Climate Change and Health Seminar: "Heat health effects: research to inform policy."

November 05, 2020

Information

Dr. Gregory Wellenius, Professor of Environmental Health, Boston University School of Public Health.

Dr. Gregory Wellenius joined the Yale Center on Climate Change and Health's seminar series to speak on his research on the human health impacts of the built environment in the context of a warming climate.

ID5844

To CiteDCA Citation Guide

00:00- Good morning, everyone.
00:03Noon.
00:04Welcome to the Yale Center on Climate Change
00:06and Health seminar.
00:08I'm your host today, Dr. Kai Chan,
00:10assistant professor at the Yale school of public health.
00:14During the presentation if you have any questions
00:17you can use the chat box
00:19and we will try to address them as the speaker finishes.
00:24As a reminder, today's seminar will be recorded.
00:27So, it is my great pleasure today to introduce our speaker
00:33professor Greg Wellenius from Boston university
00:37school of public health.
00:38So Greg is actually the 2019,
00:41recipient of the ISEE Tony McMichael award.
00:46So it is very exciting to have Greg here today because,
00:50everyone knows Tony McMichael was the pioneer
00:54that developed the connection
00:56between epidemiology and the global countries.
00:59So with that legacy,
01:01I would like to take it over to Greg and very much,
01:05looking forward to your talk.
01:08- Wonderful.
01:09Thank you, Kai.
01:10Thanks so much for the invitation to speak here.
01:12And I only wish we could meet in person,
01:14under better circumstances.
01:16I was telling Kai before
01:19a few minutes earlier that one of the great pleasures
01:22of giving seminars in places is visiting with the people
01:26in small groups.
01:28So hopefully we'll have the opportunity to do that,
01:31again shortly.
01:33So let me share my screen.
01:39Okay.
01:40So you should be able to see my slides,
01:42Kai, give me the thumbs up or somebody can see my screen.
01:48Okay, great.
01:49So we'll just go ahead and get started.
01:52So yeah, so feel free to stop me along the way.
01:55I will rely on Kai to flag me down
01:57if you wanna put questions in the chat window
02:00and then I can stop,
02:02I don't mind being interrupted
02:03and that way we can make it more interactive that's fine.
02:06I should mention that,
02:07I am currently a visiting scientist
02:11working with Google and, this
02:17nothing I say here should be interpreted
02:19as being the official position of Google.
02:22All right.
02:23So with that I will get started.
02:26So I wanted to talk today about the effects
02:29of heat on health, which is,
02:32very well described in the scientific literature
02:35and connect that to
02:37why we have sort of this disconnect between,
02:41what we know about heat and the fact that
02:44people continue to die of a heat related illness.
02:49So the problem, as I see it is that excess heat
02:51is a widely recognized threat to public health.
02:54It's often cited based on CDC statistics
02:57that in the U.S more people die
02:59of extreme heat each year
03:00than of any other meteorologic event.
03:03So despite all this knowledge,
03:06that we have about the risks of
03:09days of extreme and perhaps moderate heat,
03:12there seems to have been remarkably little progress
03:15towards preventing heat related illness and death.
03:17So we still see that heat waves
03:19are a major source of morbidity and mortality
03:22across the world.
03:23And so this got us thinking that
03:26this suggests a lack of translation
03:27of the abundance scientific knowledge about risks
03:30into public health action.
03:32And so just to highlight the point
03:35for those that may not be as familiar.
03:37So a Seminole study by Antonio Gasperini and colleagues,
03:41London school of hygiene, tropical medicine,
03:44published several years ago
03:45and have since published extensively,
03:47globally on the impacts of heat on health.
03:51And just to zoom in on a couple of locations,
03:55you could see that there's this,
03:58U shaped relationship between,
04:01daily maximum temperature,
04:02is typically used and the relative risk of
04:04some adverse outcome in this case mortality.
04:07And you can see that there is a temperature,
04:11what we'll call the temperature of minimum mortality,
04:12or the optimal temperature at which the fewest
04:15number of people die.
04:18And then as temperatures get warmer than that,
04:21you see a sharp increase, in,
04:24the relative risk of mortality and the shape of this curve,
04:28varies from location to location,
04:30but the pattern has been shown throughout the world,
04:35by Gasperini and colleagues, as well as
04:39other groups in specific locations.
04:41So this is pretty universal
04:42and pretty well understood at this point.
04:45In the U.S we additionally know,
04:48about the effects on morbidity.
04:51So as measured by hospital admissions.
04:53So this is some terrific work done by Jennifer Bob
04:56working with Francesca Dominici at Harvard and team.
05:00And, so this was in the Medicare population
05:03looking at millions of hospital admissions
05:08for a number of different causes and showing
05:11both the relative risk and the risk difference of,
05:14hospital admissions for different causes that you can see.
05:20Increased relative risk of fluid
05:22and electrolyte disorders, renal conditions,
05:23urinary tract infections, heat stroke,
05:27and other external causes.
05:29And, with the risk difference shown there as well.
05:34So, interestingly although heatstroke
05:37has the biggest relative risk
05:38because it's relatively uncommon as a diagnosis,
05:41the risk differences is smaller than for some other causes.
05:45So terrific work.
05:47So this is just a sampling.
05:48There's a huge literature now on this,
05:50and very large studies demonstrating
05:52that extreme heat is associated with higher rates of death
05:55and hospitalization all across the world.
05:58Moderate heat is associated with higher rates of death, and,
06:02building amounts of evidence suggesting also
06:05with hospitalization.
06:07And we know that the vulnerability of these effects
06:10varies by personal housing
06:12and neighborhood characteristics.
06:14Further we know that the U.S has already warmed
06:17more than a degree and is projected
06:20to warm further through the end of the century
06:21in substantially with that,
06:24regional substantial regional variation and how much,
06:27further warming we expect to see.
06:31So how do we translate this into action
06:35that actually saves lives and reduces the health impact?
