Abstract
Background
Prescription drug monitoring programs (PDMPs) have been widely adopted as a tool to address the prescription opioid epidemic in the United States. PDMP integration and mandatory use policies are 2 approaches states have implemented to increase use of PDMPs by prescribers. While the effectiveness of these approaches is mixed, it is unclear what factors motivated states to implement them. This study examines whether opioid dispensing, adverse health outcomes, or other non–health-related factors motivated implementation of these PDMP approaches.
Methods
Time-to-event analysis was performed using lagged state-year covariates to reflect values from the year prior. Extended Cox regression estimated the association of states’ rates of opioid dispensing, prescription opioid overdose deaths, and neonatal opioid withdrawal syndrome with implementation of PDMP integration and mandatory use policies from 2009 to 2020, controlling for demographic and economic factors, state government and political factors, and prior opioid policies.
Results
In our main model, prior opioid dispensing (HR 2.31, 95% CI 1.17, 4.57), neonatal opioid withdrawal syndrome hospitalizations (HR 1.55, 95% CI 1.09, 2.19), and number of prior opioid policies (HR 2.13, 95% CI 1.13, 4.00) were associated with mandatory use policies. Prior prescription opioid overdose deaths (HR 1.21, 95% CI 1.08, 1.35) were also associated with mandatory use policies in a model that did not include opioid dispensing or neonatal opioid withdrawal syndrome. No study variables were associated with implementation of PDMP integration.
Conclusion
Understanding state-level factors associated with implementing PDMP approaches can provide insights into factors that motivate the adoption of future public health interventions.
Keywords: prescription drug monitoring programs, data integration, opioid overdose, opioid prescribing, neonatal opioid withdrawal syndrome
Introduction
The opioid epidemic in the United States began in the early 1990s with year-over-year increases in prescription opioid dispensing. At its peak in 2012, more than 250 million prescription opioids were dispensed annually.1 During this time, drug overdose deaths from natural and semisynthetic opioids (hereafter referred to as prescription opioid overdose deaths) followed a similar trajectory and saw a 5-fold increase from 2749 in 1999 to 14 495 at their peak in 2017.2 Neonatal opioid withdrawal syndrome (NOWS), a postnatal opioid withdrawal syndrome that can occur in newborns whose mothers were dependent on or treated with opioids during pregnancy, also emerged as a consequence of opioid misuse and significant increases in newborn hospitalizations due to NOWS (from 9084 to 27 085) were observed from 2008 to 2017.3,4
Role of prescription drug monitoring programs to address the opioid epidemic
Prescription drug monitoring programs (PDMPs) are one tool states have implemented to address the opioid epidemic. PDMPs collect data about the prescription and dispensing of opioids and other controlled substances and make this information available to authorized health care providers to identify a patient’s prescription drug history prior to prescribing additional controlled substances, including opioids.5 This information at the point of care has the potential to improve patient safety by alerting prescribers of potential drug interactions or inappropriate doses and to identify people seeking prescription drugs for nonmedical use.6
Efforts to increase use of PDMPs have included improved technology to make it easier to access PDMP information as well as implementing mandatory use policies requiring prescribers to check the PDMP before prescribing a controlled substance or prescription opioid.7 A limiting factor preventing broad use of PDMPs by prescribers was that early PDMPs were only available as standalone websites or portals that required prescribers to log-in to a separate system outside their electronic health record (EHR).8,9 To address this issue, state PDMP administrators and health IT vendors devised systems to integrate PDMP data to be viewed directly within EHR workflows. This innovation has been shown to save prescribers time, reduce the number of mouse clicks to retrieve PDMP information, and increase the number of PDMP searches.10,11
Effectiveness of prescription drug monitoring program integration and mandatory use policies
The effects of PDMPs have been widely studied with mixed results with early studies failing to differentiate whether PDMP use was voluntary or required by states.12–14 Mandatory use policies have been associated with a 3%–10% decrease in the proportion of patients receiving opioids and a 12% decrease in overdose deaths involving prescription opioids and methadone.15–19 However, by making it more difficult for patients to obtain prescription opioids, it has been suggested that PDMPs may contribute to increases in overdose deaths from heroin and synthetic opioids.20–22 Similarly, the effects of PDMP integration on opioid prescribing and overdose deaths have also been mixed. One large health system observed a 5% decrease in opioids prescribed and patients receiving prescriptions from pre- to post-PDMP integration, while another study observed significant reductions in the prescription of high-dose opioids among clinicians with (−75%) and without (−40%) PDMP integration.11,23 Wang observed no effect of PDMP integration on opioid overdose deaths.24 Similarly, no studies have shown effects of PDMP integration or mandatory use policies on NOWS hospitalizations.25,26
Factors associated with implementation of drug and alcohol policies
Given that the current literature assessing the effectiveness of PDMP integration and mandatory use policies is mixed in terms of reducing adverse health outcomes associated with opioid use, it is important to better understand factors that motivated states to implement these approaches. Our literature search did not identify any articles examining state-level factors associated with implementation of PDMP integration or mandatory use policies. One study examined predictors of adopting naloxone access laws and found conservative political ideology and percent of Evangelical Protestants were associated with later adoption.27 However, this study did not find an association between opioid overdose deaths and adoption of naloxone access laws. Other studies examined factors associated with impaired driving and medical marijuana laws.28,29 Neither of these studies found policy adoption to be associated with the health outcome or behavior the policies were intended to address.
