 Well, good afternoon. I'm Stefan Obertozzi. Welcome to those of you who are here in the room and those of you who are following us online I'm a professor at the School of Public Health here at UC Berkeley, and it's my great pleasure today to introduce to you your Speaker, but before I introduce the speaker. Let me just tell you a little bit about this Hitchcock lecture series Whoa Now I'm hearing myself back on on YouTube So the Hitchcock lectures were established Back at the turn of the last century in in 1885 With the Hitchcock endowment fund that was established with a bequest made by Charles Hitchcock to institute a professorship at UC Which was then all of UC was here for free lectures upon scientific and practical subjects his Outer Hitchcock his long-standing interest in education inspired this contribution He was a physician and served as medical director of the Pacific Coast in the US Army Medical Corps The Hitchcock lectures began a few years later in 1909 and were later expanded and retitled After Charles and Martha Hitchcock Thanks to a generous bequest from his daughter Lily Hitchcock Coit and some of you who are familiar with the Bay Area Will recognize that name because she also paid for Coyt Tower in San Francisco In honor of all the firemen who lost their lives in the in five the five fires so today we have the pleasure of having with us as yesterday Dr. Christopher Murray and Chris is the chair of Health Metric Sciences at the University of Washington And he's the director of the Institute for Health Metrics and Evaluation often known as IHME Which I would say is unquestionably the powerhouse of global Epidemiology and houses a really important project called the global burden of disease His career has focused on improving population health worldwide through better evidence He's a physician and health economist His work has led the development of innovative methods to strengthen health measurement Analyze the performance of health systems and understand the drivers of health and produce forecasts for the future state of health his Institute houses a Various training programs including a doctoral program in health metrics to my knowledge the only one of those He's been very made very important contributions with his extensive team in Seattle To analysis of the COVID pandemic and its impact on health systems and the population But as I mentioned before he's probably best known for his contribution to the study of the global burden of disease Which I remember back, and I think it was 1991 in preparation for the world for the global for the World Bank's Global Development Report and Festing and Health Chris was then at Harvard if I'm not mistaken and he was Leading a team that developed the disability adjusted life years which really revolutionized the global measurement of burden of disease because it revealed to the whole world the importance of disability in addition to Mortality in in the global burden of disease He's he continued that work initially at Harvard then he went to the World Health Organization where he further developed those and created with Julio Frank at the time the the global health report and then moved to To Seattle where the University of Washington hosts this Institute for health metrics and evaluation The global burden of disease effort now has over 8,000 collaborators in 161 countries and is the definitive reference for Burden of disease globally He's an elected member of the National Canterbury of Medicine And he was the 2018 co-recipient of the John Dirks Canada Gardner global health award And that's a very small group of people who are in that In that club so Chris welcome again to Berkeley. It's a pleasure having you here and the stage is yours Can you hear me okay? Thank you very much Steph for the introductory comments and I'm going to jump into a very different topic than what I spoke about yesterday and This is a bit of an experiment and I'll give you a little bit of context There is going to be or about to be launched a Lancet commission on 21st century priorities or health global health priorities the idea is There's a lot of interest in pandemics is a lot of interest in climate There's a lot of interest in what obesity? Epidemic might do or what conflict might do to health and a long list of challenges that we face and the idea of this new commission is To try to look across these threats Once it gets up and running which will be coming next year and Look for some common solutions and so in the sort of preamble and I will along with Natalia Kahn. I'm at UNFPA Co-chair this commission and so I've been spending a lot of time thinking about what's coming and how the Our understanding of the past 70 years informs the next 70 years as well as Where are the main avenues for ignorance uncertainty and some of those also translate into opportunity so a first-pass work in progress view of What are the coming threats? So this diagram which I'll come back to in more detail looking from 2020 forward But this is a figure for the last 70 years from 1950 to 2020 or 2021 actually For life expectancy, so this is not healthy life expectancy. This is not dally's This is just straight life expectancy at birth and it's showing it by Region so the high-income region is the black line at the top the sort of brown line that started near the black line is Eastern Europe and Central Asia and went through Basically a 50-year period of stagnation and then has now returned to Recently some growth in life expectancy and then you see places like Southeast and East Asia that have caught up dramatically During the period of the 70 years you see sub-saharan Africa in red Which was always Behind the other regions and then dropped even farther behind particularly due to the HIV epidemic And then it's been in a period of catch-up and you can see the jagged dip due to the COVID epidemic so there's lots of interest here in the past and I think lots of people think about The future with some degree of pessimism So to jump to the conclusion or one of the first conclusions, which I'll try to back up in a little more detail our modeling of at least the next generation out to 2050 and also when we run these models out to 2100 not shown is One that is less progress than the past 70 years, but still progress and so You know spoiler alert the even though we face many challenges Our current view and I'll give you lots of reasons why that may be wrong is that we will actually continue to see a Period of progress during the rest of the century despite the many challenges we face Little deeper dive in the more recent time period because it's a little bit of the warnings that we see out there That you know, maybe we shouldn't be as optimistic as I'm saying This is change in life expectancy in the last decade Excluding COVID so 2010 to 2019 and the warning sign there's you know there's a number of countries that look red or orange where life expectancy dropped and Those are places that had conflicts and there are places like many states in the US and some states in Mexico Where life expectancy actually dropped over that interval and so there are other things going on That are adding to those declines and more on that later So it's not simply that we've been on this steady treadmill of progress We had the HIV epidemic came up with solutions and now we've had COVID We've actually start to see slowdown and reversals in other places as well This is the sort of where we stand in 2021 This is life expectancy at the country and subnational level where we do subnational monitoring Globally and so we still live in a world with enormous disparities Both within countries, but clearly across countries with low income or low resource settings particularly in sub-Saharan Africa with life expectancies still below 60 Some even down below 50 although that is now quite rare and life expectancies at the higher end that go up to 85 Forecasts so here's that same diagram. I showed earlier our forecast but now dividing before women or On the left men on the right and for both groups. We expect life expectancy after the recovery from COVID big question mark on that to grow to Improve and there should be some narrowing of disparities But despite that at least in the next generation the models suggest and I'll try to dig into why They'll will they'll be catch up for Sub-Saharan Africa, but still huge disparities So we won't see any we're not even talking about the SDG era. We're talking right out to 2050 We still see very marked disparities around the world Here's a little bit of a deeper look by males and females on the same plot now for the world high-income countries Eastern Europe North Africa in the Middle East South Asia was really marked progress Southeast Asia and East Asia the same more of a slowdown for Latin America and then you know catch up in Sub-Saharan Africa, but not enough to close the gap Here's