Identifying Opioid Overdose Predictors using EHRs Pain and effective pain management are among the most critical health issues facing Americans. In 2011, the Institute of Medicine reported that as many as one-third of all Americans experience persistent pain at an annual cost of as much as $635 billion in medical treatment and lost productivity. Prescription opioids are increasingly used to treat acute and chronic pain. To date, epidemiologic research defining opioid-related adverse drug event (ADE) risk factors has relied on broad, static categorizations of risk derived from diagnostic codes. Though important foundational work, these studies have three important limitations: (1) they focus on only the most catastrophic ADE (overdose) and thus miss the opportunity to identify less severe, prodromal ADEs (e.g. fatigue, dizziness, sleepiness, over-sedation) that may precede and predict overdose; (2) they do not reliably capture aberrant drug-related behaviors (ADRBs)?risky patterns of use that may affect overdose risk; and (3) they rely on clinician- coded diagnoses in structured data, which have notoriously weak sensitivity and specificity, and neglect rich opioid-related information from unstructured clinical narratives. To address this gap, we propose a stepwise approach that leverages the power of electronic health records and new computational methdologies to explore associations among prodromal adverse events, ADRBs, and overdose. This approach is critical to the development of next-generation opioid overdose prevention tools.
'Identifying Opioid Overdose Predictors using EHRs' Death rates among middle-aged adults, the largest demographic group in the U.S., are on the rise for the first time in several decades, largely because of the increase in drug overdose, the majority of which are due to prescription opioids. The overarching goal of this proposal is to strengthen these prevention efforts by gaining a more granular understanding of patient behaviors and clinical factors that increase overdose risk.