Study objective
Based on prior research, it is not clear whether states implemented PDMP integration and mandatory use policies in response to adverse health outcomes associated with opioid use or because of other non–health-related factors. If these PDMP approaches were intended to help address the over-prescription of opioids and identify people at-risk for opioid use disorder, we hypothesized that states with higher annual rates of opioid prescribing, prescription opioid overdose deaths, and NOWS hospitalizations would implement these approaches earlier compared to states that were less affected by the opioid crisis. The goal of this study was to better understand state-level factors associated with earlier or later implementation of PDMP integration and mandatory use policies. While both approaches seek to increase use of PDMPs by prescribers, one approach uses a state’s legal and regulatory authority while the other uses technological advancements to achieve this goal. Understanding state-level characteristics that are associated with implementation of each of these approaches could provide insights into the motivations that facilitate or impede the adoption of future evidence-based public health interventions. The primary objective of this study was to examine whether adverse health outcomes (such as prescription opioid overdose deaths and NOWS), opioid dispensing, or other non–health-related factors (such as political ideology) motivated implementation of PDMP integration and mandatory use policies in the United States from 2009 to 2020.
Materials and methods
Data sources
We performed time-to-event analysis using Cox regression to identify factors associated with earlier implementation of PDMP integration and mandatory use policies. The effective dates for mandatory use policies were published by Horwitz et al.30 The dates states began implementing PDMP integration were collected from a comprehensive review of state websites (including annual reports, press releases, data dashboards, and informational brochures) and, when necessary, state PDMP administrators were contacted directly to confirm their PDMP integration status (see Table SA1 for implementation dates). The Center for Disease Control and Prevention’s (CDC) U.S. Opioid Dispensing Rate Maps,1 the Kaiser Family Foundation (via CDC WONDER),31 and the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project (HCUP) Fast Stats4 were used to obtain state’s rates of opioid dispensing, prescription opioid overdose mortality, and incidence of NOWS hospitalizations. The American Community Survey,32 the Fording State Ideology Dataset,33 and America’s Health Rankings34 were used to obtain sociodemographic factors, political ideology, and public health funding. For our Medicaid Expansion covariate, we used the Kaiser Family Foundation,35 and for our covariates relating to other opioid policies, we used effective dates published by Lee et al.16
Outcome variables
The unit of analysis for all measures included in this study was at the state-year level. We assessed 2 outcome variables: whether a state implemented PDMP integration or mandatory use policies. A mandatory use policy was defined by Horwitz et al. as a policy “that requires a prescriber to check the PDMP before prescribing any category of controlled substance likely to contain prescription opioids.”30 We used effective dates instead of the date the mandatory use policy was passed because these policies can be enacted by states through different mechanisms, including signed legislation, state regulations, and other administrative actions (eg, medical licensing boards). We considered a state to have PDMP integration if the state reported that all authorized PDMP users could implement the technology. We did not consider states to have PDMP integration if it was only available to a limited number of organizations or pilot sites. The effective dates for both approaches were rounded to the nearest year. Delaware was the only state to implement PDMP integration (May 2020) after the onset of the COVID-19 pandemic, and no states implemented mandatory use policies during this time.
Independent variables and covariates
All independent variables and covariates were used at the state-year level, with variables updated annually and lagged to reflect values from the year prior. Our 3 independent variables represent motivational factors we believed could drive a state to implement PDMP integration or mandatory use policies. These variables were rates of opioid dispensing, prescription opioid overdose mortality, and NOWS hospitalizations. These variables were chosen based on what we perceived to have varying levels of motivational influence on policymakers. For example, we hypothesized that opioid dispensing rates would be the least convincing indicator to motivate implementation of these approaches, while drug overdose deaths and NOWS, an adverse health outcome in children, would be more convincing due to their potential to elicit an emotional response from state constituents.
Opioid dispensing was presented as a rate per 100 population. The CDC Opioid Prescribing Maps use a numerator of all new or refill opioid prescriptions dispensed at retail pharmacies. These opioid prescriptions include buprenorphine, codeine, fentanyl, hydrocodone, hydromorphone, methadone, morphine, oxycodone, oxymorphone, propoxyphene, tapentadol, and tramadol. Cough and cold formulations containing opioids as well as specific buprenorphine products used to treat opioid use disorder were excluded.1 Prescription opioid overdose mortality was presented as an age-adjusted rate per 100 000 population. International Classification of Disease, Tenth Revision (ICD-10) was used to identify drug overdose deaths and included unintentional (X40–44), suicide (X60–64), homicide (X85), and undetermined intent (Y10–Y14) causes of death. ICD-10 multiple cause-of-death codes were used to identify natural and semisynthetic opioids (T40.2).31 This class of drugs was chosen because it best reflects deaths associated with commonly prescribed opioids, and excludes heroin, methadone, and other synthetic opioids (such as illicitly manufactured fentanyl). NOWS hospitalizations were presented as a rate per 1000 newborn hospitalizations. HCUP still uses the broader definition of neonatal abstinence syndrome; however, we refer to the more specific term of NOWS in this paper given that the majority of cases occur after exposure to opioids.36–38 HCUP identifies these hospitalizations using relevant ICD-9 and ICD-10 codes indicating a diagnosis of neonatal withdrawal symptoms from maternal use of drugs of addiction.4
Other covariates were chosen based on what we believed could either facilitate or impede implementation of each PDMP approach. These factors included demographic and economic characteristics, state government and political factors, and previous adoption of other opioid policies. For demographic and economic characteristics, we examined states’ population size and the percent of population aged 19-64, non-Hispanic White, unemployed, and in poverty. Each of these variables were chosen based on their association in the literature with opioid-related adverse health outcomes.39,40 For state government and political factors, we examined state government political ideology and public health funding (per capita). State government political ideology was defined using a scale from 1 to 100, with lower numbers representing more conservative states and higher numbers representing more liberal states.33 For opioid policies, we examined total number of opioid-related policies previously adopted from the following list: pain clinic laws, prescription limit laws, Good Samaritan laws, and naloxone access laws. We also examined a separate policy indicator for states that expanded Medicaid, which has been associated with increased access to treatment and reductions in opioid overdose mortality.41,42
Study sample and statistical approach
The study sample for this analysis included 42 states with complete data for all variables of interest from 2009 to 2020. To accommodate time-varying covariates, a time-to-event dataset in counting process format was constructed with state-year observations included up until the year the outcome “event” occurred for each state. For a small number of observations in states with missing data (or data suppressed due to small numbers) between 2 years, values were imputed based on the midpoint. The District of Columbia was excluded due to not having a state political ideology score. North Dakota was excluded due to small number suppression for prescription opioid overdose deaths. Alabama, Alaska, Delaware, Idaho, Mississippi, Montana, and New Hampshire were excluded due to insufficient NOWS data available.