models around causes of death and I'm showing you now Numbers and we can of course look at these at rates We can look at these at age standardized rates But just in terms of sheer numbers of what might happen broken down over the next generation Into large causes so at the global level There's going to be a very large increase in the number of deaths due to cardiovascular disease That's mostly population growth, but even more aging of the world's population and then Modest declines in some increases in some countries in age standardized rates similar increase in cancers similar increase in neurological disorders in bright green and perhaps disturbing for those who work on maternal and child health or particularly Tropical diseases if you look at the bottom right, which is our death forecasts for Sub-Saharan Africa There's still a disturbingly large number of infectious disease deaths in the bottom three colors Bet right out a generation to the future Even though the rates are going down, but you have to deal with the fact that there's really marked Whoops Marked population growth. Thank you. I'll try not to move Now population matters a lot. We published a paper that generated quite a lot of controversy About two years ago About alternative population forecasts where we took our modeling framework both for The causes of death age specific mortality I'll explain what's in that There's an awful lot that goes into that model But we also modeled fertility and what we found out in our model for fertility was that we had a very different view Then the world population prospects produced by the UN population division About the the course of fertility in the coming century and when you work that through into population forecasts In our first pass at this, which is the dark green line the published one in 2020 We had global population peaking about 2065 in the version We are going to publish later this year or early next year Global population because there's been even faster declines in fertility in places such as India We see the world's population peaking in 2060 Now this has knock on consequences for people who are interested in things like climate of course and more on that later now When we run this model, we have a very unusual model Compared to what demographers have historically done for all cause mortality. We model Specific causes both for death and also for dallas. I'm showing you so far just the death results and We do that by first removing from the past the effect of all the known risk factors that we monitor in the GBD And they're known relative risks and then modeling what remains Which we call underlying or risk-deleted mortality as a function of some simple covariates like education income And then depending on the calls if we're thinking about road traffic accidents, we put in things like the number of cars For modeling into the future the risk attributable we forecast each of the risk factors using and I'll show you how we do that and Then we add on to the underlying mortality the forecasted future risk Associated mortality and that allows it us to make a huge array of scenarios by adjusting with the trajectory of these risk factors There's some 88 risk factors. We include in this And then we have an arena type model for what's unexplained in the background So, you know, here's a visual of all the variables that we forecast and that goes into the model Not everything in the GBD is included. So we don't have occupational exposures but everything else is modeled forward and Allows us to sort of think of how that particular factor may alter The trajectory of health now in some cases more limited than I would like we include Specific interventions we treat them like a risk factor in this analysis. We strip out from the past the effects of vaccination We do that for ART Is another example and then there's a small handful where we also look at the where the evidence on the time series of Intervention is strong enough that we can treat it like a risk factor in this framework Now how do we make those specific? Forecasts, let's say for tobacco consumption or For any of these risks and we use as is perhaps no surprise in ensemble of models Which are driven by either socio-economic development or just by time and Are weighted to some of the models in the ensemble to more recent time trends and we then do an out-of-sample predictive validity Exercise to figure out what are the weights on these different models in the ensemble and that generates our Forecasts for that particular risk And you can see an example there for a particular place for water and sanitation The black line is the true past and then our forecasts are shown with uncertainty Now there's a big debate about our model because people in the demographic literature don't like the idea of making explicit The drivers of future health they prefer to use time as a Coral it for all those future changes and You know, that's a very reasonable strategy to use it has real problems though If you think that there are as we observe for let's say obesity or for drug, you know opioid use abrupt changes in time trend So if you hope that the average correlation of all known drivers with time in the past Will give you a good approximation of all those drivers in the future That may not be a reasonable assumption or and certainly not one we think is reasonable The other advantage of modeling the explicit as opposed to the implicit drivers using, you know time as your implicit Coral it is that you can then use these models to answer specific policy questions You can try and change tobacco and see what it does and the the advantage of using the global burden of disease relative risk functions for each Risk is we're not trying to do cross-country panel regression to estimate the effect of tobacco and lung cancer We're getting that from individual level cohorts or in the case of blood pressure We're getting the relationship between blood pressure and ischemic heart disease from trials So we use higher grade evidence on causality to get the relationship between the risks and the outcome and Allow that to be built into the forecast framework So in summary our reference model says that things will improve Inequalities will narrow but stay quite large Population age structure outside of Africa inverts Meaning that we go from an age pyramid that was like this in 1950 for Portugal to like this and now inverts And so this inverted age pyramid has a lot of fiscal economic and social consequences And that'll be a theme that I'll touch on in a bit There's a huge shift towards NCDs Accelerating a shift we've already been observing for the last 30 or 40 years and Slowest progress which I'll dig into in some detail on many fronts as well as the greatest risk for some of the threats on the horizon Is in the Sahel Okay, now I'm gonna go through a whole series of factors that may make those forecasts wildly incorrect and Also create opportunities for us on a policy framework to think about ways to Hope that we can steer the course towards, you know the better scenarios in a way from the worst scenarios And I'm gonna run through these and this is going to be quite discursive because this is really Work in progress So since COVID is in many of our minds quite dominantly and even if it fades rapidly it's still something that we should be thinking about pandemics and the way our model works is we are just sampling the these events whether they're earthquakes pandemics wars From the distribution of those events for a country in the past So it's a absurdly simplistic view Which basically says your past experience gives us some information about the probabilities in your location For future events So there's a tale of the distribution if you've had a conflict or you've had a big COVID impact in our model that gets Spread over each year in the future Because of the way we do that sampling now If you want to think about ways both to modify that risk as well as what might be policy options or drivers We need to distinguish things that may be relating to the event rate in the case of pandemics Perhaps zoonotic transfer could be bioterrorism The factors that go into the initial spread The one that probably got us the worst in the globally for COVID Which is not acting when the signal became clear for a very long period of time And so there is this response delay and effectiveness of response And so if you think back to COVID was it slow detection or or slow response And I think arguably if you go and look at the pro med report for a respiratory process in Wuhan That was on the 27th of December And it took an who 35 days to require a public health emergency of international concern And it took again another 20 odd days or 21 days for Italy to lock down Lodi Province and then the rest of the world started to act in March So Detecting the COVID two or three days earlier