Climate Change and Health Seminar Series: “Interactive effects of heat and air pollution on mortality: innovative methodological approaches”

October 16, 2023

Information

Dr. Massimo Stafoggia joined YCCCH for a conversation on the interaction between air pollution and air temperature on mortality in 620 cities around the world.

Speaker:

Dr. Massimo StafoggiaSenior Researcher, Department of Epidemiology, Lazio Region Health ServiceAdjunct Senior Lecturer, Institute of Environmental Medicine, Karolinska InstitutetVisiting Researcher, Fielding School of Public Health, University of California Los Angeles

October 16, 2023

ID10864

To CiteDCA Citation Guide

00:59Alixandra Rachman: so without further you let me introduce our speak today. Dr. he's a senior researcher at the Department of Technology, at the Lazio Regional House services in Rome, Italy, he's also a jump to Senior Institute in Sweden, and he currently, he's some recent researcher at the Uci. Video.
01:27Alixandra Rachman: It's my great pleasure to have Dr. Stafford being here today is that he has really, truly one pioneers in in
01:38Alixandra Rachman: studies. If you look at some early to the 2,000 and I have collaboration, but also, and lending from in his work. So this is really great honor to have him here.
02:00Alixandra Rachman: Among many type of projects he has done one thing I want to mention is that he's has leading many hosts international but also Europe
02:21Alixandra Rachman: and the last thing I want to measure is that, as you.
02:25Alixandra Rachman: you know, we a lot of work. Our work is, you know.
02:30Alixandra Rachman: and most of those recent paper has been selected as fast amount every month. It provides the amount of international society for every month.
02:44Alixandra Rachman: So I think it's the the excitement for us to have him today and without residue. Appropriate your perfect.
02:55Alixandra Rachman: Thank you very much. Good morning, everybody. It's a big pleasure to be here. Thank you to the organizers for this opportunity.
03:02Alixandra Rachman: So in this presentation I will. As Kay said, I am environmentally ideologist, I am a statistician by train. So during the my decades of study I try to develop statistical methods. So I will try to focus a little bit on the methodological side, on the on the
03:23Alixandra Rachman: and but of course there will be a lot of results that we can discuss. So I'm very happy to get your comments on methods, results, interpretations. Whatever you do.
03:32Alixandra Rachman: you can think of.
03:34Alixandra Rachman: So the topic of the presentation will be interactive effects of heat and revolution on mortality, innovative methodological approaches.
03:42Alixandra Rachman: This is a topic which is very much.
03:44Alixandra Rachman: Let's say, though, we know that their pollution affects help. We know that string temperatures affect health. We don't know exactly how to how they interact on health, whether
03:56Alixandra Rachman: you know tha tha? There is a synergy between the 2. The independent effects are, you know, the the the interaction between the 2 is greater than the some of the individual effects, whether that is the same in different parts of the world. There is a lot of complexity there, and there is a lot of divergence results across the world.
04:14Alixandra Rachman: So in this study I will present you 2 main projects where we have been working on this on this topic. The first project, if is, I don't know.
04:25Alixandra Rachman: Stevie.
04:30Great.
04:33it's not working. So
04:36Alixandra Rachman: move them up.
04:38Alixandra Rachman: anyway. The the first project is a European project.
04:42Alixandra Rachman: Yes, the Exhaustion project and the the it's a funded by the horizon. 2020. It's a very big project involved many countries in Europe. You see, the topic is about heat, air, pollution is a side, and this project the focus of the the the project is exposure to heat and the pollution in Europe currently, in box and benefits of mitigation and adaptation. So it's not only research. There is a lot of a you know public.
05:11Alixandra Rachman: We entered the, you know, investigation here in terms of adaptation and mitigation strategies.
05:17Alixandra Rachman: And but it's a big project with different types of data that have been collected, that you can see 3 3 different levels of analysis.
05:26Alixandra Rachman: So there is data from individual cities.
05:31Alixandra Rachman: The Continent, with time series approaches to investigate the effect of heat on Cardi Bordinari outcomes at the city level.
05:40Alixandra Rachman: Then there is a
05:42Alixandra Rachman: this is which is the one I will focus today, which is a small area analyzed. So for a few of the countries we were able to collect data, not just for the main.
05:52Alixandra Rachman: for the entire country. So cities, rural areas, small towns, villages, etc. And so the purpose there was to exploit the potential of the special templeite contrasts
06:03Alixandra Rachman: in exposures and outcomes, especially to see whether there was some differences between rural, suburban and rural areas when it comes to effect of heat and people outcome. And the level 3, which is here is called data. So we have a few cohorts in Europe, and the idea was to investigate both long term effects and short term effects.
06:26Alixandra Rachman: And also we study the different aspects of the relationship between a temperature and outcomes. And we focus on what is create that blue blue circle here which is the interaction between the temperature and their pollution.
06:40Alixandra Rachman: So that's that was the the the focus of this presentation for exhaustion.
06:46Alixandra Rachman: Some details about the methodological steps that we have undertaking this this project
06:53Alixandra Rachman: so that we can go, we can go more into the details of the again. The objective was to estimate the effect of high temperatures by levels of airplane.
07:03Alixandra Rachman: So in this part, in this project.
