Prescription Drug Monitoring Programs (PDMPs) are active in all states and the District of Columbia (Prescription Drug Monitoring Program Training and Technical Assistance Center, 2022). These databases capture information about controlled substance dispensations, providing a central location for clinicians and other authorized users to review a patient’s current and past prescriptions to understand how these medications are dispensed. Nearly all states require prescribers to check the PDMP prior to opioid prescriptions, with the goal of ensuring clinicians are aware of all controlled substance prescriptions for a given patient, not just the prescriptions that they write (Prescription Drug Monitoring Program Training and Technical Assistance Center, 2022).

Many state PDMPs alert clinicians to dispensing factors or patterns that may indicate a patient has a higher risk of prescription misuse or overdose. These alerts vary in how they define high-risk, from simple measures like high opioid dosage or overlapping opioid and benzodiazepine prescriptions, to more complicated algorithms used by some PDMP vendors. The latter have been criticized because of the lack of transparency in how the models are developed and the potential for alerts to be misinterpreted by clinicians, affecting the ability of patients to obtain needed medications and in some cases leading to harmful discontinuation of opioid therapy (Szalavitz, 2021). In general, experts have called for better models for predicting adverse outcomes related to opioids and substance use disorder (Volkow and Blanco, 2021, The President’s Commission on Combating Drug Addiction and the Opioid Crisis, 2019, Tseregounis and Henry, 2021).

Pennsylvania (PA) currently uses a relatively simple alert system in its PDMP, alerting prescribers based on high opioid dosage and overlapping opioid/benzodiazepine prescriptions. The Commonwealth is working to optimize how it uses data to address the overdose epidemic and is developing a transparent risk prediction tool using PDMP data. As part of this effort, we collaborated with the Commonwealth to link PDMP data to unintentional overdose death data and de-identified hospital discharge records, overdose-related emergency department visit data identified via syndromic surveillance, and Emergency Medical Services naloxone administration data, to develop and validate an overdose risk prediction algorithm.