06:39So local public health and emergency
06:41preparedness officials
06:43need to know something a little bit different.
06:44They need to know what are the health risks
06:46associated with a given climate hazard in my location,
06:49what local actions can I take
06:52to protect the public health
06:53and do these actions actually work?
06:55So I'm gonna walk you through some of the research
06:58that we've done in this domain.
07:02And I'll start with what are the health risks
07:03associated with a given climate hazard
07:05in a particular location?
07:07So I started this work when I was in Rhode Island,
07:11actually Julia Gold at the time
07:14at the Rhode Island department of health,
07:15came to me and said,
07:17we really wanna know how many people
07:20are dying of heat and Rhode Island and how many ed visits,
07:23we have in Rhode Island.
07:25We need to know how to prioritize this.
07:27And I said, well, there's lots of literature
07:28it's a big problem you should just be worried about it.
07:30And she said, no, can you give me a number?
07:32And so I said, okay sure
07:34let's try to give a number.
07:36And then it turned out that New Hampshire and Maine
07:39were also in interested in the same question.
07:42Public health officials in those States
07:45were interested in the same question.
07:47And because this was done at small,
07:49relatively smaller populations,
07:53we all had the challenge
07:54of having sufficient statistical power,
07:57to examine the associations
07:59between heat and either mortality or ed visits,
08:02in our own communities.
08:04So we partnered with between Rhode Island,
08:09New Hampshire and Maine to pull data,
08:12do the analysis in each of the community shown here
08:15and then pull the results to have enough statistical power.
08:19And we also engage with the regional offices
08:21of the national weather service, in order,
08:25they were interested to reconsider the
08:30threshold criteria at which the,
08:33heat advisories or heat warnings were issued
08:35based on local evidence.
08:37So we were trying to provide local actionable evidence,
08:41and in particular in communities outside of
08:43the large cities of the area that would otherwise,
08:46dominate the signal.
08:49And so we found what you'd expect is that the,
08:52here we were interested in heat index,
08:54'cause we were doing this in partnership
08:57with the national weather service and heat index is
08:58this combination of temperature and humidity that,
09:01they often use for issuing heat warnings
09:03and heat advisories.
09:05And we found approximately what we expected,
09:08that there was a monotonic relationship
09:10between increasing maximum daily heat index
09:13and relative risk of emergency department admissions
09:16that you see there on the left
09:18and deaths there as you see there on the right.
09:21And, these were about of the expected magnitude.
09:24And you can see that even pooling across these 15 locations,
09:27the confidence intervals around our estimates of,
09:30for mortality relative to some mortality
09:33were somewhat imprecise.
09:35So, the, I think the key part of this is,
09:40to translate sort of relative risks
09:42and smooth curves, which are available,
09:47with standard software now,
09:49thanks in large part to work by Gasperini and colleagues,
09:54is to translate that into real numbers.
09:57So, okay.
09:59So a curve is all good but how does that translate to
10:03number of excess ed visits or excess deaths
10:07attributable to days of different heat indices?
10:11So we created this table where the bottom row here
10:14shows you on all the days of 100 degrees
10:18with a heat index of 100 degrees or higher,
10:20how many excess deaths,
10:21excess CD visits were there on the same day, or,
10:25incorporating the lag effects up to seven days.
10:29And so, across these 15 new England towns,
10:31there were 39 additional ed visits
10:34on all days over 100 degrees and 232.
10:39If you incorporate the lag structure,
10:42the fact that the next day
10:44and the next day might also have some excess ed visits
10:46and about four to eight excess deaths
10:50for the days above 100 during this time period.
10:52And, obviously there's more days that are at,
10:55or above 95 degrees.
10:57And so then, those numbers are bigger and, at,
11:01or above 95 degrees,
11:02there's close to 200 to 700 depending on,
11:06how far out in the delay you want to incorporate,
11:11excess ed visits.
11:14So we took this information to the national weather service,
11:17to the regional office for the national weather service
11:19and said, look, we think that at temperatures below that,
11:22at which you currently issue heat advisories.
11:25So during this time heat advisories were
11:27issued by the national weather service for days
11:30with a heat index forecast to be above 100 degrees.
11:33We said, look at days as low as 95 or 90,
11:37we still see excess ed visits.
11:40And you can see that in the curves too, that,
11:42it's relatively monotonic so there's no reason
11:44to pick just 100 degrees as the threshold.
11:46It could be even at 95 degrees, you could,
11:50presumably warn or prevent
11:53some excess morbidity and mortality.
11:56And the national weather service said, okay
11:59that's great.
12:00And, so let me
12:05I'm gonna skip ahead to the national weather service.
12:08Okay, sorry.
12:10So before I get to the national weather service story,
12:13so 'cause I think that's really important, but then,
12:15so I want to shout out to Kate Weinberger,
12:18who was a postdoc in my group at the time.
12:20And what she said is, okay, this is great for New England,
12:21but how many people die of
12:24deaths attributable to heat across the country?
12:28And so using data that we had a mortality through 2006,
12:32she estimated that there were 5,000 or more excess deaths
12:39per year across the U.S attributable to heat.
12:42This number is really important because
12:43it's about an order of magnitude
12:45higher than what the CDC estimates
12:51report for heat related deaths that are those
12:54that are coded as being due to heat.
12:57And so when we think of sort of the,
12:59public health burden of disease of heat related illness,
13:03the CDC estimates, are important,
13:06but we think a likely an underestimate
13:09of the true excess mortality due to heat.
13:12The other important point here is
13:15that if we separate out the extreme heat days
13:18versus the moderate heat days,
13:20so we defined extreme heat
13:22as those days above the 95th percentile
13:24for a particular location.
13:26And these 297 counties across the U.S.
13:30The burden of disease is actually bigger for,
13:34deaths due to moderate heat.
13:35And that's been reported previously,
13:38across the world and in the U.S but it,
13:41this puts concrete numbers on that that
13:44moderate heat accounts for a substantial burden of disease.