We performed a time-to-event analysis using an extended Cox regression model to identify factors associated with earlier implementation of each PDMP approach. The resulting hazard ratios reflect differences between implementing and non-implementing states at each implementation “event” time, with characteristics of states with earlier implementation dates having a greater effect on the hazard ratios. The “event” for each model was based on the first year the PDMP mandate officially went into effect or the first year the state reported that PDMP integration was available for implementation statewide. Variables were coded as 0 or 1 for each state-year indicating when a state had begun implementing PDMP integration or mandatory use policies in a given year. To account for small sample size, a Firth’s correction produced bias-adjusted estimators. Our main models included all key independent variables (opioid dispensing, prescription opioid overdose deaths, and NOWS) with a vector of time-varying covariates estimating the association with implementation of each approach. All study variables were included in this model because the Akaike information criterion (AIC) continued to decrease after adding each variable during a forward stepwise model selection procedure. Tolerance was used to assess for multicollinearity, and Pearson correlation coefficients were used to assess collinearity between individual study variables (Tables SA2 and SA3). All independent variables and covariates were updated annually and lagged to reflect values from the year prior. Hazard ratios (HR) and 95% confidence intervals (CI) were reported to express the strength of association of implementing each approach with each independent variable.
Descriptive analysis
Line charts were created to plot the number of states that adopted PDMP integration and mandatory use policies and national rates of opioid dispensing, prescription opioid overdose deaths, and NOWS. Bivariate analyses were performed to compare characteristics between implementing and non-implementing states (at any point during the study period) for the year 2009. To test for equality of variances between our adopting and non-adopting samples, we first conducted F-tests. For samples that had equal variances, we conducted pooled 2-sample t-tests, and for samples that had unequal variances, we conducted Welch’s t-tests. Significance tests were performed at the P < .05 and P < .01 levels. Analyses were performed using SAS Version 9.4 (SAS Institute, Cary NC).
Sensitivity analysis
We performed 5 sensitivity analyses. First, we assessed the robustness of our results to adjusting for state time-varying covariates by reporting unadjusted estimates as well as estimates after the addition of the other independent variables, demographic and economic factors, state government and political factors, and prior opioid policies. Second, we performed a forward stepwise model selection approach using the more conservative Schwarz’s Bayesian information criterion (SBC) to assess model fit. Third, we conducted separate regressions for each of the key independent variables (opioid dispensing, prescription opioid overdose deaths, and NOWS) to control for potential collinearity effects of including each of these variables in the same model. Fourth, we performed a separate regression including the age-adjusted rate of overdose deaths for any opioid (as opposed to only deaths involving prescription opioids). Lastly, we performed separate regressions with PDMP integration and mandatory use policies included as potential predictors in our model to assess whether prior adoption of one of these approaches was associated with earlier adoption of the other approach.
Results
Prescription drug monitoring program integration and mandatory use policy implementation and outcome trends
Figure 1 shows the cumulative number of states that implemented PDMP integration and mandatory use policies from 2009 to 2020 (among the 42 states in our sample). In 2013, Kentucky, New Mexico, Tennessee, and West Virginia were the first states to implement mandatory use policies. The cumulative number of states that implemented mandatory use policies appeared to increase steadily through 2020 where 32 states in total had adopted the policy. In 2016, Maryland, Ohio, and South Carolina were the first states to implement PDMP integration. The total number of states that implemented PDMP integration increased rapidly between 2017 and 2020 when 36 states in total had implemented the technology.
Figure 1.
Cumulative number of states implementing PDMP integration and mandatory use policies, United States, 2009–2020. Sample includes 42 states for the years 2009–2020. Alabama, Alaska, Delaware, District of Columbia, Idaho, Mississippi, Montana, New Hampshire, and North Dakota were excluded due to insufficient data available.
Figure 2 shows the rate of prescription opioid dispensing, prescription opioid overdose deaths, and NOWS hospitalizations from 2009 to 2019. The average opioid dispensing rate was 8.3 per 10 population in 2009 and remained flat through 2012 before declining to 4.9 in 2019. The average rate of NOWS hospitalizations was 3.9 per 1000 newborn hospitalizations in 2009 and increased to 9.9 at its peak in 2017 before declining to 8.0 in 2019. The average rate of prescription opioid overdose deaths was 3.6 per 100 000 population in 2009 and increased to 5.4 at its peak in 2016 before declining to 4.0 in 2019.
Figure 2.
Average rate of prescription opioid dispensing, prescription opioid overdose deaths, and neonatal opioid withdrawal syndrome hospitalizations, United States, 2009–2019. Sample includes 42 states for the years 2009–2019. Alabama, Alaska, Delaware, District of Columbia, Idaho, Mississippi, Montana, New Hampshire, and North Dakota were excluded due to insufficient data available. Data for the year 2020 were not included to reflect lagged outcomes for the year prior to PDMP integration and mandatory use policy implementation.