Probably wouldn't have changed much because the real issue is what happened in those first five to ten weeks And why did we not act in a more Coherent way Everyone wants to know for thinking about future pandemic risk about Dispreparedness help so Tom Boyke Joe dealman myself and many others at IHB and Tom Boyke said this at the council on foreign relations led the work analyzed in published in the Lancet this year relating a host of preparedness measures JEE scores the global health security index to Infections per capita or the infection fatality rate and there was no relationship at all For those now there's subsequent analysis that suggests that maybe In high-income countries with an income over I think $20,000 per capita or $15,000 per capita Maybe there's a relationship, but not in the rest of the world However in looking at things that might predict The what happened with COVID it turns out that the two clusters of things are predictive None of the preparedness measures, you know lab performance, etc But corruption predicted worse transmission And interpersonal trust turns out to be the best predictor So societies where trust is high Perhaps following the evidence available at the time Ended out with lower cumulative infection Now everybody wants to know what do we do and can have we learned from COVID? So therefore when we think about the future of pandemics, are we going to be much better off because we've had this horrific experience of probably 20 million dying so far by the latest account of excess mortality And we surely have figured out how to manage These pandemics in the future and so one of the big questions That everybody is interested in and is very hard to figure out Is did social distancing mandates have an impact? And if you do the traditional analysis, which is on the left-hand side, which is daily infections Against intensity of mandates and this is one measure of all the mandates together there is There is a statistical relationship. It's not particularly great And then if you look at cumulative which gets around endogeneity and that's a separate discussion To a substantial degree You end up saying yes mandates really had an effect if you want to ask which mandates It's very hard because the co-linearity between all those mandates is so high So lots of efforts to disentangle the mandates none of them are very compelling or convincing And we may end up only concluding that mandates can stop transmission And you're going to have to use common sense and other strategies to figure out which of those mandates you want to use For the next pandemic Okay, moving to number two, which again is very much in everybody's mind because of the invasion of ukraine Is conflicts that are have More than local consequences. There's obviously been conflicts for Over the last 40 years On the right-hand side is the number of deaths due to conflict year by year The big spike is the genocide in rwanda But you can see that the number of conflict deaths at the global level has remained below 200,000 except for the genocide And even though there have been conflicts and this is only through to 2020 on this graph Uh, we haven't seen, you know, a steady increase of conflict deaths And likely we won't see, uh, we probably won't get over the 200,000 mark, uh, even with the more recent conflicts However, the consequence of the conflicts where somebody is a major producer of food In fact, there's considerable food Net food imports into sub-Saharan Africa from the ukraine as an example Can be very substantial and so our very simplistic approach of just sampling You know your yearly death total From, uh, conflicts in the past Is not going to capture the potential for these global effects of certain types of conflicts And it's, you know, as you think ahead or we think ahead in terms of coming up with recommendations about What on earth could the health world contribute To discussions on Mitigating those risks That's a daunting task and i'm not sure That there's any obvious answers there to some groups the Conflict in the ukraine Is particularly disturbing because the un permanent members of the security council the p5 are involved You know with russia being a p5 member and it then raises the fundamental question about the effectiveness of the security council Even for its own permanent members of reducing conflict Alrighty coming back Okay climate one that I think everybody there's a strong interest here at berkeley lost strong interest globally We have built climate into the forecast. I showed you the optimistic ones And very specifically we use the goo at l forecasts of global mean surface temperature Which is pretty close to rpc 4.5 We use c-mip 5 so we're a little bit out of sync with the release of c-mip 6 And we use nasa's downscale of the c-mip 5 to figure out temperature in each locality It's sort of crazy, you know temperature for every day in each locality out to the end of the century And translate that into what it means for health Um Lots of reasons huge uncertainty in that but you know, obviously most people tend to focus on the mean value of these things And they are built into our climate modeling now First big source of uncertainty is the risk functions, which we published last year And added sub optimal temperature to the gbd as a risk factor And here's just examples of those risks For one particular well for actually six large causes pneumonia Schemic heart disease stroke c-o-p-d drowning in self-harm And then we end up with different curves Based on your mean annual temperature because it's very clear in the data that Places that are the mean temperature is higher that whole curve shifts to the right So the the bottom of the risk curve Actually tends to be very close to the mean temperature where you live Reflecting both physiological social and economic adaption But those get factored in so we've built into our forecasting The You know adaptation will slowly happen as mean temperature goes up And despite that we get quite considerable mortality from from heat in the future However, I think there's a lot of debate and discussion to be had about Uh, which causes of death Got in to this analysis met our inclusion criteria And then what do you do on when temperature goes beyond the observed Range that we had available to fit these curves to And that gets to this really key question of habitability There is not a very good statistical analysis Yet of habitability i.e There are when we analyze risk we are excluding all the places where Temperature is so high people don't even live there, right? So you have this sort of missing giant missing data problem and One of the limitations of just running a very simple habitability model Is that we don't actually currently in the world Have a good map of where people live, which is sort of extraordinary for 2022 You would think we would know where people lived around the world It's going to turn out to be there are attempts at it They're they're just not very valid in the validation exercises And we really desperately need that to strengthen both our understanding not only of habitability, but you know food stress Which communities, you know think the sahel or which specific places will be the first to be affected by climate stress Now the other thing that we want to recognize up front is that In the climate world to date There has not been an attempt to forecast Probabilistically what are the likely policies that will be in place over the next generation or two You just specify the policy and then run a model forward and say what that does And fortunately there's a group here with very strong links to berkeley the climate impact modeling group that will Hopefully fill this niche for us and we will adopt when they release their probabilistic Model of mitigation policies and build that into our future scenarios But probably that won't change our reference scenario from what we understand from them Okay, so Climates there it's in the model. It's not strong enough so far globally to stop that positive trend But certainly can be and the country specific parts we can see major effects in some places like the sahel Fertility turns out to be important through population for climate as well because the the loop from What happens to fertility and mortality from climate to population back to greenhouse gas emissions is is missing in in the modeling to date and that's something that Will help our understanding But i'm going to spend a little bit of a deeper dive on fertility because it's so interesting and Probably has such a profound effect for the future So here are some visuals these are heat maps of age specific fertility going back to 1950 and out to 22 to present And top left is korea The green color is low fertility the red is very high birth rates for women in those specific age groups And start with korea if you look at a column that is the total