Association Between Preoperative Benzodiazepine Use and Postoperative Opioid Use and Health Care Costs

Key Points Question Is preoperative benzodiazepine use associated with increased postoperative opioid use and health care costs for opioid-naive patients undergoing surgery? Findings In this US national cohort study of 945 561 opioid-naive patients undergoing 1 of 11 surgical procedures, an increased risk for long-term opioid use postoperatively was noted in patients with both long-term and intermittent preoperative benzodiazepine use compared with benzodiazepine-naive patients. Some associations between preoperative benzodiazepine use and increased postoperative opioid dosages and health care costs were also observed. Meaning The findings of this study suggest that, among opioid-naive patients undergoing surgery, preoperative benzodiazepine use may increase the risk of postoperative long-term opioid use and may increase health care costs.


Introduction
Opioid use remains a public health issue in the US.Between 1999 and 2017, over 399 000 people died of opioid-related drug overdoses in the US. 1,2Since 1999, the annual rate of overdose death involving prescription opioids has increased by a factor of 5. 1 Numerous opioid-related studies have focused on the perioperative period, which is a time with inherent risk to develop long-term opioid use postoperatively. 3Opioids are frequently used to treat acute postoperative surgical pain, with more than 80% of patients being prescribed opioids after low-risk surgery. 4[7] Benzodiazepines are another commonly prescribed class of medications that carry risk of dependence and overdose despite receiving significantly less attention than opioids, 8 and the consequences of benzodiazepine use in the perioperative period and how benzodiazepine use influences postoperative opioid use are less well understood.In 2008, it was estimated that 5.2% of adults used benzodiazepines, and that use of the drugs increased with age to as high as 8.7% among adults aged 65 to 80 years. 9search has shown that concurrent use of benzodiazepines and opioids is independently associated with an increased risk for an emergency department visit due to overdose and that preoperative use of medications from both classes is associated with increased mortality after noncardiac surgery. 10,11In addition, several studies [12][13][14] have suggested an association between preoperative benzodiazepine use and an increased risk for postoperative adverse events, and one single-center study 11 from Iceland identified an association with increased postoperative opioid use.
However, the association between long-term use and amount of benzodiazepine being used preoperatively and persistent postoperative opioid use, as well as the association with postoperative health care costs, remain not well understood.Therefore, this study used a national database of US health care claims to examine whether preoperative benzodiazepine use was associated with an increased risk for long-term opioid use and increased health care costs among patients who did not use opioids preoperatively.A secondary analysis explored the association of duration and dosage of benzodiazepines with long-term postoperative opioid use.

Data Source and Population
The data consist of deidentified administrative health claims from the US (Clinformatics Data Mart Database, Optum).The database includes approximately 17 million to 19 million annual covered lives for a total of over 57 million unique lives.These administrative claims submitted for payment by providers and pharmacies are verified, adjudicated, adjusted, and deidentified before inclusion.The data for inpatient and outpatient encounters provide diagnosis codes, procedure codes, and dates of service (eTable 1 in the Supplement).Pharmacy claims data provide the generic name of the drug, fill date, quantity supplied, and number of days supplied.This study received institutional review board approval from Stanford University, which also issued a waiver of consent because the data are deidentified.This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Outcomes
The primary outcome was whether patients continued to use opioids during postoperative days 91 to 365 and, if so, the mean daily dose prescribed.Secondary outcomes included the average daily dose of opioid prescribed during postoperative days 0 to 90 and total health care costs during the first 30 days after surgery.The time windows for postoperative opioid use were chosen to differentiate between treatment of acute postoperative pain and pain that persists beyond the expected recovery period.Health care costs were calculated as the sum of the amount billed to the patient's insurer.
Each prescription was converted to an average daily dose in oral morphine equivalents (MMEs) using the prescription date, generic name, strength, number of units dispensed, and number of days supplied provided in the database. 15