13:49And the other key point from this study is that, the risk,
13:54or the excess mortality is not distributed uniformly
13:58across the U.S and there's parts of the country,
14:01that seem much more vulnerable to,
14:05heat-related mortality than others.
14:08Again, emphasizing the importance of local knowledge
14:10and local action to prevent these.
14:14Okay, so let's turn to local actions,
14:17that can be taken to protect the public's health
14:20and evaluating if these actions actually work.
14:24So in the U.S the national weather service issues,
14:27heat, advisories, and excess heat warnings
14:30when the heat index is forecast to be high.
14:33Now, and this is for most places,
14:35there's a handful of places
14:38that use the other criteria besides heat index.
14:40But these warnings that are issued,
14:42provide information that the public can take,
14:44of actions that the public can take to protect their health.
14:48And in some places the warnings may also trigger
14:52activation of local heat response plans,
14:55that may involve things like opening cooling centers, or,
14:59reaching out to particularly vulnerable communities
15:02in addition to targeted messaging,
15:05and the optimal thresholds for issuing
15:07these heat advisories or heat warnings,
15:11remain largely unknown or unstudied,
15:15refer to heat advisories and warnings together
15:18as heat alerts.
15:19So based on the work we did in that New England study,
15:24working with the national weather service regional office,
15:27they decided to partition the Northeast, which was,
15:31had one criteria for issuing heat advisories
15:35prior to this work starting in summer 2017,
15:37they changed it so that the,
15:41new way in New England was treated separately
15:43from the rest of the Northeast,
15:45acknowledging that the vulnerability
15:47to a heat related illness might be different in New England,
15:53not just based on our study,
15:56there's other studies that have shown that as well.
15:58So this felt like a major public health victory.
16:00So following this starting of the summer of 2017,
16:04the national weather service in the region,
16:07issued heat advisories when the heat index
16:12was forecast to be greater than 95 degrees.
16:17And there was some confusion as to whether
16:18that should be for one day or for two days,
16:20it was initially for two days.
16:21And, then they subsequently revised the criteria,
16:25to be consistent across the New England region.
16:29So essentially changing the heat advisory threshold
16:32from 100 degrees heat index to 95 degrees heat index.
16:36So this felt like, to me,
16:37a major public health victory, this was, one study,
16:41one paper that, and a series of conversations
16:45that ended up changing the criteria,
16:48at which heat advisories are issued for,
16:51a region with a substantial population.
16:53So that felt very impactful,
16:55but it leads to the question of okay,
16:57so we're issuing more heat advisories now
17:00than we were before,
17:02because we've changed the threshold.
17:03Does that actually save anybody's life?
17:06So, we weren't the first or the only ones
17:10to be having this type of conversation.
17:12We followed in that research some very nice work,
17:17from New York city,
17:18where they also informed local policy
17:21through evaluation of data in New York city.
17:24And so the question we were asking is,
17:28what is the optimal threshold for issuing heat alerts,
17:32heat warnings, and heat advisories.
17:35But these conversations assume that issuing
17:38heat advisories and warnings actually
17:40reduces heat-related morbidity and mortality.
17:44And there's been relatively few studies on that question.
17:48What, again, there's a handful of studies,
17:51but one that I particularly like is this study from,
17:56Tarik Benmarhina while he was still at McGill and looking,
18:01taking a very creative approach to looking
18:03at the effectiveness of the heat action plan that including
18:06included a new heat early warning system on,
18:11heat related mortality in Montreal.
18:14And, that team reported that the,
18:20that having this heat action plan implemented in Montreal,
18:25reduced mortality during hot days
18:27by about two and a half deaths per day,
18:30and with particularly larger effects amongst the elderly.
18:34So we wanted that's exactly the question
18:37we wanted to ask is the issuing of heat warnings,
18:39heat early warning system.
18:41How much does that benefit the population?
18:44So we built this study on the advantage
18:49that heat warnings are issued by people,
18:52and they're issued on forecasts.
18:54They're not completely algorithmic.
18:56They are issued by specialists
18:58at the national weather service
19:00that are focused on heat warnings.
19:02And, they,
19:05there's a collection of days where we forecast
19:08that there will be a high degree of heat.
19:13And then it turns out to be a little bit less,
19:15and then there's other days where we forecast,
19:17lower heat levels.
19:22And it turns out to be a little bit higher.
19:23So the forecast can be wrong even just a little bit.
19:26And because they're issued by people,
19:28there's some discretion in how much they think
19:31people need to know about the upcoming heat.
19:34So for instance, we were told that on the 4th of July,
19:37you might issue a heat alert at a slightly lower,
19:40forecast heat index, then on another day,
19:43because so many people are gonna be outside.
19:45So many people are going to be exposed that maybe,
19:48we can have the flexibility to change that threshold.
19:51And that was entirely built into the system.
19:54So there should be these days with a similar heat index,
19:57right around sort of the warning threshold,
20:01some of which have a heat warning some of which do not.
20:05And so that's the
20:09paradigm we were taking advantage of.
20:11And at the time we had data on heat warnings from 20 cities
20:17that issue heat warnings regularly.
20:20And, we matched us to the mortality data we had
20:22from the CDC.
20:25So the overlap between these two data sets is 2001 to 2006.
20:31And, again, comparing days of similar heat index,
20:37with versus without a heat alert,
20:40this is the relative risk of mortality,
20:43associated with having a heat alert.
20:46And so if he'd warnings or heat advisories were,
20:49protective of the population,
20:51you would expect to see a decreased,
20:54relative risk or a decrease in the rate of mortality
20:59on days with a heat alert compared to without.
21:01So interestingly, we did not see that
21:03across these 20 cities,
21:05overall there was a null association.
21:08And the one place where we did see an association was,
21:13Philadelphia with a reduction of about 4%
21:16in mortality of about 4% on days
21:17with a heat warning versus without.
21:19So this could be for a couple of reasons.
