Do Supply-Side Prescription Drug Policies Encourage Heroin Use? Evidence from Prescription Drug Monitoring Programs
We use state-identified data from the National Survey on Drug Use and Health, a yearly survey representative of non-institutionalized US residents 12 and older, to estimate difference-in-differences models of heroin initiation, comparing the probability of initiating heroin among respondents before and after their state implements a PDMP, and using respondents living in states with no change in their PDMP policy as a comparison group. We limit our sample to respondents who report a history of any NMPR use, and, in some models, with respondents who are at high risk of heroin initiation, either because they meet criteria for NMPR abuse or dependence or they report heavy NMPR use in the prior 12 months. Although we find no evidence of increased heroin initiation among those with lower-risk NMPR use, among high-risk respondents we find that having a PDMP operational in one’s state of residence increases the probability of heroin initiation by a marginally statistically significant 2.5 percentage points (p<.10). This finding is consistent with the hypothesis that the PDMP effect stems from substitution, i.e., increased transitions to heroin use among NMPR users who have increased difficulty obtaining prescription pain medication.
Back-of-the-envelope calculations indicate that, of the 515,000 Americans meeting criteria for heroin abuse or dependence in 2013, approximately 10,000 can be attributed to PDMP enactment. Using published estimates of the cost-of-illness of a single year of heroin dependency, this corresponds to a cost of $520 million in 2013 alone. While marginally statistically significant results are often dismissed, the likelihood that our results stem from power limitations (due to the rarity of heroin initiation and a limited sample of respondents at high risk of heroin initiation), combined with the particularly high cost of Type I error suggests that these findings should be considered by policymakers, and warrant further study.