Descriptive results
Table 1 presents bivariate results examining state characteristics in 2009 by implementation of PDMP integration and mandatory use policies (at any time during the study period). On average, states that implemented mandatory use policies had a higher prescription opioid overdose death rate (3.9 vs 2.6 per 100 000 population; P = .04), higher rate of NOWS hospitalizations (3.7 vs 1.5 per 1000 newborn hospitalizations; P = .006), larger population size (1.6 million vs 800 thousand; P = .014), smaller White population (69.4% vs 81.4%; P = .001), and more previously adopted opioid policies (0.2 vs 0.0; P = .02) compared to states that never implemented mandatory use policies. States that implemented PDMP integration also had a larger population size (1.5 million vs 500 thousand; P = .002) compared to states that never implemented the technology. There were no other statistically significant differences between states that implemented PDMP integration and those that never implemented among the state-level factors assessed.
Table 1.
Bivariate analysis of state characteristics by implementation of PDMP integration and mandatory use policies, United States, 2009.
| Mandatory use policy |
PDMP integration |
|||
|---|---|---|---|---|
| Ever implemented | Never implemented | Ever implemented | Never implemented | |
| Study variables | Mean (SD) | Mean (SD) | Mean (SD) | Mean (SD) |
| Opioid dispensing and health outcomes | ||||
| Opioid dispensing rate (per 100 people) |
|
|
|
|
| Prescription opioid overdose death rate (per 100 000 people) |
|
|
|
|
| Any opioid overdose death rate (per 100 000 people) |
|
|
|
|
| Neonatal opioid withdrawal syndrome rate (per 1000 newborn hospitalizations) |
|
|
|
|
| Demographic and economic factors | ||||
| Population size (in millions) |
|
|
|
|
| Age 19-64 (%) |
|
|
|
|
| White (%) |
|
|
|
|
| Unemployment (%) |
|
|
|
|
| Poverty (%) |
|
|
|
|
| State government and political factors | ||||
| Public health funding ($ per person) |
|
|
|
|
| State government political ideology (1 = Conservative; 100 = Liberal) |
|
|
|
|
| Opioid policies | ||||
| Medicaid expansion |
|
|
|
|
| Number of opioid policies adopted (1-4) |
|
|
|
|
| Number of states | 32 | 10 | 36 | 6 |
Analysis uses 2009 data for 42 states. The “Ever implemented” and “Never implemented” groups were determined based on whether a state implemented these PDMP approaches between 2009–2020. Alabama, Alaska, Delaware, District of Columbia, Idaho, Mississippi, Montana, New Hampshire, and North Dakota were excluded due to insufficient data available.
P < .05,
P < .01.
Factors associated with implementation of prescription drug monitoring program integration and mandatory use policies
Table 2 presents HR estimates for the association of opioid dispensing, prescription opioid overdose deaths, and NOWS hospitalizations with implementation of PDMP integration and mandatory use policies controlling for each of the independent variables as well as demographic and economic factors, state government and political factors, and other opioid policies. For implementation of mandatory use policies, opioid dispensing (HR 2.31, 95% CI 1.17, 4.57), NOWS hospitalizations (HR 1.55, 95% CI 1.09, 2.19), and number of previously adopted opioid policies (HR 2.13, 95% CI 1.13, 4.00) were associated with increased likelihood of implementing a mandatory use policy (P < .05). Prescription opioid overdose deaths were not associated with implementation of mandatory use policies (HR 1.02, 95% CI 0.87, 1.20). There were no statistically significant state-level factors associated with implementation of PDMP integration (P < .05).
Table 2.
Association of year prior opioid dispensing, prescription opioid overdose deaths, and neonatal opioid withdrawal syndrome with implementation of PDMP integration and mandatory use policies, United States, 2009–2020.
| Study variables | Mandatory use policy | PDMP integration |
|---|---|---|
| HR (95% CI) | HR (95% CI) | |
| Opioid dispensing and health outcomes | ||
| Opioid dispensing rate (20 per 100 people) |
|
|
| Prescription opioid overdose death rate (1 per 100 000 people) |
|
|
| Neonatal opioid withdrawal syndrome rate (4 per 1000 newborn hospitalizations) |
|
|
| Demographic and economic factors | ||
| State population (per 5 million people) |
|
|
| Age 19-64 (1% change) |
|
|
| White (15% change) |
|
|
| Unemployment (2% change) |
|
|
| Poverty (3% change) |
|
|
| State government and political factors | ||
| Public health funding ($30 per person) |
|
|
|
|
|
| Opioid policies | ||
| Medicaid expansion (1 or 0) |
|
|
| Number of opioid policies adopted (1-4) |
|
|
| Number of states | 42 | 42 |
| Number of events | 32 | 36 |
| Number of state-year observations | 394 | 442 |
Hazard ratios were estimated using extended Cox regression. Regressions included yearly observations up until the year the event outcome occurred for each state from 2009 to 2020. Variables were scaled to approximately 1 standard deviation of the mean during the pre-adoption period. Regressions control for state population, percent of population aged 19-64, percent White, percent unemployed, percent in poverty, public health funding, state government political ideology, Medicaid Expansion, and number of opioid policies previously adopted.
P < .05,
P < .01.
