extreme version of,

08:52 of total border closure to people and goods,

08:54 which we've,

08:55 we've also seen at points around the world.

08:59 Quarantine,

08:59 of course,

09:00 is the most extreme form of social distancing,

09:02 uh,

09:03 which is really trying to impose this

09:05 corridor of at least 2 m between

09:08 all people and certainly between infected and non-infected people.

09:12 So quarantine is,

09:13 is not allowing,

09:14 uh,

09:15 any passage out of either a household if the

09:17 household's quarantined or even of entire regions as,

09:19 as was tried,

09:20 for example,

09:21 in northern Italy.

09:23 There are other forms of social distancing that are a little less severe,

09:27 um,

09:27 bans on public gathering,

09:29 public transport,

09:30 closing schools and businesses,

09:32 which we've,

09:32 we've all,

09:33 um,

09:34 seen in,

09:34 in,

09:35 in

09:36 I uh,

09:37 many countries that we're listening from today and then stay at home orders,

09:41 which are not as severe as quarantine insofar as you're

09:43 allowed to go out for essential services such as food,

09:46 uh,

09:47 procurement.

09:48 Finally,

09:49 there's hygiene recommendation and enforcement,

09:52 uh,

09:53 where possible,

09:54 uh,

09:54 promotion of preventive behaviors such as hand washing,

09:58 uh,

09:58 is certainly one tool in the arsenal

10:00 of,

10:01 uh,

10:01 reducing the spread of transmission.

10:03 Uh,

10:04 and then there's recommendations for face masks and even mandates,

10:07 um,

10:08 which is sort of an interesting subarea now with,

10:11 with more and more,

10:12 um,

10:13 uh,

10:14 evidence suggesting that it's,

10:15 it's a 11 effective tool in driving the R0 to,

10:20 uh,

10:21 below 1,

10:22 which,

10:22 of course,

10:23 is the goal of a lot of these.

10:24 So,

10:24 as soon as,

10:25 uh,

10:25 one new infection will infect less than one

10:28 going forward,

10:29 we would expect to see a decline in overall transmission.

10:33 And of course,

10:33 the effectiveness of all of these NPIs will depend on various factors,

10:37 as I said,

10:37 the duration of them,

10:39 the degree that population complies with them,

10:42 and then,

10:42 of course,

10:43 epidemiological factors as well,

10:45 such as the latency and,

10:46 and the R0 that I mentioned.

10:49 So,

10:49 thank you for bearing with me.

10:50 This is just a brief overview that will continue now with my colleague,

10:54 Damian,

10:55 I believe.

10:56 Uh,

10:56 so if I can invite Damian to continue.

10:59 Yes,

10:59 thank you,

11:00 Jed.

11:01 Um,

11:02 So,

11:03 uh,

11:03 what I,

11:04 I will do in,

11:05 in the,

11:05 uh,

11:06 the few coming,

11:07 uh,

11:08 slides,

11:09 uh,

11:09 Jed,

11:09 I think we still,

11:11 we cannot see your,

11:12 uh,

11:12 you're sharing the slides,

11:14 but we cannot see them.

11:15 Yeah.

11:16 Um,

11:17 It's uh trying to uh review um

11:22 a few of uh um

11:25 online resources for epidemiological modeling and projections.

11:29 I've reviewed some of them in a,

11:31 in a blog post that

11:32 at the end of March,

11:34 which probably would need some updating.

11:37 What's important

11:38 to know also is that the World Bank,

11:40 the International Decision Support Initiative,

11:42 the Gates Foundation,

11:44 And the World Health Organization has developed a

11:46 partnership to review and compare all models,

11:49 and

11:49 maybe David and Marli will tell us more about that.

11:53 Then one of the available tools of course,

11:56 we'll hear more about it today is the

12:00 estimates that the Imperial College COVID-19 response team provided.

12:06 Uh,

12:06 they've

12:07 providing,

12:08 they've been providing these estimates for

12:10 the world and for each country

12:12 under five scenarios going from

12:15 unmitigated epidemic to

12:17 early suppressor suppression strategy.

12:20 They'll they'll give,

12:21 I'm sure more detail about that,

12:22 but.

12:23 So

12:26 one very useful feature of their paper is that there is an online appendix,

12:29 and when you click on that,

12:31 you get access to an to an Excel table give you,

12:34 uh,

12:35 giving you

12:35 country level estimates

12:38 with even the further possibility to adjust some

12:40 of the parameters such as the baseline air knot

12:43 and the intensity of social distancing measures.

12:45 So it's a very useful tool.

12:47 Another tool that is available online is at the University of Basel.

12:52 It's an online tool that allows you to run your own projections

12:56 and enter country specific numbers.

12:59 I found it quite intuitive and flexible,

13:01 but it's important to

13:03 know that you will

13:04 need

13:05 to closely calibrate your proper

13:07 parameter estimates to the data

13:09 that you have.

13:10 Some

13:12 World Bank teams have used it to.

13:15 Make projections for the country they are working on.

13:18 Um,

13:19 it's a very important feature is that

13:22 the age structure of the population is preloaded

13:25 for each country on the online tool.

13:28 Then there is also projections available that have been

13:32 made by the Institute for Health Metrics and Evaluation.

13:35 Currently the projections are available for the US and European countries,

13:40 but it seems that uh projection for African countries will be coming soon

13:45 and the World Bank

13:48 is collaborating

13:49 for obtaining regularly updated projections for the countries

13:54 which have,

13:56 which are recipients of

13:58 COVID-19 financing from the World Bank.

14:01 It's important to know that with that tool,

14:04 the scenario projector assumes current social distancing

14:08 measures are maintained,

14:09 so it's different from the other tools where

14:11 you can go

14:13 all the way from unmitigated scenario to

14:18 mitigation measures and and suppression measures.

14:21 Also,

14:22 it's a model that is not based on a SIR model.

14:26 So SIR is susceptible infectious

14:29 recovered,

14:30 but it's a model that fits the Emperor Kelly observed COVID-19

14:34 population death curve.

14:36 So in my next slide,

14:38 I will show you an example of

14:40 how our World Bank team

14:42 has used the University of Basel tool to work on some projections.

14:47 It's a large middle income country.

14:50 they've shared it with the government,

14:51 but

14:52 the government has

14:53 preferred to uh

14:55 to keep this,

14:56 um,

14:57 not public yet,

14:58 so I'm not at liberty to give you the name of the country.

15:02 But you see how they've used the tool

15:06 in these two graphs

15:08 to predict on the left

15:11 the number of COVID-19 patients requiring hospitalization.

15:15 In red with the curve,

15:16 if there was no action,

15:18 no action.

15:19 And in blue

15:21 with

15:23 the

15:24 The shape of the epidemic

15:26 with a 2 month lockdown

15:28 and this is of course you have seen this picture before,

15:32 a nice example of flattening the curve

15:35 and on the right with the same color red for no action,

15:39 blue for a 2 month lockdown.

15:41 You see,

15:41 uh,

15:42 the COVID,

15:42 the projected COVID-19 cumulative fatalities.

15:46 I should also add that this World Bank team worked

15:49 hard on the University

15:51 of Basel T and then when the Imperial College

15:54 estimates

15:56 were,

15:57 were published,

15:57 they were very happy to see that their

15:59 numbers were very similar to the

16:02 um

16:03 suppression scenario from Imperial College.

16:08 No,

16:09 um,

16:09 I'm

16:10 moving on this slide that goes a little bit beyond

16:14 the pure epidemiology and try to talk about all testing

16:19 can be very important to inform

16:22 decisions and the effectiveness of uh non-pharmaceutical

16:26 interventions.

16:27 Um,

16:28 Jet,

16:28 Roberta,

16:29 and Aditya Matu and myself,

16:30 we've written a

16:32 short note on this,

16:34 and I'm sharing with you,

16:35 uh,

16:36 these figures.

16:37 So you'll see in blue the

16:39 epidemiological curve for the infection.

16:43 And we have been

16:44 yellow,

16:46 uh,

16:46 the fraction of the population that that is recovered.

16:50 So what we

16:52 Argue in that piece and through this figure is that

16:55 testing and two types of tests are very important to make decisions.

17:00 So one test is the test,

17:02 the test of

17:04 infectiousness,

17:05 and that's usually do

17:07 through a PCR test.

17:10 And you see that to the

17:14 left of the first red line.

17:17 That's the time when you are early in the epidemic

17:21 when

17:22 basically you can try to implement a test,

17:24 trace,

17:25 and selective

17:27 isolation,

17:27 a TTI strategy

17:29 relying mainly

17:31 on

17:33 the

17:34 infectiousness test

17:36 based on the PCR.

17:38 Similarly,

17:38 on the right of the figure to the right of the second

17:41 red line

17:43 when

17:44 the

17:45 epidemic.

17:47 The epidemic

17:48 outbreak has basically

17:51 has gone

17:52 down enough,

17:53 then we can

17:54 try to use again the same TTI test trace and selective

17:59 isolation

18:00 strategy

18:01 to try to prevent the resurgence of a new

18:05 epidemic wave.

18:07 In between these

18:09 two red lines.

18:11 When

18:12 the outbreak is really in full force,

18:14 it's time to use the suppression strategy

18:17 which we've colored in green.

18:21 We've also indicated that another test,

18:23 an antibody test,

18:25 could be used

18:27 at a certain point

18:29 first to protect essential workers because that test

18:33 detects who is immune to the.

18:37 Uh,

18:38 to,

18:38 uh,

18:38 to,

18:38 uh,

18:39 um,

18:40 COVID-19

18:41 and then eventually gradually

18:43 to allow other people to work and we also recommend that

18:47 uh

18:48 there would be a regular antibody testing

18:50 of a representative sample of the population.

18:53 As well as continuous monitoring of morbidity and mortality

18:58 to inform

19:00 these decisions on when to use the test,

19:03 trace and selective isolation strategy and when

19:07 to

19:08 enforce the suppression strategy.

19:11 No,

19:12 I'll,

19:12 I'll leave this and,

19:13 and just finish by outlining our e seminar today.

19:17 First,

19:18 um,

19:19 the Patrick Walker and Azra Ghani from the,

19:22 uh,

19:22 Imperial College COVID-19 Response Team will present their,

19:26 their model.

19:27 Uh,

19:29 they will tell us about the underpinning of this model in its main prediction.

19:34 Tell us all the effectiveness of the non-pharmaceutical information are modeled,

19:40 what are the main policy recommendations,

19:42 and they will touch upon specific issues for low and middle income countries

19:46 such as the age structure of the population.

19:49 The presence of many multi-generation households,

19:53 crowded habitat like slums,

19:56 the difficulty to enforce social distancing measures

19:59 when people are living and working in the informal sector,

20:02 and then our colleagues

20:04 from the Health,

20:05 Nutrition and Population.

20:07 Uh,

20:08 at the bank,

20:08 David Wilson and Marili Gorians

20:11 will discuss the trade-off between the need to curb the

20:15 coronavirus

20:17 and the pressure to reopen economies,

20:20 and they will go over a few key consideration in that

20:23 that will inform the trade-off.

20:26 The epidemic forces,

20:27 the capacity to manage a reduced measures,

20:31 population health,

20:33 the capability of the public health in the country,

20:36 the health system capacity,

20:38 national decision making capability,

20:40 and the role of

20:43 technical innovations.

20:45 And with this,

20:47 um,

20:48 Um,

20:49 will,

20:50 uh,

20:52 Give the floor to Patrick.

20:54 Thank you very much.

20:59 Uh

21:01 Thank you,

21:02 Damien.

21:02 This is Roberta.

21:03 I just wanted to take a moment to introduce

21:06 uh Pat uh Patrick Walker and Azrahani to our audience.

21:11 Um,

21:12 they are respectively lecturer and professor

21:15 at the infectious disease epidemiology department of Imperial College,

21:19 and

21:21 we of all,

21:22 the armchair epidemiologists,

21:24 as

21:25 Jed nicely put it,

21:27 read and learned tremendously

21:29 from the work that the Imperial College team

21:31 has done in helping the world understand the patterns

21:36 and the spread of coronavirus,

21:38 and

21:39 The potential effects of different policies.

21:41 So I understand that Patrick will start the presentation,

21:44 so thank you so much for being here and particularly thank you for the

21:47 excellent work that you are doing to um help us understand the disease.

21:51 Over.

21:55 Thank you so much,

21:55 uh,

21:56 Damien and Roberto.

21:56 I'm just,

21:57 just

21:58 trying to make sure I can master the technology,

22:01 um.

22:02 But yeah,

22:02 thank you very much for that really er

22:04 excellent introduction,

22:05 and I think.

22:07 Uh,

22:07 when it comes to actually.

22:10 It comes to

22:13 Coronavirus,

22:14 uh,

22:14 COVID-19,

22:15 epidemiology,

22:16 we're all fairly armchair cos we've only had

22:19 a few months to,

22:20 er,

22:21 to actually,

22:21 um,

22:22 I'm just gonna

22:23 move to the,

22:23 the correct slide

22:25 I picked the wrong ones.

22:42 Reviews for this.

22:53 Apologies for that.

22:55 So yeah,

22:55 so,

22:55 uh,

22:56 like I was saying,

22:56 I mean it's,

22:57 it's a newly emerging disease,

22:58 so I think we should all feel a little bit

23:01 more OK about being armchair epidemiologists.

23:03 We've only had,

23:04 um,

23:05 4 months now to,

23:07 to learn stuff about this disease and there's

23:09 not a lot of data to actually go on

23:12 and you know,

23:13 and

23:14 as we sort of know with these.

23:16 This is in when when outbreaks emerge,

23:18 there's a lot of difficulty in actually getting

23:21 representative data because

23:23 of a range of severity in terms of infections represent and

23:27 likelihood that you're going to

23:29 see them reported or they will

23:31 get into your surveillance system.

23:33 And also that is likely to change over time.

23:36 But um,

23:37 but one of the few things we do know about,

23:39 um,

23:39 COVID-19 is that there's this clear,

23:41 um,

23:42 severity pattern with age,

23:43 whereby,

23:44 uh,

23:45 the older you are,

23:46 the higher the likelihood of,

23:48 um,

23:49 having sort of severe complications,

23:50 respiratory symptoms and,

23:52 and needing

23:53 more,

23:53 uh,

23:54 you know,

23:54 advanced intervention to

23:57 a higher risk of fatality.

23:59 Um,

23:59 and this is,

24:00 these are estimates that were generated by,

24:02 um,

24:02 people in our team,

24:03 um.

24:05 and these are being used to kind of inform our

24:07 severity predictions

24:08 and by age

24:09 and also we know that there are some kind of relevant comorbidities characterized

24:13 primarily focused around

24:15 morbidities that exist in the settings where we've seen.

24:19 Uh

24:20 Large outbreaks,

24:21 so these are things like

24:22 cardiovascular disease.

24:24 Uh,

24:25 pulmonary

24:26 disease

24:27 and um diabetes.

24:29 Um,

24:30 we're also learning more as uh Damien was,

24:33 uh.

24:34 Very ably

24:36 Uh,

24:36 discussing earlier,

24:37 earlier,

24:38 and we're learning a lot more about the transmissibility of the.

24:42 So,

24:42 uh,

24:43 yeah,

24:43 so this is work,

24:44 um,

24:45 it's also been done by people,

24:46 other people in our team,

24:48 um,

24:48 Sam Bath and Seth Waxman and others,

24:51 and then what they're doing is they're taking data on the,

24:53 uh,

24:54 the deaths over time,

24:55 which even in high,

24:56 high income settings are the things that we think are most reliable

25:00 case reporting changes so quickly over

25:02 deaths are your

25:03 ways that you can most reliably get friends.

25:06 Um,

25:06 but unfortunately what makes that hard is you have this delay of around.

25:11 20 days from when you're infected to when you

25:14 I if

25:15 if

25:15 that's.

25:16 You're lucky enough that that's.

25:20 Um,

25:20 and that actually makes inferring the trends in

25:22 deaths of being in Europe quite hard,

25:24 but from this analysis,

25:26 what we are seeing is that this kind of curve is leveling off,

25:30 factual that actually if there weren't any.

25:33 That

25:34 um the curve initially is kind of

25:36 doubling times consistent with an R0 of about 3 or 4,

25:39 which is actually very highly infectious compared to something like.

25:42 So normal seasonal flu,

25:44 um,

25:44 which usually has an out of about 1.5.

25:47 Um,

25:48 and that,

25:48 um,

25:50 as we're implementing these measures,

25:51 we're seeing rather than the,

25:53 uh,

25:53 the debts double and double and double,

25:55 we're seeing that they are actually leveling off and.

25:58 that does correspond with some of the larger what we know as

26:02 non-pharmaceutical.

26:03 Uh

26:04 interventions,

26:05 um.

26:06 Things that are often called lockdown

26:08 and they are beginning to show impact so hopefully over the

26:11 we're gonna see

26:12 signs in more time.

26:14 Some of the hardest hit.

26:16 If these measures are maintained.

26:19 Um,

26:20 but even in these high income settings,

26:22 the problem is that the,

26:23 that the reasons they're needed is partially because of the,

26:26 the direct burden,

26:26 but also because they have this massive strain on the health system

26:29 and they lead to incredibly high demand for

26:32 hospitalization,

26:33 ICU,

26:34 oxygen and medical ventilation.

26:36 Also,

26:37 they,

26:37 uh,

26:38 pose a high,

26:38 high burden on the personnel who are offering care,

26:42 so.

26:43 High rates of

26:45 infection and self-isolation

26:47 problems and the need for extensive,

26:51 um,

26:51 and these are even,

26:51 these are some of the countries that have some of the strongest.

26:58 So,

27:01 Uh,

27:01 Charles

27:04 So,

27:05 um,

27:05 thinking about how modeling plays a part in this,

27:07 well,

27:08 it's,

27:08 it's

27:09 when you have a,

27:09 a,

27:09 you know,

27:10 a brand new disease,

27:11 a new disease that actually

27:13 kind of needs some sort of model in which to think about

27:16 the kind of the logical

27:18 scenarios that you might see,

27:20 and the model is just a,

27:21 it's,

27:21 it's not a,

27:22 especially where this early in October,

27:24 you can't really expect them to be kind of

27:26 completely accurate forecasts of the future,

27:28 but what they do give you is a sort of logical framework.

27:31 For

27:32 the assumptions you're making about different strategies,

27:34 and actually it kind of really helps to kind of hold you to those assumptions.

27:38 This is work that

27:39 the COVID-19 team have providing.

27:43 Neil,

27:43 the

27:44 head of our group,

27:45 uh,

27:46 has been

27:46 an AR as well,

27:47 uh,

27:47 providing to the UK government along with other modeling groups we have in the UK

27:51 um around different scenarios,

27:53 what,

27:54 what you do,

27:55 um,

27:55 what can you do and what is it likely to achieve,

27:58 where will that get you,

28:00 um,

28:00 and it does kind of

28:01 hold you to the logical assumptions.

28:03 So,

28:03 for example,

28:04 if these large suppression strategies.

28:07 We are gonna have to think about an exit strategy because we can't just go back to zero

28:12 intervention because then we do definitely get

28:15 sort of a second wave.

28:16 Um,

28:17 so,

28:18 so it,

28:18 it's helpful to really think about,

28:20 you know,

28:20 the future.

28:22 Um,

28:23 but,

28:23 a lot of countries don't have these kind of in-country modeling groups,

28:26 um,

28:27 that we have in the UK and other countries have as well,

28:29 but not many.

28:30 Um,

28:30 but they're all facing very similar questions at the moment,

28:33 um,

28:34 and

28:35 some of them might be a bit earlier in the pandemic,

28:37 but those questions are likely to be quite similar.

28:39 Um,

28:39 but what we've been trying to do in terms of our sort

28:42 of the global side of things is particularly with lower income settings,

28:45 what can we say

28:46 that kind of mimics some of these modeling approaches,

28:49 given that we probably can't do something comprehensive for every country.

28:52 Um,

28:52 and what would be useful given that

28:55 Clearly we're in a very time critical.

28:58 You have to

29:00 That needs to be useful,

29:00 it needs to

29:03 Um

29:05 So what we've,

29:05 when we've been thinking about this,

29:06 we've kind of identified 5 main factors which are

29:09 likely to vary substantially,

29:11 particularly by income.

29:13 Some of them are obvious,

29:13 some of them less so,

29:14 but kind of quite an obvious one is that

29:16 when you have a disease like this that has this kind of

29:19 um.

