One model links vaccine demand and routes across 3 Kenyan counties

Summary

The case study covers three Kenyan counties and combines predicted vaccination demand with route planning rather than treating the two decisions separately.

The approach was translated into a Microsoft Excel tool and refined through interviews with Amref programme and monitoring staff.

The source describes the project as work in progress and does not yet report a quantified field impact, so its current contribution is a usable decision-support framework rather than a measured service outcome.

Why are mobile-clinic routes difficult to plan?

People living far from fixed health facilities may rely on mobile services for vaccination and other forms of care. Planning those services requires more than finding the shortest route: teams need to estimate where demand is likely to be highest while accounting for travel time, staff availability, fuel, and vaccines that require cold storage.

If demand estimates and routing are handled separately, a route can be efficient on paper without directing capacity toward the locations where it is most likely to be needed.

Why is this a prescriptive analytics problem?

The decision is both predictive and prescriptive. The model first needs an estimate of demand, then must determine which locations to visit and in what sequence under operational constraints.

The team connected those steps directly. Instead of predicting demand first and passing that estimate into a separate routing process, the demand model is embedded in the optimisation so that route choices respond to the same information used to estimate need.

How did the team build the decision-support tool?

The team from the Amsterdam Business School developed the approach in partnership with Amref Health Africa, AstraZeneca and Kenya's Ministry of Health for work in three Kenyan counties.

The model combines machine learning (using factors including population, access to existing health services and willingness to vaccinate) with a route-optimisation formulation designed to balance travel cost with coverage of higher-demand locations.

The resulting system was built into a Microsoft Excel tool. Its design was tested and refined through interviews with Amref staff working on programme innovation and monitoring, so that the interface and assumptions reflected the way mobile services are planned in practice.

What did the analysis show?

While the project is a work in progress, the integration of demand prediction and routing in a tool that can be updated as new field data has become essential to the partners at Amref Health Africa.

From an operational perspective, the approach is designed to direct limited mobile-clinic capacity toward locations with higher predicted need while keeping routes feasible. If validated in field use, that could reduce situations in which staff time, fuel or vaccine capacity are allocated to routes that do not align with demand.

Keeping the tool in Excel also reduces the technical barrier for organisations that already use spreadsheets for operational planning, although the study does not yet quantify adoption or service-level effects.

How can the findings imapct other applications?

The same framework could be tested for other mobile health services where demand varies by location, including HPV vaccination, prenatal and postnatal care and chronic-disease services in areas far from fixed facilities.

Future research could evaluate the tool in live operations, measure how route choices affect reach and travel requirements, and examine how often demand estimates need to be updated as conditions change.