 Tako dobro, Karla. Zelo je bilo občastno, kaj je zelo početno. If you have the feeling that you don't have some kind of data before deciding not to do the study, maybe ask whether this data might be available from repositories from the web. So ask advice to... Zvrʒi? W. W. Because we understood for models, there are many sources of data potrebi se srednjih, ki je za njih prišljavačen, da je danes skupen, in da je za zelo v Europa in njih boljev, zelo to je vsega vsočnje. Tukaj sem izvahelil, da bi se predpraviti, da se zelo vsočnjih zelo vzaljič, in da se vsega zelo vzaljali, da bo nekaj nekaj neko nekaj, da so nekaj neko nekaj neko nekaj neko nekaj, zelo, da ne srečem 1,5 hrba o toga. Zelo, da sem početnja početnja izgleda italijanih data in europeanih data, ker je to, da sem nekaj bolj, zelo, da sem početnja, kaj je v kraju, v kraju, kaj je zelo, da se včetnja data o populaciju, mortalitim, morbiditim, da se početnja. Zelo, da je,robout, ki doprinjamo spotak, da bi bilo pravsi, da se predalemili, in da prišličim in adv производim med ola , ki se je tom, da se daj zelo, kako bi dasem izpraviti. Tako, tudi pa izpodnamo, prevušel poslednjenče, da tepistimo, da na poslutim svojih neizdeši nezakonči in kako objezno kaj je dnešan za ev Juniori, Morbiditi, tk. Vse to počici skupaj tosim. Prvno skupaj, da lemijo doproutovati populčnimi, danes je to vse počit. Či ki tudi tudi je zelo počit in imamo tudi doproutovati, tudi da jasno je tudi o četnjo. Kaj sem pa ne med nekaj terbi počeva, da bo načom počevi, da lahko ne malo počec se iznetno na glasbo, da bi nega nekaj bilo, da bi ne nega nega iznetno na glasbo in kako začak throwsa na glasbo. Tudi böyle izteni, da smo počevil, da v iznetnjičnik zelo na vse. Menešta what will provide us with all the information we need in terms of number of subjects in each exposure group, number of events, and so we can compute easily the relative risks which are the concentrations response functions we will speak about in the next few days, the attributable fraction and then the attributable cases. But in general of course when we do environmental improve in začnega savcha je tudi, da vse humovali na začnega časkega odliča po greba. Koliko najčeža je dohroma rečnega in zelo, nečešno način je izvil segremar, da je nekaj igraj, če jih da si je pačila na začin in začnega rečnega. Časkega je bil počaja, are to concentrated maps to administrative boundaries, which are those that generally we are able to collect from official archives in order to move from the concentration to the actual population exposure. And so this is a very simple example, this is a pollution point sort in Waste incinerator in Turinga. So it's the red dot here, there has been concentration modljnja oga, potem si prišli vse sensujske bljega, v čeži več fina, spesijalne rada, ki je, da prišli način. Začaj nekaj sljedeš, bo je, da však je počet, prišli način z obveženja data. To je izgleda, s nekaj rečen sredan. Zato, da bom počet, drugi počet je, Zelo vsem tako intujive, da nam je vse evacije, kde bo stavila pri nekaj zarenostu, ta zato je polutančref izvama Čiv ili je skonšlično model, ali je nekaj za dve spetivno celje skete. Če zelo prosimčimo, način, nekaj zaštrivamo tega nekaj poseba? Sposte, ki sounds tudi, da smejamo vstalt začin qeljena, pojeljno mupe in spremučnje zapomen. To da so imeli na dve različné vsev. Dovoril svoj med jebrič, ki jamo vsevšanje, da se vsevšanje stajnjo uklavaj, kako se predne nekaj vari prpukod nekaj, bo venomno in se veseveni, kg demesok, zemljajo, tako, kaj jaz ne večje to se ona pravって. Tako kaj če je? Četno je to, ker se bilo, ker sem se dal v odredanje in proprostnji model, da imajo vse zelo vsek, in so se nemožemo izgledati na še spasih spasih. V sej spasih, ki bomo pošli, in izgledamo v nekaj večstvo, bilo, da smo pošli naši spasih spasih. In vsek naši spasih, nemaj, da bi smo glasbi, da so smo pošli pošli vsi spasih. boš malo tudi čebno časno naprej v izbalini, v srečnih, tebi in postovične obany, in tako so ne trebimo začnevaći boljevanje z naparjav, colonizčenih odata, če tako, 4 km začnjamo. Vihli smo nekaj, vaj smo se v komantičnega brana tudi, nekaj, zbijali stranje, v neprodalizu. V primerst do stranji z nedostafilinje, In zelo to je odgaz, da smo pradljili obril prokatom, ki je tako vsi, zakon taj Žne zvom našlič, našlič, OK poč Trija. I so smo na zvom nadjese configured, kretni odlič, to je 4x4 km. Preči je reizma hroje poplusjo, iz četkih ne zelo te te vse zelo zelo poplusjo. Ko je, pravmo način o prezime, is exposed to either red level pollutants and how much is exposed to brown levels. So this is the quite a general question, but of course we also know that the population is not necessarily homogeniously distributed within the municipality. So the second question in order to try to reduce the possible exposure misclassification is whether we can account for built-up areas, which are these brown contours here. So basically from other sources of data, these small polygons here, there are data about for example the built-up areas in terms of high development, low development and anthropogenic settings as compared to natural areas. So how to incorporate this into the exposure attribution. So here the aim is to develop a methodology to obtain a map of administrative area scale so we will keep those units as our reference units and to each one of those we want to give an air pollution estimate. And we will do it in two ways, the most simple approach is just to intersect the census blocks with the grid cells and to attribute the exposure. The second approach would be to overweight the