fertility rate You just add up the fertility in each age group and you get the tfr an indicator that many people may be Familiar with but that doesn't apply to any group of women, right? It's it's an artificial construct If you look at diagonals you get the cohort completed fertility Why is this matter? Well, if you're in a place like korea where at some discrete point around about the late 1980s Women started to delay the age of birth and you can see the yellow color sort of going up And so what you had in that short window for about 15 years? You get the vertical measure the total fertility rate dropping precipitously Then those women who delayed Raise the total fertility rate, but all that's happened is delay. There is actually no change in the Complete the there is a change. There's no decrease in increase if you look at the cohort completed fertility It just goes down. There isn't this up and down phenomenon Now what's also interesting in places like china on the bottom left is we have not seen that shift to older ages You've seen a shift on the bottom right in brazil Much less impressive to older ages So there's much less of that sort of korea or western europe or north america shift to older ages And in india a minimal shift a little bit now coming out just in the last few years Put another way If you plot cohort completed fertility against women's education, you get this rather strong relationship When you throw in a few other variables, this is the most important women's education But if you throw in access to contraception Urbanicity and the the under five child death rate with nothing else in the model you explain 84 percent of all variation across all geographies from 1950 to 2020 which is extraordinary like a very simplistic regression Talent that explains most of what we see for fertility And overwhelmingly the driver here is education and then number two is access to contraception and then Under five mortality and urbanicity add a bit more That means that when we forecast into the future fertility We're taking into account the relationship shown here Where it turns out that as women get educated get to 16 years of schooling for example and have access to reproductive health services Their desire or their cohort completed fertility turns out to be below 1.4 And in some places quite a bit below 1.4. So this idea that somehow societies will naturally Women's individual choices in aggregate will lead to a stable population with a Fertility rate of 2.1 which would give you You know equal numbers or or lead to stability Is not going to happen at least in the data that we see So looking at that forecast then the cohort completed fertility for 15 year olds cohort to 15 year olds By the time we get to 2022 on here Everyone that's not green is below replacement Everyone that is red is below 1.4 radically low fertility And by the time you get out to 2050 Much of the world is in that category because of progress women's education and access to reproductive health services With only a small number of countries where fertility is above replacement by 2050 for the cohort Now this creates real threats And it's interesting. This is actually finally gotten up to the secretary general's attention in the un Helped by I think advocacy from UNFPA that Low fertility and the inverted population age structure that goes with it where you have more people in the age group ahead of you than behind you Right out to about say age 75 or 80 Uh Leads to real issues of financial stability for social insurance for health insurance Real issues for housing markets since most houses are bought by younger people Real issues for other major economic drivers and there's some interesting modeling that suggests The a major negative effect on economic growth once you get into this inverted population period And so there's really two options for countries Increased fertility, you know the right way, which is supporting women to to have families and pursue careers If they choose to Or the wrong way, which is some countries are now starting to do which is banning access to reproductive health services and or putting in some Central asian eastern european countries Marked pressure on women to have more children And then the other policy strategy is migration Where if you welcome migrants societies that do that will avoid this inverted population pyramid and its consequences You know at least for a generation or maybe well into the 2070s or 2080s At some point that's the the migration strategy doesn't work because the world starts to decline And there is no There will be many fewer people wanting to migrate and that won't solve the world's problems So I think there's an interest from the climate community in a smaller global population So people see this information and say great We'll have instead of you know peaking at You know nine and a half billion in 2060 We can um Go from that and drop down to two billion and that would be better for the planet and better for everybody on the planet The tricky part here is how do you Sort of chart a course to you know What I think some people think of as a demographic soft landing How do you get to that point where at some point you must have replacement fertility? Otherwise humans disappear as a species And so if you imagine you want that at two billion people and you think that you want to get there in 2150 or whatever it is um What's that root because what we've observed empirically is that once societies Get to very low levels of fertility It's been extremely hard to change that look at singapore korea japan eastern part of parts of china as examples And so there's a very interesting and urgent need to understand better what might be the way you get to a demographic soft landing Fifth factor that can quite substantially alter the forecast that I showed is the Intransigent systemic racism that we see in some countries. This is one of them, but there are many others as well And our forecasts are done at the state or national level and we have not been forecasting by race or ethnicity Because you if you do that you see quite different trends in some cases. Here's an illustration from ongoing work that laura linguendoir leads at ihmi uh on Looking at county race ethnicity This is actually at the national level, but her project with which is an NIH project Is to produce these analyses For overtime by cause by for county race groups And what these brightly colored figures show is that depending on the cause you have markedly different trends For different race ethnicity groups You know for the same cause and those trends are not all the same It's not a single template across them all. So if you look for example at substance abuse disorders The a i a n line is the steepest and the most dramatic For for the country whereas if you look at neonatal disorders and maternal disorders You see there is a decline, but dramatically higher for african-american populations non-hispanic black populations So very marked differences and I think you know behind those are issues of systemic racism if we really want to capture Both strategies for intervention, but also to do a better job of the forecasting. We probably need to be analyzing more groups that have more You know that reflect their own trends for those causes Number six. We've seen in the u.s. That decline of life expectancy in 2010 to 2019 uh addiction related deaths Defined very broadly here for the purposes of this talk So encompassing not just drugs, but tobacco and alcohol and some other addictions Have already reversed progress in some parts of the world And we've also seen really innovative policy strategies that open up a new window Which is for example getting rid of smoking by cohort, which was what new zealand has legislated So zero tobacco ever available to be sold To the current cohort of under six five-year-olds in new zealand So you know and and we've started to model out. What if you did that in other countries? Now in our in the reference model. I showed you that that view of optimism for the future Here's what we think is going to happen We have at the global level red is for men orange is for women We have alcohol use going up globally for men, but not women Built into our model is further progress and reducing household air pollution Slow but steady progress on improving water Some these are all scaled so bad is high Low is good for the risk factors. This is a thing called the the sev Which is a attempt to create a universal measure for continuous And dichotomous risks that allow you to compare across them That's useful for the modeling But we actually see diet getting worse not better Smoking declining at a slow pace Ambient air pollution getting worse Not better. I think there's a lot of debate about that. You know, hopefully there's that's open to regulation sanitation improving Some aspects of diet getting better like