07:05Alixandra Rachman: we are considering a third temperature as our main exposure. So we are interested in the effect of our temperature, and specifically the heat.
07:14Alixandra Rachman: So we are focusing on the part of the relationship
07:18Alixandra Rachman: on the right side of the exposure response function. So when it
07:22Alixandra Rachman: so if you look at the scale of temperature. We are not investigating this part, the effect of cold temperatures, or the effect of heat.
07:30Alixandra Rachman: and that's the first choice that we made in this project to kind of a restrict to the the size of the study. Air pollutants are considered in in this study as effect modifiers, so the task here is to investigate whether the effect of temperature changes for different levels of air pollution
07:51Alixandra Rachman: in the next project that I will talk in the second part of the presentation. Now we will do also the other way about.
07:57Alixandra Rachman: So we would see whether the effect of the pollution changes for different levels of temperature, because that is also interesting. And that's part of the the synergy between the 2 exposure. I think this part we will focus on.
08:10Alixandra Rachman: So, as I said, that we focused study on those countries which provided the small area data. And these were 5 countries that I will present in the next slides.
08:20Alixandra Rachman: So we have added the up levels.
08:24Alixandra Rachman: larger area units like counties or regions, and smaller.
08:29Alixandra Rachman: like provinces or municipalities, or in in Uk. The business distinction between Ms way, which are different size of zoom, and we will explain that
08:41Alixandra Rachman: later why we decided to have different levels. Exposure is daily in temperature as effect modifiers. We are considering these 3 pollutants. So just to have an understanding.
08:53Alixandra Rachman: all of you are more or less familiar with the air, pollution, epidemiology? Or do you need some kind of a introduction to what Pm. Is what ozone is?
09:04Alixandra Rachman: Can I assume that more or less
09:07Alixandra Rachman: so? Pm. Stands for particulate matter. 10 or 22.5 is the size, fraction, the finest, the fraction, the most, the more plausible are the particles to penetrate deep into the air wave system, and then to translocate into other blood circulation, and to to affect other other organs as well.
09:28Alixandra Rachman: So in this, that is, we are considering. Pm, 10 pm. 2.5 and and so let's focus a little bit on the analytical strategy. So first of all, we couldn't pull the country specific data, all in one big data set, each
09:45Alixandra Rachman: country referent analyze their own data. So it's country specific analysis. This meant that we had to produce a code that was standardized and that was shared among all the study participants. Each one of them run the analysis. They sent back the results we double checked, and we kind of find you under the process.
10:10Alixandra Rachman: So in each country, as I said, we defined the large areas, meaning.
10:16Alixandra Rachman: for example, just make to make it simple. In Italy we have, more or less, 8,000 municipalities, but some of them are very, very small. So analyzing each one of them individually, was not feasible.
10:30Alixandra Rachman: both, computationally speaking, but also in terms of statistical power. Many small municipalities contribute very, very few cases, so it was impossible to analyze each one of them individually. So what we did was to pull the municipality specific data into province specific data sets.
10:50Alixandra Rachman: And we analyzed each province
10:53Alixandra Rachman: per se. So we have 20 provinces. So you can understand in each province that are, I don't know, 200 300 municipalities. That is what how we clustered our data set to make it treatable. So analyze the study unit is still the municipality and the day.
11:10Alixandra Rachman: But the data set that we analyzed it one by one was the products.
11:14Alixandra Rachman: And we did the same in the other countries, different types of denominations. So Uk, the large areas are the Mso and the small areas are the way other places. We have counties.
11:27Alixandra Rachman: And yeah, but see, why is it so? Okay? So the ideas that we analyzed, for for example for Italy, each one of the 20 countries separately. Then we met, analyzed
11:39Alixandra Rachman: the 20 county specific results in order to get
11:42Alixandra Rachman: country specific result. And then we put together the 5 countries to get
11:50Alixandra Rachman: both associations. Is that clear that the process?
11:54Alixandra Rachman: Okay?
11:56Alixandra Rachman: And as I said, we focused on the world season because the project was much about heat.
12:03Alixandra Rachman: climate change, global warming. So even though there is an interest even in investigating the other side of the relationship, we had to kind of restrict our our study in some way. So we focused but the attention of the world season that we defined as a May September.
12:20Alixandra Rachman: and in sensitivity analysis, we also further restrict to June to August.
12:25Alixandra Rachman: We will see next project, which is based on global data. We couldn't just decide define, based on monthly.
12:33Alixandra Rachman: because May, September is summer, in some places is winter, in others we have to do something different, and we will see that later and in terms of models.
12:44Alixandra Rachman: It is a time series, approach. We have daily counts of specific mortality, daily mean temperature, daily air pollution. So it's a postal regression. We apply the time quasimos on regression modeling
12:59Alixandra Rachman: adjusted for time trends and other potential confounders.
13:03Alixandra Rachman: Don't think I would go too much into the details of the modeling part. But if there are specific questions on that.
13:10Alixandra Rachman: of course you can. You can ask. But the idea of the awesome regression for those of you who are not familiar with that is that they poison is based on count. So it nicely fits into the the analysis of count data. Firstly.
13:25Alixandra Rachman: second, the idea of a time series analysis is that you want to kind of a.
13:31Alixandra Rachman: let's say, focus on the day to day variability between exposure and outcome. So you want to filter out long term and seasonal trend
13:40Alixandra Rachman: seasonal trends, so that what is left to be to to to make our inference off is the short term day to day vulnerability.