fraction of the territory where there is built-up area because this is where most of the population might reside. OK, so as I said built-up areas can be obtained from several different sources. So basically what we can do is in the let's say the most simple approach top here we can just intersect our fixed grid cells from exposure to our sensors, to our polygons of the administrative scale units and basically we weight each square in terms of the size of the intersection area. So if the square is entirely inside the census block like for example this one it will have the highest weight whereas in this case we will weight this square only for this amount of intersection. And this is quite a simple approach. But the second approach which is more relevant to our purpose would be the one where we relax the assumption, the simplistic assumption that the population is homogeneously distributed. We assume that population is mostly concentrated in these small polygons which are the built-up areas. And so basically what we do in the second approach, the B approach is to basically overweight those part of the intersection where most of the population live. And this is quite a very simple GIS procedure where you can basically go from one fixed grid and bring the exposure map into the administrative scale special units, also possibly weighting different parts according to built-up areas. This has been done in the MEDIS project where basically on this map you can see the mean annual concentration of PM2.5 in 2005. This is from the mini project from the NAA at 4 km grid resolution. And then after this effort, this is a population weighted exposure to PM2.5. And then once we have this, this is just based on the intersections and the use of the Corindlan cover, we can use then an external database about nearly 1,500 municipalities in the Italian survey in order to attribute the population of interest to the map on the left side. So this was a simple approach where we gave priority to the administrative scale units and we basically moved the fixed grid of exposure into the administrative scale. This is the other approach, the opposite, where basically we do the opposite. So we really want to keep the fixed grid as our reference grid and we want to do the reverse. And so basically I will not go too much in detail but just to stress the issue that depending on which is the study objective, in this case this is what was done in the BS project, we really wanted to keep fixed the exposure map because we wanted to minimize the exposure misclassification and we basically moved all the population attributes and the mortality and morbidity data into the fixed grid by doing this other approach. Basically again we intersected the blocks with the squares but then what we did basically was to get a final estimate of population for the grid, not anymore for the polygon for the grid and we weighted this estimate of the population according to the built up areas. So these are two ways of doing basically the same thing. So now I will try to just provide a few information about where we can get data about population, about mortality, about morbidity and this is also where I will also ask you to start thinking about where you can get similar kind of data in your country, in your region for example. So this is an example where most of the population data in Italy are provided by the ISTAT which is the National Institute of Statistics in Italy. So there is this open source website where it is possible for each municipality for each single age year also distinguishing between marital status categories and gender it is possible to get for all years the resident population. So this is an example of how to download this data. So in the main page we have the most recent years but it is also possible to get data about past years from pre-sensal series of the resident population. And there are also life tables that can be used for our purposes. So my suggestion to you would be to investigate to check with maybe your colleagues in the demographic departments or in other institutions whether there is some kind of repository like this. I mean it is possible that many of this data will be publicly available at least to maybe broad details not very refined but it is possible that this data exists also in other places. And this is other examples where we can also go to census blocks which is smaller than the municipality. This is an example about the Eurostat which is a European archive. It is not just Italy. So this is Europe. I understand there are not so many people from Europe, Europe are few. So maybe this is not relevant for many of you but again it is possible that in Africa or Asia or in other places there could be something similar. So the Eurostat is a European archive where it is possible again to get information about the 2011 census and from here you can see there are many different layers of information about population by current activity status by country of citizenship and so on, by age, gender, et cetera. It is possible with a very user friendly interface to select the country and then to download the data in whatever nice format and layout you need. OK, and these are just the example tables of this kind of data. You can see here for example for the Republic we have a distinction by gender and by age groups for the year 2011. OK, just downloaded freely from the Eurostat website. OK, and there are also other pieces of information like unemployment rate, et cetera. OK, so this is when we have broad areas like country level investigation and we really want to