Background
Prescription opioids have been linked to over half of the 28,000 opioid overdose deaths in 2014, more than any year on record [
1]. Every year, hundreds of millions of opioid prescriptions continue to be written, despite almost all states having large scale prescription drug monitoring programs (PDMP) and several states having “Must access” emergency rules in place [
2]. At the end of 2013, Indiana implemented emergency rules to regulate opioid prescribing [
3].
High rates of opioid prescribing in recent years has raised renewed interest in quantifying the success of state policies, in particular the latest emergency rules, in curbing inappropriate opioid prescribing. Early studies examining the impact of such policies reported significant initial decreases in opioid-related morbidity and mortality associated with decreased prescribing [
4‐
6]. However, little is known about how the emergency rules differentially impact patient subgroups. Prior research has identified patient characteristics, such as gender, age, insurance type, and socioeconomic status as significant predictors of pain management, opioid prescribing, and drug non-medical use [
7,
8]. In addition, research has documented issues attributed to biases in prescribing, such as: 1) undertreating pain in some patient subgroups, thus worsening an existing disparity in pain management [
7,
8], and 2) failing to correct overprescribing for other patient groups, thus exposing them to the harmful effects of opioid use and non-medical use [
9].
We sought to investigate the impact of Indiana’s new emergency prescribing rules on prescription of opioids. Our study aimed to: 1) compare volumes of prescribed opioids before and after the Indiana emergency rules; and 2) stratify the changes in opioid prescribing by patient and provider subgroups. Specifically, differential impacts of the emergency rules by patient gender, age, payer and zip code level aggregate measures of socioeconomic status are considered.
Methods
Data
Data were obtained from the Indiana Prescription Drug Monitoring Program (PDMP), which is called the Indiana’s Prescription Electronic Collection and Tracking Program (INSPECT) [
10]. INSPECT records contain prescription-level, limited data set that include patient ID, patient zip code, gender, birth year, date the prescription was written, date the prescription was dispensed, quantity dispensed, number of days’ supplied, pharmacy ID, pharmacy zip code, provider ID, provider zip code, payer, National Drug Code (NDC), and drug name. All drug strengths were converted to the Morphine Equivalent Dose (MED) to facilitate comparison [
11,
12].
The Indiana emergency rules
Indiana implemented emergency prescribing rules that went into effect on December 15, 2013, and then became permanent in November 6, 2014 [
3]. The rules are only applicable if patient has been prescribed, for more than three consecutive months: 1) >60 opioid-containing pills per month; or 2) A morphine equivalent dose >15 mg/day [
3]. These rules, issued by the Medical Licensing Board, require prescribers to: (1) evaluate opioid recipients for psychiatric conditions; (2) review patients’ drug prescription history in INSPECT and (3) perform regular drug screenings; and (4) obtain a signed controlled-substance agreement from the patient [
3]. Prescribers were to query at initiation and at least yearly.
Time segments and participants
Data for all opioids dispensed in the state of Indiana between January 1, 2011 (1079 days before to the emergency rules) and November 6, 2014 (325 days after the policy) were obtained. Prescription level INSPECT data was merged with census data to characterize socioeconomic status for patients.
Statistical analysis
An interrupted time series analysis (ITSA) was utilized to investigate the association between the Indiana emergency rules and opioid prescribing, controlling for time trends and the auto-regressive nature of the time series. The aggregated ITSA results in Tables
1 and
2 are for the full group of providers or patients within the state of Indiana on a given day. Individuals within this group may vary each day. ITSA relies on OLS regression models that are more flexible than ARIMA models but at the same time broadly applicable to interrupted time series context. Thus, the difference in daily prescribing between the pre- and post-intervention period can be interpreted as plausibly associated to the rule change. The primary outcome indicator is the total MED of dispensed opioids per day in the state of Indiana. Other outcomes considered include the number of unique patients, the number of unique providers, the number of prescriptions, the MED per transaction, the MED per day, and the number of days supplied. To further check that what we are capturing is a plausible impact of the policy, we tested for interruptions by replacing the true policy start date with other pseudo-start dates along the pre-intervention continuum.
1 In addition, to control for time invariant patient and provider characteristics, we re-estimated our model using patient and provider fixed-effects. The patient and provider fixed effects analysis (Table
3) longitudinally follows a group of providers and patients before and after the policy change and the results can be interpreted as the change in opioid prescribing correlated with the change in policy.
