Opioid overdose rates have been on the rise over the past 2 decades (
1), and several factors, such as concurrent benzodiazepine and opioid use, are associated with an increased risk of opioid overdose (
2–
4). In fact, more than 30% of opioid overdoses involve concurrent benzodiazepine use (
5,
6). As benzodiazepine and opioid prescribing rates have increased over the past 2 decades (
7,
8), so has the prevalence of fatal overdoses involving this medication combination (
5,
9). This epidemic led the Centers for Disease Control and Prevention to issue new guidelines in 2016 recommending that clinicians avoid prescribing benzodiazepines and opioids together when possible to minimize the risk of fatal overdose.
A cross-sectional study by Agarwal and Landon (
7) investigating patterns in outpatient benzodiazepine prescribing in the United States found that benzodiazepine and opioid coprescribing rates quadrupled from 2003 to 2015. However, this study only reported on predictors associated with benzodiazepine use, and it did not report on predictors of concurrent benzodiazepine and opioid use specifically. Of the studies that have reported individual-level predictors associated with concurrent benzodiazepine and opioid use (
10–
12), only two have been conducted on a national level (
12,
13). Both of these studies used the National Survey on Drug Use and Health and found that, between 2015 and 2016, 8% of adults reported using both prescription opioids and benzodiazepines in the past year (
12,
13). Certain sociodemographic characteristics, such as age, sex, marital status, education, and employment, were found to be associated with the use of these medications (
12). However, other variables that may predict concurrent benzodiazepine and opioid use, such as medical and psychiatric comorbid conditions, were not analyzed.
Methods
Study Design and Sample
This cross-sectional, retrospective study was conducted with pooled data from MEPS (2011, 2013, and 2015). Alternate years of data were used to reduce the chance of the same respondent being included more than once (
14), and multiple years of data were analyzed to achieve an adequate sample size (
15). The study sample consisted of adults (age ≥18) who did not die during the calendar year. Human subjects review was not required for this study according to the University of Arizona Institutional Review Board.
Data Source
MEPS consists of large-scale surveys of families, individuals, their medical providers, and employers in the United States (
16). Data are collected on the type, frequency, and cost of health services used by Americans via two major information sources: the household component (MEPS-HC) and the insurance component (MEPS-IC) (
16). Additionally, the MEPS-HC is supplemented or replaced with information obtained from physicians, pharmacies, home health care providers, and hospitals via the medical provider component (MPC). MEPS-HC collects the following information: demographic characteristics, health status, health conditions, medical service use, charges and source of payments, health insurance coverage, income, employment, access to care, and satisfaction with care. MPC collects data on dates of visits and services, use of services, charges and sources of payments, and diagnoses and procedure codes documented for medical encounters. The primary use of MPC is to serve as an imputation source to supplement or replace household reported expenditure information, and it is not designed to yield national estimates.
Medication use was identified by using the MEPS prescribed medicines files. The following information was collected for each medication in each round: whether any free samples of the medicine were received; health problems the medicine was prescribed for; the number of times the prescription medicine was obtained or purchased; the year and month in which the person first used the medicine; and a list of the names, addresses, and types of pharmacies that filled the household’s prescriptions (
17). Information regarding comorbid conditions was obtained from the MEPS medical conditions files (
18). For our study, we used the HC files, prescribed medicines files, and medical conditions files. All files were merged by using a unique identifier (DUPERSID). We conducted our study at the individual level (unit of analysis).
Measures
Concurrent benzodiazepine and opioid use constituted the dependent variable and was identified by Multum Lexicon therapeutic class codes. We used the third level of Multum Lexicon therapeutic class codes to identify benzodiazepines (code 69) and opioids (code 60 [narcotic analgesics] and code 191 [narcotic analgesic combinations]) (
17).