29:20 Severity signal with age where

29:22 the older you are,

29:23 the more likely you are to have severe consequences.

29:26 Well,

29:27 in many lower income settings,

29:29 populations are a lot younger.

29:30 They have higher fertility rates.

29:31 They have

29:32 lowest survival rates outside of,

29:34 outside of the

29:36 window,

29:36 um,

29:37 and that

29:38 leads to,

29:38 you know,

29:39 there being

29:40 the average age of an infection

29:42 go down,

29:42 and the,

29:43 we would might anticipate that that would be a,

29:45 a force to reduce

29:47 a person's burden.

29:51 That's one thing,

29:51 there's also a pattern with

29:54 sex,

29:54 biological sex,

29:55 which is

29:56 males are slightly

29:57 more likely to have severe disease,

29:59 but that's probably more likely to be balanced across country.

30:02 Um,

30:03 also there's societal structure.

30:05 Um,

30:06 which can vary quite a lot and in very important ways.

30:09 I'll try and speak up,

30:11 but please,

30:11 yeah,

30:11 let me know if I'm,

30:12 I'm not loud enough.

30:13 Um,

30:14 that

30:15 the,

30:15 well,

30:16 what we have in,

30:17 in the UK where I am now

30:18 is that outside of the pandemic,

30:20 elderly people,

30:21 uh,

30:22 elderly people tend to be a lot more kind of isolated,

30:24 and it's something that we would see as a,

30:26 a health problem outside of the pandemic window.

30:28 But if you are an elderly person,

30:30 you,

30:31 um,

30:31 I'll,

30:31 I'll go into that in a second time.

30:33 But there's also a that's for the next slide,

30:35 then there are also these patterns of comorbidities that might change.

30:39 Um,

30:40 also,

30:40 there's clearly going to be big disparities with

30:42 both health and system strength and capacity.

30:45 Finally,

30:45 we shall get onto that there are these,

30:48 the,

30:48 the,

30:48 the extent to which countries are likely to be vulnerable

30:51 to the social and economic impact of both the disease itself

30:55 and any interventions,

30:56 especially some of these

30:57 stringent interventions,

30:58 is going to differ widely by the level of income country.

31:04 So this is what I was saying a second ago,

31:05 um,

31:06 that,

31:06 yeah,

31:07 in,

31:07 so this plot is showing the

31:09 average household size of a person aged 65.

31:12 And what we see in both these,

31:13 this,

31:14 this kind of data and,

31:15 and in

31:16 contact surveys where we ask people about the number of contacts they make

31:19 in higher income settings,

31:21 particularly in Europe,

31:22 older people make less contacts.

31:24 They live in much smaller households,

31:25 and it's this pattern of kind of social isolation,

31:28 which is generally seen as a problem,

31:30 you may live on your own.

31:31 Um,

31:32 in,

31:33 uh,

31:33 lower income settings,

31:35 uh,

31:35 older people are far more likely to share

31:37 their house with their children or their grandchildren,

31:40 so in very low income settings,

31:41 often the average household size of a

31:43 person aged 65 could be

31:46 6 or 7,

31:48 and

31:49 also there are big differences in,

31:51 um.

31:52 School attendance and work patterns

31:54 that we that we think along with the data on social

31:58 contact patterns that exists in low and middle income settings,

32:01 not much,

32:02 but actually you,

32:03 you,

32:03 you very rarely see these patterns of social isolation of elderly people,

32:07 they're far more present in terms of the

32:09 number of contacts they make in their society,

32:11 um,

32:12 and that's,

32:13 as so we anticipate that.

32:15 Older,

32:16 there will be fewer older people,

32:17 but those older people that are there will

32:19 be more vulnerable to having the disease spread

32:22 to them,

32:22 uh within their households.

32:24 So

32:25 what's that trade-off in terms of on the one hand that could make it less severe,

32:29 on the other hand,

32:30 the fact they're less isolated makes them more likely to get it.

32:33 Well,

32:33 that's something that you can kind of try to explore with modeling,

32:36 making

32:37 sure that you're,

32:37 you're very clear about the assumptions that you're making.

32:41 What we have is we have a model that mimics some of the major uh

32:46 attributes of the

32:47 larger microsimulation model

32:50 that has been used to inform.

32:51 Um,

32:52 Patrick,

32:53 sorry to interrupt.

32:54 This is Roberta,

32:55 and I'm talking to,

32:57 uh,

32:58 Ale,

32:58 you,

32:59 Ale,

32:59 and,

33:00 um,

33:01 Ryan just to indicate that I've got some messages from,

33:04 uh,

33:04 those who are following us on YouTube that your slides

33:08 cannot be seen on the YouTube channel.

33:10 Ah

33:12 So I like,

33:13 I'll give a minute to you,

33:15 uh,

33:15 Ale and,

33:16 um,

33:17 Ryan to sort it out,

33:18 but I also take this opportunity,

33:20 besides the,

33:22 uh,

33:22 small technological glitch,

33:24 to indicate to our,

33:26 um,

33:27 to those who follow us on YouTube today that they

33:30 can send us questions through the chat room in YouTube,

33:33 and I'll represent them,

33:35 uh,

33:36 on your behalf to our speakers and the same,

33:39 uh,

33:39 uh,

33:40 the Webex,

33:41 um.

33:42 Um,

33:42 uh,

33:43 audience can do,

33:44 but,

33:45 uh,

33:45 those who are following us on Webex,

33:47 and it's been a,

33:47 uh,

33:48 first come,

33:49 first serve,

33:49 a sample of the 1st 20,

33:51 I believe,

33:52 or 50 who have signed up,

33:53 uh,

33:54 they can also raise their hand,

33:56 uh,

33:56 and,

33:57 uh,

33:58 ask the question directly in the Q&A session.

34:01 So over with this small organizational announcement,

34:05 um,

34:05 and maybe

34:06 whether Ryan or Ale

34:09 can send,

34:09 give me.

34:11 Send me a text

34:12 to indicate whether

34:14 what needs to be done to have the um slides seen on YouTube.

34:19 We definitely see them on the web Webex channel.

34:21 Thank you,

34:22 over.

34:25 Thank you Robert,

34:26 uh,

34:26 yeah,

34:26 and I think I,

34:27 I saw one of the questions about how,

34:29 how we're calibrating it,

34:30 and this is what,

34:30 what this,

34:31 this slide is sort of trying to show

34:34 that

34:34 what we do is we,

34:35 we take,

34:36 uh,

34:36 this model is kind of capturing sort of the,

34:38 some of the,

34:39 the major attributes.

34:40 What we do is we take,

34:42 uh,

34:43 the,

34:43 uh,

34:43 what we know about the transmissibility of the disease in terms of.

34:48 Um,

34:48 and that's a parameter we can vary by context.

34:51 Take what we think we know about

34:54 mixing patterns by age within a setting,

34:56 and we take the severity of infection by age,

34:59 we put them all together in a model,

35:01 um,

35:02 as long,

35:02 uh,

35:03 uh,

35:03 uh,

35:03 as well as some of the more key

35:05 parameters of the epidemiology that Damien was describing,

35:08 things like

35:08 time between

35:10 one infection to pass through another infection,

35:12 and,

35:12 uh,

35:13 the time from onset to death,

35:14 things like that.

35:16 Um,

35:16 and from that we can kind of generate what we think might

35:19 be the trade-off between changing age

35:21 patterns and the changing social mixing patterns

35:24 by income strata and by country.

35:26 And

35:26 this slide here

35:28 is just showing kind of

35:29 how we think that would vary.

35:30 So this is assuming the same R0 across all settings,

35:33 here we're taking an R03 which is.

35:35 As Damien was saying,

35:36 and as,

35:37 as the data in Europe is showing us,

35:39 our sort of best estimates for what we,

35:41 what this disease is doing,

35:42 uh,

35:43 and take the sort of likely impact in it of a

35:46 total impact of an unmitigated disease,

35:49 uh,

35:49 uh,

35:49 pandemic,

35:50 and that RO 3 highly transmissible disease means that it will spread very widely

35:55 in a within a community without any action,

35:57 um.

35:58 Uh,

35:59 but that actually as you think about more the

36:01 more sort of severe severe scale of the disease,

36:04 uh,

36:04 moving from hospitalization,

36:05 requiring critical care and,

36:07 unfortunately,

36:08 sadly,

36:09 potentially,

36:10 uh,

36:10 leading to mortality,

36:11 but actually

36:12 because you do have these broader demographic patterns,

36:16 uh,

36:16 the,

36:17 um,

36:17 we like,

36:18 we do expect an unmitigated scenario with no intervention.

36:21 Actually higher income settings would be per person,

36:23 at a per person level.

36:26 More heavily affected,

36:27 experience a higher burden,

36:28 but this is making a very strong assumption at this stage,

36:31 that,

36:32 that these that these severity patterns are matching those

36:35 for the health system in China.

36:38 Now clearly China,

36:39 at least with particularly in this pandemic,

36:42 made a really large effort in making sure that that the majority of people

36:46 are getting hospitalized.

36:47 Uh.

36:49 Patrick,

36:49 sorry,

36:49 this is Roberta.

36:51 Apologies for interrupting.

36:52 One message to you is that if you can speak as close as possible to the mic,

36:56 that will help.

36:58 Second,

36:58 I'm asking

36:59 Alejandra or Ryan to please unmute themselves and let us know whether the

37:03 YouTube channel now works with the slides because I'm continuing to get,

37:07 to get messages and since I'm,

37:08 I follow everyone on Webex,

37:10 I am unable

37:11 to um.

37:14 Have the information on whether.

37:15 Yeah,

37:16 hi Roberta,

37:16 uh,

37:17 Ryan is currently with ITS.

37:18 Unfortunately,

37:19 I don't know what's going on because it seems there's no audio.

37:23 Um,

37:23 but he's working on it.

37:26 Oh excellent.

37:27 So I would appreciate um Ali,

37:29 whether uh whenever this is sort of sorted out,

37:32 whether you can um just let me know

37:35 on the Webex or on the channel.

37:38 Since we have this moment of mute,

37:40 I would like them to ask Patrick to just,

37:44 uh,

37:44 at the end of his presentation to just give like a quick

37:47 point by point.

37:49 Uh,

37:50 sort of key messages summary

37:53 for those who have not been able to follow the um.

37:57 Uh,

37:57 the conversation.

37:58 Meanwhile,

37:59 I see many questions that are popping up on Webex.

38:03 Uh,

38:03 I don't know,

38:03 Patrick,

38:04 if you see them or whether Professor

38:06 Ghani is,

38:07 uh,

38:08 answering them,

38:09 but it might be good to

38:11 maybe take a moment to,

38:13 uh,

38:14 indicate them to the audience so that,

38:16 uh,

38:17 also the audience can see the conversation that is ongoing.

38:21 Um,

38:21 just to take a pause while the

38:24 technology is being fixed.

38:25 Um,

38:26 there was a question on.

38:28 Uh,

38:29 from Dmitry on how,

38:31 uh,

38:32 R 0 was,

38:33 um,

38:34 uh,

38:34 estimated,

38:35 and I wonder whether Professor Ganni wants to amuse

38:38 herself now while we take this pause to give,

38:41 um,

38:41 a quick answer for everyone,

38:43 also those on YouTube who might be,

38:45 uh,

38:46 now,

38:47 uh,

38:48 might hear it as well.

38:49 Professor Ganni,

38:50 would you want to,

38:51 um,

38:52 address that question?

38:53 Yes,

38:54 that's fine.

38:54 I've actually just put a comment in for those who might not be able to hear.

38:58 Um,

38:58 the,

38:59 our estimates of the,

39:00 uh,

39:01 basic reproduction number come from the early growth rate of epidemics.

39:04 We actually

39:05 estimated a,

39:06 a lower,

39:07 um,

39:08 value of around 2.4 from the Chinese data originally.

39:12 Um,

39:12 the doubling times of the epidemics in Europe have been

39:16 much faster,

39:16 and our current best estimate is around 3.

39:20 clearly it's a little bit sensitive to the

39:22 context and the setting in which transmissions occurring.

39:29 Um,

39:30 excellent,

39:30 thank you.

39:31 Um,

39:32 Ale,

39:33 uh,

39:33 do you know if the problem is fixed on YouTube?

39:36 Um,

39:37 I'm

39:39 OK,

39:40 no,

39:40 Ryan,

39:41 I'm on the,

39:42 hi,

39:42 this is Ryan here.

39:43 I'm on the line with

39:45 our IT colleagues,

39:46 and they're trying to solve the problem but do not yet have a solution,

39:49 so I'll

39:50 let you know as soon as possible,

39:52 but unfortunately the slides

39:54 aren't coming through,

39:55 um,

39:56 uh,

39:57 but I,

39:57 you know,

39:57 I'm,

39:58 I'll get back to you as soon as,

39:59 as soon as they have this solved.

40:02 Very good.

40:03 Um,

40:04 well,

40:04 why don't you signal that maybe in the chat room to the YouTube,

40:07 uh,

40:07 followers,

40:08 and then

40:09 we can maybe continue with the presentation and then I'll

40:11 ask Patrick just to summarize the main messages so that,

40:15 um,

40:15 those on YouTube,

40:16 which is a much larger audience,

40:18 uh,

40:18 can still follow the presentation.

40:20 Thank you and over,

40:21 Patrick,

40:22 see you.

40:31 Patrick,

40:32 you're muted.

40:35 And that um and yeah,

40:36 and we do actually have

40:37 some additional scenarios

40:39 accompanying the report that I can also provide a link for

40:43 if,

40:43 if people want to sort of see how these may change with

40:46 in different contexts,

40:47 if or not differs

40:50 to an extent.

40:51 So yeah,

40:51 so,

40:51 so this is just to say,

40:52 well this is based upon

40:54 severity estimates by age from a place where there was relatively speaking,

40:58 at a global level,

40:59 very good access to care.

41:01 And that's gonna differ

41:03 quite substantially by context and and and change the patterns a little bit,

41:07 which I'll get onto

41:08 in a,

41:08 in a minute.

41:10 So

41:10 what we then looked at is this is kind of the,

41:12 the,

41:13 the worst possible news,

41:14 the,

41:15 what happens if we don't do anything,

41:16 but of course

41:17 as,

41:17 as a

41:18 globe we are gonna do a lot to try and prevent this burden,

41:21 given how

41:22 big it is,

41:22 given that we estimate that this could be

41:24 40 million if we don't

41:26 deaths globally if we

41:28 don't actually

41:29 do something.

41:29 But but countries will certainly do things.

41:32 And um.

41:33 But,

41:34 you know,

41:34 um,

41:35 and

41:37 Without a vaccine,

41:38 that's all gonna be these non-pharmaceutical interventions,

41:40 modifying our social contact behavior,

41:43 um,

41:44 particularly in settings that don't have the er

41:46 capacity and the technology to implement some of these

41:49 really stringent test er testing and treating and tracing mechanisms.

41:53 A lot of countries probably will,

41:54 will,

41:55 will not have that ability

41:56 in the time that it takes.

41:58 So yeah,

41:59 so,

41:59 and so here this is where we sort of looked at,

42:01 well,

42:02 let's say you can't suppress the infection,

42:05 what's the best if you didn't suppress the infection,

42:07 if you couldn't

42:08 decide not to,

42:09 what would that lead to in terms of what could you do

42:12 to reduce this burden?

42:14 And this relies upon a concept called herd immunity,

42:16 which is,

42:17 which is

42:18 quite become quite a controversial concept in the

42:20 UK but it's actually just a scientific concept,

42:21 it's not particularly an intervention.

42:23 What it all it means is that

42:25 if you can't suppress infection,

42:26 any reductions you can make

42:28 to your contact rates within your community

42:31 can reduce the spread of the infection,

42:33 and eventually that infection will lead to a level of immunity

42:37 that will then stop the er the infection

42:40 from,

42:41 make the infection peter out because you have a level of immunity in the population.

42:45 Where if you hadn't reduced your contact rates,

42:47 it would overshoot that level,

42:49 so you'd actually get much higher levels of,

42:51 of burden.

42:52 So it's actually just reducing the level of overshoot,

42:55 um that you would get if you don't do anything,

42:58 um.

42:59 So what we find is,

42:59 you know,

43:00 with a,

43:00 with an honor of about 3,

43:02 those reductions are about a 45% reduction in contact rates

43:06 that you might get some additional substantial additional benefit if

43:09 you reduce the contacts in your elderly people even more

43:13 by some of these sort of shielding strategies,

43:15 and

43:15 we can perhaps discuss

43:17 how that might be possible in some settings more than others,

43:20 like a,

43:20 like a place with a household size of 7 is

43:22 clearly going to be more problematic to do something.

43:25 But um

43:26 yeah,

43:26 so,

43:27 so clearly these reductions are uh easy to

43:30 model in a,

43:31 they're easy to model,

43:32 but they're much harder to achieve in practice and

43:35 the solutions for that will vary by setting.

43:38 But what we do find is these have,

43:39 we have these big impacts.

43:40 So here the yellow bar shows what happened.

43:42 Nothing

43:43 and the.

43:43 Uh,

43:45 yes,

43:45 Patrick,

43:46 this is Roberto.

43:47 Sorry for being um

43:50 inter for interrupting you all the,

43:51 all the time.

43:52 So now everything is working on YouTube.

43:55 Uh,

43:55 but I'm getting like

43:57 a slew of messages on my phone from

44:00 colleagues and friends who are following on YouTube saying

44:03 we can't miss this is such an important presentation,

44:06 so I'm sorry,

44:06 but I have to ask you to go back

44:09 for slides.

44:10 I think that the problem started,

44:13 maybe you can go over.

44:16 A bit

44:17 more quickly on the

44:19 demographic and social structure,

44:20 but I would like you to essentially

44:23 start,

44:24 I think,

44:24 from here and then go over again

44:26 the perfect,

44:28 go over again the structure

44:29 in um

44:30 when I was going to school in Italy,

44:32 people used to say repetitta you want.

44:34 That two front,

44:35 which is,

44:35 you know,

44:36 repetition helps,

44:37 but it's a bit boring.

44:39 I'm sure

44:40 that,

44:40 um,

44:40 the audience will go back to bear with us for a couple of minutes.

44:43 This is not boring at all.

44:44 So thank you,

44:45 Patrick,

44:46 and thank you for all of those on YouTube for,

44:49 uh,

44:49 being so enthusiastic,

44:50 such an enthusiastic audience.

44:52 Over to you.

44:54 Thank you Roberto.

44:54 OK,

44:54 yeah,

44:55 so,

44:55 so yeah,

44:55 so,

44:56 so we,

44:56 we were mainly looking at how,

44:58 uh,

44:59 the.

45:00 How,

45:01 um,

45:01 some of the sort of guidances we give to high income countries,

45:04 how that might change and how our projections of

45:07 the spread,

45:08 likely spread of the pandemic may be altered in

45:10 different contexts and to see if we can give some

45:12 kind of broad level guidance.

45:15 Um,

45:15 so we identified five factors.

45:17 One,

45:17 and they're all I think fairly um intuitive.

45:21 So one is that um

45:24 is that

45:24 er

45:26 populations in lower income towns tend to be younger

45:28 and for er

45:29 for an infection like this that's more severe with age,

45:32 that clearly is likely to have an impact.

45:34 um

45:35 er

45:35 there also there's societal structure and how we

45:38 make contacts and that varies by settings,

45:39 so I'll talk about the next slide.

45:41 Comorbidities which have been focused on the.

45:45 Been identified in the countries

45:46 where the spread has been,

45:48 which is mainly

45:50 higher income settings,

45:50 and we'll come on to that.

45:52 And clearly there's disparities in health system

45:54 strength and capacity by income settings.

45:57 What we imagine,

45:58 and also the social and economic vulnerability of

46:00 societies

46:02 to both the direct impacts of

46:04 pandemic

46:05 and

46:05 any possible impact of the,

46:07 not some of the non-pharmaceutical intervention

46:10 Damien was talking about earlier.

46:12 So I've traveled quite quickly,

46:14 but

46:15 they are,

46:15 what we're finding is that lower income countries,

46:17 yes,

46:17 they are younger on average,

46:18 so there are fewer older people.

46:20 But we think those older people are gonna be

46:21 more vulnerable to having the infection spread to them,

46:25 because in low income settings like in the UK for example,

46:28 you're,

46:28 if you're an elderly person,

46:30 on average you share your house with about one other person.

