A Bayesian Approach to Correcting Measurement Error in Estimating Childhood Malnutrition Prevalence fromPooled Demographic and Health Surveys Data

Precise assessment of malnutrition prevalence is essential for successful health policy and the monitoring of Sustainable Development Goals (SDGs). This study seeks to amalgamate repeated cross-sectional Demographic and Health Surveys (DHS) data for Cameroon within a Bayesian frame- work, rectify measurement errors in child stunting indicators, estimate adjusted temporal trends, quantify biases arising from the neglect of errors, and formulate a policy brief framework for Sus- tainable Development Goal (SDG) monitoring. We utilize a Bayesian hierarchical logistic regression model incorporating temporal random effects, applied to aggregated DHS data from Cameroon for the years 2004, 2011, 2018, and 2022. The model explicitly adjusts for misclassification in the bi- nary stunting outcome utilizing validated sensitivity and specificity metrics. Weakly informative priors are established for regression coefficients and variance components. Posterior inference is per- formed using Hamiltonian Monte Carlo. The model’s efficacy is evaluated using extensive simulation simulations that examine bias, mean squared error, and coverage probabilities. The revised model consistently produces higher prevalence estimates (by 3 to 4 percentage points) than uncorrected models across all survey years, demonstrating a systematic underestimating when measurement error is disregarded. Performance indicators indicate considerable enhancements: classification accuracy increased by 4 to 5 percentage points, the Area Under the Curve (AUC) elevated from roughly 0.808 to over 0.86, and precision markedly improved. Simulation analyses validate the model’s resilience in accurately retrieving genuine parameter values under diverse misclassification circumstances. Inte- grating measurement error correction within a Bayesian framework markedly improves the accuracy of child stunting prevalence estimations and trend analysis. This methodology yields more precise ev- idence for health policy development and monitoring of Sustainable Development Goals (SDGs). We advocate for the incorporation of these methodological modifications into national survey reporting systems to enhance data quality and policy legitimacy in resource-constrained environments.