Exposure
The independent variable of interest was whether the patient used benzodiazepines in the year before surgery.Patients using benzodiazepines were classified as either long-term users (defined as having Ն10 prescriptions filled or Ն120 days prescribed in the year before surgery) or intermittent users (defined as >0 and <10 prescriptions filled or >0 and <120 days prescribed).These patients were compared with patients who had no benzodiazepines prescribed in the year before surgery.
For our benzodiazepine analyses, we included prescriptions for alprazolam, chlordiazepoxide, clonazepam, clorazepate, diazepam, estazolam, flurazepam, lorazepam, oxazepam, quazepam, temazepam, and triazolam.8][19][20][21] However, the US Department of Health and Human Services Center for Substance Abuse Treatment has published a protocol that aggregates these data and provides conversions for the major oral benzodiazepines prescribed in the US. 22Using these data, we developed a tool-oral diazepam milligram equivalents (DMEs)-to approximately quantify benzodiazepine use and converted each prescription in the data to an average daily dose of benzodiazepine in DMEs (eTable 2 in the Supplement).

Variables
Additional variables were included to adjust for potential confounding.Age, sex, and year of surgery were directly extracted from the claims data.We used previously described methods 23 to identify medical comorbidities included in the Elixhauser index. 24 additional variable to adjust for anxiety disorders. 25Comorbidities were identified using diagnosis codes and were included if a relevant code appeared between when the patient was first enrolled in their insurance plan and the date of surgery.Variables were also included for each type of surgery in the analysis to adjust for the procedure being performed.

Statistical Analysis
We analyzed demographic and comorbidity data using descriptive statistics, including means and 95% CIs.To assess for differences in characteristics between patients who used benzodiazepines preoperatively and those who did not, we performed 2-sample t tests for continuous variables (eg, age) and χ 2 tests for categorical variables (eg, sex, comorbidities, and type of procedure).Findings were considered significant at P < .01.
We then estimated the association between preoperative benzodiazepine use and our outcomes using multivariable regression models, including adjustments for potential confounders of age, sex, year of surgery, type of surgery, and patient comorbidities.
For our primary outcome of opioid use in postoperative days 91 to 365, 812 549 of the patients in our sample (85.8%) did not have any opioid prescribed in this period.When there are many observations for which the dependent variable-in this case, opioid use during postoperative days 91 to 365-is zero, a simple linear regression analysis will tend to be downward-biased (ie, the estimated effect will be lower in magnitude than the true effect). 26Therefore, we performed a 2-step analysis.
The first analysis consisted of multivariable logistic regression in which the dependent variable was whether the patient used any opioid in postoperative days 91 to 365 and the 2 independent variables of interest were whether the patient was classified as having long-term or intermittent benzodiazepine use preoperatively, which estimates the association between preoperative benzodiazepine use and whether the patient subsequently used any opioid.
The second analysis consisted of a multivariable linear regression in which the dependent variable was the amount of opioid prescribed in average daily MMEs during postoperative days 91 to 365, and our independent variable of interest was whether the patient was classified as having long-term or intermittent use of benzodiazepines preoperatively.This analysis was restricted to patients who used at least some opioid during postoperative days 91 to 365 and estimated the association between preoperative benzodiazepine use and the amount of opioid used during that postoperative period.This 2-step approach has been used in other studies to obtain nonbiased estimates when many observations assume a value of 0. 27,28 For the secondary outcomes, we conducted multivariable linear regression analyses with the same set of covariates as used for the baseline model described above.The analyses were performed on January 9, 2020, using MATLAB, version R2015a (MathWorks Inc).

Sensitivity Analyses
To measure differences in our primary outcomes for each type of surgery, we performed a subgroup analysis, repeating the main analyses for each operation independently.Furthermore, because our definitions of benzodiazepine use are both arbitrary and dichotomous, we performed a sensitivity analysis to evaluate the robustness of the results using the DME technique described in the Exposure subsection and in eTable 2 in the Supplement.We calculated the average daily dose of benzodiazepines in the preoperative year for each patient in the sample and repeated the regression analyses described above for the primary and secondary outcomes, substituting the dichotomous long-term or intermittent benzodiazepine cohort variable for the average daily DME as a continuous variable.