46:33 You may even be by yourself and it's

46:34 outside of the pandemic window that's

46:36 seen as a problem because it's,

46:37 you know,

46:38 you,

46:38 you are socially isolated and that's

46:40 not a good thing.

46:42 In lower income settings,

46:43 the average household size of an older person is much higher.

46:47 It can be 6 or 7 in very low income countries.

46:50 Now clearly that that's going to change your contact patterns,

46:53 and it means that those contacts are sort of spread more evenly across the society,

46:57 particularly when

46:58 school attendance is maybe lower and less frequent,

47:02 so you don't get these sort of

47:04 large contacts of of younger people and work patterns are far more

47:08 spread out.

47:09 Um,

47:09 so we actually think that the contact patterns in

47:11 lower income settings are going to be far more flat

47:13 and higher income settings.

47:15 And,

47:15 and then,

47:16 so then what is this trade off in terms of

47:19 having a younger population,

47:20 but also where the older population are sort of more central in the society.

47:25 So,

47:25 um.

47:27 So that's the kind of question you can look at in a model whilst

47:30 making sure you

47:31 are very clear about the assumptions that you're gonna have to make.

47:34 So we did that looking at,

47:35 you know,

47:36 so what we do is we have a kind of a more simpler model framework,

47:40 but that incorporates most of the key elements

47:42 of the sort of more sophisticated micro simulations

47:45 that have been,

47:46 been provided to,

47:47 to the UK government and other governments.

47:50 So these are things like the

47:52 reproduction number,

47:53 the time it takes to,

47:55 for a person,

47:57 the epidemic to spread from one person to another,

47:58 known as the generation time.

48:00 And the,

48:01 all the delays between getting infected,

48:03 becoming sick,

48:04 chances of becoming severely ill,

48:05 the chances of dying,

48:07 and how long that takes and how that depends on age,

48:09 based upon the data that we have so far.

48:12 We can then

48:13 adjust this to the setting,

48:14 so we have the,

48:15 the,

48:15 the pathogen specific

48:17 parameters.

48:18 We can then take the setting specific parameters,

48:21 which are the demographics

48:22 and the the sort of representative context structure we think

48:25 within a country

48:26 and use that to sort of generate what we think might happen.

48:29 And

48:30 here,

48:30 next slide I'm going to show us what we

48:32 think would happen if nothing if nothing was done,

48:34 if we just let

48:35 the er

48:36 the the path that the um

48:38 pandemic just sweep through and didn't change our behavior or anything,

48:40 which clearly is not

48:42 what's happening and it's probably something that would never happen in reality,

48:44 it just gives us an idea of what we're dealing with.

48:47 Um,

48:49 what we would find is that,

48:51 as it has a reproduction number we're seeing in,

48:53 in 3,

48:54 that,

48:54 that

48:55 we know from the.

48:57 Epidemiological theory that that's going to lead to a

49:00 very high proportion of the population being affected.

49:03 By the end of the pandemic,

49:04 but actually if you look at the sort of the more severe aspects of infection,

49:07 which are age driven,

49:09 at least partly age driven,

49:11 hospitalization,

49:12 more,

49:13 more severe critical care,

49:14 um,

49:15 people needing critical care or

49:17 eventual,

49:18 sadly mortality,

49:19 we would,

49:20 we would um um anticipate that that would lead to a lower per person risk of death or.

49:27 In a lower income setting than a higher income setting,

49:30 but the problem here is that we make these assumptions,

49:33 that these parameters are

49:34 coming from

49:35 the early stage of the epidemic in Europe and China,

49:39 and those

49:39 settings have much higher healthcare capacity,

49:43 um,

49:43 than a lot of settings.

49:45 So clearly there's going to be that disparity that we

49:47 are going to have to try and account for,

49:48 and I'll get on to that later.

49:51 Um,

49:51 but what we did do before that is then think,

49:54 well,

49:55 what is the potential we have to mitigate the the

50:00 spread of this disease,

50:02 um.

50:03 And

50:04 first,

50:05 before we start thinking about,

50:06 you know,

50:06 closing,

50:07 locking down society,

50:08 if we suppress an infection,

50:11 what if we don't have the capacity to suppress infection?

50:14 What's the best that we could do?

50:15 Um,

50:16 and that's,

50:17 and it,

50:17 and it relies on this concept of herd immunity,

50:19 which has

50:20 become quite a sort of politicized term in the UK,

50:23 but it's actually not

50:24 any,

50:24 it's,

50:25 it's a just a scientific concept and it

50:27 comes from this idea that

50:28 as the disease is passing through,

50:30 eventually people will be requiring immunity,

50:32 and that will slow the disease down.

50:34 Um,

50:35 now,

50:35 if you lower your contact rates,

50:38 then that will,

50:39 um,

50:40 if you,

50:40 sorry,

50:40 if you don't lower your contact rates,

50:42 then actually once the level of,

50:44 uh,

50:44 immunity the population needs to prevent the outbreak taking off again,

50:48 once that's reached,

50:49 it won't just stop,

50:50 it will overshoot.

50:52 So you actually get a higher proportion

50:54 of the population affected than would be needed to get.

50:57 Herd immunity,

50:58 so lowering your contact rates

51:00 here about on average about 45%

51:02 on average,

51:04 we predict would

51:05 kind of halve the burden,

51:06 um,

51:07 not accounting for health healthcare strength,

51:10 um,

51:10 from something like 40 million deaths to 20 million deaths,

51:13 which is clearly like a a massive impact,

51:15 uh,

51:16 in terms of the burden this pandemic will cause.

51:19 But we did find that these,

51:20 these,

51:21 uh,

51:21 and with further impact if you can.

51:24 Further shield the most vulnerable,

51:26 so if,

51:26 if there are the extent to which you can

51:28 either cocoon older members

51:30 or protect them from whilst other people are,

51:33 you know,

51:33 from make their preferentially protect them.

51:36 And we could talk about how that might be more

51:38 difficult in lower income settings if you're sharing your house with

51:41 6 other people,

51:42 that might be harder than in

51:44 high income settings where that just means not leaving the house,

51:47 um.

51:48 And these are clearly quite easy things to model

51:50 in a

51:52 in a in a model,

51:53 you just change the parameter,

51:54 but in reality,

51:55 you know,

51:55 it's it's

51:56 how you,

51:57 they're best achieved,

51:58 it's probably like to vary by country.

52:00 And one of the things about these mitigation strategies

52:02 is because of these sort of flat contact patterns,

52:05 flatter contact patterns in lower income settings,

52:08 we expect them to have a lower proportional impact in

52:11 lower income settings because the disease can still spread fairly ably

52:15 to the to the older more vulnerable members of society.

52:19 So you know,

52:19 obviously

52:20 that's,

52:21 you know,

52:22 still a,

52:22 a fairly horrific burden caused by the pandemic.

52:26 Um,

52:26 but also there's,

52:27 it's,

52:27 it's gonna have a big,

52:29 a big strain on healthcare resources.

52:32 So this is just a map of what we

52:34 think the sort of under our optimal mitigation scenario,

52:37 the number of people per 1000 population

52:40 throughout the pandemic who would require a critical care bed,

52:43 in other words,

52:44 would,

52:44 would require ICU,

52:46 would require mechanical ventilation,

52:48 um.

52:49 And

52:50 that we are estimating would be higher on a per person level

52:54 of demand

52:55 in higher income settings with older populations.

52:59 But

52:59 then we need to think about how this demand for healthcare corresponds to supply.

53:04 Supply is clearly also going to be correlated with income.

53:07 This is work that was led by Charlie Whittaker,

53:09 who is one of Azra's PhD students,

53:11 um,

53:12 trying to update World Bank estimates to get more sort of fine grain estimates of

53:17 what the,

53:18 um.

53:19 What the current supply of

53:22 things like

53:23 the number of hospital beds per 1000 population

53:25 and within those hospital beds,

53:27 the proportion of those

53:29 that have the,

53:29 you know,

53:30 the capacity to treat the more severe

53:32 illness

53:33 in

53:34 more severe people with more severe consequences of infection.

53:38 And as we,

53:38 you know,

53:39 it's probably unsurprising,

53:40 there's a high correlation.

53:42 The

53:42 lower income setting,

53:44 the lower income of country,

53:45 the lower number of hospital beds and the lower number of physicians and nurses

53:49 they will have per 1000 population.

53:51 Not only are the number of hospital beds lower

53:53 of those hospital beds,

53:55 a smaller proportion of those used to

53:58 treat

53:58 severe cases,

54:00 um,

54:00 and actually in lots of lower income settings there's

54:03 near or.

54:04 We seem to be hearing a complete absence of these mechanical

54:08 ventilators which are currently saving a lot of lives.

54:12 And in these lower income settings,

54:14 these cases

54:15 aren't received this mechanical ventilation are

54:17 likely to experience a very high risk of mortality.

54:22 The absence of mechanical ventilation.

54:25 So then what we did is we took our sort of our scenarios of healthcare demand,

54:30 and we integrated them with our estimates of healthcare supply,

54:33 in terms of

54:34 both the capacity for healthcare,

54:36 er,

54:36 accounted for how long you would need a bed,

54:38 and you would need an ICU bed,

54:40 and the impact that healthcare has upon treatment.

54:43 Now I must say that we are still working with experts,

54:45 we're

54:46 like uh

54:47 uh Robert was saying,

54:48 we're,

54:48 you know,

54:48 we're we're all trying to catch up on this,

54:50 but

54:51 uh the experts to try and understand what some of these parameters are.

54:55 So

54:55 what the output I'm showing here is a bit more provisional

54:58 in terms of the quantitative,

55:00 um.

55:02 The qualitative results will be very robust.

55:05 What we on the top right is a plot showing

55:09 what we think you would need the

55:12 sort of the ideal ICU bed capacity,

55:15 100,000 in an unmitigated.

55:17 Scenario and that's the dashed

55:19 line

55:20 um

55:20 and the

55:22 the dashed blue line and then the dotted

55:24 um

55:25 the dotted gray line shows

55:26 what the typical

55:29 um ICU bed capacity is within an upper middle income.

55:33 In a low middle income or a low income country,

55:37 that should be much lower,

55:38 um,

55:40 er and what that,

55:41 what that will mean in terms of your,

55:43 uh,

55:44 the sort of,

55:45 The burden of these.

55:49 The pandemic in lower income settings is

55:51 that you're gonna see this big bump

55:54 in what I'm here calling preventable,

55:56 um burden.

55:57 I,

55:57 the nomenclature I think is something we're still working on.

56:00 But um,

56:00 but basically there's gonna be a lot of people who,

56:03 uh,

56:04 would

56:04 survive in a high income setting that won't survive in a low income.

56:08 For example,

56:09 uh we uh our Prime Minister is I think low to mid-fifties.

56:13 ICU

56:14 wouldn't be getting that ICU in a low income setting,

56:16 so

56:17 we'd be far more likely to have not survived,

56:19 so we think we think that will both push

56:22 the

56:23 per capita burden up

56:25 in these settings and also reduce the age at

56:28 which people are more likely to die from it.

56:35 There's also this this,

56:36 this question about comorbidities.

56:39 Now a lot of the comorbidities that have been identified are,

56:42 are clearly very important,

56:43 but they're also ones that are most prevalent in the,

56:46 the places where the,

56:47 the disease has spread,

56:48 and those are the more high income settings.

56:50 Um,

56:50 and what,

56:51 what we've been doing is looking at how that varies also by income strata.

56:55 Now,

56:56 it,

56:56 for some of the most

56:58 uh identified ones so far like uh coronary um like CBD.

57:03 Actually what you see is that there is a

57:04 tends to be lower prevalence and lower income settings.

57:07 Part of that might be lifestyle,

57:09 uh,

57:10 but another part is probably survivorship.

57:13 As you see in the ICU plot,

57:14 there are,

57:15 there are probably gonna be very few things

57:16 like heart bypasses that are made available.

57:19 So actually your survivorship with these comorbidities is probably a lot lower.

57:23 So

57:23 that might alter the severity by age,

57:26 but also there are other,

57:27 um,

57:28 and also those,

57:29 what we have,

57:30 the,

57:30 the commodates we found are very correlated with age,

57:33 so it gives you a good target for who to

57:35 shield,

57:36 and these are the older people,

57:37 they're also most likely to have

57:39 er the.

57:41 Um,

57:42 comorbidities.

57:43 So,

57:44 but in lower income settings there are a lot more potential comorbidities,

57:46 and we just don't know what impact they're gonna have.

57:49 Things like HIV infectious disease,

57:51 HIV AIDS,

57:52 uh,

57:52 I have particularly bias on malaria which I work on most of the time outside of

57:56 pandemic,

57:57 um,

57:57 other things like general things like malnutrition,

58:00 and these are far less correlated with age,

58:02 far more in younger age groups,

58:04 far more.

58:04 and

58:06 defining what a relevant comorbidity I think is is going to take some time

58:10 and understanding who is vulnerable and who needs to be protected the most.

58:14 It's gonna be a key thing to work out to mitigate

58:17 the onward impact of the pandemic as we go forward.

58:22 Yeah,

58:22 so,

58:23 Thinking about

58:25 what you can do here,

58:26 we've we've presented the mitigation scenarios,

58:28 and these mitigation scenarios are,

58:30 are things that we think would and have already

58:34 shown that they will strain healthcare systems

58:37 beyond capacity if

58:38 without suppression strategies in

58:41 Europe.

58:42 We also think,

58:42 given that that that will also be the case many times over in lower income settings,

58:47 um.

58:48 And that's leaving,

58:49 you know,

58:49 countries we anticipate with

58:51 to

58:52 a kind of a real big problem that doesn't really have an easy solution.

58:55 Either they can

58:57 reduce and go to suppress the disease,

59:00 suppress the disease and,

59:01 and hopefully their health system will not be overwhelmed,

59:03 but that has to be maintained until you get an exit strategy.

59:07 And an exit strategy might be a vaccine,

59:09 it it might be some,

59:11 you know,

59:11 high technology test or treat scenario,

59:13 but these are scenarios that

59:15 are probably less likely to be available

59:17 in lower income settings,

59:19 unfortunately.

59:20 Um,

59:21 and,

59:22 uh,

59:22 and one of the things is if,

59:23 if you suppress,

59:24 and then you go back to doing nothing,

59:26 um,

59:27 because of these sort of immunity patterns,

59:29 there won't be any build up of immunity.

59:30 You do risk

59:32 a second wave,

59:33 uh,

59:33 where,

59:34 uh,

59:34 eventually the health system will be overwhelmed,

59:36 but just at a,

59:37 at a later date.

59:39 So,

59:40 so it's a really hard trade-off,

59:41 um,

59:42 and

59:43 maybe sup suppressing for a short amount of

59:45 time and then going straight back to normal

59:47 is,

59:47 will,

59:47 is the the the type of scenario it's most likely to.

59:50 Like,

59:52 um.

59:53 Which would then risk

59:54 having similar impact as if you hadn't suppressed in the first place,

59:58 and might have a limited impact.

1:00:01 Patrick,

1:00:02 this is Roberta.

1:00:03 I just wanted to give you the warning that um we are

1:00:06 at the end of your presentation.

1:00:09 Because of

1:00:10 the

1:00:11 uh technological glitch,

1:00:14 I would like to ask you to

1:00:16 first speak very close to the microphone.

1:00:19 Second,

1:00:20 to

1:00:20 uh kindly summarize

1:00:23 the key hypothesis in a very sort of bullet point type of way.

1:00:27 And the conclusion in terms of

1:00:30 what,

1:00:30 uh,

1:00:31 these different suppression measures,

1:00:33 uh,

1:00:34 can deliver,

1:00:35 particular with particular,

1:00:37 um,

1:00:37 emphasis on low-income countries,

1:00:39 um.

1:00:41 And then,

1:00:41 um,

1:00:42 for all of those who have have been patient to hang with us on YouTube,

1:00:47 please rest assured that we will share

1:00:49 the recording of the event,

1:00:51 uh,

1:00:52 the,

1:00:52 uh,

1:00:52 slides.

1:00:53 I asked actually,

1:00:55 uh,

1:00:55 the technical team to see whether it's possible to share them already now,

1:00:59 uh,

1:00:59 so that you will have all of the material.

1:01:01 So before we move on to the next,

1:01:03 uh,

1:01:04 speaker,

1:01:04 the speakers,

1:01:05 David Wilson and Mali Jorgensen.

1:01:07 Uh,

1:01:08 please,

1:01:08 um,

1:01:08 Patrick,

1:01:09 summarize the main point,

1:01:11 um,

1:01:11 of your talk.

1:01:12 Thanks so much.

1:01:13 Yeah,

1:01:13 no problem.

1:01:14 Well,

1:01:14 I think,

1:01:14 I think basically what you've just said is kind of

1:01:16 the last point that we really had to make.

1:01:19 It's just that,

1:01:19 that the consequences of these suppression strategies need to be

1:01:23 thought through,

1:01:23 and they're gonna also be

1:01:25 particularly,

1:01:26 uh,

1:01:26 you know,

1:01:28 hard to,

1:01:29 they're gonna have impacts that are indirect and

1:01:32 loweringham saying,

1:01:32 so here this is just showing that.

1:01:34 There are a lot of countries that have much lower food security,

1:01:38 and

1:01:38 by suppressing or mitigating,

1:01:40 we could delay or

1:01:41 or extend the peak,

1:01:43 and that then can help to,

1:01:45 you know,

1:01:45 that could lead to disruption around things

1:01:47 like harvest seasons and things like that

1:01:50 that really need to be thought through and to balance out

1:01:53 both the indirect

1:01:54 direct effect of both the pandemic and and the interventions.

1:01:58 So,

1:01:59 yeah,

1:01:59 so,

1:02:00 so I'll just go through the last bit,

1:02:01 but,

1:02:01 but

1:02:02 quickly,

1:02:02 but yeah,

1:02:03 basically lots of countries are at the early stage.

1:02:06 We anticipate that they're not going to

1:02:08 stay that way.

1:02:09 um

1:02:10 we can hope that they won't,

1:02:11 but I think it's prudent to plan that they will.

1:02:13 The epidetitis will take overwhere.

1:02:14 We have got some tools that you can look at coming

1:02:17 in the next few days that I won't get onto but

1:02:19 hopefully go through.

1:02:21 But these are just our conclusions.

1:02:23 So um

1:02:25 yeah,

1:02:25 so these tools are available on our website and

1:02:28 follow,

1:02:28 follow the link and hopefully that will.

1:02:30 Give you,

1:02:31 you know,

1:02:31 what you need,

1:02:32 um,

1:02:33 but yeah,

1:02:33 so,

1:02:34 so a lot of lower and middle income countries have yet to see the mortality,

1:02:37 level of mortality that we're seeing in Europe.

1:02:39 That's probably just we think because they're primarily

1:02:41 at an early stage of the epidemic.

1:02:43 Um,

1:02:44 and we think that that all countries are very high risk of large

1:02:47 scale epidemics that are likely to

1:02:48 overstretch and probably overwhelm health systems,

1:02:52 um,

1:02:53 and that will lead to significant excessor morbidity.

1:02:57 Low middle income

1:02:58 countries,

1:02:59 the economies of them are far more

1:03:00 vulnerable to the indirect effects of interventions,

1:03:03 uh,

1:03:04 and shielding and things like that are things that

1:03:06 are gonna need to be very actively considered.

1:03:08 There's no quick,

1:03:09 easy solutions,

1:03:10 but we are gonna need to respond rapidly because I think it is going to be the next

1:03:14 2 or 3 months that will really determine

1:03:16 what happens in this pandemic.

1:03:18 I'll just add some acknowledgements,

1:03:19 but I wouldn't have had enough time to go through everyone

1:03:22 anyway,

1:03:22 we've had so much support from so many partners that

1:03:25 I didn't have time to put them all on the slide.

1:03:27 Sorry for going over time.

1:03:30 Thank you so much,

1:03:30 Patrick.

1:03:31 Uh,

1:03:31 Professor Ganni,

1:03:32 would you want to add a few uh quick

1:03:35 final conclusions that would hit

1:03:38 the intuitive mechanisms

1:03:40 of how you got these estimates,

1:03:41 and this will be also for the benefit of those

1:03:43 who didn't have the chance to see the whole presentation through

1:03:46 on YouTube.

1:03:47 Thank you.