Sensitivity analysis
Figures SA1 and SA2 show the robustness of our results to adjusting for covariates. For mandatory use policies, HR estimates were largely consistent for prescription opioid overdose deaths and NOWS hospitalizations after adjusting for covariates. The unadjusted HR for opioid dispensing was 1.43 (95% CI 0.95, 2.14), this estimate grew as more covariates were added and was 2.31 (95% CI 1.17, 4.57) in the full model. Due to concerns about whether overfitting was inflating this estimate, Tables SA4 and SA5 present HR estimates from models selected using the more conservative SBC criterion. Using this criterion, the covariates for state unemployment, Medicaid Expansion, and political ideology were removed from the mandatory use policy model, and the HR for opioid dispensing was 1.96 (95% CI 1.10, 3.50). For PDMP integration, HR estimates were consistent across all model specifications and no statistically significant associations were observed (P < .05).
We also assessed collinearity between individual study variables. Table SA3 presents Pearson correlation coefficients among study variables for the year 2009 and shows a statistically significant correlation between opioid dispensing and prescription opioid overdose deaths (0.45, P < .01). This may have contributed to our null findings for the association of prescription opioid overdose deaths with implementation of mandatory use policies. Therefore, Figure 3 presents results when performing separate regressions for each independent variable (opioid dispensing, prescription opioid overdose deaths, and NOWS). In these models, opioid dispensing (HR 2.35, 95% CI 1.33, 4.14), prescription opioid overdose deaths (HR 1.21, 95% CI 1.08, 1.35), and NOWS hospitalizations (HR 1.49, 95% CI 1.16, 1.91) were significantly associated with implementation of mandatory use policies (P < .01). No study variables were associated with PDMP integration when performing separate regressions for each key independent variable (P < .05).
Figure 3.
Association of year prior opioid dispensing, prescription opioid overdose deaths, and neonatal opioid withdrawal syndrome with implementation of PDMP integration and mandatory use policies in separate models for each independent variable, United States, 2009–2020. *P <.05, **P <.01. The sample included 394 state-year observations for the mandatory use policy models and 442 state-year observations for the PDMP integration models. Hazard ratios were estimated using extended Cox regression for 42 states. Yearly observations were included up until the year the event outcome occurred for each state from 2009 to 2020. Separate regressions were conducted for each key independent variable. Variables were scaled to approximately 1 standard deviation of the mean during the pre-adoption period. Regressions control for state population, percent of population aged 19-64, percent White, percent unemployed, percent in poverty, public health funding, state government political ideology, Medicaid Expansion, and number of opioid policies previously adopted.
Table SA6 presents results from our full model with age-adjusted rate of overdose deaths involving any opioid. These results were consistent with those from the model including prescription opioid overdose deaths. Lastly, Table SA7 shows that prior adoption of PDMP integration or mandatory use policies was not associated with implementation of the other approach.
Discussion
This study provides insights into state-level factors associated with implementation of 2 approaches aimed at increasing use of PDMPs by prescribers: PDMP integration and mandatory use policies. In our main model, prior opioid dispensing, NOWS hospitalizations, and previously adopted opioid policies were associated with implementation of mandatory use policies. When performing separate regressions for each key independent variable, prescription opioid overdose deaths were also significantly associated with implementation of mandatory use policies. However, no state-level factors were observed to be associated with PDMP integration.
Our findings suggest that opioid dispensing and adverse health outcomes (prescription opioid overdose deaths and NOWS hospitalizations) may have motivated policymakers to adopt mandatory use PDMP policies to address the opioid epidemic. However, these findings differed from previous studies that examined state-level factors associated with adoption of other drug and alcohol-related policies.27–29 For example, Bohler et al. did not find opioid overdose death rates to be a predictor of states adopting naloxone access laws, but instead found political and religious ideology to be significant factors.27
We did not find demographic or political factors to be associated with implementation of mandatory use policies. The differences in these findings from Bohler et al. may be due to naloxone access laws being considered a harm reduction strategy while mandatory use policies are viewed as a policy limiting or securing the supply of prescription drugs. Previous research has suggested that states with conservative political ideology and a larger Evangelical Protestant population tend to have delayed adoption of harm reduction policies.43,44 Our finding that states with more previously adopted opioid-related policies are more likely to implement mandatory use policies is consistent with Macinko and Silver’s policy appetite hypothesis that states with more previously adopted impaired driving laws were more likely to implement subsequent related policies.29
We did not find opioid dispensing or opioid-related adverse health outcomes to be significantly associated with PDMP integration. This may be because PDMP integration requires significant upfront costs associated with acquiring and installing this new technology. We included a state public health funding variable in our model to assess whether a state’s willingness or ability to fund public health programs also applied to adoption of new technologies. However, our models did not find this variable to be a significant factor associated with PDMP integration. One possible explanation could be that funding for PDMPs might come from sources that are not typically considered “public health.” For example, the public health funding variable was defined as state dollars dedicated to public health or federal dollars directed by the CDC and the Health Resources & Services Administration.34 This variable did not include funding from the U.S. Department of Justice, which has funded the Harold Rogers PDMP Grant Program since 2008, as well as other PDMP integration pilot projects funded by SAMHSA and the Office of the National Coordinator for Health IT, which could have facilitated implementation of PDMP integration.45,46
This study found that states disproportionately affected by adverse opioid-related health outcomes were more likely to implement mandatory use PDMP policies at any given time than other states that had yet to implement these policies. While policymakers must consider a variety of factors when making policy and budgetary decisions, these findings highlight the important role of data collected by public health surveillance systems in informing these decisions. Public health professionals should continue to engage with policymakers to educate them about public health issues affecting their state and to identify areas where surveillance systems can be improved to facilitate the use of public health data to inform evidence-based policies and interventions.