Table 1
Association of Indiana’s opioid prescription emergency rules to daily Morphine Equivalent Dose (MED) per patient of opioids dispensed by patient gender, age, payer type, number of scripts within MED per day per patient brackets and by number of days of supply per prescription
Gender |
Females | −2.80*** | −.0027*** | −.0042*** |
Males | −3.68*** | −.0014*** | −0.0044** |
Age |
0–20 years | - 27.26*** | −.0151*** | 0.0219*** |
20–40 years | −3.00*** | .0009*** | −0.0045*** |
40–60 years | −2.45*** | −.0017*** | −0.0023 |
60+ years | −2.04*** | −.0014*** | −0.0016 |
Payer Type |
Commercial Ins. | −3.32*** | −.0041*** | 0.0001* |
Medicaid | −3.96*** | −.0018*** | −0.0100*** |
Medicare | −4.03*** | .0037*** | −0.0080*** |
Private Pay | −2.65*** | .0007** | −0.0019 |
Worker’s Comp | −5.93*** | .0038*** | 0.0120* |
# scripts within clinically relevant MED per day brackets |
0–20 | 41.72 | .4939*** | 0.1299 |
20–40 | −836.76 *** | −.1699 | −0.6885 |
40–60 | −363.90 *** | .2325* | −0.1411 |
60–80 | −126.24*** | .0072 | −0.2214 |
80–100 | −110.34*** | .0511* | 0.0058 |
100+ | −345.37*** | .1486* | −0.2608 |
# scripts within clinically relevant # days of supply brackets |
0–15 | −635.96* | −.3354 | −1.2997 |
16–30 | −1077.80*** | 1.1207*** | 0.2100 |
> 30 | −49.59*** | −.0076 | −0.0843** |
Table 2
The association of Indiana’s opioid prescription emergency rules to daily Morphine Equivalent Dose (MED) per patient of different opioids dispensed in Indiana
All opioids | −3.17*** | −0.0021*** | −0.0044*** |
hydrocodone | −3.68*** | −0.0038*** | −0.0048*** |
oxycodone | −2.03*** | −0.0015*** | −0.0133*** |
morphine | −0.03 | −.0101*** | −0.0185*** |
methadone | −6.19*** | −.0026*** | −0.0333*** |
fentanyl | −1.15 | −.0080*** | −0.0169*** |
oxymorphone | −3.11* | −.0245*** | −0.0142* |
buprenorphine | −.47 | −.0022*** | −0.0209*** |
hydromorphone | −3.54** | .0005 | −0.0143* |
Table 3
Association of Indiana’s opioid prescription emergency rules to daily Morphine Equivalent Dose (MED) per patient of opioids dispensed, controlling for patient and provider fixed effects (n = 1404 days)
Impact of policy, instantaneous | −72.7*** | −67.2*** |
Impact of policy on trend | −0.045*** | −0.024 |
Next, analyses were stratified by patient gender, age groups, ranges of opioid dosages, and payers. We examined the significance of aggregate recipients’ and practices’ annual per-capita income by zip code level. We also investigated the policy’s plausible impact, within each decile of daily average MED of dispensed opioids per recipient. Finally, because specific drugs are associated with different extents of non-medical use and different patterns of prescribing, in addition to studying the impact of the rules on all opioids, we individually assessed the eight most commonly prescribed opioids: hydrocodone, oxycodone, morphine, methadone, fentanyl, oxymorphone, buprenorphine, and hydromorphone.
Discussion
Prescription drug non-medical use is a major public health concern. State mandated opioid prescribing emergency rules to change prescribing practices have been adopted across the United States; however, the impact of these rules on provider prescribing behavior has not been effectively studied. In this paper, we report that Indiana’s opioid emergency rules are associated to a decrease in the volume of total opioid prescriptions and in the average number of opioid MED per patient. It follows that with a reduction in inappropriate opioid dispensing lower rates of opioid misuse and non-medical use should also be realized. From the standpoint of both the population’s health and the individual patient’s care, these findings are of significant importance.
The results show that the emergency rules were associated with significant changes in prescribing behavior statewide and across all individual patient and provider characteristics, albeit with varying magnitudes. In particular, reduced opioid prescriptions were noted across genders, health care/insurance types, and age groups. In addition, reduced prescribing was not limited to certain types of prescriptions (high doses, longer durations, etc.). Regarding types of medications, prescriptions of nearly all opioids commonly prescribed for chronic pain declined. Prescribing rates were noticeably lower in all deciles, with the exception of the lower 30% of prescribing rates.
