Trends in co-prescribed opioids and benzodiazepines, non-prescribed opioids and benzodiazepines, and schedule-I drugs in the United States, 2013 to 2019

Highlights • Concurrent opioid and benzodiazepine use in United States adults decreased from 19.3% in 2013, to 9.8% in 2019.• Concurrent use decreased at a greater rate after the 2016 Opioid Prescribing Guideline by the Centers for Disease Control and Prevention and warnings from the Food and Drug Administration were released.• Schedule I drug use generally increased from 8.9% in 2013 to 13.8% through 2019, with a noticeable dip between Q3 and Q4 of 2013, after the Drug Enforcement Agency announced the rescheduling of hydrocodone.• Prescription opioid or benzodiazepine misuse decreased from 75.6% in 2013 to 59.3% by the end of 2019.


Introduction
Opioid use is associated with risks of overdose and death; these risks are substantially increased in patients concurrently taking benzodiazepines and opioids, due to the exacerbated effect of respiratory suppression.(Dowell et al., 2016) Concurrent users are at least four times more likely to overdose or experience drug-related emergencies, compared to those taking an opioid alone.(Sun et al., 2017;Park et al., 2015).
To quell these effects, the Centers for Disease Control (CDC) recommends the avoidance of co-prescribing opioids and benzodiazepines, and the use of urine drug testing (UDT) to initially assess and continuously monitor patients taking either or both drugs.(Dowell et al., 2016) The Food and Drug Administration (FDA) also requires a "blackbox" warning to be included on all opioids and benzodiazepines, cautioning the risks of concurrent use.(FDA, 2016).
Trends in co-prescribing of opioids and benzodiazepines increased from early 2000 until 2012, after which two large-scale studies show a decreasing trend through 2018.(Prescription Opioid Trends in the United States, 2022; Esechie et al., 2021) Because most studies on concurrent use have relied on prescription claims or dispensing data, it is less clear whether patient use reflects the same findings.Therefore, time trends in concurrent use positivity were examined using laboratory UDT results from a large insurance claims database, from 2013 to 2019.
National overdose death involving any opioids increased approximately 72% from 2016 to 2019, (Underlying Cause of Death, 2020) most of which have been attributed to non-prescribed opioids and illicit drugs.(Kuo et al., 2021;Scholl et al., 2018) Because it is common for patients to shift from prescription opioids to non-prescribed and illicit drug use, (Muhuri et al., 2013) trends in schedule I drug positivity (illicit drugs) and non-prescribed use of opioids or benzodiazepines were also assessed.We hypothesized that concurrent use shown in UDT results would decrease from 2013 to 2019.Schedule-I drug use and nonprescribed opioids or benzodiazepines were expected to increase as a result, to compensate for prescriptions that might have been restricted or discontinued.

Data
Optum's Clinformatics Datamart (CDM) de-identified insurance claims data were used to pull medical claims and laboratory records by current procedural terminology (CPT), Healthcare Common Procedure Coding System (HCPCS), and logical observation identifiers, names, and codes (LOINC), for UDT associated with opioids, benzodiazepines or schedule I drugs (Appendix Table 1), from January 1, 2013 through December 31, 2019.This study was determined exempt by the Institutional Review Board at the University of Texas Medical Branch at Galveston.To assess yearly match rates between UDT and results, CPT/ HCPCS codes were used to represent total UDT (Appendix Table 2), while lab results were counted using drug-specific LOINCs (Appendix Table 3).Details on match rates may be found in Appendix Table 4.Because match rates were unexpectedly low, and to avoid discarding 73.8% of UDT results that were unmatched or missing a standard CPT code, data from LOINC-pulled files were used for analysis, rather than matched data.Within each drug category, only LOINC records with interpretable alphabetic or numerical results were included, such as "pos", or "<50", respectively.Overall, the LOINCs with interpretable results in each drug category were 67%, 73%, and 70% for schedule 1, opioid, and benzodiazepine, respectively (Appendix Table 5).

