 how we can incorporate population data and other non-routine data into the choice too. So this is just a short 15 minute presentation, but it's a quite important topic because basically all your other analytics more or less relies on getting this right. So much of the analysis we're doing is based on having good denominator data. So this is sort of a small but critical element. So the goal is to talk a bit about why this non-routine data is important to include in the choice too, a bit about how we can work with the non-routine data together and how this can be set up in the best way in the choice too. So just a quick background first, why is it useful to have the non-routine data? I think Vittoria already talked a bit about data triangulation this morning, so looking at multiple data sources. One way of doing that is bringing in the non-routine data sources and looking at it together with the routine data for triangulation. It could be looking at your immunization coverage from the routine data with the immunization coverage from household service for example. Also you can bring in non-serve data like data from campaigns, which is typically handled outside of the routine system, but you can bring in the results to be able to do analytics. So we'll see an example of a vaccination campaign dashboard later. So if we think of the typical health indicator that is monitored routinely, I would say most of them are based on having some sort of population, often population but at least having some sort of denominator that is not collected routinely. So if we think for example of immunization, we have the coverage indicators using the population data as denominator. We have maternal health looking at antenatal care coverage using the expected pregnancies as the denominator. For malaria, you might be looking at the annual blood examination rate with the population at risk as your denominator. So there are all these core indicators from different health areas that rely on having good non-routine denominator data available. So just an example on the right here, we have BCG coverage indicator. For service data, it could be coming from a case-based system, it could be coming from your monthly immunization reports, but in any case, you need to have the under one population as available in DHI to be able to produce the coverage indicators. If you look at this, I don't know if any of you have seen this global reference list of health indicators from WHO. So what the WHO and partners have sort of identified as the 100 most important health indicators, including the SDGs that are health-related. Very small subset of those are actually from routine systems, a lot of it is only available through service. So if you want to use DHI to as an analysis platform where you have all your key health indicators, you would need to bring in not only the denominator data, but also survey results. Another example here, if you think of data quality, one way to assess your data quality is to do triangulation. So in this case, this is from the WHO data quality review framework, which includes an element where you look at your coverage indicators based on routine reporting versus coverage indicators from household surveys. So let's look a bit on how this could be done in DHI's. So what we generally recommend when it comes to population data and denominators is that in DHI's 2 you have one data set which is used across all the health services. So getting all the health programs, departments to agree on one set of core population indicators. This is often like the core indicator population data set is often based on census data that is then used with some projections to get estimates between the censuses. So having that as a common data set in DHI's 2 is a fairly recommended approach. Then there are some specificities here. For example, if you're doing then, if you have a new census, there are new projections. You need to make sure you not only include the future projections, but also ideally update the historical data with the updated projections so that you get, you're able to do proper year over year trends of the population based indicators. In addition to having this core population data set, it is possible and useful in many cases to include alternative denominators. So in some cases different health programs will want to have special population denominators. It could be for example that you would want population estimates based on service data. This was sort of one approach. It's important to differentiate these other population estimates from your core population data set. In countries where you have strong civil registration systems, perhaps you would want to have the census population estimates, but you would also for example want to bring in your numbers from birth registrations. So you have an additional denominator available for live births from the CRBS. So I think the third thing here, which I'll go a bit more into detail on is pulling in population estimates from other sources. That's something, there's been a lot of work in the CHICE platform itself over the last few years on making it easier to use alternative population sources and bringing those directly into the CHICE 2 for analysis. So these are just a few examples from the CHICE 2 map application, which I think is part of the CHICE 2 that is often very sort of underutilized, underestimated, but the map application of the CHICE 2 has actually become a very powerful GIS tool. And among the relatively new features here now is that you're able to pull in from different global population estimate sources into the maps application, alternative population estimates. So for example, there is this initiative called Grid 3. There is the World Population Project that provides population estimates. You can visualize those in the maps application of the CHICE 2. And with the latest version of the CHICE 2, you can also actually import those population estimates as a data set in the CHICE 2 and use them in indicator calculations, etc. Another relatively new feature is that you can actually bring in from Google Earth Engine structures from satellite maps, the individual houses from satellite images and display them within the maps application of the CHICE 2. So if you're sort of going in depth on trying to look at your indicators, sort of geographically, you can zoom in and look at where you actually have the clusters of households related to the health facilities and the indicator data for those health facilities. There is also an application that lets you calculate catchment areas for health facilities. So using this called Crosscut is a company that has developed this which is becoming available also within the CHICE 2. So based on geographical data about the health facilities, location of the health facilities, it's looking at things like travel time for the population to get to the nearest health facility, the terrain, etc. So using the geographical features to actually draw up catchment boundaries around the health facilities. So you start with the location of the health facility. The system will you look at the geography, the roads and everything, and it will figure out the sort of the boundaries, what locations, what households will have this health facility as their nearest health facility. I think we've already talked a bit about this organic profile. There's also a new feature within the maps application that lets you visualize information about the health facilities. There are also some new features, new features coming in the next one or two releases. So one is to be able to identify settlements, so not just the structures but sort of clusters of structures and look again at the travel time from these settlements to the