OSM Checkins for Malaria Eradication – Transcription
So our next speaker is Matt Berg. He’s going to talk to us about OSM and malaria eradication.
A link I can pop up quick. Do I just do it through the browser? Cool.
Sorry, I was going to do it earlier.
No worries. Let me just get into my email quick. Sorry.
We do have the coffee break swish time after this.
I know. That’s strange. Is it a screen thing? Anybody know how to select it? For the screen? Anybody know how to do this? Sure. That’s fine. Hi, everybody. Oops. Hello? Hello? All right. Good almost afternoon, I guess.
It’s great to be here. It’s my first OpenStreetMap. So it’s really exciting to meet kind of the community. I’m going to be talking today about the use of OSM checkins for malaria eradication. I work with a company, Ona, based in Kenya. We work on global health and humanitarian challenges.
So I’m going to be talking today about something called IRS, indoor residual spraying. Basically it’s one of the most effective means of controlling malaria that exists. And it basically means going into somebody’s house, spraying all the walls with DDT. And that basically helps reduce what we call the malaria vector. So basically the pooling of malaria in communities. The idea is if they fly around and touch the walls of the house, hopefully they die. And it’s really effective. It’s very, very expensive. It’s not a tool that can be used everywhere in the world.
So really, if you look at this is the map of kind of rates of malaria across the world, you see kind of less Central Africa. There’s very, very high rates. It’s not a tool to use there. Very expensive to have an effect. So we are working in can southern Africa, Zimbabwe, and the Mekong belt, which is kind of Thailand and Indonesia. Not Indonesia, but the area up there.
So IRS is very expensive. It costs between 30$50 to $90 a home. You have to buy the equipment and manage the teams. It’s a really big ordeal to spend it. And our tools for doing coverage are very, very poor. So we measure kind of success in very aggregate numbers. When I’m talking about IRS, I’m really talking about global health service deliver in general. Whether it’s vaccine coverage or ante natal care. We don’t have a push for targeting and not using data in a realtime way. But the use of OSM data has helped change that. We have been working with a group called Akros in Zambia to develop a tool called AM spray.
The first thing is we want to know where people live. So the initial round, Akros hired a bunch of grad students to do the mapping. Basically using satellite imagery to trace where people live. So the same thing as anybody contributing to OSM. It was much faster than ground enumeration, the traditional approach before. So we developed these coverage maps. What’s so cool about the work we did about 900,000 buildings I think Nate Smith did the import into OSM. We bought into OSM. But at the same time the work of the tableau, Mapbox and the OSM community have been doing a parallel community to map structures in the area we’re working.
We’re working in Zambia and work on vaccine delivery in the ministry of health in the Livingston area which happens to overlap. I was really excited kind when I saw this. The key is, if we know where people live, we can target. Targeting means can we take health data? We take malaria models and figure out where is the optimal place for us to spend our money and our time to do spraying? So we basically group up clusters of homes into what away call spray areas and plan and send out spray teams. So that’s the targeting process.
We have been doing this for four years now, we’re interesting our first season right now. In the first season we actually came up with an approach of just creating buffers around the areas. So you can see that we just I think there were 25-meter buffers around homes. And we would drop a GPS point in that area. And then within the boundary of the community, we would count the number of GPS points that are there. What I forgot to explain is the key thing with IRS, for it to be effective, we have to spray 85% of homes in that had area. If you don’t spray 85 homes go up to 70%, it’s not taking enough antibiotics. It’s a waste of money, basically.
So this approach allowed us to effectively measure overall coverage and see the hot patches of where we were able to go. So that was the first round. But then we said can we go further? And working with a tool called OpenMapKit, hey. This is a tool the Red Cross developed with a bunch of other people. And the next round is can we actually go and link spraying to the actual structure. So that’s sort of the next big step for us.
And I think it’s really, really important for global health to be able to do that. Why it’s important is we can update and create a denominator as we go. This is a simple walk through of the process. We go into a community. All the homes at first are yellow. If we spray a house, it then becomes green. If the house was marked as refused so basically saying, listen, I have to work today. I don’t want to hang around to take all my stuff around so you can spray, which is quite common, it’s red. And we come across structures that aren’t mapped. That aren’t a physical structure for a home. Might be food storage or a chicken coop, other kinds of nonliving places that you would spray. As we’re going, updating the denominator and adding new homes.