Independent variables included age (18–25, 26–34, 35–49, 50–64, ≥65); gender (female, male); race-ethnicity (white, Hispanic, black, other); marital status (married, widowed, separated-divorced, never married); region (Northeast, Midwest, West, South); insurance type (private, public, uninsured); poverty status (poor [<100% of federal poverty level], near poor [100%–<125% of federal poverty level], low income [125%–<200% of federal poverty level], middle income [200%–<400% of federal poverty level], and high income [≥400% of federal poverty level]); body mass index (underweight/normal, overweight, obese); general health (fair/poor, good, excellent/very good); mental health (fair/poor, good, excellent/very good); smoking status (current smoker, other); activities of daily living (ADL) limitations (yes, no); instrumental activities of daily living (IADL) limitations (yes, no); activities disability (yes, no); pain level (quite/extreme, little/moderate, no pain); and diagnosis of arthritis, anxiety, depression, cancer, chronic obstructive pulmonary disease, diabetes, heart disease, hypertension, hyperlipidemia, osteoporosis, and thyroid disorders.
ADL limitations were defined as MEPS participants requiring help or supervision on any ADL (bathing, dressing, eating, toileting, getting in and out of bed, and mobility into own residence). IADL limitations were defined as MEPS participants requiring assistance paying bills, using the telephone, preparing light meals, doing laundry, going shopping, or taking medications. Activities disability was identified on the basis of whether MEPS participants had any limitations with work, housework, or schoolwork. General and mental health status were determined via a survey question asking MEPS participants to rate their health as excellent, very good, good, fair, or poor (
19). A comprehensive list of definitions for the data variables described here can be found elsewhere (
19,
20).
Statistical Analysis
Chi-square tests were used to examine the differences in individual-level characteristics between the benzodiazepine and opioid, opioid only, benzodiazepine only, and no benzodiazepine or opioid groups. Multinomial logistic regression was conducted to determine the individual-level characteristics associated with the benzodiazepine and opioid use group (reference group was no benzodiazepine or opioid) among community-dwelling adults in the United States. SAS, version 9.4 was used to conduct all analyses. Complex survey structure of MEPS was adjusted with SURVEYFREQ and SURVEYLOGISTIC procedures to obtain national-level estimates.
Results
A total of 44,808 (weighted frequency=477,679,360) individuals were included in the final study sample. Sociodemographic and clinical characteristics of the benzodiazepine and opioid, opioid only, benzodiazepine only, and no benzodiazepine or opioid groups are presented in
Table 1. In total, 680 (1.6%) (weighted frequency=7,806,636) adults reported concurrent benzodiazepine and opioid use during the study period. Most MEPS participants were ages 50–64, female, white, and residing in the Southern region of the United States. Nearly half (48.1%) of the MEPS participants in the benzodiazepine and opioid group reported fair/poor general health status, whereas most participants (56.4%) in the no benzodiazepine or opioid group reported having excellent/very good general health status. Additionally, most MEPS participants (61.9%) in the benzodiazepine and opioid group reported an activities disability, which was not the case for the no benzodiazepine or opioid group (13.3%).
Findings from the multinomial logistic regression (
Table 2) revealed that several individual-level factors were significantly associated with reporting concurrent benzodiazepine and opioid use, including being white (odds ratio [OR]=1.65, 95% confidence interval [CI]=1.18–2.30), having private insurance (OR=1.68, 95% CI=1.02–2.74), having public insurance (OR=1.65, 95% CI=1.01–2.70), living in the Southern region of the United States (OR=1.71, 95% CI=1.11–2.63), having fair/poor general health (OR=1.80, 95% CI=1.16–2.79), having good general health (OR=1.59, 95% CI=1.17–2.17), having fair/poor mental health (OR=0.61, 95% CI=0.42–0.89), smoking (OR=1.71, 95% CI=1.29–2.26), having an activities disability (OR=2.68, 95% CI=1.94–3.70), having extreme pain (OR=5.11, 95% CI=2.98–8.78), having little/moderate pain (OR=1.94, 95% CI=1.22–3.09), having arthritis (OR=1.79, 95% CI=1.37–2.35), having anxiety (OR=9.61, 95% CI=7.37–12.5), and having depression (OR=1.42, 95% CI=1.05–1.92).