Results
The final sample consisted of 946 561 patients; of these, 615 065 patients (65.0%) were women and 331 496 patients (35.0%) were men; mean age was 59.8 years (range, 18-89 years).The patients were undergoing 1 of 11 surgical procedures with no opioid prescriptions in the year before surgery.
Of the 946 561 patients, 71 153 patients (7.5%) were prescribed some benzodiazepine in the preoperative year, with 23 484 patients (2.5%) meeting the criteria to be classified as long-term benzodiazepine users and 47 669 patients (5.0%) classified as intermittent benzodiazepine users (eFigure 2 in the Supplement).The remaining 875 408 patients (92.5%) were classified as non-benzodiazepine users.Patients who were prescribed benzodiazepines preoperatively were older, more likely to be women, and had a higher incidence of many comorbidities.There were also differences in the percentage of benzodiazepine users undergoing several procedures compared with non-benzodiazepine users (Table 1).

Main Analysis
Before adjustment for observed confounders, opioids were more likely to be prescribed in Similarly, for patients continuing to use opioids in this period, longterm benzodiazepine users were prescribed 44% higher opioid dosages, corresponding to an increase of 0.6 (95% CI, 0.3-0.8;P < .001)average daily MMEs compared with non-benzodiazepine users.No significant difference was observed between intermittent and non-benzodiazepine users (difference of 0.0 average daily MMEs, 95% CI, −0.2 to 0.2; P = .65MMEs) (Table 2).

Sensitivity Analyses
A subgroup analysis stratifying patients by type of surgery reinforced our findings for increased odds of continued opioid use in postoperative days 91 to 365 among patients using benzodiazepines preoperatively (Table 3).However, among patients continuing to use opioids in this period, there were not strong associations between preoperative benzodiazepine use and postoperative opioid doses within each type of surgery.
Substitution of the long-term or intermittent benzodiazepine cohort variable with the average daily dose of benzodiazepine used in the preoperative year in DMEs supported robustness of our primary and secondary outcomes.Significant associations were observed between higher preoperative benzodiazepine doses and increased odds of continuing to use opioids in postoperative days 91 to 365 (OR, 1.36 for a 10-mg increase in average daily DMEs; 95% CI, well as increased opioid dosages among patients continuing to use opioids in this period (0.7 average daily MMEs per 10 mg of average daily DMEs; 95% CI, 0.5-0.9MMEs; P < .001).
For our secondary outcomes, there was an association between the amount of benzodiazepines used preoperatively and an increase in the amount of opioids prescribed in postoperative days 0 to a Overall, patients with preoperative benzodiazepine use were older, more likely to be women, and had a higher incidence of many comorbidities.
b P values reflect the comparison between their benzodiazepine cohort and benzodiazepine-naive patients and were computed using χ 2 for categorical variables and 2-sample t tests for continuous variables..65

Secondary outcomes
Average daily opioid prescribed in postoperative days 0-90, MME (95% CI) c Unadjusted 5.9 (5.9 to 5.9) 7.8 (7.5 to 8.0) Difference, 1.9 (1.6 to 2.1) <.001 6.4 (6.2 to 6.5) Difference, 0.5 (0.4 to 0.6) <.001 Adjusted b 5.9 (5.9 to 5.9) 6.7 (6.5 to 6.9) Difference, 0.8 (0.6 to 0.9) <.001 a Both long-term and intermittent benzodiazepine use were associated with an increased likelihood to continue to use opioids after surgery and higher opioid dose requirements in the immediate postoperative period.Long-term benzodiazepine use was also associated with increased opioid doses beyond the immediate postoperative period.Intermittent benzodiazepine use was associated with increased 30-day health care costs.
b Results were adjusted for age, sex, type and year of surgery, and medical comorbidities using regression modeling.
c Patients who were not prescribed any opioid in the referenced time period were excluded from the analysis to prevent downward biasing of the results.

