Excess Mortality During the COVID-19 Pandemic in Sonora, Mexico, 2015-2022: A Time-Series Analysis of Cause-Specific...

Background/Objectives: Excess mortality (EM) is a key indicator for assessing the population-level impact of large-scale health crises, particularly when cause-of-death ascertainment is incomplete or delayed. However, EM estimates are sensitive to methodological decisions, highlighting the need for operationally feasible approaches suitable for routine public health surveillance. Methods: We conducted time series analyses of routinely collected mortality data from a subnational setting in northern Mexico. Deaths recorded between 2015 and 2022 were grouped into 28 cause-of-death categories. Expected deaths during the COVID-19 pandemic period (March 2020-July 2022) were estimated using negative binomial regression models fitted to pre-pandemic data (2015-2019), incorporating a linear trend and monthly indicators to account for long-term trends and seasonality. Newey-West standard errors were used to address serial correlation and heteroskedasticity. EM was defined as the difference between observed and expected deaths. Results: An estimated 14,482 excess deaths were observed during the pandemic period, corresponding to a 30.9% increase relative to expected mortality. Time-series models identified four distinct peaks of excess mortality coinciding with major pandemic waves. Although COVID-19 accounted for most excess deaths, among non-COVID causes, only ischemic heart disease and diabetes showed statistically significant excess mortality among major non-communicable diseases after baseline adjustment. However, additional categories-including non-transport-related accidents, ill-defined causes, and other endocrine, metabolic, hematological, and immunological diseases-also exhibited statistically significant excess mortality. For several causes, increases in crude mortality did not translate into statistically significant excess mortality. Conclusions: Negative binomial time series regression provides an implementable framework for estimating EM, underscoring the importance of expected mortality estimation for understanding population-level mortality dynamics during health emergencies.