RESUMO
OBJECTIVE: The United States is currently experiencing a public health crisis of opioid overdoses. To determine where resources may be most needed, many public health officials utilize naloxone administration by EMS as an easily-measured surrogate marker for opioid overdoses in a community. Our objective was to evaluate whether naloxone administration by EMS accurately represents EMS calls for opioid overdose. We hypothesize that naloxone administration underestimates opioid overdose. METHODS: We conducted a chart review of suspected overdose patients and any patients administered naloxone in Wake County, North Carolina, from January 2013 to December 2015. Patient care report narratives and other relevant data were extracted from electronic patient care records and the resultant database was analyzed by two EMS physicians. Cases were divided into categories including "known opioid use," "presumed opioid use," "no known opioid," "altered mental status," "cardiac arrest with known opioid use," "cardiac arrest with no known opioid use," or "suspected alcohol intoxication," and then further separated based on whether naloxone was administered. Patient categories were compared by patient demographics and incident year. Using the chart review classification as the gold standard, we calculated the sensitivity and positive predictive value (PPV) of naloxone administration for opioid overdose. RESULTS: A total of 4,758 overdose cases from years 2013-15 were identified. During the same period, 1,351 patients were administered naloxone. Of the 1,431 patients with known or presumed opioid use, 57% (810 patients) received naloxone and 43% (621 patients) did not. The sensitivity of naloxone administration for the identification of patients with known or presumed opioid use was 57% (95% CI: 54%-59%) and the PPV was 60% (95% CI: 57%-63%). CONCLUSION: Among patients receiving care in this large urban EMS system in the United States, the overall sensitivity and positive predictive value for naloxone administration for identifying opioid overdoses was low. Better methods of identifying opioid overdose trends are needed to accurately characterize the burden of opioid overdose within and among communities.