various health facilities that you have geographical coordinates for. Looking at the travel time based on the facility catchments. So from health facility, what is the travel time in different directions, enabling you to visualize that. And the last one is allowing printing of maps from the maps application to use for example if you're doing campaign planning and you prefer to have paper tools for doing that. So that was a bit around the population and the different ways that is integrated in maps in particular. I think as I mentioned initially, a big part of sort of the core health indicators is based on survey data, household surveys. And the aggregate model of DHS2 is quite well suitable for storing that information. So not actually doing the household surveys with DHS2, but once you have the results to bring those into DHS2, enabling you to do analysis to do triangulation with the routine data etc. There are often some things when you're actually going into DHS2 to set this up, some things that you need to keep in mind. One thing is of course that the survey data is perhaps available with every three years, every five years. So the way you set up the data elements indicators needs to take into account for what periods you want this survey data to be available for when you're doing an analysis in DHS2. Another issue is that the household surveys sometimes are not done and aggregated based on the same administrative hierarchy as you would have in DHS2 because of the way the households are sampled etc. You might not have the regions, the districts that you have in DHS2 for the survey results. That was the population of surveys and I think a third sort of major area where it's useful to bring in non-routine data into DHS2 is around campaigns and other sort of non-routine interventions like active case detection, formularia etc. So if you use the campaigns as an example, you can bring in your campaign data if you have immunization campaigns to address immunization gaps. You can bring it into DHS2 and then you can do triangulation with campaign data and the routine data on immunization for example. So there are some sort of best practices you need to keep in mind, for example not adding up the routine and the campaigns because then you sort of lose track of you're not able to follow the trends in the routine immunization etc. If you just lump them together. So keeping the underlying data separate but then you can use indicators written in DHS2 for the cases where you want to actually look at the combination of the campaign and the routine data. Similarly, if you're collecting routinely malaria case data and you're also doing active case surveillance for malaria, keeping those separate so you're not sort of conflating data collecting through different means. So I think at this point I will ask Vittor to just give a quick example of what campaign type dashboard can look like in DHS. Yes super quickly. This is actually a real example. These are retrospective data but nonetheless a real example from a meningitis campaign that happened in Niger and this is what they ended up visualizing for like because they started bringing in all the retrospective data for all the campaigns that have done it. For those who are not aware, Niger belongs to one of those countries is one of those countries that belong to the what it's called the meningitis belt. So they have like every year they have recurrent outbreaks of meningitis and therefore meningitis campaigns are pretty much almost routine there but nonetheless they are campaigns that can happen either a national level or a local level. This one was a national campaign and we just wanted to highlight how important it is to bring all this kind of sources of information together. Why? Because you can start triangulating the campaign data and the campaign results together with you might know for example that when you have either immunization campaigns or it could be a bed net campaign or it could be a mass drug distribution campaign. You might want to do surveys for coverage afterwards. And when you do coverage for coverage you can start triangulating those results with your routine coverage as well or update your coverage until that was there until that moment with the new coverage that you have achieved during the campaign. And just we just wanted to like show you for example the overall coverage you see here normally when you think of a coverage it shouldn't go above the 100% but here we went 105%. So why that could be? It's because of the denominators at the end of the day and once you start checking the curves of distribution here we have for example the doses by the type of site by the age groups and such here you have a map of the different coverages by the different districts as well and you see that the majority of the districts in the end are in orange which means that they have a coverage that is above 100%. And that can happen with like here is an example of immunization campaign but it can happen with any type of campaign. And you see for example here once you start disaggregating by age you start seeing for example that the main problem here was that the age group between one and five years they were more than they were expected. And these are the kind of information that when you bring in in your routine information system and you're in general you start triangulating with your routine data. This is the most important kind of information that you can triangulate why because now we have found out that in these areas where the coverage is more than 100% you chances are your children below five then you have quite a lot of children that were not registered in the national system for example. So this was just like a quick example just to give you an idea of what kind of information you can triangulate most importantly what kind of information you can extrapolate when you start triangulating when you start adding things together. So I'll give you back to you. Thank you. So that was actually the end of this short session. We are already a bit over time but if there are some burning questions specifically to the issue of population and non-routine data and PTSD. Thank you for the presentation Olaf. Just a question from previous session for the death. If the case registered as a death and we have one unified instance for all cases on unique identifier. So can we block the TI from the system to notify the user that this case or this patient is died. Just to avoid the wrong data entry while they entering so we need to keep it inside the system. But notify if it happens by. Maybe something. Can you do it in a rule program rule. Yeah that's why I can't answer from the top of my head whether I'm not sure actually. So that's something we can look at. Because I agree that's an important point because you need to keep the. Keep the enrollment on all the data there but you don't want additional data. Okay so we'll test it. It's not related to your session but in general will be in this workshop a walk through actual work or walk through through the system. Is not just only presentation or actual exercises where we enter data. And use the system because information like this is in the presentation though it's comprehensive but unless we practices or see it it's not going to stick. That's. Yeah so we had a very brief session yesterday. I think I'm looking at your jet we have an exercise later today but I'm not 100% sure how much that will be in the choice I think we'll be using the details dashboards at least to do some analysis there. Yeah, but in general this is not the sort of details to training so it's limited what will actually be doing in the tries to. There are no final questions I guess I'll leave the floor to Victor again.