Econometric Analysis - The Association Between State Prescription Drug Monitoring Programs and Fatal Opioid Overdoses
Henry Siegler
June 8, 2023
The United States is currently in the midst of an opioid epidemic, which has emerged as a significant concern over the past few decades. The number of deaths caused by opioid overdoses has been rapidly increasing since 2000. The figure below shows the median opioid overdose rate of the 50 U.S. states and the District of Columbia from 2000 to 2020.
Both the U.S. federal government and state governments have implemented several measures to fight the opioid overdose epidemic. Among the strategies employed today, Prescription Drug Monitoring Programs (PDMPs) stand out as a particularly promising state-level intervention. PDMPs are electronic databases that track controlled substance prescriptions within a state, accessible by health authorities and healthcare providers. PDMPs aim to enhance opioid prescribing practices and reduce the diversion of these substances. The figure below shows the total number of operational PDMPs in effect in each year of this study.
Between 2000 and 2020, a total of 33 states implemented operational PDMPs. The objective of this analysis is to examine the association between establishing an operational PDMP and opioid overdose rates, while also examining how these associations vary for different categories of opioids. Opioid overdose death rates were gathered from the CDC Wonder database, along with data on the year when a state implemented an operational PDMP. It is expected that the overdose death rates for all categories of opioids in a state would decrease after the implementation of an operational PDMP.
Collecting the Data
This analysis uses observational state-year-level data from 2000 to 2020. In this analysis, the four dependent variables used in the various models are the total opioid overdose rate, the heroin and synthetic opioid overdose rate, the prescription opioid overdose rate, and the methadone overdose rate. These data were obtained from the CDC Wonder Multiple Cause of Death database, which included the death counts and populations in each state-year pair for various categories of opioid overdoses.
For each of the four overdose rates, there were missing values, generally for the less populous states in the earlier years of the study. Linear imputation was applied to fill in the missing values individually for each state and overdose rate.
Using data from the Prescription Drug Abuse Policy System, an indicator variable was created to indicate whether an operational PDMP was in place in a given state and year.
Data were also obtained for various covariates in the models. Median income data were obtained for each state-year pair and were converted to thousands of dollars in the model. Using the IPUMS USA Community Survey, data were obtained on the percentage of the state that identified as white, the unemployment rate, the percentage of the state with a bachelor’s degree or higher, and the average age for each state-year pair. Finally, personal health care spending per capita in thousands of dollars was obtained as an additional covariate in the models
Creating the Econometric Models
This analysis uses a panel data model with state-year fixed effects to examine the association between having an operational PDMP in a given state and year and various opioid overdose death rates. This model treats states without an operational PDMP as control states. The panel data regression equations are of the following form:
The dependent variable is the overdose death rate for state s in time t, and there are both state and year-fixed effects. PDMP is a binary variable equal to 1 for state-year pairs with an operational PDMP and 0 otherwise. The other characteristics of a state-year pair are included as covariates in the model.
Results
The estimated coefficients of the panel regression model can be found in the table below. The four columns of the table are for the four different opioid overdose categories that were used, which are labeled in the table. The coefficient of interest is the PDMP in Operation variable.
The PDMP coefficient in the model is -0.795 and is significant at the 10% level. This estimate suggests that states with an operational PDMP are estimated to have 0.795 fewer opioid overdose deaths per 100,000 people on average, compared to states without a PDMP. The average total opioid overdose rate per 100,000 people over all states in this analysis was 9.5 overdoses per 100,000 people, so this estimated reduction is significant.
Similarly, based on the findings, states with an active Prescription Drug Monitoring Program (PDMP) are associated with an average reduction of 0.728 heroin and synthetic opioid overdose deaths per 100,000, as well as 0.329 fewer prescription opioid overdose deaths per 100,000. Both of these estimates are significant at the 10% level. However, the results do not indicate a statistically significant impact of PDMP on methadone overdose deaths.
Conclusions
The results of this study suggest that states with operational PDMPs are expected to experience reductions in opioid-related overdoses, compared to states without a PDMP. This means that for a state like California, with a population of about 39 million, we would have expected there to have been about 310 more opioid overdoses annually on average if California had never established a PDMP.
The results indicate that relative to the average overdose rates for the various categories of opioids, the reductions in overdose rates associated with PDMPs for prescription opioids and illicitly manufactured opioids are very similar.
Creating well-funded and well-functioning PDMPs is unlikely to solve the opioid epidemic, but the evidence of a significant reduction in the opioid overdose death rate after PDMP implementation is economically significant and demonstrates that PDMPs are an effective action to combat this problem.