attribute and to evaluate the impact of our exposure on a broad level, but it is also possible that we have a smaller scale study like for example residential population cohorts for which it is really necessary to derive the residential addresses. This is when, for example, we want to attribute a health impact assessment of a specific industrial plant which doesn't affect a huge area, only a smaller area, and we want to really be more careful about the exposure assessment of specific individuals living in that area. So there are a number of possibilities here either with open source or with private software to geocode the addresses and to get so x and y of individual addresses. This is an example in the Puglia region where this was done for a number of municipalities. And so you can see all the dots here represent individual residences. OK, again there are possibilities like this, also available in the Eurostat website for Europe, and again the suggestions, you can see very nice maps which are possible to produce and to download for free. And again, depending on which is your aim if you are more into this kind of small scale but very refined impact assessment study then you might check with your colleagues in other departments whether this kind of data might be available as well. OK, yes. So this is just one slide about the difference between short term and long term effect. So the question is whether we need different population estimates if we are more interested into the short term effect of air pollution. So this goes back to yesterday's question whether you are more into the conventional long term impact assessment studies where you really need to have a good characterization of space, both in terms of exposure and in terms of population and health data, versus on the other side you are more interested in investigating whether there is a short term impact coming from daily exposure into your population. The two questions are quite difficult of course, it very depends on your study aim I don't know whether there would be a possibility at some point during this week Karla to kind of bring up this issue again and trying to understand from different people here what is your main interest in terms of health impact assessment. But I mean, if the focus more on the short term effect then the special scale is not so relevant anymore, because when we study short term effect we are much more focused on the day to day variability of air pollution and we assume that this day to day variability is homogeneous across the specific city because it is more driven by let's say meteorological patterns and short term scale patterns which are assumed to be a constant over space within of course small enough areas so in this case we don't have to struggle in trying to find population data at a very refined special scale, it would be more relevant to have some sort of assessment of the temporal variability of the population in hands. Ok, so we spoke about basically about the denominator so the number of the population which live in a specific area. Now we shift a little bit into the numerators which is basically the number of events which occur in terms of either cost-specific mortality or cost-specific morbidity which occur in a specific area by always aged rate of course and this is relevant of course because in order to evaluate the impact of a specific exposure in a population we need to know how many total cases we have how many people live there in terms of information then we can estimate how many of those events can be attributed to that exposure, that's the idea starting from the total number of cases and then by attributing conventional methodologies which will be explained fully this week let's say extrapolate the few or not so few cases which are directly attributable to the exposure in under investigation ok, and so this is basically what I said. So a few slides also and this is questions also for you ok, let's assume we have understood where to collect population data by age, by gender, for whatever special scale unit you might have in your study. The next question if we want to study the impact on mortality would be to try to understand how to get mortality data and of course this mortality data should be on the same scale as the population data because they should be the numerators of those rates and so they have to be on the same scale again in this is for Italy we have this national institute of statistics which provide some information about mortality, by cause, district gender, by age, etc. So these were the causes of that which were selected in the previous impact assessment evaluation that was done in Italy and basically there was a focus on chronic effects and so these causes you can see here all causes lung cancer, myocardial infarction diseases and for the acute effects these were the causes which the researchers focused on and these are just simple maps by province in Italy we have 20 regions and I think 110 provinces with the crude rates that can be derived from this publicly available sources of data so this is for all natural causes, males and females and there are other maps for the other causes as well so of course the idea is to we really need to get this kind of information so these are crude rates which means these are maps which report the ratios between the population data as the denominator and the mortality counts for each province in this case and we have all the different causes so this is also interested because this data for example for Italy also provides information about age structure which allows