13:48Alixandra Rachman: So this is what we did in this analysis to analyze the data adjusting for time trends both to capture the long term and to capture the seasonality aspects.
14:01Alixandra Rachman: Okay, so the complex part was the the this tensors motor here. So the idea was that we wanted to capture the
14:12Alixandra Rachman: complex relationship between co-exposure to our temperature and air pollution.
14:18Alixandra Rachman: So that's the idea. And to do that, basically, what we did was
14:23Alixandra Rachman: trying to be as flexible as possible by considering products of of splines. So we we were flexible in terms of the exposure response functions, both for the air pollution and for our temperature. And we use this tensor smooter, which is
14:39Alixandra Rachman: basically a product of splines. You know what the spline is.
14:44Alixandra Rachman: It's a nonlinear function. So we relaxed the assumption of linearity, and we assume that that the relationship between temperature and mortality could be nonlinear, because this is what it is, it is nonlinear. So there is a point of temperature where the risk is minimum, and then, departing from that point, you see the risk increasing both on the right, left side for cold temperature and on the right side for the world damage. So it's a nonlinear relationship
15:13Alixandra Rachman: for air pollution. Is not that clear? I mean, there is a lot of evidence that the linearity assumption might hold
15:21Alixandra Rachman: so ideally. We might have kept air pollution. As you know. we will see in the second project, that this is exactly what we decided
15:29Alixandra Rachman: to keep our pollution as linear in order to simplify the interpretation and the presentation of the results. Here their pollution was just an effect modifier.
15:39Alixandra Rachman: Here we were interested in extrapolating the effect of temperature by levels of our pollution.
15:45Alixandra Rachman: So we use this tensor smoothing. And also we simplify the the the latency of the relationship. Considering a luxio, one luxio, one means that we are investigating the relationship between today's
15:58Alixandra Rachman: mortality and today's and yesterday exposure.
16:02Alixandra Rachman: we didn't go
16:04Alixandra Rachman: back more days.
16:06Alixandra Rachman: because
16:08Alixandra Rachman: from the literature we know that when it comes to the effect of heat, usually it is the very first days which have strongest impact.
16:16Alixandra Rachman: You don't usually go much. Thank you, Tom.
16:19Alixandra Rachman: When it comes to cold effects it is a more delayed association. Usually you you would see effect even that way. You study the exposure one or 2 weeks before the outcome.
16:31Alixandra Rachman: But for the heat it's more the heat wave that that you know the very peak in exposure today or yesterday.
16:38Alixandra Rachman: So like 0, one is basically the the average exposure between today and yesterday.
16:43Alixandra Rachman: Okay.
16:45Alixandra Rachman: okay, we have to. Since the effect of temperature is not linear. We have to define how to express this relative risk, this effect
16:57Alixandra Rachman: that are different for the possible ways in the literature. If any one of you is studying this kind of literature, you will see different ways to express the effect of heat. So you have. When I say effective, you have to define a reference point, and you have to compare the risk at some point versus the reference. That's the relative risk we have in mind.
17:19Alixandra Rachman: We've decided that since we are, we were analyzing different countries and different special units within each country, we decided
17:27Alixandra Rachman: to choose some person types which will correspond to different levels of temperature in each flavor department.
17:36Alixandra Rachman: We we do not expect that the effect would be the same at the same level of temperature because of adaptation to different climates. Even within a country you were that
17:46Alixandra Rachman: diverts, you know, temperature ranges north to south.
17:51Alixandra Rachman: So we decided to be kind of relatively
17:55Alixandra Rachman: dies. So basically, we are expressing the the effect of temperature for increments of temperature from the 70 fifth to the 99 percentile of the location specific distribution.
18:09Alixandra Rachman: Okay? And for revolution same, we have to define what is low, what is medium, what is high. And we decided to use a fifth, fiftieth and 90 fifth percentile as levels of air pollution
18:24Alixandra Rachman: and insensitivity. Analysis will be did something. Okay.
18:29Alixandra Rachman: So this is just to give you a visual
18:33Alixandra Rachman: representation of what we are doing.
18:35Alixandra Rachman: You can see this is kind of a complex relationship. You have a
18:39Alixandra Rachman: your outcome here when it's written linear, predatory. You can imagine that as the risk of mortality, you have air pollution. Here you have temperature.
18:48Alixandra Rachman: and for increased
18:50Alixandra Rachman: temperature, you have the risk which goes up in different ways, depending on where you are in this joint relationship.
18:57Alixandra Rachman: So the idea was to
18:59Alixandra Rachman: take cross sections of this relationship at different levels of pm, so let's say low, I'm sorry. Yes, low, medium and high. And take out of these 3 curves. Take the
19:13Alixandra Rachman: effect of temperature expressed top 70 fifth to the 99% time. And the nice feature of this graph is that you can reverse, and you can express the effect of air pollution
19:26Alixandra Rachman: levels of temperature from exactly the same model. That was what we did in the second project that have been.
19:34Alixandra Rachman: and the complex part was to derive standard error. So using a statistical lifestyle.
19:45Alixandra Rachman: okay.
19:46Alixandra Rachman: so these are the countries that were involved, you can see the the special units.