1:03:50 OK,

1:03:50 you have a few

1:03:52 hopes.

1:03:56 Hang on,

1:03:56 um,

1:03:57 just a few concluding,

1:03:58 um,

1:03:58 remarks really.

1:03:59 Um,

1:04:00 clearly modeling's really come to the forefront during this pandemic and people

1:04:03 will have seen these models and many of them out there.

1:04:07 I,

1:04:07 I think the main important thing to bear in mind is models themselves

1:04:11 are a tool to improve our understanding.

1:04:14 Much of what we do and have done for many

1:04:16 years has been to spend a lot of time collating.

1:04:19 Um,

1:04:19 good quality epidemiological and clinical data

1:04:22 in order to parameterize the models,

1:04:24 and I think that's a very important part.

1:04:27 Um,

1:04:27 so much of the projections we're making,

1:04:29 they will change over time and they will change because we will learn more

1:04:32 about the

1:04:33 disease and the epidemic as it progresses.

1:04:36 Um,

1:04:37 so I think I just wanted to stop

1:04:39 with a few words about where we are now because,

1:04:41 uh,

1:04:41 many countries across the world,

1:04:43 of course,

1:04:43 have

1:04:44 instigated

1:04:45 quite,

1:04:45 uh,

1:04:46 severe social distancing

1:04:48 measures,

1:04:49 and we,

1:04:50 many of us are really faced with the challenge of how do we get out of this?

1:04:53 What is the appropriate exit strategy?

1:04:56 Um,

1:04:56 I,

1:04:56 there are no simple answers to this,

1:04:58 and I don't think anybody in the world really has them,

1:05:00 and I think

1:05:01 the models are useful for in terms of illustrating

1:05:04 where we could be and what

1:05:06 the challenges are with maintaining this level of suppression because

1:05:10 without some

1:05:11 other tool,

1:05:12 um,

1:05:12 in particular a vaccine,

1:05:14 it's very unlikely given the high basic reproduction number of this virus

1:05:18 that we can lift these tools,

1:05:20 um,

1:05:20 and expect to go back to normal.

1:05:22 Um,

1:05:23 and when we think about this in a,

1:05:25 a lower income countries's context,

1:05:28 I think it is important to think now,

1:05:30 um,

1:05:30 before jumping

1:05:31 into exactly the same,

1:05:33 um,

1:05:34 patterns or strategies that have been adopted in high income countries.

1:05:37 There are many.

1:05:38 Downsides

1:05:39 um to some of these suppression strategies,

1:05:41 not least the wider impacts on the health and wellbeing of societies.

1:05:45 So this is a difficult

1:05:46 um

1:05:47 area to address um and will require a wide range of specialists to think

1:05:51 more about in the coming weeks.

1:05:54 OK.

1:05:57 Thank you so much,

1:05:58 Patrick and Asra,

1:05:59 and now

1:06:00 it's the moment to turn to

1:06:02 uh David Wilson,

1:06:03 who's a

1:06:04 a director uh in uh the health,

1:06:08 nutrition and population practice of the World Bank,

1:06:10 and

1:06:11 uh our colleague Marreli Gordon,

1:06:13 who's a senior monitoring and evaluation specialist in the same global practice,

1:06:17 and that will tell us.

1:06:19 They will share reflections exactly on

1:06:22 exit strategies.

1:06:23 So this is a tough question,

1:06:25 and I appreciate

1:06:27 the humbleness of the humility of saying we don't have

1:06:31 the answers and we need to think through in advance

1:06:34 whether these tools that have been applied in rich in advanced countries

1:06:39 can be applied.

1:06:41 Successfully,

1:06:42 incredibly

1:06:43 in low income settings.

1:06:45 So thank you for now,

1:06:47 and David and Marelli,

1:06:48 the floor is yours.

1:06:49 I will have to ask you to be as concise as possible because we

1:06:52 would like to leave some time for questions before we end at 10:30.

1:06:57 A reminder for those on YouTube,

1:06:59 you can ask questions,

1:07:00 and I know that you've been shortchanged.

1:07:03 by the technology Glitch and those on Webex,

1:07:06 please ask your questions in the chat or raise your hands.

1:07:09 Thank you.

1:07:10 Over to David.

1:07:15 I think that

1:07:16 my comments flow nicely from what's been presented and I hope I

1:07:20 can compress it to about 5 minutes and claw back time for discussion

1:07:24 because I'm building on what's been said.

1:07:26 We've tried to look at some considerations for how countries could reopen,

1:07:31 and we've tried to do that with as much of

1:07:33 an orientation as possible towards lower and middle income countries because

1:07:38 so much of the literature and experience is the

1:07:42 Northern European or East Asian and has some but limited applicability,

1:07:48 and we've really identified about 7 considerations,

1:07:51 many of which have been touched on.

1:07:53 The concept of epidemic force,

1:07:56 obviously we don't see you,

1:07:58 um,

1:07:59 this might be by design,

1:07:59 you might have chosen not to show yourself,

1:08:01 but I think you can see me now now it's so much,

1:08:03 now it's so much better.

1:08:04 Thank you.

1:08:06 So we have about 7 considerations which have been touched on.

1:08:09 The notion of epidemic force,

1:08:11 obviously primarily reproduction rate,

1:08:13 but also looking at super spread of potential and also looking at the extent to which

1:08:17 overall spread is focused as it has been,

1:08:20 for example,

1:08:21 in China,

1:08:22 Italy,

1:08:23 and the US or dispersed

1:08:26 and based on those considerations,

1:08:27 it may be possible to remove lockdowns earlier or later

1:08:32 and partially or more completely.

1:08:35 We've also,

1:08:36 as people have touched on already,

1:08:37 looked closely at population health.

1:08:39 Obviously age,

1:08:41 and obviously we think that Africa's comparative youth,

1:08:44 Uganda's average age is 16.7,

1:08:46 Italy is 47,

1:08:48 Lombardy's,

1:08:48 as you know,

1:08:49 Roberta's even older.

1:08:51 We've also looked at issues such as obesity,

1:08:54 NCDs,

1:08:55 uh,

1:08:55 smoking,

1:08:56 and other risk factors,

1:08:57 which obviously are positive for Africa,

1:08:59 but might be offset by exposure to indoor cooking pollution,

1:09:03 by widespread immunocompromisation,

1:09:06 and also,

1:09:07 um,

1:09:07 potentially by malnutrition and other factors.

1:09:10 So the picture's mixed,

1:09:11 but perhaps a little bit positive.

1:09:13 We've looked closely at what we would call the capacity to take targeted measures,

1:09:17 which is difficult,

1:09:18 but it does depend a lot on countries' epidemic experience,

1:09:22 population preparedness,

1:09:23 and population.

1:09:25 Willingness to take measures

1:09:26 and I think we see that throughout countries in

1:09:29 East Asia which have managed this extremely well,

1:09:31 Japan,

1:09:32 uh,

1:09:32 Taiwan,

1:09:33 for example,

1:09:34 South Korea,

1:09:34 Hong Kong,

1:09:35 Singapore,

1:09:36 parts of China,

1:09:37 and

1:09:38 we've looked closely at

1:09:40 the concept of public health capacity,

1:09:43 which we all understand and is going to be

1:09:44 central to manage our way out of this epidemic.

1:09:47 We've had difficulty looking at health service

1:09:49 capacity because ideally we would like to have

1:09:53 both fully protected health workers,

1:09:55 safe health facilities,

1:09:57 and

1:09:58 an adequate surplus of critical care capacity.

1:10:02 The challenge we face is defining what critical care capacity

1:10:05 means in very low resource countries such as Madagascar,

1:10:08 for example,

1:10:09 who have an estimated 7 to 12 hospital beds.

1:10:12 We focus a lot on state capacity,

1:10:15 particularly decision making capacity,

1:10:17 and we do see that higher capacity states have

1:10:20 managed their way out of this better so far.

1:10:22 But we also think the role of decision making is critical.

1:10:25 We've had to elevate decision making to the highest levels to succeed.

1:10:29 We've had to be able to take data from

1:10:31 real-time digital sources,

1:10:33 from

1:10:33 modeling,

1:10:34 from economics,

1:10:35 and make very complex decisions,

1:10:37 which are probably a function of state capacity.

1:10:40 And then we've also looked at the extent to which innovations would help,

1:10:43 and the innovations that are closest to

1:10:44 hand are probably incremental innovations and testing,

1:10:48 which is still really important.

1:10:50 Potential innovations of a simplified treatment,

1:10:53 not yet at hand,

1:10:54 but potentially not too,

1:10:55 too far away,

1:10:56 and then the obvious hope of vaccines.

1:10:59 Based on these considerations,

1:11:01 we looked at the literature and found about 6 papers on how we reopen.

1:11:06 All of them are

1:11:07 for high income contexts,

1:11:09 but they still have some useful principles.

1:11:12 And

1:11:13 we then really took a consensus for principles for how you might

1:11:18 prepare a roadmap,

1:11:19 tried to see how we could tailor it for southern context,

1:11:22 and then tried to apply it to categories,

1:11:23 which I'll discuss.

1:11:25 And

1:11:26 The obvious point is if surveillance is strong

1:11:29 and cases are either very low or falling,

1:11:32 then

1:11:33 Everything else becomes easier and we can

1:11:35 move towards relaxing measures or reopening much,

1:11:39 much more quickly.

1:11:41 So everything really hinges on this,

1:11:43 and for much of the world we simply don't have the data.

1:11:46 One striking point for the World Bank is that only

1:11:49 2% of all reported cases are from IDA countries,

1:11:53 and I suspect that IDA countries do have fewer cases,

1:11:56 but I'm certain they have much fewer testing.

1:11:59 We've also focused on the role of microdata because we think

1:12:02 managing our way out of this epidemic is going to mean

1:12:05 having surveillance,

1:12:06 having testing,

1:12:07 having

1:12:08 public health case finding,

1:12:09 tracing,

1:12:10 isolation,

1:12:11 quarantine measures

1:12:13 at the local level,

1:12:15 and

1:12:15 obviously we focus on the extent to which there's adequate public health capacity

1:12:20 to try and eliminate linked cases as one criteria for countries to open.

1:12:26 What we grappled most with,

1:12:27 as I've said,

1:12:28 is health service capacity,

1:12:29 but we've tried to focus primarily on health workers being

1:12:33 completely protected and health facilities being safe for patients.

1:12:37 We're still grappling with what critical care means.

1:12:40 And based on these considerations,

1:12:43 we've tested a very simple notion of countries that

1:12:46 may already be open or could be opened very quickly

1:12:51 with limited intermittent or targeted measures.

1:12:54 We've also tried to look at places that might be able

1:12:57 to reopen in the near future with the same limited measures,

1:13:00 and in places that may actually need a lot more information or a lot more progress.

1:13:05 And this is entirely to test a concept.

1:13:08 It's in no way to provide national advice or to contradict national advice,

1:13:13 but if we're to look at countries that are

1:13:15 already open or might open with limited measures.

1:13:19 We actually have

1:13:20 a lot of countries in East Asia already in this category,

1:13:23 um,

1:13:24 countries and territories such as China,

1:13:25 Hong Kong,

1:13:26 Taiwan,

1:13:26 Singapore,

1:13:27 South Korea,

1:13:28 Japan,

1:13:28 Vietnam,

1:13:29 Australia,

1:13:29 New Zealand.

1:13:30 We think there are a cluster of countries in Europe that meet those criteria,

1:13:34 Germany and Austria who are moving towards this already,

1:13:37 the Czech Republic,

1:13:38 Slovakia,

1:13:39 and Iceland,

1:13:39 for example,

1:13:41 uh.

1:13:41 When we look at the next category of countries that

1:13:44 actually could reopen quite soon with limited and targeted measures,

1:13:49 we see quite a lot of islands who have an extra

1:13:51 ability to insulate themselves from the wider world and reopen internally.

1:13:55 And we think that,

1:13:56 uh,

1:13:57 Sri Lanka,

1:13:57 for example,

1:13:58 is a good example of a

1:14:00 lower middle income country that could reopen substantially very soon,

1:14:04 as is the smaller and more isolated island of Mauritius,

1:14:08 and potentially also promisingly the very

1:14:10 low income isolated island of Madagascar.

1:14:13 We also think there are 20 or 30 islands in the

1:14:15 Pacific who've locked down and if testing confirms low rates,

1:14:19 could potentially rely on external circuit breakers.

1:14:22 But yesterday we got a very encouraging presentation from

1:14:26 South Africa's CapriSA modeling Group given to cabinet,

1:14:30 which actually suggested South Africa is now plateauing and peaking

1:14:33 at an earlier level than expected and it's diverging from

1:14:37 the pessimistic European scenarios that they had initially used,

1:14:41 and they're showing limited community transmission.

1:14:44 Now,

1:14:44 if this holds,

1:14:45 South Africa may be an example of a continental

1:14:48 African country that could actually progressively reopen now.

1:14:52 And just to emphasize,

1:14:53 because of its very large installed TB program

1:14:56 and gene expert capacity adapted with CAF cartridges,

1:15:00 um,

1:15:00 South Africa's tested on a large scale,

1:15:02 and other African countries are also expanding testing,

1:15:05 and some of the glimmers of data we're getting are moderately positive.

1:15:09 But then finally,

1:15:10 there's a whole swathe of the world for which we

1:15:13 need much more testing data and much more progress.

1:15:16 And I would highlight our concerns about

1:15:19 much of South Asia,

1:15:21 except Sri Lanka,

1:15:22 the large populous countries of Southeast Asia,

1:15:25 and in particular Indonesia and Philippines.

1:15:29 Much of Africa for which we have very little data and much of South America.

1:15:33 And just to note that if there's a seasonal dimension,

1:15:36 winter is approaching in

1:15:38 Southern Africa and South America.

1:15:40 And

1:15:41 this is where I'm gonna really conclude by saying,

1:15:44 I think that we started in a very different place to colleagues at DEC,

1:15:48 but I think we're converging

1:15:50 because

1:15:51 the obvious point.

1:15:53 Preceding any decisions about

1:15:55 reopening is expanded testing,

1:15:57 a point that was highlighted in all of the presentations,

1:16:00 and I think if we can do that,

1:16:02 we'll have a much better idea of how we can consider reopening,

1:16:06 but I do think we're looking at very partial,

1:16:09 very intermittent,

1:16:10 uh,

1:16:11 openings with a lot of internal circuit breakers.

1:16:14 I think circuit breakers and fire stops are going to

1:16:16 be the ways that countries can try to cautiously reopen.

1:16:20 I can't underscore the stakes here.

1:16:23 Any

1:16:23 of those of us who've seen the remarkable pictures of large scale

1:16:28 protests that verge on riots coming from Mumbai

1:16:30 today will know how explosive the situation is.

1:16:33 Let me conclude there and hopefully claw back a bit of time.

1:16:37 Thank you and over.

1:16:43 Thank you,

1:16:43 David,

1:16:44 uh,

1:16:44 is Marielisa coming on after you,

1:16:46 or have you spoken for the both of you?

1:16:49 I hope we'll be online.

1:16:51 She's,

1:16:51 um,

1:16:52 presenting the Lesotho

1:16:54 project today and was in a meeting with the country,

1:16:57 so we had to shuffle our order.

1:16:59 So can I just ask to ask if Marilli made it?

1:17:02 Um,

1:17:02 otherwise,

1:17:03 Roberto,

1:17:04 are you online,

1:17:04 Marilli?

1:17:05 I am.

1:17:05 Thanks,

1:17:06 David.

1:17:06 Over to you.

1:17:07 Thank you.

1:17:07 Excellent.

1:17:08 Thank you.

1:17:08 Thanks,

1:17:08 David,

1:17:09 for the overview and to Patrick and Azra.

1:17:11 Um,

1:17:11 I know we've had many email discussions.

1:17:13 It's,

1:17:13 it's good to be connected in person,

1:17:15 hear your voices as well.

1:17:17 Um,

1:17:18 so,

1:17:18 a few points on my end about the use of mathematical modeling in this,

1:17:22 in this space,

1:17:23 maybe just to

1:17:24 tie the two presentations together.

1:17:27 Um,

1:17:27 first,

1:17:28 to say that I think we'd all agree that mathematical modeling,

1:17:31 both epidemiological and predictive analytics,

1:17:34 um,

1:17:34 have proven essential in raising awareness of the scope of the epidemic

1:17:38 to countries that have gone through it to help

1:17:40 them prepare for it and plan for shortages.

1:17:42 Um,

1:17:43 This is,

1:17:44 you know,

1:17:44 there are many things that are unprecedented,

1:17:46 but

1:17:46 the unprecedented use of models,

1:17:48 the amount of times that

1:17:50 I've seen models being discussed and model parameters being discussed

1:17:54 in,

1:17:54 in the media,

1:17:55 we,

1:17:55 we've just never,

1:17:56 never seen this before.

1:17:58 So,

1:17:58 I,

1:17:58 I think,

1:17:59 you know,

1:17:59 I would say that

1:18:01 models obviously play an important role,

1:18:04 and at the same time,

1:18:05 and I think this is a healthy thing,

1:18:06 they have received

1:18:07 scrutiny also that maybe,

1:18:09 um,

1:18:10 you know,

1:18:10 epidemiologists and infectious disease modelers have also not,

1:18:13 not seen,

1:18:14 seen before.

1:18:15 I think one of the particular areas of scrutiny has been around the use,

1:18:18 usefulness,

1:18:19 and interpretation

1:18:20 of statistical models versus

1:18:23 these um more mechanistic SIR um Infectious disease models and clearly

1:18:28 the questions um raised in the chat today reflect,

1:18:32 reflect that or not.

1:18:33 I think it is important in all of this

1:18:34 that we recognize that there are many unanswered epidemiological questions

1:18:39 that we need to,

1:18:41 um,

1:18:42 still answer

1:18:43 and that the,

1:18:44 these will be impacted on by the models and vice versa.

1:18:47 Um,

1:18:48 the asymptomatic and pre-symptomatic cases

1:18:51 proportions in the population.

1:18:53 Um,

1:18:54 duration of infectiousness,

1:18:55 duration of infection,

1:18:57 um,

1:18:58 and duration of infectiousness after infection is cleared,

1:19:01 level of immunity,

1:19:02 duration of immunity,

1:19:03 viral load,

1:19:04 and its relationship to both infectiousness and,

1:19:07 and immunity,

1:19:08 um,

1:19:08 acquired long-term immunity.

1:19:11 Um,

1:19:11 and then most importantly,

1:19:12 as David alluded to,

1:19:14 the effect of the in interventions,

1:19:16 not just nationally

1:19:17 applied interventions,

1:19:18 but granular,

1:19:19 really detailed,

1:19:21 really detailed interventions.

1:19:24 Whilst all of these,

1:19:25 there's all of this uncertainty that we have to account for,

1:19:27 policymakers have to make decisions and revisit them frequently,

1:19:31 deciding when to

1:19:33 um

1:19:34 have stay at home orders,

1:19:35 when to release them,

1:19:36 how to release them,

1:19:37 in which phases for which populations,

1:19:39 in which areas,

1:19:40 which geographic areas they can be released for which kinds of industries.

1:19:44 And

1:19:45 so,

1:19:45 these models are going to be essential for

1:19:47 this next stage of response management as well.

1:19:50 To do that well,

1:19:52 they need to be granular enough and they need to be adaptable.

1:19:55 That means

1:19:56 that

1:19:57 Whilst model choice is important,

1:19:59 what is even more important is adapting,

1:20:02 adapting,

1:20:02 adapting,

1:20:03 recalibrating these models

1:20:05 against the data that are available.

1:20:07 Um,

1:20:08 we know,

1:20:08 for example,

1:20:09 some of the model groups does this twice a week and have a schedule

1:20:12 that is published to say we will be Iterating

1:20:15 constantly,

1:20:15 um,

1:20:16 because making decisions and deciding on,

1:20:19 on Rn effective,

1:20:20 effective reproductive rates,

1:20:21 etc.

1:20:22 that is

1:20:23 affected by the intervention itself

1:20:25 is,

1:20:25 is

1:20:26 critical in,

1:20:27 in what the entire,

1:20:28 the entire model,

1:20:29 um,

1:20:29 looks like.

1:20:30 Um.