Prior studies have assessed if patient characteristics, such as gender, age, insurance type, and socioeconomic status are significant predictors of quality of pain management, opioid prescribing, and drug non-medical use [
7,
8]. Our results show that the Indiana rules are indeed differently associated to different patient subgroups. Post emergency rules, the decline in opioid prescribing was more prominent for males compared to females, patients who have Medicare and Medicaid compared to those with private insurances, and younger patients compared to older ones. Furthermore, practicing in areas with lower incomes was associated with a greater decrease in prescribing after the policy. While it is not surprising that the new emergency rules predict steepest declines among the highest opioid recipients, other patient and provider characteristics also seem to be associated with different prescribing patterns in terms of quantities of opioids.
Fewer providers prescribed opioids after implementation of the rules, and for those still prescribing opioid analgesics, each provider wrote fewer prescriptions for shorter durations and lower doses, on average. The emergency rules are not significantly associated with changes in prescriptions with relatively low doses of 0–20 MED per day; however, they did plausibly significantly lower higher MED prescriptions. Furthermore, the largest MED reduction was observed within the 16–30 day category, with a much smaller but still significant decline in the longer-running prescriptions. Isolating such effects confirms our hypothesis that the changes in prescribing patters are possibly the result of the rules and not merely random time trends.
The emergency rules are not associated to changes in all types of opioids equally. Buprenorphine, which is used in the treatment of pain and addiction, has gained an increased popularity over the past few years. Buprenorphine, unlike methadone, is registered in INSPECT when used as buprenorphine/naloxone (Suboxone or Subutex) for addiction or as buprenorphine alone for pain management [
13]. Fentanyl was another agent that did not see a reduction in use following the new opioid prescribing emergency rules. Of note, the rules did have several patient exclusions, including that they did not apply to patients with terminal medical conditions or those in Indiana-licensed hospice or palliative care programs. Although fentanyl certainly can be subject to misuse and non-medical use, its most popular formulation in an outpatient setting, the transdermal patch, is indicated for chronic pain management in opioid-tolerant patients who have had insufficient pain relief with other treatment options [
14]. However, the continued use of fentanyl at rates similar to those before the emergency rules is a relevant observation in light of the rising concern of increased mortality due to fentanyl non-medical use [
13]. Existing literature finds that non-medical fentanyl use and resulting fatalities are largely driven by synthetic fentanyl illegally manufactured [
15]. Finally, the emergency rules predicted different magnitudes of change in the short-acting and the long-acting forms of morphine and oxycodone. Although decreases in short-acting formulations are expected post implementation of the rules, perhaps short-acting morphine remains a preferred agent in the acute pain setting (thus no reduction seen).
Our study is not without limitations. The use of time series analysis provides some evidence for causality. Stronger evidence can be identified in the presence of a comparison state. A neighboring state, such as Ohio or Kentucky, appears to be an appealing option in light of similar patient and provider demographics [
16]. In future work, we hope to gain access to PDMP data from potential comparison states and conduct comparative analyses to better infer the plausibly causal impact of the emergency rules using as a control group the state(s) that appears to be the closest to Indiana in its summary statistics, but that did not implement the emergency rules yet. While our study brings attention to several socio-demographic-economic determinants of practice, the absence of important other factors, such as race and individual patient SES indicators, have limited our ability to draw strong conclusions. Further, prescriber zip codes in our data could capture location of the prescriber’s residence or practice, which limits our ability to precisely interpret the positive association between zip code aggregate socioeconomic status and opioid prescribing changes. A qualitative evaluation of a representative sample of patients and prescribers could provide further evidence in examining these factors. Such analyses would also bring to light any unintended consequences that the implementation of the emergency rules may have. For instance, the added regulatory burden of opioid prescribing following the emergency rules may deter providers from prescribing altogether, resulting in worsening of patient pain outcomes. The impact of the emergency rules on access-to-appropriate treatment and care involving opioid alternatives could not be evaluated using our current data and is left to future work. Finally, while the daily opioid dose has been linked to patient outcomes such as overdose and death, our study did not explore those outcomes. Other important outcomes include drug diversion and crime, both of which are directly linked to opioid prescribing; they can also be explored in further studies.