Outcomes
Three non-overlapping outcomes of UDT-positivity were defined: concurrent use, schedule-I use, and non-prescribed opioid and/or benzodiazepine use (misuse).Appendix Table 5 shows how the study cohort for each outcome was generated.
Concurrent use was defined as the percent of individuals with both opioid-and benzodiazepine-positive UDT on the same day, within each quarter.In a sensitivity analysis, concurrent use was also defined as concurrent-positive UDT within 3 days.Similarly, schedule-I use rates were defined as the percent of individuals with positive UDT for any schedule-I use within each quarter."Any" prescription misuse was defined as any UDT-positive rates for an opioid, benzodiazepine, or both, outside a defined compliance window each quarter, among patient records with ≥ 180 days continuous enrollment prior to the UDT.The opioid compliance window included the prescription fill date, plus days of supply, plus 7 days to account for the time opioids remain detectable in urine.The compliance period was calculated similarly for benzodiazepines, with an additional 30 days to account for long-acting benzodiazepines.In a sensitivity analysis, prescription misuse was counted only if all opioid and/or benzodiazepine use in a quarter was misuse.
The misuse rate was the percent of individuals with any misuse among those with positive opioid or benzodiazepine UDT in a quarter.The cohort flowchart for each outcome may be found in Appendix Table 5.

Covariates
Age, sex and United States Census regions were used to stratify time trends.Region was categorized by the US Census Bureau, into the Northeast, South, Midwest and West regions.Age was categorized as < 50, 50-59, 60-69, and 70 years and above.

Analysis
Quarterly rates of UDT positivity were plotted for concurrent use, schedule-I use, and misuse by age, sex, and region.Joinpoint Regression models, with a maximum of 5 possible joinpoints, were conducted to evaluate any significant changes in time trends.A sequential application of the permutation test using 4500 possible randomly permuted data sets and a Bayesian information criterion were used to determine the optimal number of joinpoints.The slopes were estimated to represent change at a constant percentage every quarter linearly on a log scale.Joinpoint Regression Program 4.9.0.0 (National Cancer Institute), and SAS 9.4 (SAS Institute v. 9.4, Cary, NC) were used for all analyses.

Concurrent use
Fig. 1 shows trends in concurrent use in the total sample (n = 746,672 UDT) and stratified by age, sex and region.Among the total sample, concurrent use rates decreased from 19.3% in Q1 2013, to 9.8% by Q4 2019.Similar patterns were mostly observed by age, sex and region, where all groups' concurrent use decreased with time.Notably, rates were higher in females than in males, lowest in age < 50, and highest in the 50-59 and 60-69 age groups.By region, concurrent use was highest in the South in 2013 (20.3%) and lowest in the Northeast (15.8%).However, by 2019, rates in the Northeast (10.9%) slightly surpassed the South (10.5%).
Table 1 presents log-scale slopes and detected joinpoints from joinpoint regression analysis of the studied outcomes.Concurrent-positive UDT showed a 2.3% quarterly decrease from Q1 2013 to Q1 2016, non-significant increase during 2016, 2.8% quarterly decrease from Q4 2016 to Q1 2019, and 9.5% quarterly decrease during 2019.In stratified analyses, joinpoints and slopes were slightly different in females compared to the entire study cohort.However, in males, only one significant joinpoint was found, showing a 1.35% quarterly decrease from Q1 2013 to Q3 2018, followed by a steeper quarterly decrease of 7.47% through Q4 2019.

Schedule I drug use
A total n = 756,258 UDT were included in the schedule I cohort.Schedule-I use generally increased from 8.9% in Q1 2013 to 13.8% in Q4 2019, with a noticeable dip between Q3 and Q4 of 2013, from 9.4% to 5.7% (Fig. 2).Higher schedule-I use was observed in males (12.4%, Q1 2013 to 17.9%, Q4 2019) than females (6.3%, Q1 2013 to 11.1%, Q4 2019).By region, lowest use was observed in the South (7%, Q1 2013 to 11.0%, Q4 2019), followed by the Midwest.The Northeast region had the highest rates, peaking at 19.4%, Q4 2019.By age, the rate of schedule-I use was highest among those < 50 (18.3%,Q4 2019).As age group increased, the magnitude of schedule-I drug use decreased; those ≥ 70 had the lowest use (5.7%, Q4 2019).Overall, schedule I drug use increased with time in all groups.