Limitations
This study had several limitations. First, we examined implementation of PDMP integration and mandatory use policies, in general, as approaches that seek to increase use of PDMPs by prescribers. We did not account for specific variations of these approaches, such as the required frequency a prescriber must check the PDMP prior to prescribing an opioid (eg, initial prescription only vs each prescription) or the specific government entity hosting the PDMP (eg, public health department vs law enforcement agency). Second, this study was limited to 42 states due to data availability for all study variables. This could have impacted statistical power and our model fit. Third, our sensitivity analysis identified significant collinearity between our study variables, particularly between opioid dispensing and prescription opioid overdose deaths. In our full model, this resulted in prescription opioid overdose deaths not having a statistically significant association with implementation of mandatory use policies. However, when separate models were conducted for each independent variable, prescription opioid overdose deaths were significantly associated with implementation of mandatory use policies. Sensitivity analyses also showed increasing effect size for the association of prescription opioid dispensing with mandatory use policies as more covariates were added to the model, an indication of potential concerns about overfitting or confounding.
Lastly, we used binary indicators in a Cox regression model to reflect the date when states began implementing statewide PDMP integration. We did not assess the pace this new technology was adopted by health care facilities over time, which likely varied from state-to-state. There may be additional factors that could have impacted implementation of PDMP integration that we were unable to measure. For example, a 2021 national survey of office-based physicians found significant variation in use of PDMP integration by EHR developer, ranging from 4% to 56%.47 Higher market share of specific EHR developers or presence of a health information exchange may make it easier for states to implement PDMP integration. Hospital-specific factors could also influence a state’s ability to implement PDMP integration. For example, Holmgren and Apathy found teaching and system-affiliated hospitals were more likely to have PDMP integration capabilities compared to their counterparts.48 These findings could suggest that states with fewer well-resourced hospitals may be less likely to implement PDMP integration unless they receive external sources of funding. If large health systems were among the early adopters of PDMP integration prior to the technology becoming available to prescribers statewide (as seen with Ochsner Health System in Louisiana), this could bias our results towards the null.49,50 The perceived benefits and usability of PDMPs by prescribers and state policymakers could also influence whether these approaches were implemented.
Conclusion
This is the first study to examine state-level factors associated with implementation of PDMP integration and mandatory use policies. Understanding factors associated with implementing each of these approaches can provide insights into the motivations for adopting future public health interventions. Our findings show that opioid dispensing, prescription opioid overdose deaths, and NOWS hospitalizations were associated with implementation of mandatory use policies, but not with PDMP integration. Public health professionals should continue to educate policymakers on public health issues affecting their state. This includes highlighting the important role of public health surveillance systems to inform the development of evidence-based policies and interventions.
Supplementary Material
Contributor Information
Christian E Johnson, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA 52242, United States.
Elizabeth A Chrischilles, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA 52242, United States.
Stephan Arndt, Department of Psychiatry, Carver College of Medicine, University of Iowa, Iowa City, IA 52242, United States; Department of Biostatistics, College of Public Health, University of Iowa, Iowa City, IA 52242, United States.
Ryan M Carnahan, Department of Epidemiology, College of Public Health, University of Iowa, Iowa City, IA 52242, United States.
Author contributions
Christian E. Johnson conceptualized and designed the study, carried out the statistical analysis, and drafted the initial manuscript. Elizabeth A. Chrischilles, Stephan Arndt, and Ryan M. Carnahan conceptualized the study, contributed to data interpretation, and critically reviewed and revised the manuscript. All authors contributed to and have approved the final manuscript.
Supplementary material
Supplementary material is available at Journal of the American Medical Informatics Association online.
Funding
This research received no specific grant funding from any agency in the public, commercial, or not-for-profit sectors.
Conflict of interest
The authors have no competing interests to declare.
Data availability
The datasets were derived from sources in the public domain: Centers for Disease Control and Prevention U.S. Opioid Dispensing Rate Maps (https://www.cdc.gov/drugoverdose/rxrate-maps/index.html), Kaiser Family Foundation Opioid Overdose Death Rates [https://www.kff.org/other/state-indicator/opioid-overdose-death-rates/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Opioid%20Overdose%20Death%20Rate%20(Age-Adjusted)%22,%22sort%22:%22desc%22%7D], Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project Fast Stats (https://datatools.ahrq.gov/hcup-fast-stats), U.S. Census Bureau American Community Survey (https://www.census.gov/programs-surveys/acs), Richard C. Fording State Ideology Data (https://rcfording.wordpress.com/state-ideology-data/), United Health Foundation American Health Rankings (https://www.americashealthrankings.org/explore/annual/measure/PH_funding/state/ALL), and Kaiser Family Foundation Status of State Medicaid Expansion (https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions-interactive-map/). The effective dates for PDMP integration and mandatory use policies are available in this article’s online supplementary material.
