This study examines the possible clustering in the residual spatial variations of five mortality causes, classified by the tenth ICD, among 644 municipalities of the Brazilian state of São Paulo, after accounting for individuals' demographic characteristics and place of death. Mortality data by cause of death for 2014–2016 were sourced from the Brazilian Sistema de Informações sobre Mortalidade. We applied a Bayesian model that assumes a multinomial distribution for the polytomous outcome, classifying the causes of death into six categories, namely, infectious diseases, neoplasm, respiratory, cardiovascular, external, and other causes, and simultaneously estimates the spatial patterns, accounting for explanatory variables for the first five causes, taking the other causes as reference. The results show that, relative to the causes in the reference group, there is clustering of high relative risks among the municipalities for all the causes of mortality. The relative risks of death from neoplasms and cardiovascular diseases were lower among men relative to women but much higher for external causes. The identified spatial clustering can be useful in identifying appropriate cost-effective strategies suitable for addressing issues of human well-being in the specific municipalities of the state.

本研究在考虑了个人的人口特征和死亡地点后，研究了巴西圣保罗州 644 个城市中由第十个 ICD 分类的五种死亡原因的残余空间变化的可能聚类。2014-2016 年按死因分类的死亡率数据来自巴西 Sistema de Informações sobre Mortalidade。我们应用贝叶斯模型假设多项式结果为多项式分布，将死因分为六类，即传染病、肿瘤、呼吸系统、心血管、外部和其他原因，同时估计空间模式，占前五个原因的解释变量，以其他原因为参考。结果表明，相对于参照组中的原因，各市之间的所有死亡原因都存在较高的相对风险。与女性相比，男性死于肿瘤和心血管疾病的相对风险较低，但外部原因导致的死亡风险要高得多。确定的空间聚类可用于确定适合解决该州特定城市人类福祉问题的适当成本效益策略。