Table A1
List of drugs used to define opioid, benzodiazepine or schedule I drug UDT.
In the total sample, three significant joinpoints were detected; misuse decreased 0.35% quarterly from Q1 2013 to Q1 2016, decreased 1.5% quarterly in 2016-2017, decreased 3.5% quarterly from Q4 2017    1).The same joinpoints and similar slopes were observed in misuse rates of females.Among males, only two joinpoints were found, with a similar pattern of slopes: small quarterly decrease, followed by a larger rate of decrease, and shift to increasing misuse after Q4 2018.In age < 50, there was a 1.7% quarterly decrease in 2016-2018 and 2.5% increase from Q4 2018 to Q4 2019 in non-prescribed use, the highest slope among the age groups.Age 50-59 misuse decreased 0.7% quarterly Q1 2013 to Q2 2017, decreased 3.3% quarterly Q2 2017 to Q4 2018, and increased 2.3% quarterly Q4 2018 to Q4 2019.Age 60-69 had similar results to the 50-59 group.In age ≥ 70, significant slopes were found from Q4 2015 to Q4 2017 and from Q4 2017 to Q4 2018, with quarterly decreases of 1.5% and 3.7%, respectively.

Sensitivity analyses
When defining concurrent use to include opioid and benzodiazepine UDT dates up to three days apart, similar concurrent use rates were observed.Overall concurrent use trends were similar to those observed in Fig. 1, decreasing from 19.3% in Q1 2013 to 9.8% in Q4 2019.Joinpoint analysis also showed very similar results.

Discussion
In this retrospective study of national laboratory data from commercial insurance claims 2013 to 2019, we observed decreasing time trends in concurrent opioid-benzodiazepine use and in non-prescribed prescription drug use accompanied by substantially increasing use of schedule-I drugs.Understanding trends in concurrent drug use has generally been limited to prescription claims studies, associated with providers' behavior rather than patients.This study analyzed national laboratory results to gauge patient use and found that UDT trends are consistent with prior studies on co-prescribing, misuse and illicit drug use.(IQVIA Institute, 2022;Esechie et al., 2021;National Vital Statistics System, 2021).
The overall decreasing trend in concurrent drug use aligned with expectations from previous literature, (Esechie et al., 2021;Hwang et al., 2016;Koffel et al., 2020) which shows decreasing opioid and benzodiazepine co-prescriptions.Overall, concurrent use already decreased in 2013-2016, but decreased at a somewhat greater rate after the announcement and release of the 2016 CDC opioid-prescribing guideline, and decreased at a greater rate in 2019 after increased prevention measures were put in place by the Centers for Medicare and Medicaid Services (CMS).(Dowell et al., 2016; Announcement of Calendar Year (CY), 2019).
Among some groups, the change in concurrent use rates did not occur immediately after 2016, as seen in males, ages < 50, 50-59, ≥70, the West, and in the Midwest, which might indicate a delayed or lessened response to guidelines.Similarly, the linear decrease observed in the Northeast may reflect a lack of any response to the CDC guideline, although this region and the Midwest had smaller sample sizes, and patients of these regions receive UDT at lower rates than the South and West.(Taha et al., 2021) It is less clear why there was also a larger decline after 2018, however similar patterns were observed in national overdose deaths.(National Institute on Drug Abuse, 2021) Higher concurrent use in females was expected, as females are more likely to be prescribed opioids and benzodiazepines concurrently, (Sun et al., 2017) than males.People under 50 years had the lowest rates, while those 50-69 had the highest, consistent with a previous national study.(Vadiei and Bhattacharjee, 2020) Benzodiazepine use was also highest among 50-64 year-olds and may explain in part, the increased trends in this group.(Maust et al., 2019) The South had the highest concurrent use rates, which may be due to the increased prevalence of severe mental illness, (Lipari et al., 2013) UDT rates, and the large market share of CDM data in the South, and potentially indicate a lack of alternative treatments to co-prescribing in this region; (Sun et al., 2017;Taha et al., 2021;Vadiei and Bhattacharjee, 2020).
Schedule-I drug use increased across all ages, regions, and by sex, from Q2 2015 through Q4 2019.A sharp decline in schedule-I use was observed between the last two quarters of 2013, among females, those aged 18-59, and in the Northeast and South.One proposed explanation may be the effect of the Drug Enforcement Agency's rescheduling of hydrocodone in 2014.Although this change occurred late in 2014, the FDA recommended the reschedule months prior, in December 2013.The drop in schedule-I use of the current study coincides with the greater rate of change in opioid-prescribing observed at a similar timepoint in a previous study.(Raji et al., 2018) Schedule-I use increased at a greater rate from 2014 to 2015 among females, those aged 18-69, and in the South, then increased at a slower rate.The dramatic increase from 2014 to 2015 may be due to less accessibility to hydrocodone after 2014.

