Chakrabarty Aditya, Pant Mohan D
Department of Epidemiology, Biostatistics, & Environmental Health, Joint School of Public Health Old Dominion University Norfolk Virginia USA.
Health Sci Rep. 2025 Sep 14;8(9):e71235. doi: 10.1002/hsr2.71235. eCollection 2025 Sep.
Cause-specific mortality (CSM) count prediction plays a vital role in the context of public health policy. In this study, we introduce a new analytical approach, which is divided into three phases to answer specific questions regarding CSM due to 14 specific causes by computing different simple, compound, and conditional probabilities.
A multivariate time series forecasting model was developed using the CDC weekly mortality count data. A binary data matrix was constructed for 14 causes of death (COD) as a function of weeks by combining the observed and forecasted mortalities. We introduced two new concepts: Weekly Exceedance in Mortality Count () and Weekly Change in Mortality Indicator (), which were instrumental in computing various probabilities relating to all the CODs. To test the null hypothesis of no association between the COD and a chi-square test of independence was conducted whereas Cramer's V statistic was used to check the strength of the association. Wilcoxon rank sum test, and correlation indices were used to validate the method.
The results of chi-square test of independence indicated that there was no statistically significant association between COD and ( = 0.79). Furthermore, the effect size of this association between COD and was very small (Cramer's V = 0.055). The results of Wilcoxon rank sum test indicated that there was no statistically significant difference between the observed and forecasted counts ( = 0.11) confirming the consistency of our analytical method. Probabilities associated with were also computed as an illustration of the analytical method.
Utilizing this analytical approach, researchers and policymakers can compute the probabilities of any number of desired events related to different COD which can be helpful for public health interventions, resource allocation, informed decision-making and risk assessment, by controlling the underlying attributes responsible for the probabilities to surge and plummet.
特定病因死亡率（CSM）计数预测在公共卫生政策背景下起着至关重要的作用。在本研究中，我们引入了一种新的分析方法，该方法分为三个阶段，通过计算不同的简单概率、复合概率和条件概率，来回答有关14种特定病因导致的CSM的具体问题。
使用美国疾病控制与预防中心（CDC）的每周死亡率计数数据开发了一个多变量时间序列预测模型。通过结合观察到的和预测的死亡率，构建了一个二元数据矩阵，将其作为周数的函数，用于14种死因（COD）。我们引入了两个新概念：每周死亡率计数超标（）和死亡率指标每周变化（），这有助于计算与所有COD相关的各种概率。为了检验COD与之间无关联的零假设，进行了独立性卡方检验，而克莱默V统计量用于检验关联强度。使用威尔科克森秩和检验和相关指数来验证该方法。
独立性卡方检验结果表明，COD与之间无统计学显著关联（=0.79）。此外，COD与之间这种关联的效应大小非常小（克莱默V=0.055）。威尔科克森秩和检验结果表明，观察到的计数与预测的计数之间无统计学显著差异（=0.11），证实了我们分析方法的一致性。作为分析方法的示例，还计算了与相关的概率。
利用这种分析方法，研究人员和政策制定者可以计算与不同COD相关的任意数量期望事件的概率，通过控制导致概率激增和暴跌的潜在属性，这有助于公共卫生干预、资源分配、明智决策和风险评估。