21:23One Philadelphia, we know has been very proactive about,
21:27having a robust heat early warning system
21:29and taking action on days expected to have high mortality.
21:35It could also be that this was 20 estimates,
21:39and that one out of 20 was,
21:41in the direction that we expected.
21:44So clearly needs a followup study,
21:48but then we played the thought experiment of
21:52so heat alerts were effective
21:55at reducing mortality in Philadelphia.
21:57And the number of deaths we estimated,
21:59that were averted in Philadelphia
22:03each time they issued a heat alert,
22:05was about four and a half or five lives per time.
22:09And so if you extrapolate that to the,
22:13typical year in Philadelphia during this time,
22:16that meant that the heat early warning system
22:18saved about 45 lives per year.
22:21Again, lots of assumptions of causality,
22:23but it gives us a starting point that if the,
22:29if heat warnings could be as effective
22:32as they were observed to be in Philadelphia
22:35during this time then a city like New York,
22:38or Dallas or Phoenix,
22:41could potentially save avert quite a few lives per year,
22:47depending on the effectiveness of the heat warning
22:49and how often the heat alerts are issued per year.
22:53So this provides,
22:55a rough for back of the envelope calculation as to
22:58how many lives could potentially be averted each year,
23:03across the country if heat warnings, reduced,
23:09mortality by the same magnitude as we saw in Philadelphia.
23:15Okay.
23:16And, again,
23:17I want to emphasize that we're not the only ones
23:19that have considered this question.
23:20This is some great work by Kristie Ebi
23:2515 years earlier, showing that in Philadelphia, exactly.
23:31The heat warning system, she estimated,
23:33each time that a heat warning
23:36was activated at saved two and a half lives per day.
23:40So, in the same ballpark of the estimates,
23:44we were seeing but in a very different time period.
23:47Okay, so there's lots of limitations to this study.
23:50One of them is that the data we were using
23:52at the time was old, was mortality data through 2006.
23:56So, Kate Weinberger has since been updating this,
24:02sorta with more recent mortality data from,
24:06nine Northeastern cities
24:07where we found the data readily available
24:10in collaboration with Joel Schwartz and team.
24:12And, there, we, she found, that perhaps,
24:173% mortality benefit on heat warning days versus,
24:21days with versus without heat warnings.
24:24So maybe it's just that in 2006 and earlier,
24:28when most places did not yet have a heat action plan, then,
24:32we don't see very much of a benefit,
24:34but in more recent times where,
24:35many more communities do have heat action plans,
24:38tied to those heat alerts that we see,
24:43perhaps some signals so we're following that up
24:46in a broader population.
24:47And then the other question is of course,
24:50is that mortality is not the only outcome of interest that,
24:53we also want to prevent illness,
24:56as reflected through hospitalizations.
24:59And, here we saw in 97 counties
25:03in 2007 to 2012,
25:06using Medicare hospital admission data.
25:10We found no reduction
25:13in the risk of emergency hospitalization
25:16during this time point.
25:18So again, to works in progress that,
25:22we're following up on a larger scale
25:23and with more recent data.
25:27Okay,
25:29so our national weather service heat warnings effective,
25:32they may reduce the risk of death in some cities,
25:35but we don't yet see evidence of
25:37widespread health benefits.
25:40And if that's true and again it needs to be confirmed,
25:43but that would represent a missed opportunity
25:47to prevent heat-related morbidity and mortality.
25:51There's lots of limitations to the analysis I've shown here,
25:54and we're working to actively to address these limitations.
25:59So I just wanna emphasize the,
26:02that we're at the beginning of the road here not the end.
26:05Okay, so I wanna turn to talking about,
26:09how susceptibility to heat related illness
26:13might vary by age groups.
26:15And, so in one of the first studies we did in Rhode Island,
26:20we looked at emergency department visits,
26:23to the to Rhode Island over several years now,
26:27there's only a million people in Rhode Island.
26:29So again, there's an issue about statistical power.
26:35But the interesting thing is that, of course,
26:37we all think of the elderly as really vulnerable.
26:39And what we saw is that for heat related ed visits,
26:44in fact, the relative risk was a lot higher,
26:48so this is excess relative risk.
26:50So these are percents.
26:51So this would be an odds ratio of 1.6, approximately.
26:56So that the relative risk was actually higher in
26:59that study for population of adults of non elderly adults,
27:0318 to 64 and with significant for kids also
27:07or children and adolescents 18 and under,
27:11so what to follow that up.
27:15More recently we partnered with Ari Bernstein, the Harvard,
27:20center for climate health and the global environment,
27:25and using data from on ed visits from a network
27:31of standalone U.S children's hospitals.
27:34These are 47 hospitals and the recent Tara
27:37with a total of three point million ed visits,
27:39amongst children and adolescents.
27:42And you can see the location of the hospital here
27:44as well as the relative size and contribution.
27:48And so a little bit hard to see here,
27:52but so what we see is that the overall relationship between,
27:57maximum daily temperature and the relative risk
28:01of ed visits for all causes in
28:03this population is a 1.17 or about a 17% increase.
28:07And for heat related illness it's about
28:10a relative risk of 1.83.
28:12And again, you see it's interesting for all cause ed visits,
28:18there's not a lot of heterogeneity by age,
28:21but there does seem for heat related illness
28:22specifically seem to be somewhat of a stronger effect
28:25amongst the older adolescents.
28:28So that was really interesting.
28:32And then we wanted to sort
28:34of move beyond heat related illness
28:36to look at a number of potential causes.
28:39And this is a little bit hard to see.
28:41So I just wanna zoom in a little bit.
28:42So to the, we considered a number
28:45of different categories of disease,
28:47some of them that we sort of had prior hypotheses for,
28:51and some that seemed like we should just check.
28:53And these are adjusted for multiple comparisons
28:56in this sort of more agnostic analysis.
28:59And you can see that heat related illness of course
29:02comes up with a very high relative risk,
29:04but there's other interesting
29:06and much less explored associations
29:09between different causes of ed visits
29:11in children and adolescents and temperature.