Table A3 (continued ) Schedule I drugs Benzodiazepines Opioids
Schedule-I use rates of 8.8-13.9%were generally consistent with recent CDC reports, which showed illicit drug use rates ranged from 8.1% to 23.9%, in adults aged 18 and above, by 2018.(National Center for Health Statistics, 2019) Males had higher rates of schedule-I drug use than females, which was expected.(Center for Behavioral Health Statistics and Quality, 2016) The Northeast had the highest schedule-I use of all regions, while the South had the lowest; (Illicit Drug Use, 2017) these regional differences have been observed previously, suggesting primarily illicit drugs use in the Northeast, and prescription opiate abuse in the South, shown by corresponding overdose deaths in these regions.Schedule-I use was lower in age ≥ 70 and increased as age decreased, which aligned with literature showing greater illicit use in younger individuals than older.(Nationl Survey on Drug Use and Health) Given that older individuals are up to 80% less likely to receive UDT than those under 50, it may explain why no change in illicit use was observed, and trends increased linearly in ages ≥ 70.(Taha et al., 2021).
We found 75.6% any misuse in Q1 2013 and 59.3% in Q4 2019.Previous studies have shown benzodiazepine misuse rates of 6.1-82.5%,and opioid misuse rates of 9.9-58%, though sample sizes and definitions of prescription misuse varied.(Blanco et al., 2018;Koyyalagunta et al., 2018;Hosain et al., 2019;Manthey et al., 2011) The somewhat higher rates of misuse found in this study may be due to the broader definition of non-prescribed use, which included those without a prescription, and those with prescriptions that may have been used in a manner not recommended by the provider.Additionally, misuse rates may be inflated due to selection bias of the misuse cohort; if providers sense a patient is at risk for misuse, there may be a differential in UDT requests, capturing more positive results among those more likely to misuse.(DiBenedetto et al., 2019;Hausmann et al., 2013) UDT is also more common among patients that have indications for opioids or benzodiazepine use, such as those with various chronic pain indications and psychoses.(Taha et al., 2021).
The gradual decrease in misuse from 2013 to 2017 and the steeper decrease from 2017 to 2018 generally parallels the decreasing trend in co-prescribing opioids and benzodiazepines, and may indicate the shift to opioid and benzodiazepine alternatives, gabapentinoid and SSRI/ SNRIs respectively, that provide a safer drug option, especially among the older population.(Esechie et al., 2021) However, the decrease was somewhat unexpected, given the recent increased mortality associated with prescription misuse.(Kuo et al., 2021) Beginning Q4 2018, non-prescribed use began to increase, which is consistent with the recent uptick in overdose deaths involving benzodiazepines and opioids between 2019 and 2020; (Chiappini et al., 2021) the abuse of prescribed and non-prescribed drugs has continued to increase in post-covid times due to increased "pharming", "doctor-shopping", drug shortages and healthcare strain in post-covid times, emphasizing the importance early drug abuse detection, prevention and pharmacovigilance Chiappini et al., 2022.Considering the observed trends in the context of 1) findings on increased mortality from illicit use and prescription misuse, 2) few of such patients receive drug abuse treatment, and 3) the ineffectiveness of solely restricting opioid-prescribing, highlights the need for increased access to rehabilitation facilities for treatment.These findings also support the need for targeted public health initiatives, especially in males, younger individuals, and in Northeastern regions.The focus of public health efforts may require reorientation towards patient recovery rather than provider restriction.

