Modelling the future population health consequences of precipitation-induced healthcare disruptions in Malawi

Methods: Here, we provide a modelling framework to investigate these consequences, capturing the initial disruption, individuals’ decisions about whether to re-seek care afterwards, and any delays to that care. By integrating this framework into the Thanzi La Onse model -- an individual-based, multi-disease model of Malawi's whole health system -- we capture these dynamics alongside other healthcare system constraints. We used Latin hypercube sampling (200 draws) and Partial Rank Correlation Coefficients to assess the sensitivity of health-service and health outcomes to five disruption parameters, then projected outcomes under a middle-of-the-road climate future (SSP2-4.5) between 2025 and 2040.

Findings: Under a plausible ("Default") parameter set, precipitation-related disruptions reduced total health service interactions (HSIs) delivered by 6·69% between 2025 and 2040, with nearly 1-in-6 HSIs disrupted in the most-affected months and over 7·4 million people experiencing disrupted healthcare across the period. Only 0·79% of HSIs were directly disrupted; indirect effects, propagated through referrals and care chains, were around 7·5 times larger. Excess health losses were concentrated in time-critical conditions, including neonatal care and acute lower respiratory infections. Across the parameter space, the probability of disruption and the probability of re-seeking care after a disruption dominated all outcomes, identifying patient re-seeking behaviour as the key mediator of health impact.

Interpretation: Even under conservative assumptions — that individuals persist in re-seeking care after disruption — precipitation produces a measurable excess health burden in time-critical care and interrupts healthcare for millions. Because these assumptions, alongside the limited routine coverage of many services, mute impacts elsewhere, our estimates represent a lower bound on the true population health consequences of weather-related disruption. As re-seeking behaviour strongly mediates these impacts, reducing this burden will require both climate adaptation planning and better data on how patients respond when care is interrupted.