19:53Alixandra Rachman: And this is the map of yes, mean or temperature level. Kind of difficult to see the
20:00Alixandra Rachman: the range. You see, it's it's in
20:02Alixandra Rachman: Centigrade Centigrade. So I don't know in far 9. How I mean how that converts! But it's there is different, of course, variability. We have Norway, which is far north.
20:15Alixandra Rachman: much colder. We have England and Germany, which are kind of more in the center, and then Antarctica, which represents the Mediterranean climate.
20:25Alixandra Rachman: and these are maps of Pn. 2.5
20:29Alixandra Rachman: higher level. Even here you can see some vulnerability. In Italy we usually see this very high
20:37Alixandra Rachman: level in Mobile, which is part of Italy, which is one of the most pollutant polluted areas to Europe.
20:45Alixandra Rachman: And I mean different levels in different. This is, of course, London, for example.
20:50etc.
20:53Alixandra Rachman: That's awesome. That we didn't have data for Uk. So only this is quite small. But just give you an idea of the amount of data.
21:05Alixandra Rachman: So this is a number of events in terms of hundreds of thousands. So good part of the study was the power. We had a lot of statistical power, small effects.
21:17Alixandra Rachman: but I will start to the next step. Okay, so this is a summary of the main effects
21:25Alixandra Rachman: as a meta analysis of the 5 countries.
21:29Alixandra Rachman: And so this is the effect of our temperature by a revolution by Pm. Toward 5 levels, which is low, medium, and high.
21:37Alixandra Rachman: and these are expressed, as I said, change in mortality. This case you can see Co specific. So after it calls, which is the the
21:49Alixandra Rachman: some of cardiovascular and respiratory. And then you see the 2 groups separated.
21:55Alixandra Rachman: And basically what we saw, which is something that we usually see, especially from European studies. It's an increasing effect of temperature for higher levels. So this is quite a
22:08Alixandra Rachman: quite common tool to see.
22:13Alixandra Rachman: and we run also several sensitive analysis, where, as I said, we define the world period, does the 3 warmest months rather than the the 5. We tried to use a more extended log definition up to 4 days, and with different use, different types of their pollutants
22:31Alixandra Rachman: as the partiles, but basically the the name, each of the main.
22:36Alixandra Rachman: And this is the same for those.
22:39Alixandra Rachman: So this case it is less clear. I mean, we have in this case we see much higher effect in the high levels, not so great difference between low and medium. So it's not an an increasing, smooth, increasing gradient of effect. But it seems that the interaction happens especially at very extremes. And interestingly, this is also what happens in the global analysis from the the second project that would come.
23:07Alixandra Rachman: Okay, this is just some results for the country. Specific analysis also see them. There was some consistencies across countries. In some cases like, like, I guess.
23:20Alixandra Rachman: Italy, Germany, it was more pronounced in other cases. It was a little bit so. For example, in England it was not. That's from the, but mostly they would kind of suggest in the same pattern.
23:36Alixandra Rachman: and that that is below zone again for us, and we have some more. Heterogeneous sessions
23:43Alixandra Rachman: in some cases also the opposite way don't know.
23:48Alixandra Rachman: and we don't really know why. So if one of the question is, why not always the other way around. I don't know. The answer is, I don't know.
23:56Alixandra Rachman: so, in short, we found adverse effect of high temperature and mortality. But that's what that was already known from the literature, and we confirmed that in the sense
24:07Alixandra Rachman: we found stronger effects of high temperatures on the days with high P and point 5 concentrations, and the same also for for us, and with some heterogeneities across across counties.
24:20Alixandra Rachman: okay? Any question? Yes, before moving to the next. Yeah. That's really interesting. I mean, coming from from growing up in Norway myself. Ii know how cold it is.
24:33Alixandra Rachman: and I know how warm it is in Italy. And you said that you, you use these percentals right to reflect that we are.
24:40Alixandra Rachman: We are accustomed to living in a certain climate, right? Right? But did you?
24:44Alixandra Rachman: There might also be sort of these biological threshold points right where, like no one
24:50Alixandra Rachman: is dealing well with the temperature above 35°C, for instance. And like, did you? Did you look at absolute temperatures as well for this project, or just? I mean in in other projects. What we did was or investigate the the heat wave part exactly to go in the direction you are saying. I mean, if it is above some threshold. Still, location specific is still kind of use, the the absolute values, because it will
25:17Alixandra Rachman: be very difficult even to find a common support in some countries. But if it is above some very high percent time prolonged for several days, maybe that is when something triggers. So we did that as well. Not here, not here. Here we try to exploit the
25:35Alixandra Rachman: kind of the complex relationship hopefully at those extremes with the the 3D. Function. But I agree with you that it might be interesting to focus on the very extremes, because, biologically speaking.
25:49Alixandra Rachman: something different might happen. So the rough studies that show that heat waves have might add something to the
25:57Alixandra Rachman: continuous, you know, nonlinear exposure, response function. There will be something with the heat waves which goes beyond. The simple exposure response functions we are seeing here.
26:08Alixandra Rachman: Yeah. thanks.
