|
Abstract
|
INTRODUCTION: Comparing suicide mortality rates between the U.S. military and U.S. general populations is common in lay and professional literature. Standardization is required for this comparison to account for differences in the population structure, but small event counts complicate the analysis.
METHODS: We demonstrated the performance of direct, indirect, and reverse-direct standardization using U.S. military and U.S. general population suicide mortality data from 2011 to 2018. We also used simulations of direct and indirect standardization in annual comparisons, and over time for the standardized mortality ratio in Poisson regression.
RESULTS: Indirect standardization outperformed direct standardization for annual rate standardization. Direct standardization with combined subgroups can produce a biased estimate. Reverse-direct standardization was unbiased, but it generally yields incorrect interval estimates. Over 2011-2018, the U.S. military suicide mortality data were very consistent with the U.S. general population.
CONCLUSION: Indirect standardization provides more flexibility in rate standardization with rare outcomes. For comparisons between the U.S. military and the U.S. general populations, it can provide valid point and interval estimates of standardized rates and ratios both within a single year and between years without combining categories to account for sparseness.
Language: en |