An Evaluation of Death Certificate Coding and Literal Text for Overdose for the Purposes of Improved Injury Data and Poisoning Surveillance in North Carolina

BACKGROUND:  The epidemic of fatal drug overdoses continues to be a growing problem in the United States, and subsequently poisoning surveillance has become critical and essential for injury programs. Death certificate literals, or verbatim written text, could implicate “oxycodone” in an overdose death, yet some of that specific information is lost in the final ICD-10 code reported. Based on discussions with the CSTE Poisoning Surveillance Workgroup, we decided to 1) review death certificate literal text and toxicology reports for North Carolina for overdose deaths for 2010-2011; 2) compare the literal text to their coding outcomes and toxicology results; and 3) assess whether the surveillance system is reporting overdose deaths accurately.

METHODS:   We merged electronic SuperMICAR literal cause of death files with multiple cause of death files from Vital Statistics with electronic toxicology reports from the Office of the Chief Medical Examiner (OCME) and examined 2010 and 2011 resident overdose deaths in North Carolina with a primary cause of death ICD-10 code of X40-X44, X60-X64, X85, or Y10-Y14 and contributing cause codes of T40.1-T40.6, T43.6, and T50.9. Frequency analysis was performed for codes and searches within the literals for certain key words.

RESULTS: Initial analyses suggest only a handful of resident deaths for specific drugs are miscoded. For 2010 and 2011, 95% of heroin overdoses with one drug involved were coded with the correct T-code and mentioned “heroin” at least once in the literals. Findings were similar for 84% of methadone deaths. Forty-one percent of deaths with one drug involved that were coded with the “other opioid” T-code mentioned “oxycodone” in the literal text. Additionally, 61% of cases coded with T50.9 had T50.9 as the only contributing cause code (other unspecified drugs). Of these, oxycodone was mentioned in the literal text for 1.5% of cases with one drug involved, suggesting only a few oxycodone cases may have been miscoded.

CONCLUSIONS:  Literal text alone can shed insight into drugs and combinations of drugs for purposes of sentinel surveillance and exploratory analyses; however, it is unlikely that it would play a significant role in national routine surveillance until specific drugs are written on all death certificates. Utilizing electronic SuperMICAR and OCME files saves time abstracting hard copy files by hand and makes merging data files from various sources easier and less prone to error.