29:14So, more to be done there,
29:17but we're quite excited by these results.
29:21I'll make the point as in the paper
29:24I showed you at the beginning by Jennifer Bob
29:26and colleagues that not all the,
29:30those conditions with the highest relative risk
29:32don't always have the biggest sort of numeric impact.
29:36So heat related illness here,
29:39you see the attributable fraction.
29:41So of the heat related illness
29:42a substantial proportion are due to heat.
29:45And, but heat related illnesses
29:49and in frequent or uncommon diagnosis.
29:53And so the out of 100,000 ed visits,
29:56it contributes a relatively small proportion.
29:59Whereas for injury and poisonings are very,
30:01very common diagnosis amongst kids, as,
30:04so even though the attributable fraction
30:08is smaller for them the attributable number
30:10per 100,00 ed visits total
30:13is much bigger because it's much common.
30:17Okay, so I wanna share with you some,
30:20very exciting work that Darren Son in my group is,
30:24leading and working on.
30:25So this is now turning to 18 to 64 year old individuals.
30:30And this is amongst an insured population,
30:32working with data from the Optum labs.
30:39And obviously here you have the number of sorry,
30:44the average summer maximum temperature.
30:46And then this just shows you sort of the distribution
30:48of where we have information on in this population.
30:52So it tends to follow,
30:54the distribution of population
30:57focused on obviously more urban locations.
31:01But, this particular data set has a more info
31:04tends to have more information in the Southeast
31:06and in the Southwest.
31:09And, you can see here is
31:13that overall there's a relative risk of ed visits,
31:17amongst these non elderly adults an odds ratio of 1.1,
31:24let's say about a 9% increase in risk
31:26and for heat related illness
31:28it's a relative risk of about 1.9.
31:32And again, you see some variation in,
31:36the relative risk by age,
31:38some heterogeneity by age that we'll explore
31:40a little bit further to see.
31:43It's interesting though that sort of repeatedly
31:45we're seeing that although elderly are known to be,
31:49and there's good evidence
31:50that they are a susceptible subgroup,
31:53that's by no means the only part of the age distribution,
31:56where we have sensitivities and in there's,
31:58we know of from other studies, outdoor workers,
32:01children that spend a lot of time outside,
32:04perhaps children's spending time
32:05in non-air conditioned schools,
32:07can also be quite a bit at risk.
32:12Okay.
32:13So turning back to the, the bigger, framework.
32:16So on a global and national scale,
32:18we think that we understand
32:20the adverse health impacts of heat.
32:22But there's been this lack of translation
32:24of abundance scientific knowledge on the risks
32:27and to public health action in terms of prevention.
32:30And so, again,
32:32this means that there's insufficient evidence
32:35to guide the public health response
32:36to present day or future heat.
32:39If we were designing, optimal response to heat,
32:44Jeremy Hess and Kristie Ebi have written nicely about this,
32:48you'd define dangerously hot weather,
32:50you'd forecast it well,
32:52you'd identify who's at greatest risk of these effects.
32:55You'd intervene to reduce those health impacts,
32:58and you'd evaluate the effectiveness of those interventions.
33:01And you do this on a continuous cycle.
33:03You'd do this repeatedly to continue to optimize.
33:07So, our broader research agenda
33:11follows mirrors these image.
33:14So, the vision that we have is that
33:17we could provide the evidence needed for any community
33:19in the U.S to mitigate the adverse health impacts
33:22of extreme heat.
33:23And I'd probably amend that now to say
33:26both extreme and moderate heat,
33:28although we recognize
33:29that they require different strategies,
33:30the same strategies won't be effective for both,
33:33thinking about moderate and extreme heat.
33:36The concrete sort of next steps in that is
33:39to identify optimal health based and location
33:41specific metrics for issuing heat alerts.
33:44We wanna follow up our work on the benefits of
33:49heat alert's heat warnings and heat advisories,
33:53because I think there's
33:55they're probably effective in some circumstances
33:58in some places and in some populations.
33:59And if we knew where they are effective
34:02and under what conditions,
34:03then we can presumably provide information
34:06that helps other communities replicate that effectiveness.
34:09I think there's a lot of potential benefit,
34:11to investigating that further.
34:15And you, one of the shortcomings in this line of research
34:20is that we don't actually have
34:21a centralized database of which,
34:24what local health departments are,
34:26what actions local health departments are taking
34:29in response and preparation for,
34:31and in response to days of extreme heat.
34:33And so one of our goals is to try to,
34:36catalog that we're working with Jeremy has and Nicole era,
34:40at university of Washington.
34:43And then if we can identify again,
34:45the key elements of these interventions and
34:48where they're most effective,
34:50then we can share this information back
34:51with local health departments and say,
34:53"hey, if you have limited resources and you,
34:55"here's what has worked in other settings
34:58"that are similar to your settings
35:00"in terms of whatever characteristics,
35:02"we wanna have about the community.
35:06Okay, so I wanna acknowledge also that,
35:09heat doesn't happen alone.
35:11This is some great work done by Keith Spangler,
35:14who is currently a post-doc in working in my group.
35:17And this was part of his doctoral dissertation at Brown.
35:19And what you see here is different hazards across different,
35:26across New England, sorry.
35:27So, this is a probability of one or more days
35:32with the heat index above 95 degrees.
35:35And so you could see the distribution of that.
35:37So there's parts of New England that are more prone
35:40to getting really hot days.
35:42The distribution of getting an inch or more of rainfall
35:46is quite different.
35:48And similarly, the distribution of the
35:51risk of high ozone days is again different.
35:54And we don't have high PM 2.5 levels in New England.
35:58But, if you were to look at where they are highest,
36:02you can see the distribution again is quite different.
36:05And so if you integrate those into the percent of days with,
36:10one or more hazards during this time period,
36:13you see that there's an interesting distribution where,
36:17parts of the Connecticut river Valley
36:19and Southern Connecticut are particularly,
36:23high risk of being exposed to one or more hazards.
