Assessment of Prescriber and Pharmacy Shopping Among the Family Members of Patients Prescribed Opioids

IMPORTANCE Most prescription opioid misuse involves opioids prescribed to others—a form of opioid diversion. However, few indicators of diversion risk exist. Because family members can often accesspatients’opioids,onesuchindicatormaybethefrequencywithwhichopioidprescriptionsare filled by patients when their family members are engaged in opioid prescriber and pharmacy shopping (“doctor and pharmacy shopping”). To date, this frequency has not been estimated. OBJECTIVE To estimate the proportion of opioid prescription fills for which family members meet prescriber and pharmacy shopping criteria. DESIGN, SETTING, AND PARTICIPANTS A cross-sectional analysis of 2015-2016 claims from a national commercial insurer was conducted from August to October, 2018. The sample included patients without cancer who were covered by family insurance plans and had 1 or more opioid prescription fill in 2016, as measured by prescription drug claims. Fills were the unit of analysis. MAIN OUTCOMES AND MEASURES For each fill in 2016 by the patient and each family member enrolled in the same plan (eg, spouse or child), the number of prescribers and number of pharmacies in the prior 12 months were counted. Prescriber and pharmacy shopping was defined as 4 or more prescribers and 4 or more pharmacies, following a National Quality Forum–endorsed measure. The proportion of fills for which 1 or more family member met criteria and the proportion for which the patient met criteria were calculated. RESULTS


Introduction
][3][4][5][6][7][8][9][10][11] In one study, the odds of opioid overdose death were 6.5 times higher when patients received opioid prescriptions from 4 or more vs fewer than 4 prescribers during the prior year, and 6.0 times higher when patients filled opioid prescriptions at 4 or more vs fewer than 4 pharmacies. 4Obtaining opioids from high numbers of prescribers and pharmacies, often referred to as prescriber and pharmacy shopping ("doctor and pharmacy shopping"), has been reported to be associated with greater risk of overdose death than high numbers of prescribers or pharmacies alone. 10,12To mitigate this risk, 49 states have implemented prescription drug monitoring programs (PDMPs), which allow clinicians and pharmacists to identify patients engaged in prescriber and pharmacy shopping by providing information on the prescribers and pharmacies of recent controlled substance prescriptions. 13,14e link between opioid overdose death and prescriber and pharmacy shopping is at least partially mediated by prescription opioid misuse, a behavior that includes using one's own opioids in a manner other than prescribed. 15To facilitate efforts to prevent this type of misuse, the Pharmacy Quality Alliance recently developed a National Quality Forum-endorsed measure assessing use patterns consistent with prescriber and pharmacy shopping among patients without cancer who are prescribed opioids. 12However, prescription opioid misuse also includes taking someone else's opioids, which are the source for misuse more than half of the time among the 4.2% of American individuals who misuse prescription pain relievers each year. 15Efforts by governments, payers, and clinicians to prevent this other type of misuse-a form of opioid diversion-would be facilitated by the development of risk indicators that are feasible to measure using existing data.
One indicator of diversion for the purpose of misuse may be the prevalence of prescriber and pharmacy shopping among the family members of patients prescribed opioids.Families are a natural unit of analysis because opioid use among patients and their family members can often be observed concurrently in data such as claims or electronic medical records.6][17] To our knowledge, no study has estimated the frequency with which opioid prescriptions are filled by patients when their family members are engaged in prescriber and pharmacy shopping.The objective of this study was to estimate this frequency using a national commercial claims database and linkages between family members enrolled in the same health insurance plan.

Study Design and Data Source
From August to October 2018, we conducted a cross-sectional analysis of the 2015-2016 Clinformatics Data Mart (OptumInsight).This commercial claims database represents approximately 12 million patients each year who are covered by a large national insurer.Key data elements include unique family identifiers for enrollees in the same plan, as well as encrypted pharmacy identifiers and national provider identifier (NPI) data in prescription drug claims.Family plans cover the policyholder and dependents, such as spouses and children.
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. 18The institutional review board of the University of Michigan Medical School exempted this study from review because of its use of deidentified data.

Opioid Prescription Fills
We defined an opioid prescription fill as a prescription drug claim for an opioid analgesic.We identified such claims using national drug codes compiled by the Centers for Disease Control and Prevention. 19We additionally included national drug codes identified by searching the IBM

JAMA Network Open | Substance Use and Addiction
Prescriber and Pharmacy Shopping Among the Family Members of Patients Prescribed Opioids

Sample
The sample included patients who ( Without Cancer). 12 used opioid prescription fills as the unit of analysis to capture the number of occasions in which information on family members' opioid use patterns could inform prescribing decisions.We initially included all fills by patients in the sample during 2016 (index fills; the date was the index date).We excluded index fills if there was missing NPI data for any opioid prescription fill by the patient or any family members during a 12-month look-back period.This period began 366 days before the index date and ended 1 day before the index date.