1:20:31 So,

1:20:31 whilst modeling has been useful upfront

1:20:34 in estimating the overall epidemic potential,

1:20:36 estimating,

1:20:37 therefore,

1:20:37 the

1:20:37 healthcare needs,

1:20:39 um,

1:20:40 the

1:20:41 The important need is now shifting to assessing the potential magnitude

1:20:45 of the different measures,

1:20:46 and these are measures that are going to be

1:20:48 implemented,

1:20:49 not implemented,

1:20:50 implemented,

1:20:50 not,

1:20:51 not implemented,

1:20:52 right?

1:20:52 So,

1:20:53 models

1:20:53 can and should be essential to answer a wider range of efforts,

1:20:57 a wider range of questions.

1:20:59 In that,

1:21:00 this is Roberta.

1:21:01 Let me just give you,

1:21:02 unfortunately one.

1:21:03 Minute warning because we have run

1:21:05 vastly over time today.

1:21:07 Thank you.

1:21:07 Absolutely.

1:21:08 So there's 3 things the bank is doing to help with this.

1:21:10 First of all,

1:21:10 we're co-convening with IDSI,

1:21:13 WHO and the Gates Foundation,

1:21:14 a global COVID modeling,

1:21:16 um,

1:21:17 rapid assessment effort

1:21:18 to help countries understand which

1:21:21 parameters go in and how they can better use models.

1:21:24 Secondly,

1:21:24 is we're developing

1:21:25 Social distancing and stay at home impact dashboard for every

1:21:29 country that will be available on the World Bank website.

1:21:32 And thirdly,

1:21:33 a vulnerability dashboard

1:21:35 that will help countries to understand at a granular

1:21:38 local level where are the areas of highest,

1:21:40 highest vulnerability,

1:21:41 and we hope that this will be useful

1:21:43 and impact on these models themselves.

1:21:46 Over.

1:21:47 Thanks,

1:21:47 Roberta.

1:21:51 Thank you,

1:21:51 Mali,

1:21:52 and

1:21:53 thank you everyone.

1:21:54 It's a moment,

1:21:54 we have about 8 minutes for

1:21:57 Q&A sessions with our speakers.

1:21:59 Just by way of summary,

1:22:02 uh,

1:22:02 Damien and Jed gave us a nice

1:22:06 overview of the key concepts

1:22:08 uh related to the epidemiology of uh COVID-19 and

1:22:13 an overview of the mathematical models that are out

1:22:16 there that can help us make these predictions.

1:22:19 Um,

1:22:20 Patrick and Azra

1:22:22 presented

1:22:24 the predictions of their model,

1:22:26 and just to give you a quick

1:22:27 snapshot of their conclusion,

1:22:29 they show that

1:22:31 mitigation strategy that

1:22:34 is based on partially shielding the elderly,

1:22:36 every year 60% of reduction in their social contacts and slowing down

1:22:42 without completely interrupting transmission in the general population,

1:22:45 so with a 40% reduction.

1:22:47 Would have the burden of death in the world from 40 million to 20 million.

1:22:52 They find out,

1:22:53 however,

1:22:53 that in low income countries,

1:22:55 the impacts are likely to be much stronger,

1:22:58 and a data point that I see here in their paper

1:23:02 is that in a typical low income setting,

1:23:06 the demand

1:23:07 of critical care beds would outstrip the supply by a factor of 25,

1:23:13 while this factor is only 7 for high income countries.

1:23:17 Their conclusion was that there are no easy answers,

1:23:20 and then essentially we should,

1:23:22 uh,

1:23:22 buckle up for,

1:23:23 uh,

1:23:24 long periods of measures of containment if we don't want that,

1:23:27 that the virus comes back.

1:23:30 Um,

1:23:30 David and Marlise,

1:23:31 uh,

1:23:32 have reflected on,

1:23:33 uh,

1:23:34 a sort of matrix of,

1:23:35 uh,

1:23:36 the intersection of country capacity and

1:23:40 possibility and options for.

1:23:41 For reopening.

1:23:43 Clearly this depends on

1:23:45 the epidemiology and type of spread of the virus because one

1:23:49 of their conclusions was that islands that are able to maintain

1:23:53 the spread low within

1:23:56 the country could reopen relatively

1:23:59 sooner

1:24:00 internally while still maintaining their closure of the border to avoid

1:24:05 new infections.

1:24:06 Let me start

1:24:08 the Q&A session with a question for,

1:24:10 from myself,

1:24:12 uh,

1:24:12 which is also related to one of the questions that we got

1:24:15 via YouTube from George.

1:24:18 So,

1:24:19 The question from,

1:24:21 um,

1:24:21 I believe,

1:24:22 a question to Patrick was,

1:24:24 does your model assume long-term immunity,

1:24:27 and does it consider the slim possibility that the mortality rate

1:24:31 might be overestimated because of that?

1:24:34 My related question was,

1:24:37 uh,

1:24:37 as epidemiologists,

1:24:41 What do we,

1:24:42 what do you understand about the acquired immunity to COVID once,

1:24:47 uh,

1:24:47 somebody has contracted the disease and then becomes negative?

1:24:52 Many of the

1:24:53 reasonings that are being done these days is that um

1:24:56 we hope that with antibody testing we'll be able to

1:25:00 identify who is not contagious anymore and who could be

1:25:04 safely

1:25:05 employed to interact with the vulnerable and elderly,

1:25:09 but is the immunity real?

1:25:11 So let me

1:25:12 pass it over to Patrick and Azra

1:25:14 with these initial questions,

1:25:16 and then I'll come back with an extra round.

1:25:18 Over.

1:25:21 Excuse me,

1:25:22 yeah,

1:25:22 so I'm,

1:25:22 I'm afraid,

1:25:23 uh,

1:25:23 Azra had to leave,

1:25:24 uh,

1:25:25 but,

1:25:25 um,

1:25:25 I'll do my best in her absence.

1:25:27 Um,

1:25:28 in terms of,

1:25:29 uh,

1:25:29 uh,

1:25:30 let me try and remember them in order,

1:25:31 in terms of severity,

1:25:33 um,

1:25:34 I think that that,

1:25:35 uh,

1:25:35 has been a key question because like I said at the beginning of the talk,

1:25:39 there's

1:25:39 the data that you have to parameterize that is quite limited,

1:25:42 and I think.

1:25:44 We,

1:25:44 along with a lot of the world have been waiting for some really good serology,

1:25:49 um,

1:25:49 surveys to,

1:25:50 to start coming through.

1:25:52 I think we're starting to see them and,

1:25:53 and I think the picture is um

1:25:55 broadly consistent with the level of severity,

1:25:59 um,

1:25:59 that,

1:25:59 that,

1:26:00 that was estimated.

1:26:01 Um,

1:26:02 it could still be some,

1:26:03 some

1:26:04 recalibration and readjustment that needs to happen

1:26:06 once we do some formal fitting.

1:26:08 But,

1:26:09 um,

1:26:09 but yeah,

1:26:10 that will certainly be

1:26:11 updated as and when,

1:26:12 as and when we can.

1:26:14 Um,

1:26:14 but,

1:26:14 but clearly I think,

1:26:15 I think some of the,

1:26:17 some of the,

1:26:18 uh,

1:26:19 kind of,

1:26:19 uh,

1:26:20 more

1:26:20 hopeful estimates of this idea that there might be

1:26:23 really high proportion of the population who've already been infected in London,

1:26:27 for example,

1:26:27 I think,

1:26:28 uh,

1:26:28 and in the UK I think are,

1:26:30 are,

1:26:30 are not,

1:26:31 sadly I think aren't

1:26:33 bearing fruit,

1:26:34 and then it does look like we have a way to go.

1:26:36 Um,

1:26:37 and the second is the affected population immune?

1:26:42 Oh

1:26:44 yeah,

1:26:44 oh,

1:26:44 so,

1:26:45 so that's a shame that I've just gone because I have to confess that I'm,

1:26:47 I'm not a virologist,

1:26:49 and we have a big,

1:26:50 what we do have is,

1:26:51 is a wider,

1:26:52 um,

1:26:52 group,

1:26:53 a much larger response group,

1:26:54 and also that group,

1:26:55 um,

1:26:56 we,

1:26:57 they go to,

1:26:57 um,

1:26:58 in the,

1:26:58 at least in the UK we go to a scientific advisory group,

1:27:01 uh,

1:27:01 where you have a lot of expert virologists and people like that,

1:27:04 and that's where we get our assumptions from the modeling and parameterization.

1:27:08 I think it's,

1:27:09 it's worth saying that,

1:27:10 that,

1:27:10 um.

1:27:12 But yeah,

1:27:12 it's,

1:27:12 it's,

1:27:13 it's those data again you can hopefully reparameter and could be important

1:27:17 um

1:27:18 like anything at the population level you

1:27:20 can see some levels of reinfection but then

1:27:24 the,

1:27:24 you know,

1:27:24 if they kind of outlying cases that don't necessarily

1:27:26 mean that people aren't broadly making immune responses,

1:27:29 but

1:27:30 yeah,

1:27:30 again it's,

1:27:30 it's,

1:27:31 it's,

1:27:31 it's um

1:27:32 yeah,

1:27:33 early days.

1:27:38 Thank you,

1:27:38 Patrick.

1:27:39 Let me see if anyone from Webex uh has questions or wants to raise their hand.

1:27:45 Um,

1:27:45 Alec can unmute you if you would like.

1:27:50 Yes,

1:27:50 there's a question from

1:27:51 Dimitri.

1:27:53 Um,

1:27:54 how is contact rate related to our,

1:27:56 uh.

1:27:57 Um,

1:27:58 OK,

1:27:58 let me just,

1:27:58 Dimitri,

1:27:59 would you like to,

1:28:00 I just muted you.

1:28:02 Would you like to?

1:28:03 To ask you a question?

1:28:06 Uh,

1:28:07 yes,

1:28:07 uh.

1:28:09 So,

1:28:09 I'm trying to figure out how to model this,

1:28:11 uh,

1:28:12 how to calibrate this model as well.

1:28:14 So,

1:28:14 I,

1:28:15 as I understand,

1:28:16 the contract rate is related,

1:28:18 directly related to RR0.

1:28:22 And so,

1:28:23 in the,

1:28:23 in the model,

1:28:24 I assume that uh the countercreatives are not divided by the length of the disease.

1:28:31 I,

1:28:31 I assume there is 18 days.

1:28:34 Is it a fair assumption to make?

1:28:37 Oh,

1:28:37 OK,

1:28:38 so,

1:28:38 uh,

1:28:38 I think,

1:28:39 uh,

1:28:40 well,

1:28:40 it,

1:28:40 what you,

1:28:40 what we can do is you,

1:28:42 you and I can,

1:28:43 can talk offline and I can take you through exactly every assumption we're making.

1:28:47 Um,

1:28:48 so the,

1:28:49 so what you're saying is,

1:28:51 is pretty much in a,

1:28:51 in a microcosm,

1:28:53 correct?

1:28:53 You,

1:28:53 you,

1:28:54 your,

1:28:54 your R0 is the,

1:28:56 uh,

1:28:57 is something is the rate at which you infect people per day

1:29:00 multiplied by the number of days you stay infectious.

1:29:03 So you

1:29:04 want to sort of account for how long you're infectious for,

1:29:07 um,

1:29:08 and,

1:29:09 and so what generally what you do is you,

1:29:11 there are two components that go into your R0 estimation.

1:29:14 There is the

1:29:15 exponential growth rate,

1:29:16 so sort of the doubling time,

1:29:17 so

1:29:18 a different doubling time of 3 days or 4 days or 5 days will have a different.

1:29:22 And

1:29:22 also the duration of infectivity,

1:29:24 cos that tells you the number of generations you're going through

1:29:27 within that,

1:29:28 so,

1:29:28 so

1:29:28 if you're seeing much faster growth,

1:29:31 but fewer going through fewer generations,

1:29:34 that means more infections per generation,

1:29:37 a higher R00.

1:29:38 It's getting a bit technical,

1:29:39 but,

1:29:40 but,

1:29:40 um,

1:29:41 but yeah,

1:29:41 we can,

1:29:41 we can walk through that,

1:29:42 it's with you,

1:29:43 your basic

1:29:44 in terms of your duration,

1:29:46 there's also the,

1:29:47 well,

1:29:47 at least with Rmoni you have to account for the

1:29:49 fact that not everyone makes contact at the same rate.

1:29:52 That's where our age dependent comes in,

1:29:54 and that's where,

1:29:55 yeah,

1:29:56 you,

1:29:56 you have to calibrate to account for that,

1:29:57 and that also has onwards impact or severity

1:30:00 if

1:30:01 your severity depends on age,

1:30:02 which it does,

1:30:04 um.

1:30:05 The

1:30:05 one thing is the duration of infectivity,

1:30:08 we assume a lot shorter than 18 days.

1:30:10 Now 18 days is how long it takes you to,

1:30:12 you can have if you have severe disease,

1:30:14 you can

1:30:15 experience the um.

1:30:17 Experience the.

1:30:19 That's a keyla of the disease for a long time,

1:30:22 but that doesn't necessarily mean that you are infectious and

1:30:24 it doesn't necessarily mean you're representative of all infections,

1:30:26 so

1:30:27 I think at the moment we're assuming a sort of an

1:30:29 average infectious period of somewhere in the order of 3 days,

1:30:32 but with a lot of variation.

1:30:34 Um,

1:30:35 and that depends a bit longer if you have a more severe long lasting.

1:30:39 You have to cancel the

1:30:40 much less,

1:30:41 you know,

1:30:42 less severe cases as well as the most severe.

1:30:45 Thank you.

1:30:48 Patrick,

1:30:48 and thank you for the excellent question.

1:30:50 We have arrived at 10:30 and it's time for us to conclude.

1:30:54 Um,

1:30:55 some of you have asked whether this is published and all the

1:30:58 parameters and the assumptions are available for people to work with.

1:31:02 Patrick,

1:31:02 would you want to indicate where

1:31:05 people could find all of this?

1:31:06 I understand you,

1:31:07 they are in detail.

1:31:10 Yeah,

1:31:10 there's a GitHub,

1:31:11 yeah,

1:31:11 which,

1:31:11 and we're making things,

1:31:13 yeah,

1:31:13 it's,

1:31:14 it's,

1:31:14 um,

1:31:14 yeah,

1:31:15 we're

1:31:15 going as quick,

1:31:16 but yeah,

1:31:16 the GitHub's there,

1:31:17 so that then

1:31:18 you can just recreate everything in our report.

1:31:20 I will,

1:31:21 I don't know where,

1:31:22 where's the best place to put that Roberta,

1:31:24 um,

1:31:25 the link.

1:31:27 Well,

1:31:27 we will circulate to everyone

1:31:30 the recording.

1:31:31 Yeah,

1:31:31 what we can do is that we'll circulate to everyone the recording of the event,

1:31:35 uh,

1:31:36 your presentations,

1:31:37 and with that we could probably circulate not only your paper but also the link

1:31:42 where people can,

1:31:43 uh,

1:31:43 access.

1:31:43 I remember that reading your paper that was within the paper an Excel sheet link

1:31:49 where I could

1:31:51 just

1:31:52 in

1:31:53 there and.

1:31:55 Yeah,

1:31:57 play around.

1:31:58 Yeah,

1:31:58 and just,

1:31:59 just to say that everything will be posted,

1:32:01 the recording and the documents and the presentations were

1:32:03 going to be posted in the e-seminar's website.

1:32:07 So

1:32:09 they're already some of them posted.

1:32:11 And we will have the recording,

1:32:13 um,

1:32:13 hopefully tomorrow.

1:32:16 Perfect.

1:32:17 This is excellent.

1:32:18 So I wanted to thank

1:32:19 um our speakers,

1:32:21 uh,

1:32:22 Damien De Valke,

1:32:23 Jeff Friedman from the World Bank,

1:32:25 who gave us an,

1:32:26 an overview of

1:32:28 the models that are uh available.

1:32:30 Also in their presentation,

1:32:31 you will find links to

1:32:33 the modernization of the AHME Institute,

1:32:35 the University of Basel,

1:32:37 to Patrick Walker and Professor

1:32:39 Azragani for

1:32:41 uh their overview of this influential model,

1:32:44 which is the one.

1:32:44 which we've all have based our

1:32:47 fears and sort of conclusions,

1:32:49 and thank you also for the

1:32:51 sobering and

1:32:53 humble

1:32:55 approach that you're taking to the mobilization.

1:32:58 Thank you to David Wilson,

1:32:59 Madli Gargens,

1:33:01 because

1:33:01 they helped us reflect

1:33:03 about

1:33:04 potential like these strategies,

1:33:06 and thanks to all of you for your interest,

1:33:08 enthusiasm,

1:33:09 and patience.

1:33:09 The fact that I got so many text messages about I can't listen to the.

1:33:13 Presentation,

1:33:14 I want to see it.

1:33:15 It means it was really

1:33:17 um providing an important um

1:33:20 added value for all of us

1:33:22 who are trying to think about how best to

1:33:25 suppress and stop the spread of this disease.

1:33:27 So thank you everyone,

1:33:29 and uh looking forward to the

1:33:31 uh

1:33:31 other upcoming uh great seminars that uh the Development Research Group is

1:33:36 putting together for us to understand better the impact of coronavirus.

1:33:40 Thank you very much.

1:33:41 Bye.

