Abstract

Estimating and monitoring indicators such as incidence rates of infectious diseases and infant mortality rates are important for the continuous assessment of the health conditions of a population. Such indicators represent essential aspects of the international development agenda related to the United Nations’ Sustainable Development Goals for 2030. However, the underreporting of cases constitutes a barrier to identifying the true magnitude of mortality levels or the burden of a disease, preventing reliable decision-making, especially in less involved regions. In this context, the temporal analysis of health-related indicators can collaborate in development of more efficient public policies aimed at eradicating poverty and promoting good health and well-being. Although there is a growing demand for evaluation of the spatial distribution of official health statistics over time, there is a lack of adequate statistical methods for this purpose in scenarios with evident data misreporting. We extend a hierarchical Bayesian model that allows correction of underreporting of clustered data to a spatiotemporal context. By allowing areas to share information through their spatio-temporal structure, accuracy of disease/mortality incidence estimates can be increased in poor data quality scenarios thus having a great practical appeal. The model was applied to data on early neonatal mortality rates in the microregions of Minas Gerais State, Brazil, and also to model the tuberculosis incidence rate in the Brazilian microrregions, from 2000 to 2022. The statistical software R was used for its computational implementation. From the maps constructed with the corrected mortality/incidence rates, we provide a more realistic distribution of the associated risks and their evolution over time. Estimates of the probabilities of underreporting were also mapped, providing an assessment of advances in the quality of the data collection process. The proposed Bayesian method can be applied in many underdeveloped countries where data tends to be underreported, depending on the availability of adequate prior information. The use of statistical methods that addresses the issue of underreporting in a spatio-temporal perspective is an important tool for the surveillance and monitoring of various epidemiological events, guiding the definition of more reliable intervention and control actions by public agents. We method can play an important role in more reliable monitoring of health aspects of sustainable development, which is a concern on political agendas worldwide. This is a joint work with Rosangela Helena Loschi (UFMG) and Vinícius Lara Fonseca (UFMG).