Discussion
Our findings estimate that during 2011, 2013, and 2015, more than seven million adults in the United States reported concurrent benzodiazepine and opioid use, and several individual-level factors were associated with an increased risk of using this combination. Many of these risk factors were modifiable, which is significant considering that the rate of prescribing this combination has quadrupled between 2003 and 2015 (
7). Additionally, our study highlights that certain treatable health conditions may be significant predictors of concurrent benzodiazepine and opioid use along with other previously identified sociodemographic characteristics (
12,
13).
The only nonmodifiable factor associated with an increased risk of concurrent benzodiazepine and opioid use was being white. This is similar to Li et al.’s (
12) study in which non-Hispanic white survey participants were more likely than other racial-ethnic groups to report concurrent benzodiazepine and opioid use. Previous studies have also found that white patients are more likely to be prescribed opioids each year (
21) and are more likely to experience opioid overdose (
22). However, patients from other racial-ethnic minority groups are less likely to receive prescription opioids than are white patients when presenting with the same symptoms (
23,
24). One could infer that this pattern puts white Americans at a greater risk of developing an opioid use disorder and subsequent overdose.
Yet, although the prevalence of deaths due to opioid overdose has been higher among whites, the rate of opioid overdose deaths has actually been higher among blacks (
25). It is therefore dangerous to label the current opioid epidemic as one predominantly affecting white communities because black individuals may simply be more likely to use illicit opioids over prescription opioids. Furthermore, racial-ethnic minority groups are often not portrayed the same as whites when discussing the tragedies surrounding addiction and overdose (
26). For example, it is possible that white Americans are more likely to use benzodiazepines and opioids together because they are told more often that ongoing use is medically indicated (even if dependency has developed) compared with members of racial-ethnic minority groups.
Of all modifiable risk factors analyzed, anxiety was associated with the highest risk of reporting concurrent benzodiazepine and opioid use. This finding may be because individuals with chronic pain often develop anxiety (
27), which benzodiazepines are commonly used to treat. However, there are safer medication alternatives for the treatment of anxiety, such as selective serotonin reuptake inhibitors (SSRIs) and serotonin and norepinephrine reuptake inhibitors (SNRIs), that do not increase the risk of respiratory depression when combined with opioids (
28). Because SSRIs and SNRIs can take weeks to months to exert their clinical effect (
29), it is not uncommon for health care providers to prescribe benzodiazepines, which provide rapid relief, adjunctively during the first few weeks of initiating an SSRI or SNRI medication (despite a lack of evidence to support this practice) (
30).
It is also not uncommon for benzodiazepines to end up being prescribed longer than the recommended 2–4 weeks (
30), indicating the difficulty that patients have in discontinuing them. This trend can be attributed to rebound anxiety symptoms that often occur with ongoing benzodiazepine use (
31). Therefore, health care providers are faced with the challenge of helping patients with anxiety achieve timely symptom relief while minimizing the risk for benzodiazepine dependence. Our study findings reveal that patients with anxiety are not only at risk of being prescribed benzodiazepines long term (
30) but are also at higher risk of reporting concurrent benzodiazepine and opioid use compared with those without anxiety. Although the opioid epidemic has led to substantial nationwide efforts on improving pain management, these findings demonstrate that a similar effort is critically needed in improving the current approach for treating anxiety disorders.
Those with extreme as well as little/moderate pain were also at higher risk of reporting concurrent benzodiazepine and opioid use compared with those with no pain. Research has shown that having chronic pain and comorbid anxiety is associated with more disabling pain compared with having chronic pain without comorbid anxiety (
32). Chronic pain is complex because it has both sensory and cognitive dimensions (
33). In fact, prior research has demonstrated that the brain regions involved in both emotional and cognitive modulation of pain are altered in those with chronic pain (
34). As a result, individuals who experience psychological distress as a result of their pain are also at increased risk for central amplification of pain (
33). Thus, although pain can lead to anxiety, anxiety can also worsen pain levels.