1:33:52 Thank you,

1:33:52 Patrick

showAllTimestamps
no
transcript
Good morning everyone. My name is Roberta Gasti. I'm the chief economist for human development, and it is my pleasure today to be here and chair this important event on behalf of Art CR, our acting chief economist, who's very sorry not to be here to listen to what promises to be a very interesting and informative conversation. Today we will be talking about Understanding the epidemiology of coronavirus. Um, I think this is a very positive, hopeful title, and, uh, this is something that I think everyone since uh this virus has, has, uh, emerged and wrecked, uh, the world's life, uh, would like to do. We are fortunate to have today a prominent epidemiologist who will go over some of the implications of different nonpharmaceutical interventions on the spread of the disease, and we also have Our bank colleagues who will provide us with an overview of the different tools that are available to us as practitioners in economics and health to understand better the spread of the disease. Uh, without further ado, let me start by inviting Damien de Valke and Jed Friedman, uh, to present a bird's-eye view of the instruments available to understand the epidemiology of coronavirus. Uh, both, uh, Damien and Jed are senior economists in the Development Research Group of the World Bank. Um, they are both working on health. Damien with, uh, an emphasis on, um, Uh, the interaction between health and education and on the epidemiology of HIV AIDS and Jed, uh, on the, um, impact of, um, uh, health on productivity, among other things. Uh, why don't I turn to you, uh, Damien and Jed, uh, for your, uh, introductory presentation, and then we'll pass on, we'll move on to the next speakers. Thank you. Over. Hi, Roberto. Thank you very much. This is Jed. Um, Damien and I are just here to give a, uh, a really brief orientation to the speakers to come that we look forward to from, uh, we look forward to hearing. Um, so we will only take a few minutes to discuss, um, a few issues related to of uh non-pharmaceutical interventions. Um, what was the motivation for this? Um, e seminar, well, as we all know, uh, our institution has reoriented much of staff effort now towards, uh, counteracting and ameliorating the effects of the pandemic, and, uh, we all believe it would be useful to review the underpinning epidemiological models, um, so that we can understand the Anticipated courses of illness in the countries we work in. Um, and so what we hope to discuss in the seminar, as Roberta just said, is the epidemiology of the SARS-COV-2 virus, uh, as well as the effectiveness of strategies to reduce, uh, the spread in the population. So, uh, Dommy and I will begin with a short primer. Um, which is largely a brief overview of key terms and some ideas. So here's just one slide on some terminology that I think we've all become familiar with. We've all become chair epidemiologists in the last month or so, and I'm glad that we have actual epidemiologists on the line today. Uh, to give us a much more informed view. Um, but just, uh, to bring everyone up to speed, um, let's, let's just talk about a few terms such as the reproduction or are, are not. Um, we've all seen this. It's defined as the average number of secondary infections from an initial infected person. Uh, the basic R0 is the sort of, sort of the maximum epidemic potential of a pathogen, uh, in the absence of any distancing or other types of interventions. And, uh, estimates for the SARS-CoV-2, um, vary a bit, but they essentially in the 2 to 3.5 range, which means that for each infected person, we should expect, uh, 2 to 3.5 additional infections. Uh, in the absence of any, any intervention. Um, and then, of course, the effect of R0, which, um, is the empirical, uh, R0 of what happens in the population as we begin, uh, to try to control transmission. And of course, this depends on susceptibility, so the percent that, um, uh, are, have yet to be exposed to the virus as well as the effectiveness of interventions. And I just saw a piece yesterday, uh, estimating that The effective R0 for Washington DC has, is, has fallen from 2.5 to 1 in the last few weeks as a result of the, the stay at home measures and other things. So that's a bit of good news if that's true. Um, the latent period, uh, is the time from infection to onset of infectiousness. Uh, this, of course, is important in understanding the, the spread of the disease, and then that, that's followed by the infectious period. Again, measured usually in days of the time a person's, uh, is infectious. Um, and I, I, we've been told by Imperial College colleagues now that maybe the best guess of the infectious period is that it begins roughly half a day before the onset of symptoms. Um, and then there's the serial interval, which is often used in, in modeling, which is the time between the onset of symptoms in the index case and then the onset of symptoms in their context. Seasonal forcing, of course, uh, this is one way to get at differences in how climate affects transmission. Not much is known about this. This is obviously another critical factor, uh, that I don't think we'll be speaking about too much today, but there might be questions. And one way to model seasonal forcing is to vary the amplitude of variation and transmission. And of course, there are some familiar statistics to all of us now, such as the case fatality ratio, which is the number of deaths divided by the total number of people diagnosed. Uh, for a certain period of time, ideally it's for that period of time to represent the duration of the course of illness in the vast majority of cases. And one way to account for the possibility of asymptomatic infections is to switch from the case fatality ratio to the infection fatality ratio, where the denominator also includes estimates for the number of asymptomatic and undiagnosed infections. And of course, what's very relevant uh for policy and a lot of the health teams now working in coronavirus preparation are various features of the health system. Hospital beds, staffing, ICU beds. This all relates, of course, to treatment of severe cases, the duration of treatment, and then the presence of key equipment, um, such as oxygen, ventilators, and then the personal protective equipment that staff need, health staff need to try to prevent, uh, exposure themselves. And then a large topic of this talk will be non-pharmaceutical interventions. Uh, there's various types I'll discuss in a second, and, um, Uh, what's relevant though for their effectiveness is the degree of compliance or enforcement with these types of interventions and of course their duration. Um, so just continuing on this, uh, a brief overview here in this slide of the types of non-pharmaceutical interventions at, at our disposal. Um, the first would be surveillance. Uh, this is important throughout, but it's particularly important at points in the epidemic, um, when transmission is at a relatively low level and you can hopefully try to contain it solely through the strategy of testing, tracing, and isolation. Which of course is having widespread surveillance in the, in the population where you're testing either widely or, or, or potential cases and then tracing all their contacts and, and isolating them. Um, from the international perspective, uh, there are trade and travel restrictions. Many countries adopted these, uh, at the start of March. They, they range in severity from traveler screening, which is allowing borders to be open, I'd say screening for, for symptoms, and then if found to say, have a fever, then forcing traveler quarantine. Uh, and they run all, all the way to the extreme version of, of total border closure to people and goods, which we've, we've also seen at points around the world. Quarantine, of course, is the most extreme form of social distancing, uh, which is really trying to impose this corridor of at least 2 m between all people and certainly between infected and non-infected people. So quarantine is, is not allowing, uh, any passage out of either a household if the household's quarantined or even of entire regions as, as was tried, for example, in northern Italy. There are other forms of social distancing that are a little less severe, um, bans on public gathering, public transport, closing schools and businesses, which we've, we've all, um, seen in, in, in I uh, many countries that we're listening from today and then stay at home orders, which are not as severe as quarantine insofar as you're allowed to go out for essential services such as food, uh, procurement. Finally, there's hygiene recommendation and enforcement, uh, where possible, uh, promotion of preventive behaviors such as hand washing, uh, is certainly one tool in the arsenal of, uh, reducing the spread of transmission. Uh, and then there's recommendations for face masks and even mandates, um, which is sort of an interesting subarea now with, with more and more, um, uh, evidence suggesting that it's, it's a 11 effective tool in driving the R0 to, uh, below 1, which, of course, is the goal of a lot of these. So, as soon as, uh, one new infection will infect less than one going forward, we would expect to see a decline in overall transmission. And of course, the effectiveness of all of these NPIs will depend on various factors, as I said, the duration of them, the degree that population complies with them, and then, of course, epidemiological factors as well, such as the latency and, and the R0 that I mentioned. So, thank you for bearing with me. This is just a brief overview that will continue now with my colleague, Damian, I believe. Uh, so if I can invite Damian to continue. Yes, thank you, Jed. Um, So, uh, what I, I will do in, in the, uh, the few coming, uh, slides, uh, Jed, I think we still, we cannot see your, uh, you're sharing the slides, but we cannot see them. Yeah. Um, It's uh trying to uh review um a few of uh um online resources for epidemiological modeling and projections. I've reviewed some of them in a, in a blog post that at the end of March, which probably would need some updating. What's important to know also is that the World Bank, the International Decision Support Initiative, the Gates Foundation, And the World Health Organization has developed a partnership to review and compare all models, and maybe David and Marli will tell us more about that. Then one of the available tools of course, we'll hear more about it today is the estimates that the Imperial College COVID-19 response team provided. Uh, they've providing, they've been providing these estimates for the world and for each country under five scenarios going from unmitigated epidemic to early suppressor suppression strategy. They'll they'll give, I'm sure more detail about that, but. So one very useful feature of their paper is that there is an online appendix, and when you click on that, you get access to an to an Excel table give you, uh, giving you country level estimates with even the further possibility to adjust some of the parameters such as the baseline air knot and the intensity of social distancing measures. So it's a very useful tool. Another tool that is available online is at the University of Basel. It's an online tool that allows you to run your own projections and enter country specific numbers. I found it quite intuitive and flexible, but it's important to know that you will need to closely calibrate your proper parameter estimates to the data that you have. Some World Bank teams have used it to. Make projections for the country they are working on. Um, it's a very important feature is that the age structure of the population is preloaded for each country on the online tool. Then there is also projections available that have been made by the Institute for Health Metrics and Evaluation. Currently the projections are available for the US and European countries, but it seems that uh projection for African countries will be coming soon and the World Bank is collaborating for obtaining regularly updated projections for the countries which have, which are recipients of COVID-19 financing from the World Bank. It's important to know that with that tool, the scenario projector assumes current social distancing measures are maintained, so it's different from the other tools where you can go all the way from unmitigated scenario to mitigation measures and and suppression measures. Also, it's a model that is not based on a SIR model. So SIR is susceptible infectious recovered, but it's a model that fits the Emperor Kelly observed COVID-19 population death curve. So in my next slide, I will show you an example of how our World Bank team has used the University of Basel tool to work on some projections. It's a large middle income country. they've shared it with the government, but the government has preferred to uh to keep this, um, not public yet, so I'm not at liberty to give you the name of the country. But you see how they've used the tool in these two graphs to predict on the left the number of COVID-19 patients requiring hospitalization. In red with the curve, if there was no action, no action. And in blue with the The shape of the epidemic with a 2 month lockdown and this is of course you have seen this picture before, a nice example of flattening the curve and on the right with the same color red for no action, blue for a 2 month lockdown. You see, uh, the COVID, the projected COVID-19 cumulative fatalities. I should also add that this World Bank team worked hard on the University of Basel T and then when the Imperial College estimates were, were published, they were very happy to see that their numbers were very similar to the um suppression scenario from Imperial College. No, um, I'm moving on this slide that goes a little bit beyond the pure epidemiology and try to talk about all testing can be very important to inform decisions and the effectiveness of uh non-pharmaceutical interventions. Um, Jet, Roberta, and Aditya Matu and myself, we've written a short note on this, and I'm sharing with you, uh, these figures. So you'll see in blue the epidemiological curve for the infection. And we have been yellow, uh, the fraction of the population that that is recovered. So what we Argue in that piece and through this figure is that testing and two types of tests are very important to make decisions. So one test is the test, the test of infectiousness, and that's usually do through a PCR test. And you see that to the left of the first red line. That's the time when you are early in the epidemic when basically you can try to implement a test, trace, and selective isolation, a TTI strategy relying mainly on the infectiousness test based on the PCR. Similarly, on the right of the figure to the right of the second red line when the epidemic. The epidemic outbreak has basically has gone down enough, then we can try to use again the same TTI test trace and selective isolation strategy to try to prevent the resurgence of a new epidemic wave. In between these two red lines. When the outbreak is really in full force, it's time to use the suppression strategy which we've colored in green. We've also indicated that another test, an antibody test, could be used at a certain point first to protect essential workers because that test detects who is immune to the. Uh, to, uh, to, uh, um, COVID-19 and then eventually gradually to allow other people to work and we also recommend that uh there would be a regular antibody testing of a representative sample of the population. As well as continuous monitoring of morbidity and mortality to inform these decisions on when to use the test, trace and selective isolation strategy and when to enforce the suppression strategy. No, I'll, I'll leave this and, and just finish by outlining our e seminar today. First, um, the Patrick Walker and Azra Ghani from the, uh, Imperial College COVID-19 Response Team will present their, their model. Uh, they will tell us about the underpinning of this model in its main prediction. Tell us all the effectiveness of the non-pharmaceutical information are modeled, what are the main policy recommendations, and they will touch upon specific issues for low and middle income countries such as the age structure of the population. The presence of many multi-generation households, crowded habitat like slums, the difficulty to enforce social distancing measures when people are living and working in the informal sector, and then our colleagues from the Health, Nutrition and Population. Uh, at the bank, David Wilson and Marili Gorians will discuss the trade-off between the need to curb the coronavirus and the pressure to reopen economies, and they will go over a few key consideration in that that will inform the trade-off. The epidemic forces, the capacity to manage a reduced measures, population health, the capability of the public health in the country, the health system capacity, national decision making capability, and the role of technical innovations. And with this, um, Um, will, uh, Give the floor to Patrick. Thank you very much. Uh Thank you, Damien. This is Roberta. I just wanted to take a moment to introduce uh Pat uh Patrick Walker and Azrahani to our audience. Um, they are respectively lecturer and professor at the infectious disease epidemiology department of Imperial College, and we of all, the armchair epidemiologists, as Jed nicely put it, read and learned tremendously from the work that the Imperial College team has done in helping the world understand the patterns and the spread of coronavirus, and The potential effects of different policies. So I understand that Patrick will start the presentation, so thank you so much for being here and particularly thank you for the excellent work that you are doing to um help us understand the disease. Over. Thank you so much, uh, Damien and Roberto. I'm just, just trying to make sure I can master the technology, um. But yeah, thank you very much for that really er excellent introduction, and I think. Uh, when it comes to actually. It comes to Coronavirus, uh, COVID-19, epidemiology, we're all fairly armchair cos we've only had a few months to, er, to actually, um, I'm just gonna move to the, the correct slide I picked the wrong ones. Reviews for this. Apologies for that. So yeah, so, uh, like I was saying, I mean it's, it's a newly emerging disease, so I think we should all feel a little bit more OK about being armchair epidemiologists. We've only had, um, 4 months now to, to learn stuff about this disease and there's not a lot of data to actually go on and you know, and as we sort of know with these. This is in when when outbreaks emerge, there's a lot of difficulty in actually getting representative data because of a range of severity in terms of infections represent and likelihood that you're going to see them reported or they will get into your surveillance system. And also that is likely to change over time. But um, but one of the few things we do know about, um, COVID-19 is that there's this clear, um, severity pattern with age, whereby, uh, the older you are, the higher the likelihood of, um, having sort of severe complications, respiratory symptoms and, and needing more, uh, you know, advanced intervention to a higher risk of fatality. Um, and this is, these are estimates that were generated by, um, people in our team, um. and these are being used to kind of inform our severity predictions and by age and also we know that there are some kind of relevant comorbidities characterized primarily focused around morbidities that exist in the settings where we've seen. Uh Large outbreaks, so these are things like cardiovascular disease. Uh, pulmonary disease and um diabetes. Um, we're also learning more as uh Damien was, uh. Very ably Uh, discussing earlier, earlier, and we're learning a lot more about the transmissibility of the. So, uh, yeah, so this is work, um, it's also been done by people, other people in our team, um, Sam Bath and Seth Waxman and others, and then what they're doing is they're taking data on the, uh, the deaths over time, which even in high, high income settings are the things that we think are most reliable case reporting changes so quickly over deaths are your ways that you can most reliably get friends. Um, but unfortunately what makes that hard is you have this delay of around. 20 days from when you're infected to when you I if if that's. You're lucky enough that that's. Um, and that actually makes inferring the trends in deaths of being in Europe quite hard, but from this analysis, what we are seeing is that this kind of curve is leveling off, factual that actually if there weren't any. That um the curve initially is kind of doubling times consistent with an R0 of about 3 or 4, which is actually very highly infectious compared to something like. So normal seasonal flu, um, which usually has an out of about 1.5. Um, and that, um, as we're implementing these measures, we're seeing rather than the, uh, the debts double and double and double, we're seeing that they are actually leveling off and. that does correspond with some of the larger what we know as non-pharmaceutical. Uh interventions, um. Things that are often called lockdown and they are beginning to show impact so hopefully over the we're gonna see signs in more time. Some of the hardest hit. If these measures are maintained. Um, but even in these high income settings, the problem is that the, that the reasons they're needed is partially because of the, the direct burden, but also because they have this massive strain on the health system and they lead to incredibly high demand for hospitalization, ICU, oxygen and medical ventilation. Also, they, uh, pose a high, high burden on the personnel who are offering care, so. High rates of infection and self-isolation problems and the need for extensive, um, and these are even, these are some of the countries that have some of the strongest. So, Uh, Charles So, um, thinking about how modeling plays a part in this, well, it's, it's when you have a, a, you know, a brand new disease, a new disease that actually kind of needs some sort of model in which to think about the kind of the logical scenarios that you might see, and the model is just a, it's, it's not a, especially where this early in October, you can't really expect them to be kind of completely accurate forecasts of the future, but what they do give you is a sort of logical framework. For the assumptions you're making about different strategies, and actually it kind of really helps to kind of hold you to those assumptions. This is work that the COVID-19 team have providing. Neil, the head of our group, uh, has been an AR as well, uh, providing to the UK government along with other modeling groups we have in the UK um around different scenarios, what, what you do, um, what can you do and what is it likely to achieve, where will that get you, um, and it does kind of hold you to the logical assumptions. So, for example, if these large suppression strategies. We are gonna have to think about an exit strategy because we can't just go back to zero intervention because then we do definitely get sort of a second wave. Um, so, so it, it's helpful to really think about, you know, the future. Um, but, a lot of countries don't have these kind of in-country modeling groups, um, that we have in the UK and other countries have as well, but not many. Um, but they're all facing very similar questions at the moment, um, and some of them might be a bit earlier in the pandemic, but those questions are likely to be quite similar. Um, but what we've been trying to do in terms of our sort of the global side of things is particularly with lower income settings, what can we say that kind of mimics some of these modeling approaches, given that we probably can't do something comprehensive for every country. Um, and what would be useful given that Clearly we're in a very time critical. You have to That needs to be useful, it needs to Um So what we've, when we've been thinking about this, we've kind of identified 5 main factors which are likely to vary substantially, particularly by income. Some of them are obvious, some of them less so, but kind of quite an obvious one is that when you have a disease like this that has this kind of um. Severity signal with age where the older you are, the more likely you are to have severe consequences. Well, in many lower income settings, populations are a lot younger. They have higher fertility rates. They have lowest survival rates outside of, outside of the window, um, and that leads to, you know, there being the average age of an infection go down, and the, we would might anticipate that that would be a, a force to reduce a person's burden. That's one thing, there's also a pattern with sex, biological sex, which is males are slightly more likely to have severe disease, but that's probably more likely to be balanced across country. Um, also there's societal structure. Um, which can vary quite a lot and in very important ways. I'll try and speak up, but please, yeah, let me know if I'm, I'm not loud enough. Um, that the, well, what we have in, in the UK where I am now is that outside of the pandemic, elderly people, uh, elderly people tend to be a lot more kind of isolated, and it's something that we would see as a, a health problem outside of the pandemic window. But if you are an elderly person, you, um, I'll, I'll go into that in a second time. But there's also a that's for the next slide, then there are also these patterns of comorbidities that might change. Um, also, there's clearly going to be big disparities with both health and system strength and capacity. Finally, we shall get onto that there are these, the, the, the extent to which countries are likely to be vulnerable to the social and economic impact of both the disease itself and any interventions, especially some of these stringent interventions, is going to differ widely by the level of income country. So this is what I was saying a second ago, um, that, yeah, in, so this plot is showing the average household size of a person aged 65. And what we see in both these, this, this kind of data and, and in contact surveys where we ask people about the number of contacts they make in higher income settings, particularly in Europe, older people make less contacts. They live in much smaller households, and it's this pattern of kind of social isolation, which is generally seen as a problem, you may live on your own. Um, in, uh, lower income settings, uh, older people are far more likely to share their house with their children or their grandchildren, so in very low income settings, often the average household size of a person aged 65 could be 6 or 7, and also there are big differences in, um. School attendance and work patterns that we that we think along with the data on social contact patterns that exists in low and middle income settings, not much, but actually you, you, you very rarely see these patterns of social isolation of elderly people, they're far more present in terms of the number of contacts they make in their society, um, and that's, as so we anticipate that. Older, there will be fewer older people, but those older people that are there will be more vulnerable to having the disease spread to them, uh within their households. So what's that trade-off in terms of on the one hand that could make it less severe, on the other hand, the fact they're less isolated makes them more likely to get it. Well, that's something that you can kind of try to explore with modeling, making sure that you're, you're very clear about the assumptions that you're making. What we have is we have a model that mimics some of the major uh attributes of the larger microsimulation model that has been used to inform. Um, Patrick, sorry to interrupt. This is Roberta, and I'm talking to, uh, Ale, you, Ale, and, um, Ryan just to indicate that I've got some messages from, uh, those who are following us on YouTube that your slides cannot be seen on the YouTube channel. Ah So I like, I'll give a minute to you, uh, Ale and, um, Ryan to sort it out, but I also take this opportunity, besides the, uh, small technological glitch, to indicate to our, um, to those who follow us on YouTube today that they can send us questions through the chat room in YouTube, and I'll represent them, uh, on your behalf to our speakers and the same, uh, uh, the Webex, um. Um, uh, audience can do, but, uh, those who are following us on Webex, and it's been a, uh, first come, first serve, a sample of the 1st 20, I