We have done this a few rounds and now we have a really, really detailed map of the communities we work in. Which is key. But this is probably the most important aspect about why having a denominator really, really matters. Is it actually helps us ensure that we’re doing the job we’re supposed to do. Development or global health has what I call a main street problem. We walk down the main street, whatever is easy. And then we spray the homes that are easy to see.
So in this case we walk down the street. We spray five of the six homes, the ones in green. Miss one in yellow. We pat ourselves on the back and report to the donor, we hit 86% coverage. But in reality, because we did the OSM mapping, we know that’s not true. And we actually see that we’re actually at 50% coverage. This is a huge issue in global health and vaccine coverage where we work a lot. The idea of fifth child. One out of five children are missing from vaccine coverage because of issues like this. We don’t know where people live. We don’t go the last mile. If you have been to a rural community, you see there’s great disparity within a community of wealth distribution.
So the whole area might be poor, but there’s parts of it that could be even more poor. So my wife has worked in vitamin A campaigns. And she was in a place in Cameroon where they’re doing a postevent coverage survey. Going housetohouse to see who received vitamin A. They found a garbage patch and walked through it and found a whole bunch of kids that hadn’t received this vitamin A dose. This is important.
Getting back to the spray. We create these kinds of maps available on a tablet. Here you can see at the end of the day that they sprayed all the homes. Everything is green, which is what we want. Here is a case where there’s a bunch of homes on the bottom that were missed. So at the end of the day, do mop up. Do planning. We have to go back to that place and finish the job for this to be effective.
Here is a case where on the bottom right there was a bunch of refusals. So we can send a community mobilizer back to say, guys, we need to do the spraying. If you don’t do it, you’re going to increase the risk of malaria in your community. So using this approach we also do these kinds of reports. So at the end of the day, the kind of spray team managers, the field coordinators, can see kind of what’s happening. And in this season, it’s actually going on right now, we were actually creating a bot that pushes these reports out to all the spray teams. And it forces them to kind of say whether they’re going to do a response or not.
So all this data is being pushed to their phones and that’s helping kind of manage the process. But it’s been really cool in that we went from the first year, 7% of areas achieving kind of the 85% spray. So it doesn’t mean we only sprayed 7% of homes. I meant we got to our target in 7%. And last season was over 70%. We also did an independent evaluation from PMI, the presidential malaria initiative, and discovered that our findings our coverage rates were within 1% of what was reported. And the nonM spray sites, which is the the U.S. government spends a lot of money on malaria control. It’s about $250 million a year. And for these kinds of programs. And we’re showing that they potentially aren’t that effective because of data quality issues and coverage issues.
So what I’m showing you has the potential to really improve things. So just shifting gears really quickly, we also work in vaccines. So this is an application we developed for Path and the Ministry of Health in Zambia. It’s the electronic registration in the country. We’re rolling it out. And I’m excited about the potential of this approach for vaccine coverage.
We do coverage rates based on demographics. So we know the population in this area is X. Therefore, we think the number of kids in this communities is X. Which works, but these demographic surveys are done every couple years. They’re mixed surveys and really expensive. What if we could change that what if we selected randomly 30 homes around a health facility every month and sent a health worker to go to their house. Check the household to see if the kids are vaccinated. The next month, and using spatial sampling come up with a method to estimate coverage rates in that community.
What’s happening is health workers at a clinic, she sees a whole bunch of kids. Thing they’re doing their job. But they don’t know if kids are missing. Because they don’t know true coverage rates. So having a map, which you’re helping to build, and having a way of measuring this more effectively using things like spatial sampling could be powerful tools for improving coverage for kids.
This is another, I think, really interesting project. This is from the polio campaigns from Gates are doing. They are working with Oakridge Labs to do remote sensing. And they have mapped millions and millions of buildings in northern Nigeria. They give health workers phones. This is a polio campaign. They’re doing mass distribution of a specific drug. But they actually leave the phone in their pocket and they’re taking a GPS point every two minutes. And what it does is it basically tracks them. And if they’re not going above certain speeds, that means they’re walking, and they walk through these hamlets, these buffers, it will turn that hamlet yellow, which is marked as complete. And at the end of the day it allows them to see if they missed any areas where they’re walking.
