Spatiotemporal shifts in respiratory cause-of-death assignment in routine death records in Ghana

Abstract Background Routine death registration provides valuable insight into population-level mortality patterns in many parts of the world, including sub-Saharan Africa, yet the quality of cause-of-death recording and changes in diagnostic labelling over time remain poorly understood in many settings. This study examined temporal, spatial, and labelling shifts in respiratory cause-of-death assignment across 296,093 registered deaths in Ghana from 2015 to 2024, spanning the pre-pandemic, COVID-19 pandemic, and post-pandemic periods. Methods Indicator variables for pneumonia, respiratory failure, acute respiratory distress syndrome (ARDS), and explicit COVID-19 terminology were derived by case-insensitive string matching on a standardised free-text cause-of-death field, extracted from Ghana’s routine civil death register. Logistic regression models with district-clustered standard errors and fixed effects for region and month of death quantified period-associated changes in labelling, expressed as odds ratios (ORs) with 95% confidence intervals (CIs). Spatiotemporal analyses described regional and district-level patterns. Anomaly detection using seasonality-adjusted one-sided exact binomial tests identified regions with statistically elevated counts beyond historical baselines. Comprehensive sensitivity analyses tested robustness to population restriction, construct definition, district composition, and registration lag. Results ARDS and respiratory failure labelling increased markedly and significantly during 2020–2022 relative to 2015–2019 (ARDS: OR 3.02, 95% CI 2.05–4.45; respiratory failure: OR 1.64, 95% CI 1.34–2.00), alongside a smaller increase in the author-derived composite indicator combining all respiratory construct (termed “any-respiratory”) (OR 1.27, 95% CI 1.07–1.51) and a pneumonia association that was weaker and not statistically significant in the primary model (OR 1.28, 95% CI 0.98–1.67). Within the respiratory-cause subgroup, the relative share of pneumonia fell (OR 0.78, 95% CI 0.68–0.89) as ARDS and respiratory failure shares rose, indicating a compositional shift away from pneumonia. The ARDS and respiratory-failure associations were robust across sensitivity analyses, whereas the pneumonia association attenuated toward the null under several specifications. Anomaly detection identified four of sixteen regions (Central, Greater Accra, Ashanti, and Volta) with respiratory-failure counts significantly exceeding historical baselines. Elevated ARDS and respiratory-failure labelling persisted, in attenuated form, into 2023–2024. Conclusions COVID-19 was associated with measurable shifts in respiratory cause-of-death labelling in Ghana’s routine death records, particularly for ARDS and respiratory failure, consistent with a combination of genuine disease burden and changing diagnostic practice, underscoring the need for continuous auditing of routine mortality data in low- and middle-income country settings.