Study Variables
We defined prescriber and pharmacy shopping as filling opioid prescriptions from 4 or more prescribers at 4 or more pharmacies during the 12-month look-back period, following the specification of the Pharmacy Quality Alliance measure. 12][3][4][5][6][7][8][9][10] Measure validation occurred by assessing face validity among experts. 12To assess how results changed when using different criteria, we performed sensitivity analyses using more and less stringent criteria (Ն4 prescribers and Ն4 pharmacies in 6 months, Ն3 prescribers and Ն3 pharmacies in 12 months, Ն4 prescribers in 12 months, and Ն4 pharmacies in 12 months).

Statistical Analysis
For categorical variables, we calculated counts and frequencies.For continuous variables, we calculated means with SDs or medians with interquartile ranges (IQRs) when the distribution was skewed.Because of the study's large sample size, the width of 95% CIs was narrow (eg, typically less than a few hundredths of a percentage point for proportions); consequently, we do not report these intervals.
We calculated the proportion of index fills for which 1 or more of the patient's family members met criteria for prescriber and pharmacy shopping (primary outcome).We then independently repeated this analysis for the patient (secondary outcome).Family members were linked to patients using the unique family identifier and were eligible to meet prescriber and pharmacy shopping criteria only if they were 2 years or older at the end of 2016, had no cancer diagnosis in 2015 or 2016, and were continuously insured throughout 2015 and 2016.As an example, for index fills occurring on January 1, 2016, we identified prescription fills by each eligible family member that occurred between January 1, 2015, and December 31, 2015, and determined whether these fills were associated with 4 or more unique NPIs and 4 or more unique pharmacy identifiers.We then repeated this analysis for the patient.
Among index fills for which 1 or more family member met criteria, we calculated the proportion for which the patient did and did not also meet criteria.PDMP reports would lack evidence of prescriber and pharmacy shopping by patients, even though their family members engage in this behavior.
Among index fills for which 1 or more family member met criteria, we determined the type and specialty of prescribing clinicians using taxonomy codes.Among index fills for which 1 or more family member did and did not meet criteria, we calculated median total morphine milligram equivalents (MMEs), a standardized measure of opioid amount. 19To assess whether the former group of index fills typically had a higher amount of opioid potentially available for diversion than the latter group, we assessed the significance of differences in total MMEs using a nonparametric test (Wilcoxon signed rank test), as the distribution of total MMEs was right skewed.We repeated this analysis for index fills for which the patient did and did not meet criteria.
To construct patient-level prevalence estimates, we calculated the proportion of patients who had 1 or more index fill in 2016 for which 1 or more family member met criteria, and the proportion who had 1 or more index fill in 2016 for which the patient met criteria.We also calculated the characteristics of family members who met criteria during 2016.
We conducted analyses both in the overall sample and by subgroups defined by age (index fills by adults Ն18 years vs children 2-17 years).We considered P values <.05 to be statistically significant and used 2-sided tests.Analyses were conducted using SAS, version 9.4 (SAS Institute Inc).

Sensitivity Analysis
In addition to using different criteria for prescriber and pharmacy shopping, we performed a sensitivity analysis in which patients and family members with cancer were not excluded.We also repeated analyses that included buprenorphine and methadone as index fills.

Sample
The database included 4 644 898 patients 2 years or older who had family insurance coverage throughout 2015 and 2016 and had no cancer diagnoses during this period.Among these patients, 554 417 individuals (11.9%) had 1 or more opioid prescription fill during 2016 and were included in the sample.Among patients in the sample, 301 297 (54.3%) were female, 506 370 (91.3%) were adults, and 48 047 (8.7%) were children (

Main Results
For 8485 of the 1 471 971 (0.6%) index fills, 1 or more family member met prescriber and pharmacy shopping criteria.For 44 547 of the index fills (3.0%), the patient met criteria (Table 2).For 1538 (0.1%) of the index fills, both the patient and 1 or more family member met criteria; these fills are reflected in both the 0.6% and 3.0% estimates above.Among 8485 index fills for which 1 or more family member met criteria, the patient did not meet criteria for 6947 (81.9%) of the fills, but did meet criteria for 1538 fills (18.1%).
Among 8485 index fills for which 1 or more family member met prescriber and pharmacy shopping criteria, 5782 (68.1%) of prescribing clinicians were physicians, 667 (7.9%) were physician

JAMA Network Open | Substance Use and Addiction
Prescriber and Pharmacy Shopping Among the Family Members of Patients Prescribed Opioids  3).