Furthermore, evidence suggests that benzodiazepines may increase the reinforcing and rewarding effects of opioids (
35). Therefore, individuals reporting higher pain levels may seek to amplify the effects of opioids by using them with benzodiazepines. This practice may make it even more difficult for patients to taper off benzodiazepine and opioid regimens when transitioning to alternative treatments for pain and anxiety. Although the risk of overdose with concurrent benzodiazepine and opioid use is highest during the initial few weeks of use (
9), the longer use of these medications is continued, the more slowly they must be tapered off to avoid uncomfortable and unsafe withdrawal (
36). Future research should investigate trends in the prevalence of short- versus long-term concurrent benzodiazepine and opioid use.
Those with fair/poor or good general health compared with those who reported excellent/very good health were also more likely to report concurrent benzodiazepine and opioid use. Because greater medical comorbidity burden is associated with pain and mood disturbances (
37), it makes sense that this category was associated with an increased risk of using this combination. In line with this finding, those who reported an activities disability were also more likely to report concurrent benzodiazepine and opioid use. A previous systematic review found that early opioid use for musculoskeletal disorders was associated with prolonged work disability (
38). Additionally, Gebauer et al. (
39) found that those receiving disability benefits were more likely to be continued on opioids long term and at higher morphine equivalent doses. This finding suggests that those with an activities disability are not only at increased risk of being continued on opioids but they are more likely to have prolonged disability. This outcome is significant when considering the high economic costs of pain resulting from lower economic productivity in the United States (
40).
Individuals residing in the South were more likely to report concurrent benzodiazepine and opioid use compared with other U.S. regions. This finding is in line with the study by Sun et al. (
8), who found that concurrent benzodiazepine and opioid use sharply increased between 2001 and 2013 among privately insured individuals in the United States. Because this study used MarketScan data, which are more likely to include individuals from the Southern region of the United States, the reported increase in concurrent benzodiazepine and opioid use may have also been more prevalent in this region. Prescription opioid overdose rates have also reportedly been highest in the Southern region of the United States (
41). Additionally, according to the National Survey on Drug Use and Health from 2012 to 2014 (
42), most regions with the highest estimates of serious mental illness in the United States were in the South. Prior research has shown that socioeconomically disadvantaged individuals with comorbid serious mental illness are frequently prescribed concurrent benzodiazepine and opioids and that this combination is associated with an increase in suicidal ideation (
43). This trend may be due to limited community mental health resources available for these individuals (
44), resulting in ongoing use of high-risk pharmacologic treatment strategies instead.
Previous studies have also shown that people with psychiatric disorders are at increased risk of opioid abuse and overdose (
45–
47). This finding highlights the connection between mental health and benzodiazepine and opioid use, and how it is important for policy makers to consider differences in mental health care needs across different U.S. regions to tackle this issue. A surprising finding from our study was that having fair/poor mental health compared with excellent/very good mental health was a protective factor in reporting concurrent benzodiazepine and opioid use. It is difficult to interpret why this might be, but it is possible that prescribers are more careful to avoid this medication combination in individuals with poor mental health status because they are aware that it may worsen the risk of suicidal ideation and overdose. Another possibility is that individuals reporting more positive mental health are being effectively treated with both benzodiazepines and opioids.
This study had several limitations that should be taken into consideration. The rationale for benzodiazepine and opioid use cannot be extracted from the MEPS database (e.g., insomnia vs. anxiety), and descriptions regarding which individual medications constituted the classifications “benzodiazepine” and “opioid” were not provided. Data regarding dosing frequency were not available, and reliable estimates could not be generated for whether a prescription was “as needed” because of underreporting of this variable. It also was not possible to determine whether concurrent benzodiazepine and opioid users were being tapered off their regimen or whether they may have been obtaining either medication illicitly. MEPS is a self-reported survey and as such has the possibility of recall bias. However, this recall bias is minimized by conducting the interviews regularly at an interval of 4–5 months. Lastly, the most recent MEPS data used in our study were from 2015, and it is likely that changes in benzodiazepine and opioid use patterns have occurred since then.