Skip to content
AI4M Docs

Data sources

AI4M is built entirely on openly available data. The risk model does not use health-facility case reports.

Used by the model

Section titled “Used by the model”
Source What AI4M takes from it Updated
CHIRPS, Climate Hazards Center, UC Santa Barbara Monthly rainfall for every LGA Monthly
WorldPop Population, for density and for weighting state averages About yearly
The DHS Program Malaria test results from four national surveys (2010, 2015, 2018 and 2021). They train and test the model, and what the latest of them measured around each area (malaria positivity, childhood anaemia, net use, household wealth and electricity) is among its inputs Every few years

Used for transmission dynamics

Section titled “Used for transmission dynamics”
Source What AI4M takes from it
CHIRPS Usual rainfall by calendar month, for the transmission season
The DHS Program Positivity at each survey round, for the survey trend
Bakare et al. (2025), from Nigeria’s National Malaria Data Repository Monthly confirmed malaria cases for Kwara, Nasarawa and Zamfara, 2015 to 2024. They check the season and fit the three-state transmission model
Malaria Atlas Project seasonality records Published monthly records of malaria from Nigerian studies, used to check the season
TerraClimate Monthly temperature, for the three-state transmission model

Used for maps and area codes

Section titled “Used for maps and area codes”
Source What AI4M takes from it
GADM, version 4.1 State and LGA boundaries, and the lga_code identifiers

Shown beside the score

Section titled “Shown beside the score”

These do not feed the model. They appear in the AI4M dashboard and in the context endpoints, to show what surrounds a risk score.

Source Figures
Malaria Atlas Project Bed-net use and access, indoor spraying coverage, effective treatment, travel time to healthcare
WorldPop Population and area
The DHS Program State-level survey figures, such as net use among children under five, shown in the dashboard

What AI4M does not use

Section titled “What AI4M does not use”
  • Routine health-facility data (DHIS2). AI4M does not have access to it. The one public extract, for three states, is used only for transmission dynamics.
  • Temperature, humidity or vegetation. Rainfall is the only climate input. Temperature and humidity were tested and did not help.
  • The Malaria Atlas Project’s malaria maps. They are built from the same surveys the model is tested against, so they are not model inputs.
  • Reported cases or deaths as an input to the risk score.

Using these sources yourself

Section titled “Using these sources yourself”

Each source has its own terms of use and its own preferred citation. If you publish work that uses AI4M data, check and credit the original sources as well as AI4M. See Citing AI4M.