Patient-Level Prevalence Estimates
Among the 554 417 patients in the sample, 2455 individuals (0.4%) filled 1 or more opioid prescription in 2016 for which 1 or more family member met criteria for prescriber and pharmacy shopping.Of these 2455 patients, 2242 (91.3%) were adults, 213 (8.7%) were children, and 1043 (46.5%) were female.Among 2363 family members who met criteria at any point during 2016, 2339 (99.0%) were adults and 1472 (62.3%) were female.Among the 554 417 patients in the sample, 6582 individuals (1.2%) filled 1 or more opioid prescription for which they met criteria themselves.

Subgroup Analysis by Age
Among index fills by adults, the proportion for which the 1 415 483 adults met prescriber and pharmacy shopping criteria (44 416 [3.1%]) was higher than the proportion for which 1 or more family  2).
Among the 506 370 adult patients, 2242 individuals (0.4%) filled 1 or more opioid prescription in 2016 for which 1 or more family member met criteria, while 6564 adults (1.3%) filled 1 or more opioid prescription in 2016 for which the patient met criteria.Among the 48 047 child patients, 213 individuals (0.4%) filled 1 or more opioid prescription in 2016 for which 1 or more family member met criteria, and 18 children (0.04%) filled 1 or more opioid prescription in 2016 for which the patient met criteria.Among children with fills for which 1 or more family member met criteria, the mean (SD) age of these family members was 40.1 (7.6) years.

Sensitivity Analyses
When prescriber and pharmacy shopping criteria were defined as 4 or more pharmacies and 4 or more pharmacies over a 6-month look-back period, the proportion of fills for which 1 or more family member met criteria decreased from 0.6% (n = 8485) to 0.2% (n = 2394).When criteria were defined as 3 or more prescribers and 3 or more pharmacies over 12 months, this proportion increased to 1.9% (n = 27 578).When requiring 4 or more prescribers over 12 months (prescriber shopping alone), this proportion was 2.6% (n = 38 543), compared with 1.4% (n = 20 003) when requiring 4 or more prescribers over 12 months (pharmacy shopping alone).Findings were virtually identical when including patients and family members with cancer, as well as when including buprenorphine and methadone as index fills (Table 4).

