132 Tracking the Opioid Epidemic Through the OHDSI Collaborative

Emergency departments in the United States (US) are in the midst of an opioid epidemic (OE), with more than half of overdose deaths in 2014 attributable to opioids. The US consumes 80% of the world’s supply, and contributes to growing concern of an international crisis. A lack of standardization and inadequate drug monitoring complicate the comparison of opioid overdose (OO) rates between countries. This study serves as a proof of concept on single site data, that in the future could compare the rate of OO through sites connected by the Observational Health Data Sciences and Informatics (OHDSI) collaboration. OHDSI is a network of sites sharing patient data standardized to the OMOP Common Data Model from over 600 million patients in 11 countries. This allows for the comparison of OO rates from various sites that participate with OHDSI, potential characterization of the true nature of the OE, and may elucidate whether concerns of a global opioid crisis are founded. By demonstrating the feasibility of these methods, we hope to then characterize the OE through international results. The OHDSI collaborative at a single urban US hospital was queried to establish the ratio of OO compared to annual census. Poisson regression was used to determine significance of a positive trend. A study cohort was generated for patients who had clinical documentation for the 90 days before and after their OO event. Similar criteria were used to create a control cohort for patients with any new condition, excluding OO. Demographics and frequencies of conditions associated with OO and controls were calculated. To maintain patient privacy, only counts were included in results and limited to counts greater than 10. A disproportionality analysis (DA) was performed between OO’s and controls to determine odds ratios (OR) of OO associated with each condition. From 2006 to 2015 there are a total 502 OO visits compared to 9,498,646 total patient visits. Poisson Regression indicates that the increase in the annual rate of OO is statistically significant (p<0.001) with an average rate ratio of 1.09 per year. There are 379 (75.5%) OO patients and 281,474 (2.96%) control patients that meet cohort inclusion criteria. 61.7% of OO are male, while 40.9% of controls are male (X2=66.376, p<0.001). Of OO cases, 75% presented to the ED and 54.8% of these were treated and discharged. Of the 4095 conditions associated with our cohorts, ORs based on DA were calculable for 127, of which 125 were significant. The highest odds of overdose were associated with patients diagnosed with poisoning by analgesic drug, who were 811 (95% CI 470-1399) times the odds to overdose than patients without this condition. Conditions of interest include poisoning by cocaine (OR 225, 95% CI 136-373), gastroparesis (OR 65, 95% CI 38-109) and bipolar disorder (OR 30.77, 95% CI 23-41). Rates of OO and characterization of overdose can be investigated through ODHSI. Single US site data shows overdose is increasing approximately 10% annually. DA has demonstrated prior overdose increases odds of OO; however, additional conditions including gastroparesis and bipolar disorder also demonstrate increased odds of OO. In the future, we will generate similar DA’s for medications and procedures, all of which could be compared internationally. Further studies through ODHSI will aid in the understanding of the US OE, generate potential indicators for OO, and allow for the comparison between the US and international sites to better understand the risk of a global crisis.