36:28Interestingly, if you connect this with the
36:31social vulnerability index,
36:33this is the CDC social vulnerability index
36:35that is also not homogeneously distributed.
36:38And interestingly, those high vulnerability locations,
36:47also tend to have a higher probability
36:51of having more than one hazard.
36:55This is primarily driven by the distribution of,
36:58the hazard of excess heat,
37:00and somewhat by the excess ozone.
37:02So really interesting to think about
37:06how the hazards overlap with each other
37:10and with social vulnerability
37:13and Keith created a climate risk index,
37:16based on this which looks different
37:19depending on the spatial scale that you look at.
37:22So again, if you combine the hazards
37:24and the social vulnerability, again,
37:25the Connecticut river Valley at Southern Connecticut,
37:28coastal Connecticut show up as places of particularly,
37:34potential pretty high impact.
37:36And if you were to look instead at the,
37:39Boston metropolitan area here,
37:41you can see that on a very fine spatial scale.
37:44There's tremendous heterogeneity as well in this.
37:49Okay, so to close.
37:51So in order to adapt to current and future climate hazards,
37:55local officials need to know what's the current health risk
37:57associated with a given hazard,
37:59what local actions can be taken
38:01to protect the public health.
38:03Do these actions actually reduce the risk of the hazard?
38:07How has the risk likely to change into the future?
38:10I didn't go into that today,
38:11but obviously we have very good projections of future,
38:16temperature changes under different concentration pathways,
38:21so we can predict into the future
38:25under different potential alternative realities.
38:27And we can do this in a repetitive way
38:29to continue to optimize.
38:32And so this just Zooming way out,
38:34highlights the needs and challenges
38:36of translating scientific research
38:38into public health benefits.
38:40So, this none of this would be possible
38:45without a fantastic team local team in my group,
38:51as well as, fantastic collaborators.
38:53Kate Weinberger was a former post-doctoral fellow
38:56that worked with me and is now
38:58at the university of British Columbia.
38:59We have a terrific team at Boston university and formerly,
39:06people were still connected with at Brown
39:08and then fantastic collaborators at Harvard,
39:10university of Michigan,
39:11university of Washington and Mount Sinai.
39:16And of course we all need funding,
39:17and I'm very grateful to the funding from NHS
39:20and Wellcome trust.
39:21So I will stop there and a welcome your questions.
39:29- Great, thanks, Greg, for the very, insightful presentation
39:33and also sharing with us your latest research.
39:38Before we go to the question from the attendees,
39:41we actually, have already pre collected questions
39:45from the our students who attend the
39:49Climate Change and Health seminar.
39:51I'm happy to see actually doing your presentation.
39:55A lot of questions has been answered.
39:57So just, pick some of the questions remaining.
40:01One the heat topic that the students are wondering is
40:04about the effectiveness of the heat index system.
40:08So they're wondering,
40:10like why there's no standard index
40:14in different places, and why there can be some, action of,
40:21why there can be some other matrix
40:23that can be considered like the wet bulb temperature,
40:29which may shows, more spatial rate disperse,
40:34varied effect rather than that or temperature.
40:39- Yeah, it's a great question.
40:41So the national weather service sets up, actually
40:45the national level of the national weather service
40:47makes recommendations of criteria that could be used,
40:52to issue heat alerts and then encourages regional offices
40:56and even local offices to come up
40:58with their own criteria that,
41:01are most appropriate for the populations that they serve.
41:05And so there isn't exact, it's not,
41:08a top-down sort of you must use this,
41:11here's a standardized threshold, which,
41:12some countries have taken that approach.
41:14This is a much more decentralized approach.
41:17So many, many, locations do use the heat index.
41:21And for approximately, Northern location
41:26sort of Northern half of the country
41:29uses a heat index of 105 as a threshold for
41:32issuing heat warnings and,
41:37a threshold of 100 degrees heat index
41:40for issuing, heat advisories,
41:42and then the Southern half of the country, approximately,
41:45each of those is five degrees set at five degrees higher,
41:49but there's a number of locations,
41:50they use their own system, including,
41:53Philadelphia is notable for using
41:59a predictive model of sort of
42:01how many people are at risk from this heat.
42:04New York city has done some terrific work on,
42:08changing the threshold.
42:10So there a number of examples around the country where,
42:14local health departments have worked with the community
42:17to identify what's the most appropriate metric
42:22and threshold for issuing heat alerts.
42:26But the challenge with that approach is that,
42:29it's not a systematic investigation
42:30of what would be work the best.
42:33So one of our goals is to think of,
42:37well, let's look everywhere in the country
42:39and see what either by region or by community
42:42or by climate zones,
42:44what would be the optimal metric for predicting,
42:48which are the most dangerous days of extreme heat,
42:52keeping in mind that it's in nobody's interest to issue,
42:58a very high number of heat alerts each year.
43:01So you really wanna focus each summer on like,
43:03what are going to be the worst days,
43:05how do we identify those
43:06and sort of using a health based perspective
43:09rather than a weather based perspective?
43:10So it's not necessarily the hottest days, but rather,
43:13we know from the work of others that, the,
43:18vulnerability to heat varies by location,
43:22by population and by time of year,
43:25as well as it's been shifting over the years.
43:27And so taking all that into consideration,
43:29can we sort of have a health based metric
43:31for issuing heat alerts heat warnings,
43:36and heat advisory's.
43:36Wet bulb globe temperature is a really interesting one.
43:40There's,
43:43I think that it's potentially very interesting,
43:46and I know that in some occupational settings,
43:49a wet bulb globe temperature is used as the guiding metric.
43:55It has not been to my knowledge been widely used,
43:58in sort of population level, heat warning work.
44:04But I think it'd be really interesting
44:05to look at that as well.
44:08- Great, thanks.