believe, or 50 who have signed up, uh, they can also raise their hand, uh, and, uh, ask the question directly in the Q&A session. So over with this small organizational announcement, um, and maybe whether Ryan or Ale can send, give me. Send me a text to indicate whether what needs to be done to have the um slides seen on YouTube. We definitely see them on the web Webex channel. Thank you, over. Thank you Robert, uh, yeah, and I think I, I saw one of the questions about how, how we're calibrating it, and this is what, what this, this slide is sort of trying to show that what we do is we, we take, uh, this model is kind of capturing sort of the, some of the, the major attributes. What we do is we take, uh, the, uh, what we know about the transmissibility of the disease in terms of. Um, and that's a parameter we can vary by context. Take what we think we know about mixing patterns by age within a setting, and we take the severity of infection by age, we put them all together in a model, um, as long, uh, uh, uh, as well as some of the more key parameters of the epidemiology that Damien was describing, things like time between one infection to pass through another infection, and, uh, the time from onset to death, things like that. Um, and from that we can kind of generate what we think might be the trade-off between changing age patterns and the changing social mixing patterns by income strata and by country. And this slide here is just showing kind of how we think that would vary. So this is assuming the same R0 across all settings, here we're taking an R03 which is. As Damien was saying, and as, as the data in Europe is showing us, our sort of best estimates for what we, what this disease is doing, uh, and take the sort of likely impact in it of a total impact of an unmitigated disease, uh, uh, pandemic, and that RO 3 highly transmissible disease means that it will spread very widely in a within a community without any action, um. Uh, but that actually as you think about more the more sort of severe severe scale of the disease, uh, moving from hospitalization, requiring critical care and, unfortunately, sadly, potentially, uh, leading to mortality, but actually because you do have these broader demographic patterns, uh, the, um, we like, we do expect an unmitigated scenario with no intervention. Actually higher income settings would be per person, at a per person level. More heavily affected, experience a higher burden, but this is making a very strong assumption at this stage, that, that these that these severity patterns are matching those for the health system in China. Now clearly China, at least with particularly in this pandemic, made a really large effort in making sure that that the majority of people are getting hospitalized. Uh. Patrick, sorry, this is Roberta. Apologies for interrupting. One message to you is that if you can speak as close as possible to the mic, that will help. Second, I'm asking Alejandra or Ryan to please unmute themselves and let us know whether the YouTube channel now works with the slides because I'm continuing to get, to get messages and since I'm, I follow everyone on Webex, I am unable to um. Have the information on whether. Yeah, hi Roberta, uh, Ryan is currently with ITS. Unfortunately, I don't know what's going on because it seems there's no audio. Um, but he's working on it. Oh excellent. So I would appreciate um Ali, whether uh whenever this is sort of sorted out, whether you can um just let me know on the Webex or on the channel. Since we have this moment of mute, I would like them to ask Patrick to just, uh, at the end of his presentation to just give like a quick point by point. Uh, sort of key messages summary for those who have not been able to follow the um. Uh, the conversation. Meanwhile, I see many questions that are popping up on Webex. Uh, I don't know, Patrick, if you see them or whether Professor Ghani is, uh, answering them, but it might be good to maybe take a moment to, uh, indicate them to the audience so that, uh, also the audience can see the conversation that is ongoing. Um, just to take a pause while the technology is being fixed. Um, there was a question on. Uh, from Dmitry on how, uh, R 0 was, um, uh, estimated, and I wonder whether Professor Ganni wants to amuse herself now while we take this pause to give, um, a quick answer for everyone, also those on YouTube who might be, uh, now, uh, might hear it as well. Professor Ganni, would you want to, um, address that question? Yes, that's fine. I've actually just put a comment in for those who might not be able to hear. Um, the, our estimates of the, uh, basic reproduction number come from the early growth rate of epidemics. We actually estimated a, a lower, um, value of around 2.4 from the Chinese data originally. Um, the doubling times of the epidemics in Europe have been much faster, and our current best estimate is around 3. clearly it's a little bit sensitive to the context and the setting in which transmissions occurring. Um, excellent, thank you. Um, Ale, uh, do you know if the problem is fixed on YouTube? Um, I'm OK, no, Ryan, I'm on the, hi, this is Ryan here. I'm on the line with our IT colleagues, and they're trying to solve the problem but do not yet have a solution, so I'll let you know as soon as possible, but unfortunately the slides aren't coming through, um, uh, but I, you know, I'm, I'll get back to you as soon as, as soon as they have this solved. Very good. Um, well, why don't you signal that maybe in the chat room to the YouTube, uh, followers, and then we can maybe continue with the presentation and then I'll ask Patrick just to summarize the main messages so that, um, those on YouTube, which is a much larger audience, uh, can still follow the presentation. Thank you and over, Patrick, see you. Patrick, you're muted. And that um and yeah, and we do actually have some additional scenarios accompanying the report that I can also provide a link for if, if people want to sort of see how these may change with in different contexts, if or not differs to an extent. So yeah, so, so this is just to say, well this is based upon severity estimates by age from a place where there was relatively speaking, at a global level, very good access to care. And that's gonna differ quite substantially by context and and and change the patterns a little bit, which I'll get onto in a, in a minute. So what we then looked at is this is kind of the, the, the worst possible news, the, what happens if we don't do anything, but of course as, as a globe we are gonna do a lot to try and prevent this burden, given how big it is, given that we estimate that this could be 40 million if we don't deaths globally if we don't actually do something. But but countries will certainly do things. And um. But, you know, um, and Without a vaccine, that's all gonna be these non-pharmaceutical interventions, modifying our social contact behavior, um, particularly in settings that don't have the er capacity and the technology to implement some of these really stringent test er testing and treating and tracing mechanisms. A lot of countries probably will, will, will not have that ability in the time that it takes. So yeah, so, and so here this is where we sort of looked at, well, let's say you can't suppress the infection, what's the best if you didn't suppress the infection, if you couldn't decide not to, what would that lead to in terms of what could you do to reduce this burden? And this relies upon a concept called herd immunity, which is, which is quite become quite a controversial concept in the UK but it's actually just a scientific concept, it's not particularly an intervention. What it all it means is that if you can't suppress infection, any reductions you can make to your contact rates within your community can reduce the spread of the infection, and eventually that infection will lead to a level of immunity that will then stop the er the infection from, make the infection peter out because you have a level of immunity in the population. Where if you hadn't reduced your contact rates, it would overshoot that level, so you'd actually get much higher levels of, of burden. So it's actually just reducing the level of overshoot, um that you would get if you don't do anything, um. So what we find is, you know, with a, with an honor of about 3, those reductions are about a 45% reduction in contact rates that you might get some additional substantial additional benefit if you reduce the contacts in your elderly people even more by some of these sort of shielding strategies, and we can perhaps discuss how that might be possible in some settings more than others, like a, like a place with a household size of 7 is clearly going to be more problematic to do something. But um yeah, so, so clearly these reductions are uh easy to model in a, they're easy to model, but they're much harder to achieve in practice and the solutions for that will vary by setting. But what we do find is these have, we have these big impacts. So here the yellow bar shows what happened. Nothing and the. Uh, yes, Patrick, this is Roberto. Sorry for being um inter for interrupting you all the, all the time. So now everything is working on YouTube. Uh, but I'm getting like a slew of messages on my phone from colleagues and friends who are following on YouTube saying we can't miss this is such an important presentation, so I'm sorry, but I have to ask you to go back for slides. I think that the problem started, maybe you can go over. A bit more quickly on the demographic and social structure, but I would like you to essentially start, I think, from here and then go over again the perfect, go over again the structure in um when I was going to school in Italy, people used to say repetitta you want. That two front, which is, you know, repetition helps, but it's a bit boring. I'm sure that, um, the audience will go back to bear with us for a couple of minutes. This is not boring at all. So thank you, Patrick, and thank you for all of those on YouTube for, uh, being so enthusiastic, such an enthusiastic audience. Over to you. Thank you Roberto. OK, yeah, so, so yeah, so, so we, we were mainly looking at how, uh, the. How, um, some of the sort of guidances we give to high income countries, how that might change and how our projections of the spread, likely spread of the pandemic may be altered in different contexts and to see if we can give some kind of broad level guidance. Um, so we identified five factors. One, and they're all I think fairly um intuitive. So one is that um is that er populations in lower income towns tend to be younger and for er for an infection like this that's more severe with age, that clearly is likely to have an impact. um er there also there's societal structure and how we make contacts and that varies by settings, so I'll talk about the next slide. Comorbidities which have been focused on the. Been identified in the countries where the spread has been, which is mainly higher income settings, and we'll come on to that. And clearly there's disparities in health system strength and capacity by income settings. What we imagine, and also the social and economic vulnerability of societies to both the direct impacts of pandemic and any possible impact of the, not some of the non-pharmaceutical intervention Damien was talking about earlier. So I've traveled quite quickly, but they are, what we're finding is that lower income countries, yes, they are younger on average, so there are fewer older people. But we think those older people are gonna be more vulnerable to having the infection spread to them, because in low income settings like in the UK for example, you're, if you're an elderly person, on average you share your house with about one other person. You may even be by yourself and it's outside of the pandemic window that's seen as a problem because it's, you know, you, you are socially isolated and that's not a good thing. In lower income settings, the average household size of an older person is much higher. It can be 6 or 7 in very low income countries. Now clearly that that's going to change your contact patterns, and it means that those contacts are sort of spread more evenly across the society, particularly when school attendance is maybe lower and less frequent, so you don't get these sort of large contacts of of younger people and work patterns are far more spread out. Um, so we actually think that the contact patterns in lower income settings are going to be far more flat and higher income settings. And, and then, so then what is this trade off in terms of having a younger population, but also where the older population are sort of more central in the society. So, um. So that's the kind of question you can look at in a model whilst making sure you are very clear about the assumptions that you're gonna have to make. So we did that looking at, you know, so what we do is we have a kind of a more simpler model framework, but that incorporates most of the key elements of the sort of more sophisticated micro simulations that have been, been provided to, to the UK government and other governments. So these are things like the reproduction number, the time it takes to, for a person, the epidemic to spread from one person to another, known as the generation time. And the, all the delays between getting infected, becoming sick, chances of becoming severely ill, the chances of dying, and how long that takes and how that depends on age, based upon the data that we have so far. We can then adjust this to the setting, so we have the, the, the pathogen specific parameters. We can then take the setting specific parameters, which are the demographics and the the sort of representative context structure we think within a country and use that to sort of generate what we think might happen. And here, next slide I'm going to show us what we think would happen if nothing if nothing was done, if we just let the er the the path that the um pandemic just sweep through and didn't change our behavior or anything, which clearly is not what's happening and it's probably something that would never happen in reality, it just gives us an idea of what we're dealing with. Um, what we would find is that, as it has a reproduction number we're seeing in, in 3, that, that we know from the. Epidemiological theory that that's going to lead to a very high proportion of the population being affected. By the end of the pandemic, but actually if you look at the sort of the more severe aspects of infection, which are age driven, at least partly age driven, hospitalization, more, more severe critical care, um, people needing critical care or eventual, sadly mortality, we would, we would um um anticipate that that would lead to a lower per person risk of death or. In a lower income setting than a higher income setting, but the problem here is that we make these assumptions, that these parameters are coming from the early stage of the epidemic in Europe and China, and those settings have much higher healthcare capacity, um, than a lot of settings. So clearly there's going to be that disparity that we are going to have to try and account for, and I'll get on to that later. Um, but what we did do before that is then think, well, what is the potential we have to mitigate the the spread of this disease, um. And first, before we start thinking about, you know, closing, locking down society, if we suppress an infection, what if we don't have the capacity to suppress infection? What's the best that we could do? Um, and that's, and it, and it relies on this concept of herd immunity, which has become quite a sort of politicized term in the UK, but it's actually not any, it's, it's a just a scientific concept and it comes from this idea that as the disease is passing through, eventually people will be requiring immunity, and that will slow the disease down. Um, now, if you lower your contact rates, then that will, um, if you, sorry, if you don't lower your contact rates, then actually once the level of, uh, immunity the population needs to prevent the outbreak taking off again, once that's reached, it won't just stop, it will overshoot. So you actually get a higher proportion of the population affected than would be needed to get. Herd immunity, so lowering your contact rates here about on average about 45% on average, we predict would kind of halve the burden, um, not accounting for health healthcare strength, um, from something like 40 million deaths to 20 million deaths, which is clearly like a a massive impact, uh, in terms of the burden this pandemic will cause. But we did find that these, these, uh, and with further impact if you can. Further shield the most vulnerable, so if, if there are the extent to which you can either cocoon older members or protect them from whilst other people are, you know, from make their preferentially protect them. And we could talk about how that might be more difficult in lower income settings if you're sharing your house with 6 other people, that might be harder than in high income settings where that just means not leaving the house, um. And these are clearly quite easy things to model in a in a in a model, you just change the parameter, but in reality, you know, it's it's how you, they're best achieved, it's probably like to vary by country. And one of the things about these mitigation strategies is because of these sort of flat contact patterns, flatter contact patterns in lower income settings, we expect them to have a lower proportional impact in lower income settings because the disease can still spread fairly ably to the to the older more vulnerable members of society. So you know, obviously that's, you know, still a, a fairly horrific burden caused by the pandemic. Um, but also there's, it's, it's gonna have a big, a big strain on healthcare resources. So this is just a map of what we think the sort of under our optimal mitigation scenario, the number of people per 1000 population throughout the pandemic who would require a critical care bed, in other words, would, would require ICU, would require mechanical ventilation, um. And that we are estimating would be higher on a per person level of demand in higher income settings with older populations. But then we need to think about how this demand for healthcare corresponds to supply. Supply is clearly also going to be correlated with income. This is work that was led by Charlie Whittaker, who is one of Azra's PhD students, um, trying to update World Bank estimates to get more sort of fine grain estimates of what the, um. What the current supply of things like the number of hospital beds per 1000 population and within those hospital beds, the proportion of those that have the, you know, the capacity to treat the more severe illness in more severe people with more severe consequences of infection. And as we, you know, it's probably unsurprising, there's a high correlation. The lower income setting, the lower income of country, the lower number of hospital beds and the lower number of physicians and nurses they will have per 1000 population. Not only are the number of hospital beds lower of those hospital beds, a smaller proportion of those used to treat severe cases, um, and actually in lots of lower income settings there's near or. We seem to be hearing a complete absence of these mechanical ventilators which are currently saving a lot of lives. And in these lower income settings, these cases aren't received this mechanical ventilation are likely to experience a very high risk of mortality. The absence of mechanical ventilation. So then what we did is we took our sort of our scenarios of healthcare demand, and we integrated them with our estimates of healthcare supply, in terms of both the capacity for healthcare, er, accounted for how long you would need a bed, and you would need an ICU bed, and the impact that healthcare has upon treatment. Now I must say that we are still working with experts, we're like uh uh Robert was saying, we're, you know, we're we're all trying to catch up on this, but uh the experts to try and understand what some of these parameters are. So what the output I'm showing here is a bit more provisional in terms of the quantitative, um. The qualitative results will be very robust. What we on the top right is a plot showing what we think you would need the sort of the ideal ICU bed capacity, 100,000 in an unmitigated. Scenario and that's the dashed line um and the the dashed blue line and then the dotted um the dotted gray line shows what the typical um ICU bed capacity is within an upper middle income. In a low middle income or a low income country, that should be much lower, um, er and what that, what that will mean in terms of your, uh, the sort of, The burden of these. The pandemic in lower income settings is that you're gonna see this big bump in what I'm here calling preventable, um burden. I, the nomenclature I think is something we're still working on. But um, but basically there's gonna be a lot of people who, uh, would survive in a high income setting that won't survive in a low income. For example, uh we uh our Prime Minister is I think low to mid-fifties. ICU wouldn't be getting that ICU in a low income setting, so we'd be far more likely to have not survived, so we think we think that will both push the per capita burden up in these settings and also reduce the age at which people are more likely to die from it. There's also this this, this question about comorbidities. Now a lot of the comorbidities that have been identified are, are clearly very important, but they're also ones that are most prevalent in the, the places where the, the disease has spread, and those are the more high income settings. Um, and what, what we've been doing is looking at how that varies also by income strata. Now, it, for some of the most uh identified ones so far like uh coronary um like CBD. Actually what you see is that there is a tends to be lower prevalence and lower income settings. Part of that might be lifestyle, uh, but another part is probably survivorship. As you see in the ICU plot, there are, there are probably gonna be very few things like heart bypasses that are made available. So actually your survivorship with these comorbidities is probably a lot lower. So that might alter the severity by age, but also there are other, um, and also those, what we have, the, the commodates we found are very correlated with age, so it gives you a good target for who to shield, and these are the older people, they're also most likely to have er the. Um, comorbidities. So, but in lower income settings there are a lot more potential comorbidities, and we just don't know what impact they're gonna have. Things like HIV infectious disease, HIV AIDS, uh, I have particularly bias on malaria which I work on most of the time outside of pandemic, um, other things like general things like malnutrition, and these are far less correlated with age, far more in younger age groups, far more. and defining what a relevant comorbidity I think is is going to take some time and understanding who is vulnerable and who needs to be protected the most. It's gonna be a key thing to work out to mitigate the onward impact of the pandemic as we go forward. Yeah, so, Thinking about what you can do here, we've we've presented the mitigation scenarios, and these mitigation scenarios are, are things that we think would and have already shown that they will strain healthcare systems beyond capacity if without suppression strategies in Europe. We also think, given that that that will also be the case many times over in lower income settings, um. And that's leaving, you know, countries we anticipate with to a kind of a real big problem that doesn't really have an easy solution. Either they can reduce and go to suppress the disease, suppress the disease and, and hopefully their health system will not be overwhelmed, but that has to be maintained until you get an exit strategy. And an exit strategy might be a vaccine, it it might be some, you know, high technology test or treat scenario, but these are scenarios that are probably less likely to be available in lower income settings, unfortunately. Um, and, uh, and one of the things is if, if you suppress, and then you go back to doing nothing, um, because of these sort of immunity patterns, there won't be any build up of immunity. You do risk a second wave, uh, where, uh, eventually the health system will be overwhelmed, but just at a, at a later date. So, so it's a really hard trade-off, um, and maybe sup suppressing for a short amount of time and then going straight back to normal is, will, is the the the type of scenario it's most likely to. Like, um. Which would then risk having similar impact as if you hadn't suppressed in the first place, and might have a limited impact. Patrick, this is Roberta. I just wanted to give you the warning that um we are at the end of your presentation. Because of the uh technological glitch, I would like to ask you to first speak very close to the microphone. Second, to uh kindly summarize the key hypothesis in a very sort of bullet point type of way. And the conclusion in terms of what, uh, these different suppression measures, uh, can deliver, particular with particular, um, emphasis on low-income countries, um. And then, um, for all of those who have have been patient to hang with us on YouTube, please rest assured that we will share the recording of the event, uh, the, uh, slides. I asked actually, uh, the technical team to see whether it's possible to share them already now, uh, so that you will have all of the material. So before we move on to the next, uh, speaker, the speakers, David Wilson and Mali Jorgensen. Uh, please, um, Patrick, summarize the main point, um, of your talk. Thanks so much. Yeah, no problem. Well, I think, I think basically what you've just said is kind of the last point that we really had to make. It's just that, that the consequences of these suppression strategies need to be thought through, and they're gonna also be particularly, uh, you know, hard to, they're gonna have impacts that are indirect and loweringham saying, so here this is just showing that. There are a lot of countries that have much lower food security, and by suppressing or mitigating, we could delay or or extend the peak, and that then can help to, you know, that could lead to disruption around things like harvest seasons and things like that that really need to be thought through and to balance out both the indirect direct effect of both the pandemic and and the interventions. So, yeah, so, so I'll just go through the last bit, but, but quickly, but yeah, basically lots of countries are at the early stage. We anticipate that they're not going to stay that way. um we can hope that they won't, but I think it's prudent to plan that they will. The epidetitis will take overwhere. We have got some tools that you can look at coming in the next few days that I won't get onto but hopefully go through. But these are just our conclusions. So um yeah, so these tools are available on our website and follow, follow the link and hopefully that will. Give you, you know, what you need, um, but yeah, so, so a lot of lower and middle income countries have yet to see the mortality, level of mortality that we're seeing in Europe. That's probably just we think because they're primarily at an early stage of the epidemic. Um, and we think that that all countries are very high risk of large scale epidemics that are likely to overstretch and probably overwhelm health systems, um, and that will lead to significant excessor morbidity. Low middle income countries, the economies of them are far more vulnerable to the indirect effects of interventions, uh, and shielding and things like that are things that are gonna need to be very actively considered. There's no quick, easy solutions, but we are gonna need to respond rapidly because I think it is going to be the next 2 or 3 months that will really determine what happens in this pandemic. I'll just add some acknowledgements, but I wouldn't have had enough time to go through everyone anyway, we've had so much support from so many partners that I didn't have time to put them all on the slide. Sorry for going over time. Thank you so much, Patrick. Uh, Professor Ganni, would you want to add a few uh quick final conclusions that would hit the intuitive mechanisms of how you got these estimates, and this will be also for the benefit of those who didn't have the chance to see the whole presentation through on