Article Text

Download PDFPDF

115 Using national vital statistics system (NVSS) mortality data to estimate county-level trends in drug–specific overdose mortality: considerations and limitations
  1. Holly Hedegaard,
  2. Margaret Warner
  1. National Center for Health Statistics


Statement of Purpose Researchers often use National Vital Statistics System (NVSS) mortality data to examine county-level opioid-involved overdose mortality rates. Limitations such as small counts and county-level variation in reporting the specific drugs involved are not always considered. This study describes county-level death counts and variation in drug reporting, and possible implications when making county-level comparisons.

Methods/Approach NVSS mortality data from 2015–2017 were analyzed. Drug overdose deaths were identified using the International Classification of Diseases, 10th Revision underlying cause-of-death codes X40-X44, X60-X64, X85 and Y10-Y18. Drug specificity was calculated by determining the percent of drug overdose deaths with a multiple cause code of T36-T50.8.

Results Of 3,149 counties, only 1,574 (50%) had 10 or more drug overdose deaths in 3 years, 1,261 (40%) had fewer than 10, and 314 (10%) had zero. Drug specificity was 86% over all. For counties with at least one death, drug specificity was 0–50% for 469 counties (17%), 51–85% for 644 (23%), 86–99% for 730 (26%) and 100% for 992 (35%). Compared to counties with high specificity (>85%), counties with lower specificity were more likely to be rural, located in Public Health Regions 5–9, have county coroners as the state system for medicolegal death investigation, and less likely to have a state medical examiner.

Conclusions Even with 3-year aggregate data, half of the counties had fewer than 10 overdose deaths, and therefore possibly unstable rate estimates. Counties with lower specificity in drug reporting differed from counties with high specificity. These differences might lead to biases when reporting county-level drug-specific overdose rates.

Significance and Contributions to Injury and Violence Prevention Science Low counts and differences in specificity of drug reporting should be considered when comparing county-level drug-specific overdose rates. Aggregation, imputation, modeling and other statistical techniques to account for these limitations may be warranted.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.