26:13Alixandra Rachman: Yes, it's a very great project. And my question is about so tensor, smooth product. I mean, when you form the tensile product, for example, Pm. To move wine and a high temperature. Have you ever adjusted for ozone and the Pm. Tap in the tensor app? So this is a good
26:36Alixandra Rachman: question definitely, not for piano, because usually when we investigate it, when we investigate. Pn. 2.5, adjusting for P. And 10 is kind of a counterintuitive one is part of the other.
26:49Alixandra Rachman: So you would change the interpretation of the coefficient, because usually, since one is contained in the other, adjusting for one would mean kind of make your inference on the residual part, which is the course fraction.
27:02Alixandra Rachman: This was not something we really wanted to do here, so we didn't adjust
27:07Alixandra Rachman: for PM. 10 when we investigated Pn. 2.5 or the other way around
27:11Alixandra Rachman: for Rosalinda. I think in some of the sensitivity analysis. We did that. And it didn't change much. So we we didn't do. We didn't change. It didn't change our our relationships.
27:24Alixandra Rachman: definitely, we did that in the next project that I will show. So there was not a lot of residual confounding from mutual go. Pollutant textbooks.
27:38Alixandra Rachman: Is it? Pm. To Pm.
27:45Alixandra Rachman: PM. 2.5.
27:47Alixandra Rachman: So what people do if they want to focus on the course fraction
27:55Alixandra Rachman: thing you can do is to subtract Pm. 2.5 from P. And 10 ahead. So to get a sort of estimate of the course fraction. I say it's naive, because to do that you should have exactly the same monitors make sure that they are monitoring exactly the same way at the same time, and then you get some kind of realistic estimates of the course fraction.
28:19Alixandra Rachman: But even so, you are mixing together measurement which are taken with some errors. So you are adding up errors in the Pm. 2.5 estimates errors in the Pn. Pn. 10. And so that what usually happens is that the signal you see in the course fraction
28:36Alixandra Rachman: this way, you have to find it this way is more noise because of this exposure measurement, arrow, which comes from both
28:43Alixandra Rachman: fraction, but conceptually investigating the coarse fraction. is is very interesting, because course fraction is
28:52Alixandra Rachman: possibly
28:54Alixandra Rachman: by different sources of other pollution. And so it might be interesting. For example, natural sources in Southern Europe we have lot of intrusions that are dust from the Sahara. That impacts more the cores fraction and less than the fine fraction. So investigating different fractions might be interesting
29:13Alixandra Rachman: efficiently.
29:17Alixandra Rachman: Okay.
29:19Alixandra Rachman: So we move to the next.
29:22Alixandra Rachman: Okay? So the Ncc collaborative network for those of you who haven't heard about is a collaborative network of investigators from all around the world. people repairing data
29:35Alixandra Rachman: on a voluntary basis. Basically.
29:37Alixandra Rachman: it's Nona. It's a non. It's anonymous data. It's a a
29:42Alixandra Rachman: aggregated data. It's daily counts of mortality. Daily mean temperature, daily mean air pollution for all cities around the world.
29:52Alixandra Rachman: It is led by the London School of Egyptian and tropical medicine as a price.
29:58Alixandra Rachman: It's a big effort. But the nice thing is that people can volunteer to to run up. I mean to conduct a study. So there is protocol to feel like you send it to the steering committee. Usually they are okay. Or maybe they can put you in touch with people doing something similar within the same network.
30:18Alixandra Rachman: You get the data, you run the analysis. So it's it's a good opportunity for all of you that might be interested in working on this time series data to to know about the
30:30Alixandra Rachman: yes, this is summarized here. Of course, the representation is not perfect.
30:35Alixandra Rachman: There are areas like most of Africa
30:39Alixandra Rachman: which is not represented. Other areas in Asia not represented only a few locations from Australia. But that's one of the limitations of a voluntary based, you know.
30:51Alixandra Rachman: and collaborations.
30:55Alixandra Rachman: And okay.
30:59Alixandra Rachman: so all this. So what we did here was to investigate whether the association between our temperature
31:06Alixandra Rachman: and mortality is modified by our position
31:09Alixandra Rachman: exactly as before.
31:11Alixandra Rachman: but also the other way around, so to evaluate whether the association between air pollutants and mortality is modified by our temperature. and then we try to see whether there was any. I mean differences
31:24Alixandra Rachman: across different countries across different continents. Try to to to have a
31:31Alixandra Rachman: perception of the virginity of the results.
31:36Alixandra Rachman: Very briefly, this is the the distribution of cities. This is temperature disease.
31:43Alixandra Rachman: pn, 10,
31:44Alixandra Rachman: average. So we use daily averages. But of course, for the map. We use the average across the entire time series.
31:51Alixandra Rachman: and this is piano owns on Mpm. To 45, and we have an end of tourism here. She's not here. Hello.
32:02Alixandra Rachman: small, very big table with a lot of numbers. We will not go into them. Just give you a flavor of the variety of the countries generally.
32:12Alixandra Rachman: Any one of you know about the Ncc network.
32:16Alixandra Rachman: Okay, let's see.
32:21so much.
32:22Alixandra Rachman: So thank you. Some different, some similarities with the previous studies and some differences.
32:28Alixandra Rachman: Again, here we focus the attention on the world. Video.
32:32Alixandra Rachman: one of the comments of the reviewers, why are you only doing world wide. Not the cold period.
32:38Alixandra Rachman: because it was too big.