44:09Another kind of very detailed technical question
44:14is one students is wondering,
44:16the previous paper,
44:21where you choose the control days,
44:25because if you have a very higher threshold,
44:28then it's likely that you don't have enough control days.
44:34- That's a great question.
44:35So this refers I believe to Kate's study
44:39of looking at the effectiveness of heat warnings.
44:43And so what we did is we compare days,
44:46of the similar heat index
44:48and with or without a heat warning.
44:50And you're right, that for very hot days,
44:54like if a day is 110 degrees, heat index,
44:55that there's not going to be any days
44:58in that same location of 110 degrees,
45:00that didn't have a heat warning.
45:03So, by so we had to limit ourselves to those days in which,
45:09we sometimes saw a heat warning but not always.
45:13And if, a 90 degree day,
45:16nobody's issuing heat alerts and on 110 degree day,
45:19everybody's issuing heat alerts.
45:20And so we had to focus on the middle.
45:22So one of the limitations of this work is that
45:25it is there's no counterfactual,
45:28there's no information about the counterfactual of like,
45:31what would have happened had we not issued a heat alert
45:33on a very, very hot day?
45:35There's just, there's no data is conditional on location.
45:38So that is one of the challenges.
45:40So we should, our results are generalizable
45:42to those days on which you might,
45:45or sometimes issue heat alerts.
45:47And not outside of that relatively narrow band
45:51of temperatures.
45:54- Thanks.
45:55I think we do have a question from the audience,
46:00one of the first, so,
46:02the question from Stephan Lessen is asking
46:06about one third of the Medicaid population
46:10has no access to the internet.
46:12So how, the heat alerts commonly distributed within cities.
46:17- Yeah, that's a really great question.
46:19And again, it varies a little bit by location.
46:22The several or many of the national weather service,
46:26local offices are actually on social media now, and you,
46:29you could follow them on Twitter, there's, also,
46:34you can sign up for their email newsletters,
46:37that'll warn you of particular, threats,
46:41and you're right that those channels,
46:44while they might reach some segments of the population,
46:47they, probably are focused
46:50on those segments of the population
46:52that are particularly engaged
46:53and maybe not particularly at risk,
46:56for heat specifically.
46:58So, traditionally this was all through TV and radio,
47:03where you would say, national weather service has
47:06issued a heat alert for the next two days, or for,
47:09this region for tomorrow and advises you to,
47:13drink lots of water avoid exposing yourself to
47:16your kids to high heat, et cetera.
47:19So I think they use a combination of traditional
47:24and digital media, channels,
47:28but I think it raises a good question of,
47:31are we reaching the most vulnerable populations,
47:33with these alerts?
47:35And even if we inform people that there's a risk
47:38that doesn't necessarily mean that people are able,
47:41to protect themselves from that risk.
47:44So for instance
47:45when we think of the most vulnerable populations,
47:48you're amongst them sort of perhaps outdoor workers,
47:52so outdoor workers, there are guidelines,
47:56in temperatures above which outdoor workers shouldn't work,
48:00but your roofers and landscapers and construction workers,
48:03they're not getting paid if they're not doing the work.
48:05So sort of the opportunity for not just
48:10reaching and informing people,
48:12but actually giving them options
48:13of how to protect themselves,
48:15is I think a really hard challenge.
48:18You see this also with agricultural workers
48:20and other settings.
48:21So I think that there's we have to move from a model
48:25where we're just trying to reach people,
48:27to give them information to discovering, understanding,
48:32and addressing the hurdles
48:35to actually protecting themselves,
48:37or helping them protect themselves,
48:40rather than sort of just an information deficit model.
48:44- Yeah thanks.
48:45I think, kind of follow up on these detailed questions
48:50one of the students is asking like,
48:53behind this (indistinct) system exactly.
48:57Kind of mixture of all multiple different intervention
49:01matters such as you said, some including TV,
49:05some including other informing approaches.
49:09So, kind of further question is how to,
49:14evaluate the cost and effectiveness
49:17of different approaches when people, when
49:20the public health officials want to inform,
49:25want to intervene.
49:27- Yeah, I think it's a really interesting question.
49:29And so there's two questions.
49:30There is sort of what,
49:32how do you evaluate the effectiveness
49:34of these different channels?
49:37And I think the broader question is,
49:39can we move away from thinking that
49:43a channel of communication or a series
49:47works on the population as a whole?
49:48So, for example, if we,
49:50if you wanna try to reach and protect outdoor workers,
49:54there's probably channels of communication
49:56and engagement that are different
49:58than if you're concerned about seniors
50:01in institutional facilities,
50:03or if you're thinking about kids in school
50:05based environments or summer camp environments.
50:08So I think we probably in our communication strategies
50:11and engagement strategies need to move away
50:13from thinking that if only we use channel X,
50:17we'll reach more people,
50:18it's not about reaching more people,
50:20it's about reaching specific segments of the population
50:23that in specific ways that are amenable to their needs
50:29and the resources available to them.
50:32So I think working with school nurses is a great way
50:34to reach kids in school.
50:36I think working with organized kids activities
50:40is a great way to, reach again,
50:43vulnerable children and adolescents.
50:46But those strategies aren't gonna work in other settings.
50:49So I think it has to be much more targeted
50:51than we're doing now.
50:54- Thanks, yes, those words are insightful.
50:58I do have another question from the audience,
51:01from Alexi, is asking,
51:04is there evidence of political inference,
51:07determining the implementation of the warning system?
51:12- It's a great question.
51:13I actually don't know enough to,
51:16so I haven't seen political influence in that, but,
51:19I haven't worked with,
51:22too many national weather service offices directly.
51:27So I think there's probably others involved
51:32that can answer that more.
51:34One of the interesting linkages is that sort of the
51:38whether these heat alerts trigger local action
51:44varies across locations.
51:46So in New York city,
51:47I understand that every time
51:49the national weather service issues a heat warning,
51:52that triggers a certain number of activities.
