2026-11-11 –, Multifuncional (Remoto)
Introduction. Arterial hypertension is the leading modifiable risk factor for cardiovascular and premature mortality. This condition is multifactorial, with a population distribution that responds to both individual characteristics and contextual environmental conditions that amplify or mitigate its effect.
Objective. To estimate municipal-level mortality rates from hypertensive diseases (HD) in Veracruz from 1990 to 2024, along with projections for 2030, using a hierarchical Bayesian spatio-temporal model for total, premature, and senescent deaths.
Methodology. The methodological strategy began with the cleaning and merging of mortality datasets extracted from the interactive dashboard of the Registered Death Statistics (EDR) from INEGI's open data, along with annual municipal population estimates obtained from CONAPO. Statistical processing consisted of estimating HD mortality rates at the national and state levels, exploratory statistical analysis of the rates, and hierarchical Bayesian spatio-temporal modeling with Knorr-Held interaction at the municipal level for the 1990-2030 period.
Results. In 2024, Veracruz ranked sixth nationally in HD mortality, recording a rate of 38.3 deaths per 100,000 inhabitants, above the national rate. In all five-year periods, the spatial autocorrelation of total mortality was significant and positive (p-value <0.01). For premature death rates, Moran's Index was not significant for the 1990-1994 and 2005-2009 periods. In contrast, spatial autocorrelation was present across all periods for senescent death rates, albeit with varying levels of statistical significance. The values obtained were related to the LISA clusters, since a higher Moran's Index corresponded to a greater number of municipalities with significant associations. A concentration of these clusters was identified primarily in the central region of the state for total, premature, and senescent deaths. Model selection was based on the DIC and WAIC metrics, where lower values indicated a better fit. Accordingly, a hierarchical Bayesian spatio-temporal model with Type IV Knorr-Held interaction was selected for total and senescent mortality. Based on these metrics, the best-fitting model for premature mortality used a Type II Knorr-Held interaction. The Relative Risk (RR) remained below unity until before 2005 across all three mortality types. After that year, an increase was observed, surpassing unity and remaining above it throughout the rest of the period. Premature and senescent rates were distributed differently across time and space. The central region was characterized by high rates in both age groups. The Sotavento region had the highest concentration of elevated premature HD mortality rates throughout the study period. Cotaxtla and Manlio Fabio Altamirano stood out for consistently high premature HD mortality rates across the entire period, with values of 20.2 and 19.3 deaths per 100,000 inhabitants, respectively. Ilamatlán, on the other hand, maintained the lowest rate in this population group across all years. The municipalities with the highest senescent mortality were Ayahualulco (478), Acajete (470), Atzacan (453), Chiconquiaco (416), and Las Vigas de Ramírez (412). Unlike premature mortality, which increased in a relatively uniform pattern, these municipalities did not follow the same trend.
Conclusion. The hierarchical Bayesian spatio-temporal model proved effective in smoothing the evolution of HD mortality rates by integrating space, time, and their interaction. This helped avoid rate distortions caused by municipal population size. The methodological strategy implemented can be replicated by following each of its steps to study other health conditions.
8th-semester student in the Bachelor's Degree in Statistics at the Universidad Veracruzana (UV). I completed my Social Service at the Statistical Research and Consulting Laboratory (LINAE). I have a strong interest in statistical research, through which I have developed and disseminated various projects. I founded the Semillero de Investigación en Estadística y Ciencia de Datos (SIECID) at the School of Statistics and Informatics (FEI) to motivate and support students in joining research projects alongside professors and researchers. I took part in a virtual research internship at UV, where I applied machine learning techniques and wrote an outreach article. I have participated in academic events such as the Foro Internacional de Estadística Aplicada, the Expo Estadística en el Entorno, and the Foro Nacional de Estadística. My areas of interest include applied statistics, geospatial analysis, multivariate techniques, and statistical visualization.