Antimalarial Pharmacotherapy Gaps in Nigerian Children Under Five: A Predictive Machine Learning Analysis of Care-Seeking, Testing, and ACT Treatment Using NDHS 2023-24
This study utilizing 2023-24 NDHS data and four machine learning models reveals that household and maternal characteristics poorly predict antimalarial care gaps in Nigerian children under five, though financial barriers consistently emerge as a significant determinant of suboptimal ACT treatment, suggesting cost reduction at the point of sale as a key intervention target.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
In the vast landscape of global health, few challenges are as persistent or as deadly as malaria, a disease that thrives in warm climates and strikes hardest at the most vulnerable. For children under five in Nigeria, the stakes are particularly high, as the country bears the heaviest burden of the disease in the world. To combat this, health organizations have long promoted a three-step strategy known as "Test, Treat, and Track." This approach is simple in theory: when a child develops a fever, a caregiver should seek medical help, a health worker should confirm the diagnosis with a blood test rather than guessing, and if malaria is confirmed, the child should receive a specific, modern medicine called an artemisinin-based combination therapy, or ACT. However, the path from a sick child to the right cure is often broken. A child might be taken to a shop but not tested, or tested but given the wrong medicine. Understanding where these breaks happen is crucial, because fixing a broken step requires knowing exactly which step is failing.
Researchers recently turned their attention to this exact problem, using a massive dataset from the 2023–24 Nigeria Demographic and Health Survey to map the journey of nearly 4,000 children who had a fever. Instead of looking at the whole process as one big block, they split it into two distinct gaps. The first gap, or "access," asked a simple question: did the child get seen by a provider and actually get a blood test? The second gap, or "quality," looked only at those children who had already received some form of medicine, asking whether that medicine was the correct, modern ACT or an older, less effective alternative. To find the answers, the team did not just rely on traditional statistics. They employed four different types of computer modeling, ranging from standard statistical equations to more complex machine learning algorithms, to see if they could predict which children would make it through each stage of the process. They wanted to know if factors like a mother's education, the family's wealth, or where they lived could explain why some children got the full, correct care while others did not.
The results painted a picture of a system struggling to move children through the necessary steps. Only about 16 percent of the febrile children in the study managed to clear the first hurdle; they were taken to a provider and received a diagnostic blood test. This means that for the vast majority, the journey stopped before a diagnosis was even made. Among the children who did receive some form of antimalarial medicine, about 60 percent were given the recommended ACT, while the rest received older or inappropriate drugs. When the researchers tried to use their computer models to predict these outcomes, they found something surprising: the models were not very good at guessing who would get the right care. In fact, none of the eight models they tested performed significantly better than a random guess. The complex machine learning tools, which are often expected to find hidden patterns in data, did not outperform the simpler statistical methods. In some cases, the most advanced algorithm actually performed the worst, a finding that mirrors similar studies done in other countries. This suggests that the standard information collected in household surveys—such as income, education, and location—does not tell the whole story of why children miss out on proper care.
Despite the models' inability to predict the outcomes perfectly, the analysis did uncover a consistent and important clue regarding the second gap: the quality of the medicine received. Across different modeling approaches, a clear pattern emerged linking money to the type of treatment. Caregivers who reported that they found it difficult to afford treatment were significantly less likely to give their child the recommended ACT, even if they had already bought some medicine. This suggests that when families face financial barriers, they are often steered toward cheaper, older alternatives rather than the more expensive, modern drugs. The study indicates that the issue is not necessarily that caregivers do not know which medicine is best, but that the cost at the point of sale is the deciding factor. This is a critical distinction, as it points to price, rather than a lack of knowledge, as the primary driver of inappropriate treatment.
The researchers also noted that the first gap—getting a child tested—remains a massive bottleneck. While the study could not pinpoint exactly why so few children were tested using the available data, it highlighted that the places where most Nigerians first seek care, such as small community pharmacies and local medicine vendors, are rarely equipped to perform blood tests. These vendors are the first stop for more than half of the families, yet they lack the tools to diagnose before they treat. The study concludes that simply telling people to get tested is not enough if the places they visit cannot provide the test. Instead, the path forward likely involves bringing diagnostic tools directly to these community vendors and addressing the cost of the correct medicine. The findings suggest that to save more children, interventions must focus on the specific barriers of cost and the availability of testing at the very first point of contact, rather than assuming that a single solution will fix the entire chain of care.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.