The addition of cannabis to prescription drug monitoring programs and medication fills in Medicaid.

Abstract

To date, there is considerable evidence of the medical applications of cannabis, however concerns regarding the safety of cannabis are also mounting. To improve the safety of cannabis, nine states have added medical cannabis to their state PDMPs, helping providers to take patient cannabis use into consideration when making prescribing decisions. Across a variety of models using Medicaid State Drug Utilization claims data, I find statistically significant reductions in severely and moderately contraindicated medication fills across two outcomes. In my main specification, adding cannabis to a state PDMP is associated with a 14.4% (p < 0.01) and 7.74% (p < 0.001) decrease in the units per prescription, for severely and moderately contraindicated medications, respectively, as compared to states with legal medical cannabis dispensaries open. An interesting spillover effect of adding cannabis to PDMPs is an apparent decrease in the prescribing of scheduled narcotics, with Schedule II medications seeing a moderately significant decrease and Schedule IV medications seeing a 11.4% decrease (p < 0.01) in the prescribing rate and a 16.2% decrease (p < 0.001) in the units per prescription. The main analysis was conducted using the Borusyak et al. (2023) Imputation Estimator with a robustness check using the Callaway and Sant'Anna (2021) difference-in-difference. This work presents evidence that adding cannabis to a state PDMP impacts provider prescribing decisions involving medications that are contraindicated for use with cannabis as well as controlled substances. This paper demonstrates that adding cannabis to PDMPs influences prescribing and thus has the potential to make medical cannabis use safer. Conversely, this work indicates providers may have bias against patients who use cannabis and deny them life-improving medications (like controlled ADHD medication or opioids) on the basis of medical cannabis use. This paper also contributes to the literature comparing DD outcomes estimated using Borusyak, Jaravel, and Spiess' (2023) Imputation Estimator and Callaway and Sant'Anna's (2021) DD Estimator.