References
- 1. Centers for Disease Control and Prevention. U.S. opioid dispensing rate maps. Accessed August 27, 2022. https://www.cdc.gov/drugoverdose/rxrate-maps/index.html
- 2. Hedegaard H, Miniño AM, Warner M. Drug Overdose Deaths in the United States, 1999–2019. NCHS Data Brief, No. 394. National Center for Health Statistics; 2020. [PubMed] [Google Scholar]
- 3. McQueen K, Murphy-Oikonen J. Neonatal abstinence syndrome. N Engl J Med. 2016;375(25):2468-2479. [DOI] [PubMed] [Google Scholar]
- 4. Agency for Healthcare Research and Quality. Healthcare Cost and Utilization Project fast stats: neonatal abstinence syndrome among newborn hospitalizations. Accessed August 27, 2022. https://datatools.ahrq.gov/hcup-fast-stats
- 5. Centers for Disease Control and Prevention. Prescription drug monitoring programs. Accessed February 12, 2021. https://www.cdc.gov/opioids/providers/pdmps.html
- 6. Office of the National Coordinator for Health Information Technology. Health IT playbook: opioid epidemic and health IT. December 18, 2019. Accessed May 22, 2022. https://www.healthit.gov/playbook/opioid-epidemic-and-health-it/
- 7. Strickler GK, Zhang K, Halpin JF, et al. Effects of mandatory prescription drug monitoring program (PDMP) use laws on prescriber registration and use and on risky prescribing. Drug Alcohol Depend. 2019;199:1-9. [DOI] [PubMed] [Google Scholar]
- 8. Rutkow L, Turner L, Lucas E, et al. Most primary care physicians are aware of prescription drug monitoring programs, but many find the data difficult to access. Health Aff (Millwood). 2015;34(3):484-492. [DOI] [PubMed] [Google Scholar]
- 9. Hildebran C, Cohen DJ, Irvine JM, et al. How clinicians use prescription drug monitoring programs: a qualitative inquiry. Pain Med. 2014;15(7):1179-1186. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.United States Government Accountability Office. Prescription drug monitoring programs: views on usefulness and challenges of programs. October 2020. Accessed March 7, 2023. https://www.gao.gov/products/gao-21-22
- 11. Weiner SG, Kobayashi K, Reynolds J, et al. Opioid prescribing after implementation of single click access to a state prescription drug monitoring program database in a health system’s electronic health record. Pain Med. 2021;22(10):2218-2223. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Finley EP, Garcia A, Rosen K, et al. Evaluating the impact of prescription drug monitoring program implementation: a scoping review. BMC Health Serv Res. 2017;17(1):420. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Fink DS, Schleimer JP, Sarvet A, et al. Association between prescription drug monitoring programs and nonfatal and fatal drug overdoses: a systematic review. Ann Intern Med. 2018;168(11):783-790. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Wilson MN, Hayden JA, Rhodes E, et al. Effectiveness of prescription monitoring programs in reducing opioid prescribing, dispensing, and use outcomes: a systematic review. J Pain. 2019;20(12):1383-1393. [DOI] [PubMed] [Google Scholar]
- 15. Puac-Polanco V, Chihuri S, Fink DS, et al. Prescription drug monitoring programs and prescription opioid-related outcomes in the United States. Epidemiol Rev. 2020;42(1):134-153. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lee B, Zhao W, Yang KC, et al. Systematic evaluation of state policy interventions targeting the U.S. opioid epidemic, 2007-2018. JAMA Netw Open. 2021;4(2):e2036687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Bao Y, Wen K, Johnson P, et al. Assessing the impact of state policies for prescription drug monitoring programs on high-risk opioid prescriptions. Health Aff (Millwood). 2018;37(10):1596-1604. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Dowell D, Zhang K, Noonan RK, et al. Mandatory provider review and pain clinic laws reduce the amounts of opioids prescribed and overdose death rates. Health Aff (Millwood). 2016;35(10):1876-1883. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Wen H, Schackman BR, Aden B, et al. States with prescription drug monitoring mandates saw a reduction in opioids prescribed to Medicaid enrollees. Health Aff (Millwood). 2017;36(4):733-741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Kim B. Must-access prescription drug monitoring programs and the opioid overdose epidemic: the unintended consequences. J Health Econ. 2021;75:102408. [DOI] [PubMed] [Google Scholar]
- 21. Compton WM, Jones CM, Baldwin GT. Relationship between nonmedical prescription-opioid use and heroin use. N Engl J Med. 2016;374(2):154-163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Pitt AL, Humphreys K, Brandeau ML. Modeling health benefits and harms of public policy responses to the U.S. opioid epidemic. Am J Public Health. 2018;108(10):1394-1400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Calcaterra SL, Butler M, Olson K, et al. The impact of a PDMP-EHR data integration combined with clinical decision support on opioid and benzodiazepine prescribing across clinicians in a metropolitan area. J Addict Med. 2022;16(3):324-332. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Wang LX. The complementarity of drug monitoring programs and health IT for reducing opioid-related mortality and morbidity. Health Econ. 2021;30(9):2026-2046. [DOI] [PubMed] [Google Scholar]
- 25. Gihleb R, Giuntella O, Zhang N. Prescription drug monitoring programs and neonatal outcomes. Region Sci Urban Econ. 2020;81:103497. [Google Scholar]
- 26. Anwar T, Jayawardhana J. The association between pill mill legislation and neonatal abstinence syndrome. J Pharm Health Serv Res. 2021;13(1):41-47. [Google Scholar]
- 27. Bohler RM, Hodgkin D, Kreiner PW, et al. Predictors of U.S. states’ adoption of naloxone access laws, 2001–2017. Drug Alcohol Depend. 2021;225:108772. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Bradford AC, Bradford DW. Factors driving the diffusion of medical marijuana legalisation in the United States. Drugs Educ Prev Policy. 2016;24(1):75-84. [Google Scholar]
