Artificial Intelligence: An Essential Tool in the Global Health Toolbox

Over the past 18 years, I’ve been part of many different types of teams addressing health system challenges. One global health project took me to Mozambique where we focused on strengthening maternal and child health services across the country. 

While visiting a rural health center, I noticed the nurses filling out large, paper health registers (think of a large newspaper sized ledger) with maternal, newborn and child health (MNCH) data that would again be aggregated and sent to the district as a paper report at the end of the week. This was one of many layers of data capture, translate, and aggregate required before it reached decision makers.  I asked the nurses: “Wouldn't it be nice if MNCH data could be entered once electronically so that it could be available at the district and national level immediately and avoid all this paper?” 

One of the nurses with whom I was speaking looked at me quizzically and asked that I follow her.  She led me to a back room where three young people sat in the air-con busily entering HIV data into computers that were connected to 3G routers. The HIV data was being sent to an NGO and directly to a national government server, entirely missing the MNCH and other health input at the facility level.

I asked if the maternal and child health data was also included, but she said that unfortunately it was only for the national donor funded HIV program to track progress against targets. Adding further reporting wasn’t possible, because it wasn't part of the HIV program. 

I was struck by this thought: If the maternal and child health teams could have their patient reports supplied directly into the district system, there would be fewer errors and the data could be aggregated by the system, saving precious lives and time. 

AI can change all of this – breaking down data silos, collecting information from massive amounts of different sources, and even look into the future and accurately predict what’s next – without adding stress to workers’ days or a strain on existing resources. 

Let me explain.

Data has a story only AI can tell

One set of data on its own will not tell the whole story, and creating new data collection systems (for the sole purpose of collecting newer, more ‘perfect’ data) is not the answer. The way forward is to use as much past and present data as possible and predict what will come next. Each set of information is a single perspective and takes many perspectives to understand the entire story. 

Humans are not able to do this at scale on their own because there is simply too much to analyse and ground truth is constantly changing. However, AI, and more specifically machine learning (ML), can gain insight from these vast amounts of data. The more data you feed into the model, the better it becomes at answering questions. This unique ability of AI to take health and other data from the past and present to make accurate predictions of the future is the exciting part of the work at Pendulum.

AI in many other industries has shown itself to be a game changer. From the automotive industry to space exploration, AI is powering these areas into hyperdrive. In the health field, AI and machine learning have the ability to change the way leaders interact and benefit from data.  

Much the same way, before the inventions of the automobile or the internet, people could not imagine how those things could work or how they would impact their lives. Now, both of these innovations are essential parts of our lives. 

AI has the same potential.  

Think of Henry Ford, or Steve Jobs of Apple, visionaries who were able to see what people needed before they could see it themselves  –  that’s what those who are exploring the power of AI are tinkering with. It’s a revolution that can add efficiency to our lives, and the ability to solve vast and complex challenges. 

“If I had asked people what they wanted, they would have said faster horses” - Henry Ford

AI can analyse vast amounts of old and new data from systems such as District Health Information Systems (DHIS2), Electronic Medical Records (EMR), Logistics Management Information Systems (LMIS), publicly available satellite imagery from Maxar, population data from Facebook and news reports from CNN and local media to not only tell us what is next, but also to learn and get better with its predictions as it processes more data. We can use machine learning to get a deep understanding of the health system, and work with our partners to deliver more care using existing resources.

And soon, gone will be the days when project data sits on a computer serving only one project or goal. AI combines data from different sources and provides deep insights while promoting an integrated data system that helps governments understand and improve health outcomes. 

I have always maintained that if there was any chance of making development programs sustainable or resilient, our primary focus needs to be on working with those on the front-line to improve what is already there and make it easy to do. Efforts to improve health outcomes need to be built into existing systems, and donors need to identify (and fund) approaches that fit into local systems as seamlessly as possible. These changes should be transformative, but they also need to be easily adoptable by the final users of the systems.

"Much the same way, before the inventions of the automobile or the internet, people could not imagine how those things could work or how they would impact their lives. Now, both of these innovations are essential parts of our lives."