YouTube. Thank you. OK, you have a few hopes. Hang on, um, just a few concluding, um, remarks really. Um, clearly modeling's really come to the forefront during this pandemic and people will have seen these models and many of them out there. I, I think the main important thing to bear in mind is models themselves are a tool to improve our understanding. Much of what we do and have done for many years has been to spend a lot of time collating. Um, good quality epidemiological and clinical data in order to parameterize the models, and I think that's a very important part. Um, so much of the projections we're making, they will change over time and they will change because we will learn more about the disease and the epidemic as it progresses. Um, so I think I just wanted to stop with a few words about where we are now because, uh, many countries across the world, of course, have instigated quite, uh, severe social distancing measures, and we, many of us are really faced with the challenge of how do we get out of this? What is the appropriate exit strategy? Um, I, there are no simple answers to this, and I don't think anybody in the world really has them, and I think the models are useful for in terms of illustrating where we could be and what the challenges are with maintaining this level of suppression because without some other tool, um, in particular a vaccine, it's very unlikely given the high basic reproduction number of this virus that we can lift these tools, um, and expect to go back to normal. Um, and when we think about this in a, a lower income countries's context, I think it is important to think now, um, before jumping into exactly the same, um, patterns or strategies that have been adopted in high income countries. There are many. Downsides um to some of these suppression strategies, not least the wider impacts on the health and wellbeing of societies. So this is a difficult um area to address um and will require a wide range of specialists to think more about in the coming weeks. OK. Thank you so much, Patrick and Asra, and now it's the moment to turn to uh David Wilson, who's a a director uh in uh the health, nutrition and population practice of the World Bank, and uh our colleague Marreli Gordon, who's a senior monitoring and evaluation specialist in the same global practice, and that will tell us. They will share reflections exactly on exit strategies. So this is a tough question, and I appreciate the humbleness of the humility of saying we don't have the answers and we need to think through in advance whether these tools that have been applied in rich in advanced countries can be applied. Successfully, incredibly in low income settings. So thank you for now, and David and Marelli, the floor is yours. I will have to ask you to be as concise as possible because we would like to leave some time for questions before we end at 10:30. A reminder for those on YouTube, you can ask questions, and I know that you've been shortchanged. by the technology Glitch and those on Webex, please ask your questions in the chat or raise your hands. Thank you. Over to David. I think that my comments flow nicely from what's been presented and I hope I can compress it to about 5 minutes and claw back time for discussion because I'm building on what's been said. We've tried to look at some considerations for how countries could reopen, and we've tried to do that with as much of an orientation as possible towards lower and middle income countries because so much of the literature and experience is the Northern European or East Asian and has some but limited applicability, and we've really identified about 7 considerations, many of which have been touched on. The concept of epidemic force, obviously we don't see you, um, this might be by design, you might have chosen not to show yourself, but I think you can see me now now it's so much, now it's so much better. Thank you. So we have about 7 considerations which have been touched on. The notion of epidemic force, obviously primarily reproduction rate, but also looking at super spread of potential and also looking at the extent to which overall spread is focused as it has been, for example, in China, Italy, and the US or dispersed and based on those considerations, it may be possible to remove lockdowns earlier or later and partially or more completely. We've also, as people have touched on already, looked closely at population health. Obviously age, and obviously we think that Africa's comparative youth, Uganda's average age is 16.7, Italy is 47, Lombardy's, as you know, Roberta's even older. We've also looked at issues such as obesity, NCDs, uh, smoking, and other risk factors, which obviously are positive for Africa, but might be offset by exposure to indoor cooking pollution, by widespread immunocompromisation, and also, um, potentially by malnutrition and other factors. So the picture's mixed, but perhaps a little bit positive. We've looked closely at what we would call the capacity to take targeted measures, which is difficult, but it does depend a lot on countries' epidemic experience, population preparedness, and population. Willingness to take measures and I think we see that throughout countries in East Asia which have managed this extremely well, Japan, uh, Taiwan, for example, South Korea, Hong Kong, Singapore, parts of China, and we've looked closely at the concept of public health capacity, which we all understand and is going to be central to manage our way out of this epidemic. We've had difficulty looking at health service capacity because ideally we would like to have both fully protected health workers, safe health facilities, and an adequate surplus of critical care capacity. The challenge we face is defining what critical care capacity means in very low resource countries such as Madagascar, for example, who have an estimated 7 to 12 hospital beds. We focus a lot on state capacity, particularly decision making capacity, and we do see that higher capacity states have managed their way out of this better so far. But we also think the role of decision making is critical. We've had to elevate decision making to the highest levels to succeed. We've had to be able to take data from real-time digital sources, from modeling, from economics, and make very complex decisions, which are probably a function of state capacity. And then we've also looked at the extent to which innovations would help, and the innovations that are closest to hand are probably incremental innovations and testing, which is still really important. Potential innovations of a simplified treatment, not yet at hand, but potentially not too, too far away, and then the obvious hope of vaccines. Based on these considerations, we looked at the literature and found about 6 papers on how we reopen. All of them are for high income contexts, but they still have some useful principles. And we then really took a consensus for principles for how you might prepare a roadmap, tried to see how we could tailor it for southern context, and then tried to apply it to categories, which I'll discuss. And The obvious point is if surveillance is strong and cases are either very low or falling, then Everything else becomes easier and we can move towards relaxing measures or reopening much, much more quickly. So everything really hinges on this, and for much of the world we simply don't have the data. One striking point for the World Bank is that only 2% of all reported cases are from IDA countries, and I suspect that IDA countries do have fewer cases, but I'm certain they have much fewer testing. We've also focused on the role of microdata because we think managing our way out of this epidemic is going to mean having surveillance, having testing, having public health case finding, tracing, isolation, quarantine measures at the local level, and obviously we focus on the extent to which there's adequate public health capacity to try and eliminate linked cases as one criteria for countries to open. What we grappled most with, as I've said, is health service capacity, but we've tried to focus primarily on health workers being completely protected and health facilities being safe for patients. We're still grappling with what critical care means. And based on these considerations, we've tested a very simple notion of countries that may already be open or could be opened very quickly with limited intermittent or targeted measures. We've also tried to look at places that might be able to reopen in the near future with the same limited measures, and in places that may actually need a lot more information or a lot more progress. And this is entirely to test a concept. It's in no way to provide national advice or to contradict national advice, but if we're to look at countries that are already open or might open with limited measures. We actually have a lot of countries in East Asia already in this category, um, countries and territories such as China, Hong Kong, Taiwan, Singapore, South Korea, Japan, Vietnam, Australia, New Zealand. We think there are a cluster of countries in Europe that meet those criteria, Germany and Austria who are moving towards this already, the Czech Republic, Slovakia, and Iceland, for example, uh. When we look at the next category of countries that actually could reopen quite soon with limited and targeted measures, we see quite a lot of islands who have an extra ability to insulate themselves from the wider world and reopen internally. And we think that, uh, Sri Lanka, for example, is a good example of a lower middle income country that could reopen substantially very soon, as is the smaller and more isolated island of Mauritius, and potentially also promisingly the very low income isolated island of Madagascar. We also think there are 20 or 30 islands in the Pacific who've locked down and if testing confirms low rates, could potentially rely on external circuit breakers. But yesterday we got a very encouraging presentation from South Africa's CapriSA modeling Group given to cabinet, which actually suggested South Africa is now plateauing and peaking at an earlier level than expected and it's diverging from the pessimistic European scenarios that they had initially used, and they're showing limited community transmission. Now, if this holds, South Africa may be an example of a continental African country that could actually progressively reopen now. And just to emphasize, because of its very large installed TB program and gene expert capacity adapted with CAF cartridges, um, South Africa's tested on a large scale, and other African countries are also expanding testing, and some of the glimmers of data we're getting are moderately positive. But then finally, there's a whole swathe of the world for which we need much more testing data and much more progress. And I would highlight our concerns about much of South Asia, except Sri Lanka, the large populous countries of Southeast Asia, and in particular Indonesia and Philippines. Much of Africa for which we have very little data and much of South America. And just to note that if there's a seasonal dimension, winter is approaching in Southern Africa and South America. And this is where I'm gonna really conclude by saying, I think that we started in a very different place to colleagues at DEC, but I think we're converging because the obvious point. Preceding any decisions about reopening is expanded testing, a point that was highlighted in all of the presentations, and I think if we can do that, we'll have a much better idea of how we can consider reopening, but I do think we're looking at very partial, very intermittent, uh, openings with a lot of internal circuit breakers. I think circuit breakers and fire stops are going to be the ways that countries can try to cautiously reopen. I can't underscore the stakes here. Any of those of us who've seen the remarkable pictures of large scale protests that verge on riots coming from Mumbai today will know how explosive the situation is. Let me conclude there and hopefully claw back a bit of time. Thank you and over. Thank you, David, uh, is Marielisa coming on after you, or have you spoken for the both of you? I hope we'll be online. She's, um, presenting the Lesotho project today and was in a meeting with the country, so we had to shuffle our order. So can I just ask to ask if Marilli made it? Um, otherwise, Roberto, are you online, Marilli? I am. Thanks, David. Over to you. Thank you. Excellent. Thank you. Thanks, David, for the overview and to Patrick and Azra. Um, I know we've had many email discussions. It's, it's good to be connected in person, hear your voices as well. Um, so, a few points on my end about the use of mathematical modeling in this, in this space, maybe just to tie the two presentations together. Um, first, to say that I think we'd all agree that mathematical modeling, both epidemiological and predictive analytics, um, have proven essential in raising awareness of the scope of the epidemic to countries that have gone through it to help them prepare for it and plan for shortages. Um, This is, you know, there are many things that are unprecedented, but the unprecedented use of models, the amount of times that I've seen models being discussed and model parameters being discussed in, in the media, we, we've just never, never seen this before. So, I, I think, you know, I would say that models obviously play an important role, and at the same time, and I think this is a healthy thing, they have received scrutiny also that maybe, um, you know, epidemiologists and infectious disease modelers have also not, not seen, seen before. I think one of the particular areas of scrutiny has been around the use, usefulness, and interpretation of statistical models versus these um more mechanistic SIR um Infectious disease models and clearly the questions um raised in the chat today reflect, reflect that or not. I think it is important in all of this that we recognize that there are many unanswered epidemiological questions that we need to, um, still answer and that the, these will be impacted on by the models and vice versa. Um, the asymptomatic and pre-symptomatic cases proportions in the population. Um, duration of infectiousness, duration of infection, um, and duration of infectiousness after infection is cleared, level of immunity, duration of immunity, viral load, and its relationship to both infectiousness and, and immunity, um, acquired long-term immunity. Um, and then most importantly, as David alluded to, the effect of the in interventions, not just nationally applied interventions, but granular, really detailed, really detailed interventions. Whilst all of these, there's all of this uncertainty that we have to account for, policymakers have to make decisions and revisit them frequently, deciding when to um have stay at home orders, when to release them, how to release them, in which phases for which populations, in which areas, which geographic areas they can be released for which kinds of industries. And so, these models are going to be essential for this next stage of response management as well. To do that well, they need to be granular enough and they need to be adaptable. That means that Whilst model choice is important, what is even more important is adapting, adapting, adapting, recalibrating these models against the data that are available. Um, we know, for example, some of the model groups does this twice a week and have a schedule that is published to say we will be Iterating constantly, um, because making decisions and deciding on, on Rn effective, effective reproductive rates, etc. that is affected by the intervention itself is, is critical in, in what the entire, the entire model, um, looks like. Um. So, whilst modeling has been useful upfront in estimating the overall epidemic potential, estimating, therefore, the healthcare needs, um, the The important need is now shifting to assessing the potential magnitude of the different measures, and these are measures that are going to be implemented, not implemented, implemented, not, not implemented, right? So, models can and should be essential to answer a wider range of efforts, a wider range of questions. In that, this is Roberta. Let me just give you, unfortunately one. Minute warning because we have run vastly over time today. Thank you. Absolutely. So there's 3 things the bank is doing to help with this. First of all, we're co-convening with IDSI, WHO and the Gates Foundation, a global COVID modeling, um, rapid assessment effort to help countries understand which parameters go in and how they can better use models. Secondly, is we're developing Social distancing and stay at home impact dashboard for every country that will be available on the World Bank website. And thirdly, a vulnerability dashboard that will help countries to understand at a granular local level where are the areas of highest, highest vulnerability, and we hope that this will be useful and impact on these models themselves. Over. Thanks, Roberta. Thank you, Mali, and thank you everyone. It's a moment, we have about 8 minutes for Q&A sessions with our speakers. Just by way of summary, uh, Damien and Jed gave us a nice overview of the key concepts uh related to the epidemiology of uh COVID-19 and an overview of the mathematical models that are out there that can help us make these predictions. Um, Patrick and Azra presented the predictions of their model, and just to give you a quick snapshot of their conclusion, they show that mitigation strategy that is based on partially shielding the elderly, every year 60% of reduction in their social contacts and slowing down without completely interrupting transmission in the general population, so with a 40% reduction. Would have the burden of death in the world from 40 million to 20 million. They find out, however, that in low income countries, the impacts are likely to be much stronger, and a data point that I see here in their paper is that in a typical low income setting, the demand of critical care beds would outstrip the supply by a factor of 25, while this factor is only 7 for high income countries. Their conclusion was that there are no easy answers, and then essentially we should, uh, buckle up for, uh, long periods of measures of containment if we don't want that, that the virus comes back. Um, David and Marlise, uh, have reflected on, uh, a sort of matrix of, uh, the intersection of country capacity and possibility and options for. For reopening. Clearly this depends on the epidemiology and type of spread of the virus because one of their conclusions was that islands that are able to maintain the spread low within the country could reopen relatively sooner internally while still maintaining their closure of the border to avoid new infections. Let me start the Q&A session with a question for, from myself, uh, which is also related to one of the questions that we got via YouTube from George. So, The question from, um, I believe, a question to Patrick was, does your model assume long-term immunity, and does it consider the slim possibility that the mortality rate might be overestimated because of that? My related question was, uh, as epidemiologists, What do we, what do you understand about the acquired immunity to COVID once, uh, somebody has contracted the disease and then becomes negative? Many of the reasonings that are being done these days is that um we hope that with antibody testing we'll be able to identify who is not contagious anymore and who could be safely employed to interact with the vulnerable and elderly, but is the immunity real? So let me pass it over to Patrick and Azra with these initial questions, and then I'll come back with an extra round. Over. Excuse me, yeah, so I'm, I'm afraid, uh, Azra had to leave, uh, but, um, I'll do my best in her absence. Um, in terms of, uh, uh, let me try and remember them in order, in terms of severity, um, I think that that, uh, has been a key question because like I said at the beginning of the talk, there's the data that you have to parameterize that is quite limited, and I think. We, along with a lot of the world have been waiting for some really good serology, um, surveys to, to start coming through. I think we're starting to see them and, and I think the picture is um broadly consistent with the level of severity, um, that, that, that was estimated. Um, it could still be some, some recalibration and readjustment that needs to happen once we do some formal fitting. But, um, but yeah, that will certainly be updated as and when, as and when we can. Um, but, but clearly I think, I think some of the, some of the, uh, kind of, uh, more hopeful estimates of this idea that there might be really high proportion of the population who've already been infected in London, for example, I think, uh, and in the UK I think are, are, are not, sadly I think aren't bearing fruit, and then it does look like we have a way to go. Um, and the second is the affected population immune? Oh yeah, oh, so, so that's a shame that I've just gone because I have to confess that I'm, I'm not a virologist, and we have a big, what we do have is, is a wider, um, group, a much larger response group, and also that group, um, we, they go to, um, in the, at least in the UK we go to a scientific advisory group, uh, where you have a lot of expert virologists and people like that, and that's where we get our assumptions from the modeling and parameterization. I think it's, it's worth saying that, that, um. But yeah, it's, it's, it's those data again you can hopefully reparameter and could be important um like anything at the population level you can see some levels of reinfection but then the, you know, if they kind of outlying cases that don't necessarily mean that people aren't broadly making immune responses, but yeah, again it's, it's, it's, it's um yeah, early days. Thank you, Patrick. Let me see if anyone from Webex uh has questions or wants to raise their hand. Um, Alec can unmute you if you would like. Yes, there's a question from Dimitri. Um, how is contact rate related to our, uh. Um, OK, let me just, Dimitri, would you like to, I just muted you. Would you like to? To ask you a question? Uh, yes, uh. So, I'm trying to figure out how to model this, uh, how to calibrate this model as well. So, I, as I understand, the contract rate is related, directly related to RR0. And so, in the, in the model, I assume that uh the countercreatives are not divided by the length of the disease. I, I assume there is 18 days. Is it a fair assumption to make? Oh, OK, so, uh, I think, uh, well, it, what you, what we can do is you, you and I can, can talk offline and I can take you through exactly every assumption we're making. Um, so the, so what you're saying is, is pretty much in a, in a microcosm, correct? You, you, your, your R0 is the, uh, is something is the rate at which you infect people per day multiplied by the number of days you stay infectious. So you want to sort of account for how long you're infectious for, um, and, and so what generally what you do is you, there are two components that go into your R0 estimation. There is the exponential growth rate, so sort of the doubling time, so a different doubling time of 3 days or 4 days or 5 days will have a different. And also the duration of infectivity, cos that tells you the number of generations you're going through within that, so, so if you're seeing much faster growth, but fewer going through fewer generations, that means more infections per generation, a higher R00. It's getting a bit technical, but, but, um, but yeah, we can, we can walk through that, it's with you, your basic in terms of your duration, there's also the, well, at least with Rmoni you have to account for the fact that not everyone makes contact at the same rate. That's where our age dependent comes in, and that's where, yeah, you, you have to calibrate to account for that, and that also has onwards impact or severity if your severity depends on age, which it does, um. The one thing is the duration of infectivity, we assume a lot shorter than 18 days. Now 18 days is how long it takes you to, you can have if you have severe disease, you can experience the um. Experience the. That's a keyla of the disease for a long time, but that doesn't necessarily mean that you are infectious and it doesn't necessarily mean you're representative of all infections, so I think at the moment we're assuming a sort of an average infectious period of somewhere in the order of 3 days, but with a lot of variation. Um, and that depends a bit longer if you have a more severe long lasting. You have to cancel the much less, you know, less severe cases as well as the most severe. Thank you. Patrick, and thank you for the excellent question. We have arrived at 10:30 and it's time for us to conclude. Um, some of you have asked whether this is published and all the parameters and the assumptions are available for people to work with. Patrick, would you want to indicate where people could find all of this? I understand you, they are in detail. Yeah, there's a GitHub, yeah, which, and we're making things, yeah, it's, it's, um, yeah, we're going as quick, but yeah, the GitHub's there, so that then you can just recreate everything in our report. I will, I don't know where, where's the best place to put that Roberta, um, the link. Well, we will circulate to everyone the recording. Yeah, what we can do is that we'll circulate to everyone the recording of the event, uh, your presentations, and with that we could probably circulate not only your paper but also the link where people can, uh, access. I remember that reading your paper that was within the paper an Excel sheet link where I could just in there and. Yeah, play around. Yeah, and just, just to say that everything will be posted, the recording and the documents and the presentations were going to be posted in the e-seminar's website. So they're already some of them posted. And we will have the recording, um, hopefully tomorrow. Perfect. This is excellent. So I wanted to thank um our speakers, uh, Damien De Valke, Jeff Friedman from the World Bank, who gave us an, an overview of the models that are uh available. Also in their presentation, you will find links to the modernization of the AHME Institute, the University of Basel, to Patrick Walker and Professor Azragani for uh their overview of this influential model, which is the one. which we've all have based our fears and sort of conclusions, and thank you also for the sobering and humble approach that you're taking to the mobilization. Thank you to David Wilson, Madli Gargens, because they helped us reflect about potential like these strategies, and thanks to all of you for your interest, enthusiasm, and patience. The fact that I got so many text messages about I can't listen to the. Presentation, I want to see it. It means it was really um providing an important um added value for all of us who are trying to think about how best to suppress and stop the spread of this disease. So thank you everyone, and uh looking forward to the uh other upcoming uh great seminars that uh the Development Research Group is putting together for us to understand better the impact of coronavirus. Thank you very much. Bye. Thank you, Patrick
showAllTranscripts
no
duration
PT1H33M53S
scene7File
worldbank/Understanding the Epidemiology of COVID19
scene7Domain
https://worldbank.scene7.com/
scene7FileAvs
worldbank/Understanding the Epidemiology of COVID19-AVS
title
Understanding the Epidemiology of COVID19
description
Understanding the Epidemiology of COVID19
showTimestampAndTranscript
yes
col-xs-12
col-sm-12
col-md-3
col-lg-3
col-xs-12
col-sm-12
col-md-7
col-lg-7
  • add-style
  • lp-body-content
According to estimates published by Imperial College London, in the absence of any intervention COVID-19 would have resulted in 7 billion infections and 40 million deaths globally this year. In this e-seminar on April 14, 2020, staff from the World Bank and Imperial College London discussed the basic underpinnings of the epidemiology of COVID-19, current epidemiological models of COVID-19 transmission, and the effectiveness of various non-pharmaceutical interventions in reducing transmission—with a focus on lessons for low- and middle-income countries. Timestamps: 2:13 - Jed Friedman (World Bank) and Damien de Walque (World Bank); 21:54 - Patrick Walker (Imperial College London); 127:10 - David Wilson (World Bank); 137:04 - Marelize Gorgens (World Bank)
lp-heading-top-medium
lp-heading-bottom-medium
col-xs-12
col-sm-12
col-md-2
col-lg-2