fruit and then critically obesity just going up now We've seen a lot in the gvd Also in the forecast about the role of air pollution It is one of those drivers that we think may be getting worse at least for ambient pm 2.5 even if it's getting better for household and When we dig more deeply into the trends for ambient air pollution The big driver of things getting worse at the global level Is on the bottom right, which is the south the trends for south asia Meaning as this both is uncertainty for the forecast, but it's also opportunity You know a concerted effort around an mpm 2.5 in south asia would actually change the global forecast Quite substantially and we do see rising ambient air pollution in sub-saharan africa as well in the modeling I'll won't go through the detail here But as you start to think about the physical environment There's a much more prominent role for regulation Then in many of the behavioral factors that we have to worry about for the future of health So it's a very attractive area for changing the course Of this those risks Including you know the links to physical activity, which is another risk that we model We have not built in into the reference scenario any information either positive or negative I'll how rising urbanization And the nature of the urban environment might affect some of the other drivers in the model So that's a clear gap and something that we will want to address So Well, I don't know what the question why the wrong version of the slides on here But I'll show you the graphs in a moment. So diet obesity and physical activity Any historic analysis of looking at the gbd of the last 30 years obesity And to some extent diet leap out As major risks particularly obesity that are progressively getting worse And there's no single example globally of reversal of obesity There's some places where the trend is slow, but it's still getting worse There are some places where obesity still remains quite low, but nobody's Decreasing obesity Now part of the the issue around diet is How sure are we about the gbd type results around diet? We get a lot of critiques around that John Ionides has written some very powerful critiques of diet epidemiology as have others And so to help address that We are launching next week a series A new approach to star rating the strength of evidence around these risk outcome relationships And then that and it turns out that except for some specific things, you know salt and high blood pressure If at a high level of salt consumption Or vegetables and ischemic heart disease Many of the diet relationships are scored as weak evidence But in aggregate It's hard to imagine that Even if you have 25 things where all the evidence looks weak that there's by chance They're all going to turn out to be on that weak side But it is an area Where we have to be cautious because they're you know these relationships may be quite different once we hopefully have better evidence in the future Yet in aggregate, there's a very clear opportunity through food systems and changing national diets to likely improve health outcomes and certainly contribute to climate change And I think there's a lot of interesting work that's out there That we're Would like to grapple with more explicitly about what are the leaders that work for changing diet And then to more explicitly model the linkage between local production And food systems as well as what it gets consumed. Obviously that's mitigated by imports But in some environments that may be a much smaller factor For obesity just to reinforce our forecasts out. This is the past and our forecasts just continue them out into the future These universal increases in obesity around the world by region with still a range across regions, but still going up everywhere And this is for the Sorry, I think this was overweight and this is obesity. So even more marked increases for obesity So on the policy strategy before I leave obesity Not a lot on offer To think about to deal with obesity as a risk And so I think that'll be an interesting question. Is there really anything given the complete global failure on obesity That we can imagine To mitigate those risks this century AMR so we did extend the gbd To encompass antimicrobial resistance this year. It was a five-year effort A joint effort with the university of oxford a large consortium of researchers around the world And that's given us an insight into a problem that probably is getting worse Our analysis was a snapshot Five million deaths associated with amr One and a quarter million deaths that are directly attributable or you know in the sense of excess deaths from amr Now we think in the time trend analysis is in progress But there's a good suspicion that in aggregate amr is going up And there was the lord o'neill report that suggested alarmingly large numbers of deaths by 2050. I think 17 million From amr. I doubt we will have numbers that large in our forecasts But amr is definitely missing from our forecasts. And so as that comes online That may bring down Some extent, uh, uh, you know add a little more pessimism to our optimistic view Now that time trend analysis, uh, the the preliminary results of that is Quite patchy in the sense that some drug bug combinations things have been getting better Many have been getting worse. It's quite local based on, uh, antibiotic use patterns And stewardship programs, uh, and the interaction with, uh, animal antibiotic use in animals And so there's really an interesting set, uh, of there's a there's a body of work to be done to get a handle On how big a risk this is for the future even as we have a better handle on what it is today Now the response I hear from lots of people to many of the threats that I've been trying to run through That could swing us from our reference scenario, which is already somewhat optimistic, um, to being much more optimistic Is in the whole space of innovation Where innovation can make a difference For things that are distant drivers of health or or or distal drivers like, um, global mean surface temperature and greenhouse gases So, you know having fusion would make a big difference probably over the course of the century to health Uh, to more specific things like reducing pm 2.5 more more cheaply and so making it easier to adopt regulation in south asia as an example To, uh, the push for vaccination for new targets We were very happy that the amr work has led at least one pharma company to start looking for new interventions for acinetobacter Um, and then of course the the array of drugs and there's some interesting ones that maybe people haven't thought about as As people started to see a glimmer of hope on pharmacotherapy for high bmi Maybe new drugs will come that will be more tolerable At a wider spread level that open up new strategies for dealing with that particular threat Last let me end with this issue for the sahel Uh, I think we need to pay a lot of attention to the sahel because it's sort of this convergence of factors That mean the forecast there could be really much worse than we are estimating So we've done a lot of geospatial work This is work led by uh, simon hay at ihmi and a large team that have done geospatial analyses And i'm borrowing from a series of papers that he has uh, he and colleagues have published in nature This is wasting and you can just visually see on the maps 2000 on the top left 2010 and then 2015 that while there's been huge progress in places like drc for wasting We still have a very marked belt of wasting Right across the sahel Uh, and I we're i'm using the sahel in an extended sense Not just west africa about through south sudan and over into somalia So the some call people call this the sahel plus The same thing for child mortality bottom right is 2015 child mortality And while there's been huge progress in most of africa through the other time slots on the maps on the left Um, there's still High rates there's been a lot of progress, but there's still this band of high rates across the sahel countries And then you get to this remarkable shift in global fertility Which says that now even though it's a relatively small fraction of the world's population in the sahel We have 16 percent of all births Are in the sahel and given what we know about trends in women's education access to reproductive health services And what's happening everywhere else in the world That number by 2050 goes to a quarter of all births are in the sahel and could go up to almost 40 percent Of global births at the end of the century in the sahel This is the same populations which will be the first populations to face marked climate stress And reductions in food output due to rises in mean temperature So, you know a marked convergence of high child mortality high birth rates low progress in for women's education political fragility in a number of those countries and Now representing this