College of Pharmacy and Pharmaceutical Sciences Research Day, Spokane, WA, August 9, 2024
Several USTUR faculty members attended the College of Pharmacy and Pharmaceutical Sciences’ (CPPS) Research Day, where USTUR faculty member, Xirui Liu, gave a seminar presentation. Ms. Liu’s research simulates the impact of death certificate misclassification errors on the significance of findings in epidemiological studies. The CPPS Research Day is held annually in August on the WSU Spokane campus. It includes several presentations by CPPS faculty, and provides an opportunity for students to present posters of their research.
Impact of death certificate misclassifications on epidemiological risk models
Xirui Liu (USTUR), Stacey McComish (USTUR), Sergey Y. Tolmachev (USTUR), Joey Zhou (DOE)
Death certificates are commonly used as a primary source of information in epidemiological studies investigating the relationship between radiation exposure and health outcomes. However, death certificates may misclassify the underlying cause of death. At the United States Transuranium and Uranium Registries, these misclassification errors have been observed at an overall rate of 25.5% for a group of 275 individuals with internal deposition of actinide elements. This simulation study aims to evaluate whether there is a statistically significant impact on risk estimates resulting from misclassifications. For the analysis, the logistic regression model was used as the risk model. Dose datasets were generated using a log-normal distribution with predefined geometric mean and geometric standard deviation. Subsequently, outcomes were randomly generated using a predefined odd ratio and baseline prevalence. Over- and under-classification rates varied from 0 to 30% were evaluated to assess the impact of misclassification on the risk estimate results. With a forced odd ratio close to 1, misclassification errors on death certificates can result in statistically significant odds ratios from 48% to 52% of the time. With a forced p-value close to 0.05, misclassification errors on death certificates can result in statistically significant odds ratios from 8% to 45% of the time. [USTUR-0686-24A]