us to standardize these rates and so to check differences across space between crude and adjusted rates by province and generally this kind of information is at least on a broader special scale like this for example is available in many other countries so again check whether you have this kind of information and if you don't maybe just contact us and we can try to understand whether from other sources which might be non-European maybe for example WHO sources or other kind of global sources whether there are at least some kind of information for countries outside Europe so same issue goes with morbidity rates we might want to investigate the impact not just on mortality but maybe we want to investigate the impact or we want to compute this ability adjusted life years et cetera you will hear about these concepts and this terminology during these days so this is just to say we are not only focusing on mortality because air pollution and climate change might impact not just mortality but also a wider range of outcomes so the question here there are available data on hospitalizations that we might download again for investigating either acute or long term effects of air pollution ok and this is again those outcomes which were evaluated in Italy for the morbidity outcomes ok and it is possible again to derive different maps of rates ok so again how to get morbidity in Italy we have several multi center studies some of those were focused on specific cities like for example the APR or MISA project or specific areas like the Sidria study at the national level we have also the availability the ministry of health is not so easy to get this data but there are data about hospital discharge files at the national level so all the hospitals both public and private by law they have to provide the ministry of health with all the discharge records and in some cases it is possible to get this data at the national level for impact assessment evaluations whereas it is more easy at the regional level because each region is in charge of collecting and monitoring and the surveillance of the population and so to get the hospital discharge files at the regional level so what about health data in other countries here I am just reporting a few slides I am not an expert about this data sources because I have been using mostly Italian data but I know that for example WHO provides information about health status of worldwide population there might be disease registries specific there might be local health information systems in your region, in your county whatever there might be information systems or maybe also there could be a DOC service that you might want to use for a more local or specific investigation and yes this is just a few examples from the World Health Organization archive so again there is the website you can just try to see what is available for your country, you might want to use your specific service ok, so basically I finished so I think I was ahead of time and I would more open this kind of discussion if you have questions of course you can ask me but I would more open the discussion to the audience I think Carla for opening this kind of reasoning for example it could be interesting to share now maybe some problems that you have, it can be common to others guy because what Massimo presented is mainly focused on what we have in Europe so it's probably not the same situation for the majority of the Isabella wanna say something? Ok, just for example could you raise your hand if you or your department has availability of health data at least mortality data in computer based file I mean not how many, mortality so I'm not going to ask for morbidity no, no, I mean just if you work in a department that cope both with environment and health how many? ok, so few all the other comes both from university or place where just the exposure assessment is conducted, I'm correct in say that who wanna speak? ok yes ok, thank you, thank you first for your presentation it was very interesting so we see that this impact assessment of air quality for example is a big issue and it depends on many parameters and I think that there are a lot of approximations so my question is how to validate the results and how much confidence can we give to the when we make the assessment and we say that we have this part of mortality is due to air quality or is due to another effect so how to validate the results and how much confidence can we give to our estimates ok, I mean you are right, there are several assumptions which are made both in the exposure model and in the way it is attributed to population and the way population is distributed over the territory and the mortality and morbidity counts etc there is I think Carla specific session led by Andrea, maybe about uncertainty Andrea there are ways to quantify the uncertainties behind this chain of assumptions and to report them, I mean the best thing we can do is to be sincere and clear about the uncertainties we have because there are several so not just provide 20% is due to more to air pollution and that's it but to provide some range of confidence within which we lie which take into account if I'm not wrong all single aspects of uncertainties which is not just statistical error in the statistical model that's a smaller part but it's more uncertainties in the exposure models and in all the steps along the way so yes it's a good point to be clear about the doubts and the uncertainties we have because as you say when we do this kind of efforts generally it's big wide scale effort and so the more wide you go the less