So it’s a way of painting a map as you walk with coverage. And using this approach, they’re able to dramatically increase their coverage rates of the hamlets which are like those little pockets of homes, quite dramatically. Which I think is really exciting.
So taking this one step further, you know, we want to try to digitize health, but it’s also a pain. It’s a lot of extra work. So why don’t we look at if we can map communities in a detailed way, we could also start to trace where health workers work. If we give them a GPS logger, a basic smartphone that can do this. And if we start mapping where people live, so we say, hey, this family lives in this house. We could potentially start to do some proxies for health delivery.
So we can say, listen, you know, with training data we know that if you spend around 30 minutes in a house, and there’s a woman of reproductive age there, maybe she got an ante natal visit for her pregnancy. If you’re there for five minutes, that’s a quick check in. We could monitor health checkups and service delivery through geospatial. But none of this is possible without having the base map and knowing where people live.
So, you know, we’re really excited by the idea of, you know, data living on a shared map. So I really enjoyed the talk before. So I see OSM not as much as like a base map with roads and, you know I see it really as a point of interest database. That’s the thing for me for coordinating care. If we can check if we can link different service deliveries to a house, we can link a delivery to a specific community in Somalia. Which is a huge issue because you can’t get UNICEF and WFG agree on a name for a village.
If we click on it in OSM, that’s a common community and we can start to do better things with it. And lastly, we think that using databases or system like OSM can unlock new opportunities for coordinated aid and better service delivery. So here is a scenario. One is, you know, UNICEF could build a school and put in a water point. A different NGO, Save the Children could visit and notice it’s broken and notify them. And through Facebook and all these apps here, we don’t have a lot of the infrastructure in the world. Facebook provides an open approach to developing it so we don’t have to rely Facebook does good things, Google do good things on proprietary databases to look at that. They’re ignoring their areas of the world.
There’s great potential for using OSM for these kinds of services. The work, Missing maps and all the tools coming out dramatically help us find where kids live and need help and the work there. So your work does save lives in a meaningful way. I hope this helps bridge the gap between the mapping and how you can deliver services to the ground. So thank you.
[ Applause ]
Do we have any questions for Matt? That’s great work that you’re doing. Very excited about it. In fact, I have somebody I need to tell you about. That you need to connect with. Any questions?
AUDIENCE: A quick question regarding how long the DDT lasts and how often you have to respray?
Yeah. So lasts somewhere between three to six months. And luckily, that’s sort of the rainy season. There’s dry seasons where the malaria rates drop. So covers most of the transmission period of the malaria season. That’s why it’s quite effective. But you have to do it annually. That’s one of the nice things about this approach is we know exactly how many homes are in these areas. Once you have that thing, it’s going back, the planning is ready to go. And we don’t have to spend a lot of time on preenumeration. Which is key. Yep.
AUDIENCE: So what about the longterm use of the DDT in this area? Is that a concern? Or I understand that it’s very effective in reducing malaria, but the longterm use of DDT is pretty scary as well.
Yeah. I don’t I misspoke, it’s not DDT anymore.
But they’ve shown though, that they’re using DDT in the States again some places. It’s just effective. We used to crop dust with DDT over the crops. If you’re spraying the base of a building where there’s no food and nobody is there, it’s not as big of a deal. So we just misused it heavily in the past. That’s why there’s kind of a bad thing toward there are definitely tradeoffs.
Any other questions? Nope? Oh.
AUDIENCE: Thanks. That was really interesting. Have you talked to any local governments or international organizations that work in the same mace that might have open data they might contribute like roads or other stuff?
Yes. So I mean so this work in Zambia has been coordinated with the geospatial lab at USGAD. And I forgot his name. He was at the HOT Summit. It’s through the malaria control board of every government. They’re very involved in doing that. Unfortunately we usually have to do the mapping of the actual structures. The governments don’t usually have that in place. Zambia has a lot and Zimbabwe has a lot now because of the community’s work.
All right. Great. There is also always this delay on the microphone and I forget about it. Thank you so much for your talk.
[ Applause ]
Yeah, let’s give him another round.
So we have ten minutes before the next session starts. Again, if you look at your program, it’s not connect in there. Because we’re doing the satellite portrait, we are starting the next session at 11:10. So we want to make we get out on to that terrace where the satellite portrait is going to be after the next session. So get ready for the satellite. And we have a real camera, too.