huge fraction of of the world's birth cohort So let me end with uh, sort of the where on as we think about this commission and What could be useful in facing these, you know, not only Trying to get a a more credible set of forecasts That are neither pessimistic nor optimistic, but help us explore what is going to bend that curve What are some of the strategies that might be used to come up with Solutions that will address more than one of those issues. So clearly innovations on that list There's an argument that says uh, that we're in an era where regional Cooperation as opposed to strictly global mechanisms through the un or breton woods Is already playing a greater role think of the role of africa cdc in in the covid pandemic in africa And is there strategies around enhanced regional cooperation? We saw how failures of health systems really made a difference for covid so on the on the positive side Can we reimagine what a more resilient health system? um Could be And then there are the sort of failings that we saw during covid and in previous rounds of the challenge of Communicating the evidence to decision makers in a in a way that's on their schedule And is delivered in forms that is useful for decision makers So those are there's a real set of issues around communication I think as we see the places like the sahel where you will first see the impacts of climate We need to think about focused strategies for adaptation and Um Managing migration risk And then there is questions around the social and fiscal stability For societies that are going to meet this inverted population pyramid And lastly a clear need to finance these strategies and we need new financing Or at least to re-explore some of the financing options that exist So, uh, i'm going to stop there ask questions. Um, this is Intentionally being very much early Work in progress And it would be great to get people's thoughts or reactions Steph has a phone So chris very elegant and very sobering My question is Steph told you I teach an introduction to epidemiology course To mph students and I currently get quite a bit of pushback about obesity being a problem A health problem and and getting into issues of fat shaming And and the notion that obesity itself is in fact not a health problem But it's the fat shaming That causes ill health and i'm wondering if in your work on obesity You've confronted similar pushback about even labeling it as a problem You know, we're happy to adopt any name that people prefer We actually in the gbd called high bmi not obesity So we have never actually used obesity as the label But the evidence on high bmi and a range of cancers and cardiovascular outcomes is utterly overwhelming So there isn't any question that high bmi is bad for you. There is one risk outcome that I think gets a little bit of debate and that's around esophageal cancer And then the second one is this slightly strange Bottom of the risk curve for dementia, which is probably just reverse causation So those are the two little outcomes, but for the big outcomes from high bmi You know, these are some of them are quite high star rated Relationships in our star rating system The idea that that risk is driven by you know fat shaming I have seen no You know compelling evidence that would suggest the specific cancers driven by high bmi or the cardiovascular disease Is related to that so You know that's its own separate problem that the society needs to manage But it is a one of the bigger drivers as you look back the last you know when you do that risk deletion It is one of the more important determinants of the last 25 years. It really is Thanks chris for a whirlwind but really interesting take on where we may be going I want to consider continue the Bmi high bmi theme, but also learn a little more about your project to predict the probability of policies and Specifically i'm i'm curious as to You know how you might go about that and with a case study for the united states where we subsidize foods that Contribute to the high bmi problem versus as you suggested subsidies for vegetables and fruits so there's uh, thanks claire great great question and First on the climate side, we're very happy that the climate impact modeling group is going to take that task on And we're not but in general we do this like we we face this for covet a lot right which is If your only audience is a policy maker Then you don't have to worry about this because you're you're not talking to anybody else But when you make these forecasts, there are many policy takers So it's not useful for them to say well if the government does this it might be this and if the government does that It might be that they actually want to know what's my most likely outcome So think economic forecasts right which are clearly driven or impacted by policy But imf world bank ocd make economic forecasts You know, this is what we think will happen because they build in Some probabilities around which what will be the action of central banks for example into their their economic forecasts So, you know, there's a whole branch of quantitative political science that that Study modeling of of policy and and policy adoption And you know some of those models we've tried for example for covet policies You know, what's the likelihood of a red state or a blue state imposing? You know a mask mandate turns out that you can predict that So, you know, there are determinants on policy and so it's a very interesting Sort of bridge here because for our what we call a reference scenario Which is what we think is most likely to account happen We need to embed either implicitly or explicitly a policy forecast And then we want to then tell the policy maker that line assumes you're doing x and if you do y It's going to look like this. So we're sort of trying to to do both, you know But it is a Often a source of great debate and then in some cases The the predictors for policy seem hard, you know migration policy is famously difficult to predict Because it just suddenly swings from You know open to closed in certain settings and it's very hard to do a long-range prediction of that It's why the out-of-sample predictive validity for migration from everybody not just us is super poor It's a really hard one because it's policy driven the On the climate side you talked about the effects of heat in particular on the Sahel Recent experience in florida suggests that the other problem is extreme weather events And I just wonder in your baseline There is somebody here at berkeley sol shang that that argues that extreme weather events Have a much bigger impact than simply the numbers that we we look at and the reason we're not too worried about extreme weather events Is that the official numbers that you get Are obviously concerning if you're somebody who's affected but in the you know 50 to 80 million deaths a year space that we're thinking about for the future they are Quite small If it turns out there's bigger long-run effects from hurricanes or extreme weather events You know mediated through many mechanisms including like depressed economic growth Then that may be something that we will want to much more explicitly try to capture But it's it's missing from our reference forecast Other people to come up, but the other one You wouldn't you mind going back to the life expectancy by sex um Sure in different places The double curves going up. No the the ones that show the male female. Sorry next to each other. You have that one um The lower left One is south asia Yep, and it just this is a from me But this just made me think for the first time that the relative life expectancy between women and women Is an indicator of the status of women And that the separation in those curves in south asia says something about the status of women in south asia Changing over time But the difference is still nowhere close to what it is in most other places And that's at the regional level. I just wondered the granular level. Have you ever looked at that as a potential Well, I mean this this is sort of the Modern take on amarchus ends, you know analysis of missing women and his book on that So, yeah, they're the the When women's life expectancy, you know back in the 60s 70s 80s is the same as men. That's very unusual and reflects You know something that's going wrong And uh, you know, there have been interesting efforts to look at Um And we've looked at it by development status and it's very, uh, you know income and education And it's pretty interesting because you see um, the a big gap Uh, that gap may actually get larger between men and women in some phase of development And then it gets to be very small at high education and income And so yes, there's a pretty predictable pattern of that relationship So you could try to look at that conditional on sort of where you are in development as a as a marker of the status of women So the idea being