precise you can be in all the steps you are doing but this is the price to pay so yeah there will be a session on that thank you, my question is for example if we use only the survey approach to have an idea about morbidity rates for example for city which is divided on 5 or 6 municipalities is sufficient or not I mean you have a survey and you divide a city into 5 or 6 macro areas this is this question I mean this is a very approximation of course but it is I would say as long as you declare your uncertainties it is better than nothing but of course having so crude approximation you divide so much information in terms of the impact I would say if you have a big city and you divide it into 5 or 6 based on a few points you have from a survey I guess that your uncertainty estimate will be very wide but I mean it's better than nothing for sure I think if I can comment that it's probably a starting point to describe the situation in that area I mean you cannot generalize to all the population but it's still valid for inside your survey the results can be used maybe to decide if it's the case or not to start a more deeper approach let's say in epidemiological evaluation or something like that but it's still valid as a description of what is going on in that city because also conduct a survey is a very tiring and a very good initiative in public health but you have to take into account that you cannot say something not to extrapolate your results to the whole continent or the whole area Any other? Thank you for your talk Actually I have a question about the population data Does these records include any population who doesn't reside in the city for example in my city there are around 3 million people coming to work but we don't have any records about them that they reside in the city but they are the population that is very affected by the air pollution so what do you usually do for this kind of situation? Do you remember Karla in this National Archives there is the difference between resident and present population generally we rely on the resident population because this comes from census data which report the actual residents of the people so there might be this is one of the sources of uncertainty we were speaking about we use it as a proxy of the population who live there but then we know that there is a lot of commuting and a lot of people which spend most of their daily life somewhere else compared to where they reside so in principle we would like to have both resident and present population I would say in practical terms generally it's more easy to get the resident population and not the present one Also because all the health indicators are available for the residents so unless you don't have a very huge availability of health indicators because what about for the people present everyday then come back in another place what about their hospitalization rates their morbidity rates so when you use the residents you are assuming that people stay at home 24 hours which is not true but you can have more advantages but still you have to consider that now I know that with this availability of smartphones you could follow people because all the smartphones have a GPS system so if available they could be a new frontier for this kind of research because you have your personal dosimeter and you can move during the day and Google knows exactly where you are but it's still starting now so let's be a little more traditional any other comment? because it would be nice to listen from your own experience in using this data something please Miriam I'm just sorry, I've just interrupted you two seconds is someone opposed to be filmed or not? because I would be happy to see all the room when asking questions and talking here for the people from the webinar and yourself is someone opposed to have it because maybe it will be on YouTube and you would be on YouTube but like that is someone opposed to that? since you asked about our feedback the best setting that we may get but in countries like I'll be presenting the case of Beirut on Thursday we don't have such privilege and such database to work on it so this is just what I wanted to say and let's say for census it's not easy always to get this information I don't know if in other countries they have this at least I can't record the date but it's too old let's say for Lebanon and for data on mortality we tried hard in BAF study to get data on mortality and it was really difficult and even in the registries we don't have adequate information because when we fill the certificate of death we end up with heart attack or something like that so we don't have the steps that usually we use in Europe so for us it's really very interesting and very advanced but we need to go into small steps to reach such stage I can add that you don't have yet because it's still valuable to think about starting this process will take 10 years to implement and validate and everything which is very quick time you can have... I'm not being pessimistic, we already started in 2010 but it's not really easy to reach this level thank you it was me Yes, I didn't understand the last part of your presentation concerning the mortality data so the utility of this information the utility I think it will be discussed a lot in the next few days the methodology to do the health impact assessment but basically what you need from this data you need to get rates of mortality overall in order to extrapolate the amount which is attributable to pollution the background rates nothing from the... so I think we can thank Massimo thank you so much