32:41Alixandra Rachman: would be another paper doing also the call effects. It's very interesting.
32:47Alixandra Rachman: but it was another. It was matter for another. In this case we couldn't just define. Based on calendar months we had to adapt to each
32:56Alixandra Rachman: different country, and the best
32:59Alixandra Rachman: could think of was to to consider a running average of the monthly means.
33:04Alixandra Rachman: and to define as the warm period of the
33:08Alixandra Rachman: the the 6 consecutive war response in each in each country.
33:14Alixandra Rachman: You need to see it in sort of.
33:16Alixandra Rachman: And this is what we did. So we have a continuous time period, because it's consecutive months. It might shift a little bit from one place to the other, but that was exactly the point.
33:27Alixandra Rachman: and in sensitive analysis, with strict
33:31work as possible.
33:33Alixandra Rachman: Again, it's time series data by city. So in this case we don't have a big area, small area. It's just the city, the country, and the and the country. Yes, so we have a time series for each one of these 620 cities.
33:49Alixandra Rachman: and we analyze them with using again that person aggression followed by country specific, and then some global meta analysis
34:00Alixandra Rachman: again. Both. The exposure in this case we don't have just one exposure. All of them are exposures. All of them are effect modifiers.
34:09Alixandra Rachman: So because we study both sides of that 3D relationship and all of them over analyzed at the lak 0 one for the city
34:17Alixandra Rachman: I mentioned before.
34:20Alixandra Rachman: In this case we adopted a different approach. So
34:25Alixandra Rachman: we apply that and a product term between a cubic spline for temperature and a linear term for air pollution.
34:33Alixandra Rachman: And the reason for that is that
34:35Alixandra Rachman: one of the nice feature of the Ncc. Is that before going to external review you go to internal review, and in this collaboration there are a lot of big. you know, epidemiologists giving giving you insights.
34:50Alixandra Rachman: George Schwartz is Chappelle ben Armstrong. All big names. and many of them were not convinced about using their pollution as nonlinear. Originally I had used the tensors motor exactly. Yes.
35:04Alixandra Rachman: showed before.
35:05Alixandra Rachman: But then they say, Okay, it's difficult to explain. It's difficult, because one thing is to report the effect of temperature by
35:12Alixandra Rachman: when you express the effect of revolution. What does that mean? Going from the I don't know. From the 20 fifth to the 70 fifth person time, it's totally different levels.
35:23Alixandra Rachman: So the the adaptation
35:27Alixandra Rachman: think, you know, we can kind of easily understand it for temperature, that it it is relative because of adaptation. And so you want to keep it relative to each location.
35:37Alixandra Rachman: But when it comes to a pollution, it's not very that that easy to understand that you should be relatively, I mean 20 micrograms is 20 micrograms.
35:48Alixandra Rachman: although there might be still some adaptation there as well. There are places where, on average they have 50 or 60 micrograms per cubic meter, and they do not die.
35:58Alixandra Rachman: So there is some adaptation. But still it's not like Earth temperature.
36:03Alixandra Rachman: and also all of the literature uses, you know, usually linear turns, 10 micrograms, etc.
36:10Alixandra Rachman: So because of all these comments, I switched to this approach.
36:14Alixandra Rachman: which is nonlinear temperature, linear air pollution still exploiting the 3D relationship.
36:21Alixandra Rachman: but with some further constraint.
36:24Alixandra Rachman: And then we have usual co-founders, which are a long-term and seasonal time trends and day of okay.
36:31Alixandra Rachman: And it was not all cause. Mortality here didn't have cause specific mortality in this study, but we did cause specific mortality in another study
36:41Alixandra Rachman: from the Ncc.
36:43Alixandra Rachman: So here they're pollutants are expert. The relative risk of mortality are expressed per 10 microgram per qubrelled.
36:53Alixandra Rachman: Okay? And we express that overall to see what was the effect of revolution per se, and then by changing temperature levels from the first to the one on the
37:03Alixandra Rachman: 100% time. So we
37:06Alixandra Rachman: we're kind of more continuous here. And the reason was that the internal of the viewers told us, why are you just picking up
37:14Alixandra Rachman: a few percent. Why don't you explore the full
37:18Alixandra Rachman: range of of the effect modifier?
37:21Alixandra Rachman: And the same we did, for
37:23Alixandra Rachman: we express the relative risk for increments of our temperature software temperature
37:28Alixandra Rachman: on the 70 fifth to the 99% title of city specific distribution.
37:33Alixandra Rachman: But then we changed exposure levels from 10 to 150. But this is pollutant specific, and the range of the pollution was based on the common support of all the cities involved
37:47Alixandra Rachman: when temperature was between the 70, fifth and 99.
37:51Alixandra Rachman: So it's it's complicated. But the idea is that if we are expressing the effect of temperature by 70 fifth to 99. Maybe in that range you never had air pollution below some values or above some like values.
38:05Alixandra Rachman: you are restricting. Your effect modifier range. And so we try to find ranges which accommodated most of the cities in the analysis.
38:16Alixandra Rachman: Is that clear for us?
38:22Alixandra Rachman: Okay, so these are the main effects. So before going into the effect modification. Let's just see what kind of main effect we found. So this top part is temperature. So in the main model, we analyze the 620 cities.