Discussion
In this national study of 554 417 patients who were enrolled in private family insurance plans and filled opioid prescriptions during 2016, 0.6% of the fills (1 of 167) occurred when at least 1 of the patient's family members met criteria for opioid prescriber and pharmacy shopping.For most of these fills, patients did not meet criteria.Family members who are engaged in prescriber and pharmacy shopping may have both the motive and opportunity to divert patients' opioids.
Consequently, findings suggest the potential for opioid diversion within families.
To our knowledge, this study is the first to assess prescriber and pharmacy shopping among the family members of patients prescribed opioids.If the study's prevalence estimate of 0.6% is nationally representative, then family members met prescriber and pharmacy shopping criteria for approximately 1.2 million opioid prescriptions in 2016, a year in which 210 million opioid prescriptions were filled in the United States. 21However, this estimate likely represents a lower bound.First, the study's database did not capture opioid prescriptions paid for by cash, which is a common source of payment among patients engaged in opioid prescriber and pharmacy shopping. 1Second, the study only included commercially insured patients with family coverage for 24 months.Commercially insured patients have lower rates of prescriber and pharmacy shopping than other populations, 12 while the ability to maintain family coverage for a prolonged period may be a marker of unobserved factors that decrease the likelihood of this behavior (eg, economic stability).Finally, the study used a somewhat conservative definition of prescriber and pharmacy shopping to be consistent with a national quality measure and avoid misclassification of legitimate use patterns. 8,12Based on these considerations, the absolute number of opioid prescriptions associated with prescriber and pharmacy shopping by family members is likely to be substantial.Prescriber and pharmacy shopping indicates a fairly extreme example of opioid misuse-related behavior.Less-extreme misuse-related behavior likely occurs in the family members of patients prescribed opioids, suggesting that this study's findings may also underestimate the potential for opioid diversion within families.
In this study, 0.4% of patients had at least 1 opioid prescription fill in 2016 for which family members met prescriber and pharmacy shopping criteria, even though family members met these criteria for 0.6% of opioid prescription fills.This finding indicates that some patients had multiple such fills in 2016, suggesting that there were multiple opportunities to detect prescriber and pharmacy shopping among their family members.However, such detection may be difficult for clinicians.For 81.9% of opioid prescription fills for which family members met prescriber and pharmacy shopping criteria in this study, the patient did not meet criteria, implying that his or her PDMP reports would not contain evidence of this behavior.Therefore, the absence of this behavior on patients' PDMP reports does not reliably indicate the absence of this behavior among their family members.Furthermore, findings suggest that prescriber and pharmacy shopping occurs in the family members of patients with both acute and chronic pain.Among opioid prescription fills for which family members met criteria in this study, prescribers included clinicians who often prescribe opioids for chronic pain (eg, family medicine and pain medicine physicians), as well as clinicians who often prescribe opioids for acute pain (eg, surgeons and dentists).
The amount of opioids potentially available for diversion was typically higher when fills were associated with prescriber and pharmacy shopping by family members than when they were not.
One explanation is that the former were more likely to represent high-dose opioid prescriptions for chronic pain, in which case counseling on safe opioid storage and disposal may be an important intervention to prevent diversion. 22,23An alternative explanation is that family members engaged in prescriber and pharmacy shopping might sometimes accompany patients to visits and manipulate clinicians by suggesting a higher-than-needed opioid requirement.Yet another explanation is that some patients may attempt to obtain high-dose opioid prescriptions on behalf of family members engaged in prescriber and pharmacy shopping.In either of these latter 2 situations, counseling on safe storage and disposal will likely not prevent diversion.A more effective intervention may be to rely on nonopioid therapy when feasible and avoid opioid prescriptions with amounts that exceed clinical need, thus decreasing the amount of opioids available for diversion regardless of family dynamics.
Among opioid prescription fills by children in this study, family members met prescriber and pharmacy shopping criteria for 0.7%.Although the study's database did not contain information on family relationships, the mean age of these family members was 40.1 years, suggesting that many may have been parents.Furthermore, 0.2% of fills by children met prescriber and pharmacy shopping criteria.Children lack the ability to obtain opioid prescriptions from multiple clinicians and fill them at multiple pharmacies by themselves, suggesting that apparent prescriber and pharmacy shopping behavior in children is likely driven by an adult family member.Collectively, these findings potentially support concerns among some pediatric clinicians that parents and other family members may misuse opioids prescribed to children or may feign or exploit pain in children to obtain opioids for themselves. 24Our analysis of PDMP statutes in October 2018 showed that 39 states mandate health care professionals to review PDMP reports under some circumstances before prescribing controlled substances to patients, including children. 13However, for opioid prescriptions to children in this study, family members were 3 times more likely than the child to meet prescriber and pharmacy shopping criteria.Thus, for children prescribed opioids, interventions focused on preventing misuse among family members may be more useful than interventions focused on children.
The challenges and potential importance of detecting prescriber and pharmacy shopping among family members raises the question of whether policy makers should increase clinicians' access to information on family members' opioid use patterns.In theory, for example, clinicians could be explicitly allowed to check family members' PDMP reports if amendments were made to the Health Insurance Portability and Accountability Act, 25 state privacy laws, and state PDMP laws that impose criminal penalties for wrongful use of PDMP data, defined typically as access under circumstances that do not pertain to a patient's care or fall within an established clinician-patient relationship. 26However, such amendments are hard to justify given the relative rareness of prescriber and pharmacy shopping in family members and the importance of maintaining patient privacy to avoid stigma and other negative consequences.Consequently, interventions at the clinical level, such as avoiding excessive opioid prescribing, are likely the most feasible and appropriate method to prevent opioid diversion within families.

Limitations
This study has several limitations.First, prescriber and pharmacy shopping was defined in part based on the number of unique NPIs associated with opioid prescription fills, but some opioid prescribers may have been assigned multiple NPIs.Second, the database did not capture opioid use patterns by family members who were not enrolled in the same insurance plan.Third, this study did not examine the consequences of prescribing opioids to patients whose family members are engaged in prescriber and pharmacy shopping, such as increased overdoses in these family members.

JAMA Network Open | Substance Use and Addiction
Prescriber and Pharmacy Shopping Among the Family Members of Patients Prescribed Opioids

Table 1
). Mean (SD) age was 41.4(16.4)years.Patients in the sample represented 469 913 family plans.The mean (SD) number of family members per plan was 3.3 (1.3), the mean number of adults per plan was 2.5 (0.9), and the mean number of children per plan was 0.8 (1.1); 224 742 of the plans (47.8%) had at least one 1 adult member and at least 1 one child member.

Table 1 .
Demographic Characteristics of Study Sample

Table 3 .
Prescribers of Opioid Prescription Fills for Which 1 or More Family Member or the Patient Met Prescriber and Pharmacy Shopping Criteria Prescriber and Pharmacy Shopping Among the Family Members of Patients Prescribed Opioids d Excludes physicians in these specialties who are also listed as pain medicine specialists.JAMA Network Open | Substance Use and AddictionJAMA Network Open.2019;2(5):e193673.doi:10.1001/jamanetworkopen.2019.3673(Reprinted) May 10, 2019 6/11 Downloaded From: https://jamanetwork.com/ on 11/03/2023

Table 4 .
Sensitivity Analyses Number of index fills is higher than the main analysis because index fills were excluded if there was missing national provider identifier information on any opioid prescription fills by the patient or family members in the 6 months before the index date instead of 12 months, resulting in fewer exclusions. a