Limitations
There are several limitations.First, insurance claims data limit generalizability to insured individuals, and therefore misses an important population affected by the opioid crisis-uninsured individuals-, which may underestimate drug use trends.Second, reliable race/ ethnicity information is not available in CDM data and therefore could not be studied, though racial disparities in opioid compliance monitoring have been observed (Hausmann et al., 2013).Third, only independent laboratories were used, missing any tests that were done in a hospital setting.However, 116 labs were included; approximately 49% of results came from one large lab, 25% from another large lab, and 12% from a third lab.Fourth, some results were excluded for having uninterpretable values, which may have biased the sample, though it is uncertain how this could impact trends.However, the percent of annual UDT with interpretable results increased over time, indicating improvement of UDT results.Finally, match rates between procedure codes and results were unexpectedly low (Appendix Table 4).However, few UDT results included in the study were missing a CPT (<1%); instead, a nonstandard CPT code was often used, specific to benzodiazepines, opioids, and specific schedule-I drugs, though it's unknown whether these codes link to standard CPT codes in a system outside of f Although few LOINCs had a missing CPT (<1%) in each drug category, a large portion of LOINCs included in the cohorts (54.6% of schedule I UDT, 64.9% of benzodiazepine UDT, 67.5% of opioid UDT) were associated with unexpected CPT codes.These included invalid CPT codes such as '99999′ or '1111′, valid CPT codes that were broad (82540 "chemistry procedure"), or non-standard CPT values that may be laboratory-specific, such as "OPI_2" for opioids, or "BZD_2" for benzodiazepines, which led to a non-match.
CDM data.

Conclusion
Concurrent opioid and benzodiazepine use decreased 2013-2019, and at a greater rate after the 2016 CDC guideline and FDA warnings against concurrent use, and in 2019 after CMS implemented increased safety measures against concurrent use.There was an increase in schedule-I use from 2013 to 2019 and a decline in prescription misuse, which began to increase after 2018.The continued increase in schedule-I drug use, while concurrent opioid and benzodiazepine use decreased, indicates a potential shift from prescribed to illicit use and emphasizes the need to support addiction recovery programs, and focus public health interventions on patient recovery and prevention.c Any positive UDT test in a quarter was selected.For those without positive results, the first UDT per person, per quarter was selected.d Any misuse for an opioid or benzodiazepine in a quarter was selected, which may overestimate misuse.For those without any misuse, the first positive UDT per person, per quarter was selected.In the case of multiple tests where some indicated misuse and others indicated compliance, any misuse in a quarter was chosen which overestimated the rate of misuse.e Misuse was counted if all opioid and/or benzodiazepine use in a quarter was misuse, which underestimated the rate of misuse.

(
57.9%) were opioid UDT.b Tests were consolidated by UDT date; whether a patient received a panel for multiple drugs within a drug use category (schedule I, benzo, opioid), or a single test for any of these, it was counted once.The median (interquartile range) for number of UDT per person, per date increased from 3(2,4) in 2013 to 4(2,6) in 2019 in the schedule I category, decreased from 11 (4,16) in 2013 to 8(3,24) in 2019 among opioid UDT, and decreased from 6(1,8) to 1(1,12) in benzodiazepine UDT.Decreases in median tests among opioids and benzodiazepines occurred in 2015-2019, where LOINCs for panels such as "benzodiazepine panel" (LOINC 3390-2), or "opiates panel" (LOINC 3879-4) were more frequently use, than individual LOINCs for multiple drugs.

Table 1
Joinpoints in trends and slopes of quarterly concurrent use, schedule I drug use, and any non-prescribed use, 2013-2019.

Table A2
CPT and HCPCS codes used for all drug tests related to Schedule I drugs, opioids and benzodiazepines.

Table A3
LOINC codes for Opioid, Benzodiazepine and Schedule I UDT.

Table A4
CPT/HCPCS and LOINC Urine Drug Test Match rates.All records that included CPT/HCPC codes related to UDT for schedule I drugs, opioids or benzodiazepines.b CDM only includes lab results from independent laboratories, therefore tests (CPT/ HCPCS) codes were limited to those from independent labs.c Independent lab records were consolidated by person, date and CPT/HCPCS code.d Because a single CPT could be associated with many LOINCs (due to drug panels) the lab records were consolidated based on person, date, and test type (opioid, benzo, schedule 1) prior to matching.e Indicators for schedule I drug, opioid or benzodiazepine were assigned to CPT/HCPCS codes and LOINCs.Distinct patient records that had both a valid CPT/HCPCS code and LOINC on the same drug test date in each respective category were considered a match. a