38:37Alixandra Rachman: we found that when temperature increased from the 70 fifth to the 99% time of city specific distribution on average, globally of the 620 cities, mortality increased by 8.8%,
38:52Alixandra Rachman: which is quite a strong association, which is very similar to what is found in the literature also from the Mcc.
39:01Alixandra Rachman: And this is what happens where we start adjusting for air pollutants.
39:06Alixandra Rachman: So if we adjust for. And you see the number of cities changes, because not all of them, not all the 620 cities and data on all meetings.
39:16Alixandra Rachman: So in total, we had 620 cities, but the combinations were different, depending on the availability of one for the other. But, interestingly, despite you know, the different mix of cities the results required
39:30Alixandra Rachman: who must upset
39:32Alixandra Rachman: here we have to use Lauxido for piano 2.5, because Pm. 2.5 pm. 2.5 was mostly contributed by the Us.
39:42Alixandra Rachman: But unfortunately, the Us. Doesn't provide them. It provides data every third day
39:48Alixandra Rachman: for most of the years, including in the data center.
39:50Alixandra Rachman: and so it was impossible to have a 0 one package. I mean, we could have it, but at the cost of losing a lot of information. so we prefer to use to preserve as much information as possible.
40:06Alixandra Rachman: But still you see similar results. And then these are sensitivity, analysis, different
40:12Alixandra Rachman: ways of so seasonal, defined as the 3 warmest months seems more effective.
40:17Alixandra Rachman: Time trends analyzed with different degrees of freedom. I mean, these are details.
40:22Alixandra Rachman: You can see here that when you start expanding the lug
40:27Alixandra Rachman: the the effect
40:29Alixandra Rachman: drops, and this is expected. As I was saying before, it is, the very most of the effect is concentrated in very first days, if you dilute it, using a more delayed time window, the effect will go down.
40:45Alixandra Rachman: And yeah, this is, instead of using the 70 fifth as as reference. What if we use the fiftieth person type
40:53Alixandra Rachman: and the results?
40:54Alixandra Rachman: It's a little bit bigger, because the 70 fifth is already in this increasing part of the curve.
41:03Alixandra Rachman: and these are the effects of pm. 10. Time micro rapper to a bit of increments. Pian 2.5 ozone and and talk.
41:11Alixandra Rachman: The magnitude of the association is much smaller. Again, this is totally in line with the literature.
41:19Alixandra Rachman: but we we found value the results which are very much line. We bought as we published by the Ncc collaborative network, very important papers around.
41:30Alixandra Rachman: So 0 point 4 0 point 6 0 point 6. So small effects. But you can see very precise. So very narrow confidence.
41:40Alixandra Rachman: Okay, so
41:42Alixandra Rachman: that this is very simple. A a example of what happens when we start investigating the effect of temperature by 11 s.
41:52Alixandra Rachman: Of pollutants.
41:54Alixandra Rachman: And what we see here is that basically apart, some functioning here, but with very large confidence intervals we see quite a clear, increasing gradient of the effect. So effect of temperature is more and more high, more
42:11Alixandra Rachman: high foot higher level of Pn. 2.5 4 pm. 10 at the flattening seats. Kind of more for now. So maybe the disability something going on.
42:21Alixandra Rachman: So at some point you have some maximum effect in that.
42:25Alixandra Rachman: It's also to be said that at this point the number of cities start to become very small.
42:31Alixandra Rachman: So the case mix of cities included in which
42:34Alixandra Rachman: I knew of you know, different concentrations is different. So that is also why you see larger competency problems.
42:43Alixandra Rachman: Olson and Panoto, so pretty much the same, the same feature.
42:50Alixandra Rachman: while when we see the other way around. So these are the effects of different air pollutants by levels of air temperature.
43:01Alixandra Rachman: we are still in the world season. So even at the 0 10% time, it's not very cold. We are still in the 6 warmest months.
43:11Alixandra Rachman: But what we can see here is that art sum up and down. There is a sharp increase. In the effect of a pollutants when
43:21Alixandra Rachman: temperature person ties are the highest. So it's not an increasing gradient. It's kind of kept on some up and down, or maybe some more or less stable effect, and then there is a sharp increase at the experience.
43:37Alixandra Rachman: This is also the case for ozone more pronounced
43:42Alixandra Rachman: while forever told. It's not that clear. Actually. So it goes down. But here it's pretty much similar. So we don't see. So we see interroginous effect effects across levels.
43:55Alixandra Rachman: But we don't really see trend that we can. Now
43:59Alixandra Rachman: that we can.
44:02Alixandra Rachman: Okay.
44:04Alixandra Rachman: these are just a few figures of what apps into different countries. We couldn't go much deeper into that, because it's a huge study. I mean, it would be interesting to investigate
44:18Alixandra Rachman: why there are these differences. What kind of a country or city specific characteristics are responsible for these different slopes. We didn't go into that panel of the dates, but what we see here is that there are some places
44:32Alixandra Rachman: Europe is kind of popping up, always very sharp, increases in the effect of of temperature by increasing level of our blue dots. So the red one
44:44Alixandra Rachman: and Australia is only 3 cities. It's not really Australia. So one of the limitations of the study is that it's not representative of the whole count representative of the cities contributing for that count.
44:58Alixandra Rachman: So.