that As women get more educated and can achieve their potential the gap widens And then as men do less dangerous work, exactly it narrows You get the two phenomena as men stop, you know Fighting drinking, you know Doing those things you narrow the gap so it widens and then it narrows Now, you know another good indicator of the status women is the Uh total fertility rate under age 25 Because it turns out that if women Have access to education and access to pursuing careers They almost have no children before 25 And so that turns out when that doesn't occur that turns out to be pretty good marker You know, it's it's an indirect measure But it turns out to to correlate and be super predictive of many outcomes for women So the fertility rate, you know cumulative fertility rate by age 25 is another marker I've never been in a meeting when joe hasn't asked a question so I keep looking over at your So uh some of these factors like fertility are dependent on Uh per capita income I was wondering how sensitive the models are overall or the scenarios are overall To forecasts about per capita GDP Over the course of 70 years and how confident are you of those forecasts? Uh not very but Come back to that so for fertility it turns out that uh income drops out of the models It it doesn't actually uh have a independent explanatory factor It matters enormously for climate it matters enormously for A host of health care type related factors. So it definitely matters We joe dealman at iHME for his work on health finance does produce a big ensemble of models for Each country to forecast GDP They're the usual sort of growth type models that are out there And there's other groups that have not more recently been producing Driven by the interests by the climate community of these long range forecasts All of them are theirs other people's have in you know, uh very large on certain intervals now it turns out that As we learned during covid and uh while in academia we're very interested in certain intervals the decision makers who act on these Don't care They're really only interested in the central tendency of of the distribution, but There's definitely this wild card on economic growth So earlier you were talking about the that trust in government and public health institutions It leads to better responses to public health crises. I was wondering if you had Or if there were any certain determinants of trust in public health That's a great question and there are people who study trust i'm not one of them, but um Part time with us has Heidi Larson who's done a huge amount of work on vaccine confidence And also like what can you do to intervene to build trust and I think there is Uh, there are proposals about how you do that. Uh, I don't know if there's data on whether they work but it's Given this very unusual like it's a really strong predictor it comes through all sorts of models and so uh Hard to imagine that uh, it's not something that will turn out to be important if we can find ways to build trust That's really interesting also, um for this hell and the um I guess the cat or speeding up of births her capita and As in relation to um global births. Does that take into account child mortality rate possibly decreasing or It does does so that still leads to it's already built into that forecast. Yeah I mean the the It's the only part of the world where the decline in fertility has not started But it's not unusual because there's been Very little progress in education in those countries for women. Are you touched on? The correlation between education and fertility. Yes There's got to be There's an there's an ambiguity in my mind. Um A woman might say I can now afford to have more kids But there's got to be a whole host of reasons cultural etc Why she's having more children? You know, there's definitely a cultural component but the Extraordinary thing is that it's only 15 percent of the variation and 85 percent is women's education and access to reproductive health services Okay, yeah Thanks chris. Um This is on the communication front and thinking about how Given sort of the uncertainty we're all interested in and communicating that to stakeholders Whether that's policy makers of the media how you all kind of approach Changing estimates and evolving science and communication strategies. It's been In the household air pollution world as our estimates have kind of bounced around a little bit It's proven to be one of the things that people ask a lot about Why was the number this many million this year and that many million the next year and My take has always been what it's a bunch of people Let's not worry about the exact figure and the uncertainty bounds always overlap and they their eyes close it over So I'm just curious how you all approach some of those conversations You know, I've had lots of conversations with decision makers about you know distributions of Results or you know What's the benefit of one policy over another? And it usually boils down to the fact that they will simply say what's your best Estimate what's your best number? Because and there's there's good theoretical reasons if you're risk neutral. That's the rational behavior You know if the social welfare function is is linear then it's super rational, right? So There's very little interest in uncertainty it turns out that uncertainty is largely an academic version of you know cya So that if you can't somebody can't come back and say you were wrong He said no no no I had these giant uncertainty intervals So it encompassed every possible future And so, you know, there is this huge disconnect between us on the academic side and really For for sound reasons care a lot about trying to characterize uncertainty and being transparent about that and then the users Don't and I don't think we're going to change that We just need to be cognizant of That you know, they have to take a choice and they're probably going to act on The mean estimate So thank you. This is really a pleasure to see something so wide ranging Um and not to make you switch to a totally different topic than what everyone else has asked about but one of the areas of focus You mentioned is amr. Um, and one of the things I really liked about the 2019 the prediction through 2019 was the separation of Amr attributable and amr associated since for many of the pathogens, you know death from invasive bloodstream infection. It's getting you know happen. Anyway, um It also highlights sort of a gap. I think in some of the pathogens new macaques, you know, being one of the Ones that, you know, a really big estimate Of amr attributable deaths was shown. That's not necessarily borne out in many epidemiologic studies. Um, you know, we're maybe, you know, in low-income countries There might be some, you know issues around antibiotic access that are contributing to why you put a really high attributable Mortality on some of these and I'm just wondering Um being close, you know, especially to a lot of decision makers around, you know, research Priority setting and things like that how much, um enthusiasm you've seen in association with, you know, correcting the o'neill idea of amr is going to kill us all And also studying, you know, really how many deaths especially in, you know, sub-saharan Africa south asia Place where we have recognition of, you know, growing an extreme problem of amr like how much interest there's been in real, um, you know, I'm thinking, you know, uh Gems or champs scale large Gates kind of studies of, you know, the attributable burden of amr in mortality Uh, you know, there there is a very active group advocating for amr, and I think the global champion is sally days at trinity and, um They've been At least with with the last round of work Pretty good at getting amr on the sort of g7 agenda Whether and then Sally was also very instrumental in getting the uk government to invest through the fleming fund in a lot of new surveillance data Um, other other types of studies. So that's all being good. I think they'll that'll continue at least for the next three four years um in the uk Uh, welcome is also pretty committed. I haven't seen much else on on investing on the amr surveillance front there's Looks to be even more data than has been there before Being collected by some of the pharmaceutical companies. So there there is some You know, I think there there's some hope that we'll see more data The big challenge that we had on the amr analysis is that most of the data for sub-saharan africa comes From vast majority from hospitals and a good chunk of that from referral hospitals And then you have to sort of correct for that bias And you know, maybe that wasn't You know, you never know right because there there is a real paucity of community level studies for amr um We've more recently in stuff that has yet to be published used more champs data and we've been looking at other sources