51:54Like there's no intermediate decision,
51:56whereas in the city of Boston I understand that
51:59it's when the mayor declares a heat emergency,
52:02which is informed by the national weather service forecast
52:04and heat warnings,
52:05but it's not automatically triggered by.
52:07So I think there's some differences in,
52:10or quite a bit of differences actually around the country
52:13as to whether the national weather service heat alerts
52:17automatically trigger action,
52:19or are they informational,
52:21but the action is triggered by some other mechanism.
52:23And that's one of the things that we need
52:26to get a better handle on across the country is
52:28this the right trigger for local heat action plans to,
52:32and heat responds plans to be activated.
52:35And, I don't have a preconceived notion
52:38as to what the right answer there is.
52:40Maybe this is the optimal trigger
52:42or maybe something that it's appropriate
52:44to have an intermediate step of somebody else sort
52:46of making a judgment call for that local population.
52:50So I think that's an exciting area of research.
52:53- Thanks.
52:54We do have another question from, Rob Tuber.
52:57He's asking,
52:58have you ever looked into the effectiveness
53:00of cooling centers?
53:03- I love cooling centers
53:04because they seem like such a great idea.
53:06Oh, people are know dying or or being hurt by heat
53:11let's provide them a cool place to go.
53:13And the anecdotal evidence is that,
53:17you open cooling centers and very few people go.
53:19And so again, understanding the hurdles of that.
53:22And I think, again,
53:25I've worked somewhat with people in New York city
53:28and I understand that they provide
53:30transportation assistance for vulnerable populations,
53:35because I think one of the hurdles they found was that,
53:40not everybody can get themselves to a cooling center,
53:43so you opened a cooling center and that assumes that
53:44somebody can go.
53:46Okay, so there's cultural barriers to or
53:53barriers in terms of like, well,
53:55what am I going to do there?
53:57Is this a place where I'm actually welcome?
53:58How do I get there?
53:59Can I actually afford, like,
54:01if I work, again,
54:03can I take the time to go do that?
54:05Or if I have, medication needs will I be able to,
54:11treat my medical condition while I'm there?
54:13So I think that cooling centers are really
54:15intuitively attractive option.
54:17And I think with so much of what we do in response to heat,
54:21there is not a body of evidence as to what works.
54:23And I think that's really where we need to
54:26sort of move the field is starting to think
54:28about what works in what settings and for whom,
54:30so that we can really provide evidence-based guidance
54:33for developing solutions.
54:37- Thanks very well said.
54:39We do need a lot of these evidence-based research
54:41on these policy actions.
54:43I do have another follow-up question from the students,
54:48is that actually within your next steps?
54:50So the students is kind of wondering
54:54how do you actually verify the causal assumption
54:59in evaluating the heater systems?
55:03- Yeah, that's great.
55:04So, the best we can do is use the data,
55:12this isn't a randomized, these aren't randomized studies.
55:14So the best we can do is,
55:17use observational data to the best of our ability.
55:20So, can we ever prove that we understand
55:23the causal effect of heat alerts?
55:24No, but I think we can do,
55:27more detailed, more insightful analysis
55:32of the existing observational data.
55:34And I think this idea of there are a range of days.
55:39So going back to the heat warnings,
55:41there's these days where we say,
55:42we're always going to issue a heat warning,
55:44'cause it's just so hot that we just take it for granted
55:46that it's dangerous and we need to do something,
55:49so we're going to do it.
55:50And then there's this other bucket,
55:52a days on the other end where like, it's just,
55:55issuing key warnings is just not likely to be effective,
55:57but there's this middle range where you're like,
55:59should I issue a heat warning?
56:01Yes or no.
56:02And so what we're doing is providing information
56:05on that part, the spectrum, and where we say,
56:09should we issue somewhat more heat alerts
56:11because we can do it right around this threshold,
56:14would that save lives?
56:16And, that's it's not the entire picture.
56:20It would be so interesting to know
56:22on these very hot days when we issue heat warnings,
56:25do they actually prevent deaths?
56:29And the problem is as we said before,
56:31that there's no data on the counterfactual,
56:33like what would have happened
56:35had you not issued a heat alert?
56:37So, there's probably other creative ways to do it,
56:40but we haven't figured that out yet.
56:41So this is really about at the margin,
56:44would you do better issuing say 10%
56:46more heat alerts each year,
56:49or 15% more heat alerts each year?
56:50'Cause you don't wanna issue them if they're not,
56:54there's risks of warning,
56:56fatigue of people not taking it seriously.
56:58Because there are too often and there's some costs
57:01associated with each time you issue it,
57:03if it triggers actions.
57:05So it's again, it's like, no, should we issue a few more?
57:08And in that question, we,
57:10so far our evidence suggests
57:12that there's not widespread benefit of them, but,
57:17sort of with the asterisk that more work is needed on that.
57:22- Okay, thanks, yeah.
57:24I think we have the final comment or question
57:28from Donna Spellman.
57:31I've been struggling to see how implementation science
57:34might promote environmental health.
57:38This project is a perfect example of the connection.
57:40Thanks.
57:41- Thanks Donna.
57:43I think that's a great point.
57:44And I think that there I have not seen a large amount
57:47on implementation science,
57:48specifically oriented towards solutions
57:53in environmental health.
57:56We're really great at describing problems
57:58and less good at figuring out and implementing solutions
58:03and then evaluating their effectiveness.
58:05So I think that this is right for that
58:07because we know there's a risk there.
58:08We just don't actually know exactly what to do about it.
58:11And there are lots of good ideas,
58:12but we need to move from good ideas to,
58:15good evidence supporting specific ideas.
58:20- Great.
58:21I think with that we will conclude, this seminar
58:25and thank you Greg, for this wonderful presentation
58:27on the science-based actions.
58:30And, this seminar will be recorded
58:34and will be posted later.
58:37So thank you all for coming and thanks again Greg.
58:40- Wonderful thanks for the opportunity, bye bye.
58:42- Bye.