Discussion
In this retrospective analysis of 946 561 opioid-naive patients aged 18 to 89 years undergoing 1 of 11 surgical procedures, long-term or intermittent benzodiazepine use in the year before surgery was associated with increased odds of continuing to use opioids in postoperative days 91 to 365.Among patients continuing to use opioids in this period, long-term preoperative benzodiazepine use was associated with an increase in the amount of opioid prescribed compared with patients who did not use benzodiazepines preoperatively, although the difference in doses was small and no significant difference was observed for intermittent benzodiazepine use, suggesting limited clinical significance.
A secondary analysis that directly quantified preoperative benzodiazepine dosages also showed a positive association between the dose of benzodiazepine used preoperatively and the amount of opioids prescribed in postoperative days 91 to 365.
For our secondary outcomes, preoperative benzodiazepine use was associated with a small increase in the amount of opioids prescribed in postoperative days 0 to 90 among patients who used any opioid in this period compared with benzodiazepine-naive patients.An association between intermittent preoperative benzodiazepine use and a small increase in total health care costs during the first 30 days postoperatively was also observed.However, this association was not found for patients who used benzodiazepines over a long period, nor was this association replicated in our sensitivity analysis.
Our analysis adjusted for multiple potential demographic and medical confounders, including comorbidities related to benzodiazepine use or misuse.The results were robust in a sensitivity analysis that quantified preoperative benzodiazepine use in DMEs rather than including number of prescriptions or prescription-days as a surrogate for benzodiazepine use in the main analysis.
Recent studies have suggested that preoperative benzodiazepine use is associated with worsened surgical outcomes. 11,12One study used a nationwide sample from the US with the same primary data source as our study and suggested that preoperative benzodiazepine use is associated with increased postoperative adverse events. 12Another single-center study based in Iceland also demonstrated an increased risk for postoperative opioid use and dose-dependent effect on the risk of adverse events associated with preoperative benzodiazepine use. 11However, the generalizability of the second study to the US population is unclear given lower rates of comorbidities and lower mean age in their sample, and potential differences in practice habits related to narcotic prescribing between the 2 countries.
To our knowledge, no previous study has assessed how preoperative benzodiazepine use relates to the duration and amount of opioids used, as well as health care costs in the perioperative period.Our study expands on previous work by suggesting an association between the intensity of preoperative benzodiazepine use and an increase in both probability of continuing to use opioids long term and the opioid dose being prescribed.This finding may suggest overlap in the exposures and triggers leading to psychotropic drug dependence.However, the economic implications of preoperative benzodiazepine use may be limited, as the association between preoperative benzodiazepine use and increased postoperative costs was small and inconsistent.

Limitations
This study has several limitations.First, we found many differences in the sample characteristics between patients who used benzodiazepines preoperatively and those who did not: patients using benzodiazepines were older, had a substantially higher comorbidity burden, and underwent the included procedures at different rates.While some differences are expected because the comorbidity is directly associated with benzodiazepine use or misuse (eg, neurologic, psychiatric, and substance use disorders), differences were also found for unrelated conditions.This finding implies that patients using benzodiazepines are sicker and that, despite adjusting for numerous potential observable confounders, the possibility of additional hidden confounders associated with benzodiazepine use is substantial.The limited nature of claims data also prevents adjustment for a The robustness of the main results of this study was evaluated by measuring the association between the specific dose of benzodiazepine being used preoperatively in DMEs and our outcomes.This sensitivity analysis was mostly consistent with the primary analysis, finding direct associations between preoperative benzodiazepine dose and an increased likelihood of persistent opioid use as well as increased doses of opioids prescribed both in the immediate postoperative period and long term.However, no association between the dose of benzodiazepine and 30-day postoperative health care costs was observed.

Table 1 .
Sample Characteristics a (continued)

Table 2 .
Outcomes a

Table 3 .
Subgroup Analysis of Primary Outcomes by Type of Procedure a Association Between Preoperative Benzodiazepine and Postoperative Opioid Use and Health Care Costs Abbreviations: NA, not applicable; OR, odds ratio.aIncreased odds of opioid use in postoperative days 91 to 365 for both long-term and intermittent preoperative benzodiazepine users across most surgery types in our sample.However, additional associations between preoperative benzodiazepine use and the dose of postoperative opioid used in the same period was not found.b P < .001.c P < .01. d Regression models were unable to converge to generate coefficients for the variables of interest.JAMA Network Open | Anesthesiology JAMA Network Open.2020;3(10):e2018761. doi:10.1001/jamanetworkopen.2020.18761(Reprinted) October 27, 2020 8/12 Downloaded From: https://jamanetwork.com/ on 05/05/2021

Table 4 .
Sensitivity Analysis Results a