- 29. Macinko J, Silver D. Diffusion of impaired driving laws among U.S. states. Am J Public Health. 2015;105(9):1893-1900. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Horwitz JR, Davis C, McClelland L, et al. The importance of data source in prescription drug monitoring program research. Health Serv Res. 2021;56(2):268-274. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Kaiser Family Foundation. Opioid overdose death rates and all drug overdose death rates per 100,000 population (age-adjusted). February 26, 2021. Accessed March 7, 2023. https://www.kff.org/other/state-indicator/opioid-overdose-death-rates/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Opioid%20Overdose%20Death%20Rate%20(Age-Adjusted)%22,%22sort%22:%22desc%22%7D
- 32. U.S. Census Bureau. American community survey. March 11, 2022. Accessed March 7, 2023. https://www.census.gov/programs-surveys/acs
- 33. RC Fording. State ideology data. June 18, 2018. Accessed March 7, 2023. https://rcfording.wordpress.com/state-ideology-data/
- 34. United Health Foundation. American’s health rankings. 2020 Annual Report. Accessed August 10, 2021. https://www.americashealthrankings.org/explore/annual/measure/PH_funding/state/ALL
- 35. Kaiser Family Foundation. Status of state Medicaid expansion decisions: interactive map. February 24, 2022. Accessed March 7, 2023. https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions-interactive-map/
- 36. Warren MD, Miller AM, Traylor J, et al. ; Centers for Disease Control and Prevention (CDC). Implementation of a statewide surveillance system for neonatal abstinence syndrome—Tennessee, 2013. MMWR Morb Mortal Wkly Rep. 2015;64(5):125-128. [PMC free article] [PubMed] [Google Scholar]
- 37. Levinson-Castiel R, Merlob P, Linder N, et al. Neonatal abstinence syndrome after in utero exposure to selective serotonin reuptake inhibitors in term infants. Arch Pediatr Adolesc Med. 2006;160(2):173-176. [DOI] [PubMed] [Google Scholar]
- 38. Nyakeriga AM, McDonald M. Neonatal abstinence syndrome surveillance annual report 2020. Tennessee Department of Health, Nashville, TN. 2020. Accessed March 7, 2023. https://www.tn.gov/content/dam/tn/health/documents/nas/NAS-Annual-Report-2020.pdf
- 39. Unik GJ, Ciccarone D. U.S. regional and demographic differences in prescription opioid and heroin-related overdose hospitalizations. Int J Drug Policy. 2017;46:112-119. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Lippold KM, Jones CM, Olsen EO, et al. Racial/Ethnic and age group differences in opioid and synthetic opioid–involved overdose deaths among adults aged ≥18 years in metropolitan areas—United States, 2015-2017. MMWR Morb Mortal Wkly Rep. 2019;68(43):967-973. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Saloner B, Levin J, Chang H, et al. Changes in buprenorphine-naloxone and opioid pain reliever prescriptions after the Affordable Care Act Medicaid expansion. JAMA Netw Open. 2018;1(4):e181588. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Sharp A, Jones A, Sherwood J, et al. Impact of Medicaid expansion on access to opioid analgesic medications and medication-assisted treatment. Am J Public Health. 2018;108(5):642-648. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Cloud DH, Castillo T, Brinkley-Rubinstein L, et al. Syringe decriminalization advocacy in red states: lessons from the North Carolina Harm Reduction Coalition. Curr HIV/AIDS Rep. 2018;15(3):276-282. [DOI] [PubMed] [Google Scholar]
- 44. Ezell JM, Walters S, Friedman SR, et al. Stigmatize the use, not the user? Attitudes on opioid use, drug injection, treatment, and overdose prevention in rural communities. Soc Sci Med. 2021;268:113470. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. U.S. Department of Justice. Office of justice programs. Harold Rogers prescription drug monitoring program fact sheet. Accessed March 7, 2023. https://www.ojp.gov/ncjrs/virtual-library/abstracts/harold-rogers-prescription-drug-monitoring-program-fact-sheet
- 46. Centers for Disease Control and Prevention. Integrating & expanding prescription drug monitoring program data: lessons from nine states. February 2017. Accessed March 7, 2023. https://www.cdc.gov/drugoverdose/pdf/pehriie_report-a.pdf
- 47. Richwine C, Everson J. Electronic Prescribing of Controlled Substances and Use of Prescription Drug Monitoring Programs among Office-Based Physicians, 2019-2021. ONC Data Brief, No. 63. Office of the National Coordinator for Health Information Technology; 2023.
- 48. Holmgren AJ, Apathy NC. Evaluation of prescription drug monitoring program integration with hospitals electronic health records by U.S. county-level opioid prescribing rates. JAMA Netw Open. 2020;3(6):e209085. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49. Vaidya A. Ochsner to integrate state prescription drug monitoring data into its Epic EHR. Becker’s Health IT. January 10, 2018. Accessed March 7, 2023. https://www.beckershospitalreview.com/ehrs/ochsner-to-integrate-state-prescription-drug-monitoring-data-into-its-epic-ehr.html
- 50. Louisiana Board of Pharmacy. Prescription monitoring program annual report: fiscal year 2020-2021. July 1, 2021. Accessed March 7, 2023. https://www.pharmacy.la.gov/assets/docs/PMP/PMP_AnnRpt_2021_Pkg.pdf
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets were derived from sources in the public domain: Centers for Disease Control and Prevention U.S. Opioid Dispensing Rate Maps (https://www.cdc.gov/drugoverdose/rxrate-maps/index.html), Kaiser Family Foundation Opioid Overdose Death Rates [https://www.kff.org/other/state-indicator/opioid-overdose-death-rates/?currentTimeframe=0&sortModel=%7B%22colId%22:%22Opioid%20Overdose%20Death%20Rate%20(Age-Adjusted)%22,%22sort%22:%22desc%22%7D], Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project Fast Stats (https://datatools.ahrq.gov/hcup-fast-stats), U.S. Census Bureau American Community Survey (https://www.census.gov/programs-surveys/acs), Richard C. Fording State Ideology Data (https://rcfording.wordpress.com/state-ideology-data/), United Health Foundation American Health Rankings (https://www.americashealthrankings.org/explore/annual/measure/PH_funding/state/ALL), and Kaiser Family Foundation Status of State Medicaid Expansion (https://www.kff.org/medicaid/issue-brief/status-of-state-medicaid-expansion-decisions-interactive-map/). The effective dates for PDMP integration and mandatory use policies are available in this article’s online supplementary material.