Estimating the local spatio-temporal distribution of disease from routine health information systems: the case of malaria in rural Madagascar

Abstract

Background: Reliable surveillance systems are essential for identifying disease outbreaks and allocating resources to ensure universal access to diagnostics and treatment for endemic diseases. Yet, most countries with high disease burdens rely entirely on facility-based passive surveillance systems, which miss the vast majority of cases in rural settings with low access to health care. This is especially true for malaria, for which the World Health Organization estimates that routine surveillance detects only 14% of global cases. The goal of this study was to estimate the unobserved burden of malaria missed by routine passive surveillance in a rural district of Madagascar to produce realistic incidence estimates across space and time, less sensitive to heterogeneous health care access. Methods: We use a geographically explicit dataset of the 73,022 malaria cases confirmed at health centers in the Ifanadiana District in Madagascar from 2014 to 2017. Malaria incidence was adjusted to account for underreporting due to stock-outs of rapid diagnostic tests and variable access to healthcare. A benchmark multiplier was combined with a health care utilization index obtained from statistical models of non-malaria patients. Variations to the multiplier and several strategies for pooling neighboring communities together were explored to allow for fine-tuning of the final estimates. Separate analyses were carried out for individuals of all ages and for children under five. Cross-validation criteria were developed based on overall incidence, trends in financial and geographical access to health care, and consistency with geographic distribution in a district-representative cohort. The most plausible sets of estimates were then identified based on these criteria. Results: Passive surveillance was estimated to have missed about 4 in every 5 malaria cases among all individuals and 2 out of every 3 cases among children under five. Adjusted malaria estimates were less biased by differences in populations' financial and geographic access to care. Average adjusted monthly malaria incidence was nearly four times higher during the high transmission season than during the low transmission season. Geographic distribution in the adjusted dataset revealed high transmission clusters in low elevation areas in the northeast and southeast of the district that were stable across seasons and transmission years. Conclusions: Understanding local disease dynamics from routine passive surveillance data can be a key step towards achieving universal access to diagnostics and treatment. Methods presented here could be scaled-up thanks to the increasing availability of e-health disease surveillance platforms for malaria and other diseases across the developing world.

Integrating Health Systems and Science to Respond to COVID-19 in a Model District of Rural Madagascar

“…These quantitative data are triangulated with other information sources: anecdotal reporting and programmatic updates from front line health workers, quantitative data on other indicators of the health system (availability of tests, health system utilization, geography, stock outs, and dynamics of other infections/health services), and analytical methods including mathematical modeling that are all material for assessing COVID-19 epidemiology. Such methods of combining complex data sources at different spatial and temporal scales for projecting dynamics of malaria (“nowcasting”) have been recently employed for Ifanadiana District ( 31 ).…”