particularly Breaking down different age groups for under age five And not surprisingly the patterns are very different after about the first month of life So more to come But I don't know if we'll see a big jump in data quality I can't help but hope that An improvement in surveillance for Outbreak events and fevers of unknown origin etc Can also as a secondary benefit help on the amr front. I wanted to ask you about the You talked about increases in spillover events, which I think is very clear Whether it's deforestation or whatever I think deforestation doesn't explain nearly as much as it should Or that we think it might But the escalator going in the opposite direction is our improvement in detection of containment And I'm just wondering where that comes out in terms of your projections Does the increased spillover End up dominating or does the improved Detection and containment end up dominating So the the in the reference scenario I've got is is Super naive and so it's basically saying what has occurred in the past will continue in the future So that or you put in your language those two forces will balance each other out So um That well the net has been increasing epidemics and pandemics not decreasing so Over the last hundred years. We have not been keeping up With spillover events. Oh for spillow. Yes. No, no, I first I'll agree on spillover events I thought you were referring to the sort of final outcome. Yeah We haven't barcode seen much of a big impact on pandemics So we have a small signal and a very big signal, right? So well HIV was an important signal and zika was an important signal and You know, we've had major events Okay, I I would class the uh By mode of transmission, right so Yeah, but yeah, uh hiv is an interesting one, right because The modeling out of what happens to Incidents or of or the incidence hazard for hiv Is also not reflecting either behavioral change in the up or down direction, right? And so I think there's definitely scope for hiv worse than perhaps we think if people Relax behavior I wanted to hear you explore a little further the way that policymakers might be convinced to Consider the uncertainty a little more Accurately and specifically I've been involved in the california climate policy area where They do some fairly sophisticated modeling of road maps to get to a greenhouse gas emission strategy Which appear to give these sort of clear answers as to what will happen, but Does not incorporate any of the underlying uncertainties or inadequacies of the modeling Particularly impressive in the natural and working lands emission area And so I've sort of been playing with I mean some of it is the way you describe Using your modeling where you can look at alternative outcomes With different assumptions or different policies, but I'm also wondering if You know, maybe policymakers like to get a clear answer like This is the way it is but I I'd be more comfortable if they knew this was You know, you basically almost gave a star rating to the level of your policy recommendation You know the in one discussion I had with uh, I won't name them but policymaker the The the question was turned back to me Which was okay here are these two lines And you know wearing masks or whatever it was looks better than that one And then you give me two lines that look like this. What do you want me to do that's different? And I don't think there's an answer that's different Right, so then it really at least in that context became It's it's our comfort level not theirs, right? So unless you can come back and and I I keep thinking about this because there are specific cases you can But in most cases it's pretty hard to come back and say You should do something different because the uncertainty intervals are wider And that's sort of the the test right is there an actionable thing for them that goes along with the wider uncertainty Or is it just us sort of saying, you know, don't blame us if you make that choice because we're not exactly sure I think with the one exception being that Sometimes it speaks to the value of information And if you could reduce the uncertainty interval you would reduce the likelihood of making the wrong decision And that speaks to the value of information of I think there's a completely different audience on the research funders Where uncertainty is probably the most interesting thing, right? Where if you don't really know that's where you want to put your money for But I would box up the research funders in that different category So I completely agree that that's a group that is sort of in ask, right? You get that question like where are we least sure? One thing that people haven't picked up in the questions is the sort of Shocking overweight and obesity trends. I mean it's just you know My parochial view of mexico in the u.s Thinking about it as we're the sort of extreme versions of the obesity trends, but in fact, it's universal and and I just wonder It seems to me that we're making so little progress on understanding the determinants thereof that Do you see any sort of major movement of You know massive research efforts on that. I certainly don't see it from the gates foundation. You're You're nearest neighbor, but I'm I find that to be the most striking thing that you presented today You know, I I think amongst the risks that are big It's the most alarming trend, right? It's the one big risk in terms of Tributable mortality that's really you know air pollution is getting worse But it's getting better in most places and it's concentrated in certain parts of the world But bc is getting worse everywhere and so I think it it is There's probably less discussion of it now than there was five years ago well, but also I'm optimistic that south asia will Not follow that path And that it will change its trajectory at least there's a future that I can imagine of that I can't imagine the future that reverses obesity because none of us know how to do it No, no we I mean that's why the the the ease of regulation for air pollution Is an attractive target, but for obesity it's pretty darn hard and You know, there's been various reports that make recommendations But if you dig into what's behind them They're usually grounded on some very small studies where there was success in one school with one intervention and You know you've got to try something in a setting where it's just overwhelmingly getting worse But I do think there should be a whole lot more You know effort to find solutions Going once going twice um, so I was wondering if uh water quality or water scarcity Is built into the models for in respect to migration or conflict or health Uh fecal content in water. Yes Right, uh, but not Other, you know heavy metals or or some other type of contaminant So I mean leds in the model, uh, but nothing else So there's there's a series of other heavy metals that could be there's been a quite an argument case made for trying to add those and so The water scarcity no And I think that gets to where I was trying to talk about habitability because that'll be part of it, right We desperately need to figure out. What are the determinants of Or the the threshold over which people will just leave And that's a very doable analysis Once we know where people are And you would think that's That is something that we should be able to figure out I mean, you know when I was thinking of extreme weather events versus direct heat effects You have like the heat for the farm worker and the direct health effects But the desertification effects are going to swamp that I would assume right and that's not that's I think of that probably in the heat side of it rather than on the extreme weather side of it But it's somewhere in between So the models for california and mexico for example are pretty catastrophic in terms of the sort of places that become Unhabitable the the places that Stop having water and therefore stop producing agricultural products and you know We still can live in arizona But but you can't grow much in arizona. Yeah, and those kinds of effects I think the the food stress part due to heat and drought are Super important. They're not explicitly modeled, but that is something that I think that We would really like to capture that, you know the whole sort of if you want to put a big rubric over at the whole food system aspect to You know what will happen with climate change? because I think that will be quite large I totally agree and Not well represented here. Yeah, and what I haven't seen discussed before is I mean heat makes water evaporate more so presumably there's more rain net and the question of what are the changing patterns of other places that benefit from yes, Canada Canada Russia Did I see somebody else raise their hand? No All right, well chris. Thank you very much from the online audience and the in-person audience We really appreciate your coming. Thank you