Poster Paper: How Much Do We Really Know about Drug Overdose Deaths? Correlates of Incomplete Toxicology Reporting in the United States, 2010-2016

Saturday, November 9, 2019
Plaza Building: Concourse Level, Plaza Exhibits (Sheraton Denver Downtown)

*Names in bold indicate Presenter

Katherine Tote1, Heather Bradley2, Erika Martin1, Recai Yucel1 and Eli Rosenberg1, (1)State University of New York at Albany, (2)Georgia State University

Background: The drug overdose death rate in the United States more than tripled from 1999 to 2017. Having precise and timely measures on the local regions and populations with the highest rates of opioid-related overdose mortality is critical for targeting resources, adjusting prevention programs, and conducting ongoing monitoring and evaluation. However, standardized toxicology data are not consistently available in existing death certificates, and there are known differences across states and medicolegal death investigation systems. Demographic and geographic disparities in drug-specific overdose deaths may be biased by differential patterns of missing toxicology data. To address this gap, we evaluate the prevalence of missing toxicology data from 2010 to 2016, assess how demographic and geographic characteristics of drug overdose decedents are associated with missingness, and measure changes in the rates of missingness and their associations with decedent characteristics over time.

Methods: We used the Centers for Disease Control and Prevention’s National Vital Statistics System’s cause of death mortality microdata from 1999 to 2016. Our sample includes all accidental, intentional, and undetermined intent drug poisoning deaths in the 50 states and District of Columbia (N=351,345). Drug overdose deaths were classified using the International Classification of Diseases 10th Revision cause of death codes (ICD-10; X40-X44, X60-X64, Y10-Y14). The outcome of interest, toxicology code missingness, is indicated by ICD-10 T-code 50.9 (drug poisoning by other and unspecified drugs, medicaments and biological substances) or by a missing T-code value. A multilevel generalized linear model estimated adjusted prevalence ratios for missing toxicology code information by demographic and geographic characteristics for each year, including interaction terms for each variable and death year. The model accounted for clustering within states and counties.

Results: Overall, 20.3% of deaths were missing toxicology information, ranging from 24.4% in 2010 to 14.6% in 2016. Deaths were more likely to have missing toxicology information if they occurred in non-metro areas (24.9% for very rural counties versus 15.4% for large metro counties) and in counties with coroners (27.8%) versus medical examiners (14.6%). Deaths among women were more likely to have missing toxicology information than those among men (24.0% versus 18.0%, respectively) as were those among non-Hispanic whites compared to Hispanics and non-Hispanic blacks (21.4%, 16.9%, 15.2%, respectively). The percentage of deaths with missing toxicology information declined over time, but these demographic and geographic differences in missingness persisted.

Policy Implications: We found a high degree of missing toxicology reporting, with one-fifth of overdose deaths having missing data. Although it is promising that differences in missingness by demographic and geographic characteristics attenuated over time, these differences persist. As shown repeatedly in the literature, statistical analyses that ignore missing data adversely affect the validity of inferences, and comparisons of drug-specific overdose deaths across time and populations are complicated by changing patterns of missingness. Overall and differential toxicology missingness will likely continue to decline with increasing resources for toxicology assessment for drug overdose deaths, but creative data solutions are needed in the interim to facilitate valid comparisons in drug-specific overdose deaths across populations and time.