44:59Alixandra Rachman: but we have to keep some labels here. So the 3 cities in Australia present these results, while in the Us. For example, which is here
45:09Alixandra Rachman: there is some
45:11Alixandra Rachman: increasing effect here, not much going on, so there are differences.
45:16Alixandra Rachman: and the same is true here for ozone and the admin.
45:20Alixandra Rachman: So this is interesting to launch.
45:25Alixandra Rachman: while yes, that's
45:27Alixandra Rachman: yes.
45:29Alixandra Rachman: So in conclusion, we
45:31Alixandra Rachman: deep fire find adverse effects of high temperatures, and and so we confirm what is known from the literature. We also saw stronger effects of attack, versions of days with high pollutant concentrations, and that was kind of
45:47Alixandra Rachman: sort of smooth, increasing, increasing gradients of effect for higher levels. For
45:54Alixandra Rachman: while we, we also found stronger effect of our pollutants when temperature was high, is that was highest, but with a totally different pattern. So the effect was totally concentrated. The very extreme days of damage.
46:09Alixandra Rachman: and also some high levels of interrogating of the originality. the results across regions and the cross-country.
46:20Alixandra Rachman: Yeah, I think that's that's all.
46:30Alixandra Rachman: Yeah, I think I would have have time for maybe 2 or 3 questions.
46:35Alixandra Rachman: I know. So you'll submit to questions through Congress. But that would be good time to ask.
46:44Alixandra Rachman: Was it clear with some part that I totally lost?
46:51Alixandra Rachman: Yeah, I guess one question I had for the seconds of both studies and the sensitivity analysis. You considered the lag from 0 for 10 days.
47:00Alixandra Rachman: potentially
47:02Alixandra Rachman: or accurate, in a sense, because it's maybe considering for mortality displacement of the first couple of days, or if that was less relevant to the study because you were looking at the association between heat and air pollutants rather than like the displacement. We haven't really analyzed that
47:21Alixandra Rachman: if it was true that after a peak in the effect. On the first days there is a drop, even a protective in the next days, because of the depletion of the susceptible population.
47:34Alixandra Rachman: I could say that maybe a 0 10 is more realistic, you know feature.
47:39Alixandra Rachman: but that is not
47:42Alixandra Rachman: certain. It it can also happen that there is a very sharp increase at the beginning, and then the effect is not there anymore. It's flat and having a 0 10 will dilute. You would not see what's what's happening when there are peaks in air temperature for a few days, so to me it is more the second than the first.
48:02Alixandra Rachman: I wouldn't consider the 0 1 0 3 are more accurate
48:07Alixandra Rachman: picture of what's happening when it's hot, because of what literature tells us about. You know, heat waves and how mortality increase exponentially immediately. When you have such a triggering. You know environmental trigger rather than going. But of course, that it would be interesting to investigate the displacement, maybe even over a longer time period.
48:30Alixandra Rachman: In addition to that, yes, here the focus was on the interaction. and for revolution there is.
48:37Alixandra Rachman: I mean I. There is not much in the literature in terms of mortality displacement. while there is something to temperature. So I think that is.
48:46Alixandra Rachman: we had to keep it simple for the ease of interpretation
48:57Alixandra Rachman: question. Yes, I think if we're talk very clear and my question is in the Mcc project. Did you ever try to analysis. Analyze the relationship. I mean the interactive effect in different climate zones. I mean.
49:14Alixandra Rachman: maybe in some very hot areas, hot rooms, they may have different. You know the interactive effect.
49:22Alixandra Rachman: You know the hot area compared to the colder area. Maybe they have a different relationship. As far as I know, in the Ncc. The only project which dealt with the interaction with our temperature. The pollution is this one.
49:35Alixandra Rachman: because if there was another one, the steering committee would have put us
49:38Alixandra Rachman: an action, and we would have done that together in this one. We didn't go into that level of detail, but that would be definitely an interesting development. So we see the recent originating.
49:51Alixandra Rachman: why is that we have 620 cities. So ideally, we might collect for each city some city specific characteristics, for example, the climatic zone.
50:03Alixandra Rachman: or other characteristics that might explain that virginity, and see once you account for those, how much the virginity uses.
50:12Alixandra Rachman: or how they modify that. So that would be a good development. Yeah, like building off of that. That was one observation I had is that
50:21Alixandra Rachman: Europe is kind of a more homogeneous climate than
50:26Alixandra Rachman: Asia or North America or South America like, you can even see just by latitude where the cities are.
50:34Alixandra Rachman: Yeah. So I wonder if just sort of subgrouping within consonants, you could
50:41Alixandra Rachman: get a stronger signal there. Yeah, yeah, that is one of the reasons for the exhaustion. So we were in Europe. We had the Norway and Greece, and we tried to do found specific analysis, and we did see very large differences between the Norway and Italy, for example.
50:58Alixandra Rachman: Here we might do the same.
51:00Alixandra Rachman: one problem also is in the labels. Why, we say, you know Australia, 3 cities. even in Asia. It's mostly it's the Eastern Asia. So it's
51:11Alixandra Rachman: not very well represented. So that is one of the limitations that we don't have a lot of our representation of some areas to explore these kind of regions. But yes, we could do that.
51:24Thank you.
51:26Alixandra Rachman: Great. I think you were perfect on time. Thank you once about.
51:33Alixandra Rachman: Thank you.