Using a 20% representative Medicare sample, we conducted a retrospective analysis of muscle relaxant prescribing to patients ≥ 65 years of age. For the 5% and 20% samples, beneficiaries are selected for inclusion in the database on the basis of the last two digits of their health insurance claim number (which, in the vast majority of cases, is their social security number). Information on all beneficiaries included in the 5 or 20% Standard Analytic Files (SAF) are provided for all years for which they have received Medicare benefits (until death or disenrollment) within the time period included [3]. We merged patient data from Medicare Carrier, MedPAR, and Outpatient Files with Medicare Part D for the years 2013–2018. We used the Master Beneficiary Summary File (MBSF) base file to determine cohort composition regarding age, gender, race, and comorbidity score. We identified both hospitalizations and fee-for-service claims at free-standing ambulatory surgical centers for specifically identified procedures. We tracked both muscle relaxants and postoperative opioid prescribing over time including prescription strength and quantity.

We included patients undergoing one of the 14 most common non-cataract surgeries performed in older adults (eAppendix 2) [8, 9], who were ≥ 66 years at time of the procedure to allow for a prior year for calculating comorbidities using the Charlson comorbidity score [10], calculated using an updated 17-disease version for use in administrative databases [11]. We only included patients who had at least one prescription filled in Medicare Part D 3 months prior to surgery and who had continuous Part D coverage for 3 months before and 6 months after the procedure date to ensure they were using Part D. For patients who had multiple procedures over the time period, we included only their most recent procedure. We excluded patients whose discharge disposition was death or hospice [12], who died within 30 days after discharge, and who had ≥ 3 procedures on the same day (eAppendix 3).

The 14 surgical procedures included represent a wide range of anatomic regions and specialties. We defined inpatient procedures using ICD9-CM or ICD10-PCS codes and outpatient procedures using HCPCS/CPT codes (eAppendix 4). We created specific groupings for procedures commonly performed together. We defined race using the Research Triangle Institute (RTI) race code, which is an algorithm providing an expanded definition of race to the Medicare data [13].

We defined patients who had a new postoperative prescription for a muscle relaxant at the time of surgery, which excluded patients already on muscle relaxants prescribed in the 3 months prior to surgery (excluding the 7 days prior to surgery) [8]. We linked the NDC codes from Part D claims with Medispan crosswalk files to identify generic drug names and prescription information. We considered a postoperative prescription as any fill between 7 days before and 7 days after the surgery (or discharge for inpatients) [14, 15], as some surgical practices prescribe medications preoperatively so that patients can have the medication ready at home. We created a variable to assess chronic preoperative opioid use, including any patient who received a prescription of opioids for either 60 continuous days before surgery or had filled three or more prescriptions of at least 28 days duration for opioids within 180 days before surgery. We excluded patients discharged to SNF as prescribing information in SNFs is unavailable in Medicare data.

Our primary outcome was fills of postoperative muscle relaxant prescribing (eAppendix 5). We then evaluated fills of postoperative muscle relaxant prescribing in several ways. First, we identified associations with muscle relaxants by defining which types of patients and which procedures most commonly had fills of muscle relaxants prescribed postoperatively. Then, we analyzed the rate of procedural prescribing of muscle relaxants after surgery over time. Of note, we also evaluated the prevalence of prolonged use of fills of muscle relaxants, which we defined as a prescription refilled at 90–180 days after discharge from surgery, a time period based on definitions of prolonged use of opioid after surgical procedures [8, 16, 17].

We also assessed opioid prescribing as the premise that postoperative muscle relaxant use could decrease the need for opioids [4]. We measured opioid prescribing (eAppendix 6) using the same methods we used for muscle relaxant prescribing. We also evaluated concomitant prescribing of opioids in the postoperative period, as well as oral morphine equivalents (OME) to assess overall trends of the amount of opioid prescribing. OME is used as a tool to compare the amount of different opioids using an equianalgesic dose chart to calculate opioid dosage in a consistent and systematic way [18].

To identify associations with overall prescribing, we constructed logistic regression models adjusted for procedure characteristics (surgery type), patient characteristics (age, sex, race/ethnicity, Charlson comorbidity score), length of stay, disposition location, and care complexity [19,20,21] (number of physicians seen in prior 6 months). We defined which procedures most commonly had muscle relaxants prescribed postoperatively, and the unadjusted risk of prescribing for each medication category and for type of surgery. Additionally, we assessed concomitant prolonged use of opioids since that can increase the risk of adverse drug events.

We then analyzed prescribing trends over time. We adjusted for age, race, gender, and type of procedure. We analyzed the proportion of postoperative prescribing of new muscle relaxants, opioids and average OME across each year from 2014 to 2018. We used 2013 to calculate comorbidities for people in the first cohort year. To analyze overall prescribing trend over year, we constructed multivariate logistic regression models for muscle relaxant (MR) and opioid prescribing, including procedure year as a categorical variable and adjusting for age, sex, race and ethnicity, and procedure types. Linear trends in the log odds of both MR and opioid prescribing over time were analyzed by comparing a linear contrast of regression coefficients across all levels of procedure year to the null value zero. We also analyzed trends over time by each procedure group by collapsing procedures into similar groups including laparoscopic, open, orthopedic, spine, and vascular (eAppendix 1).

To assess risk factors for prolonged use, we first defined the unadjusted risk of prolonged use for each medication category and for type of surgery. We then constructed logistic regression models, adjusted for age, gender, race, facility type, and procedure type. We managed competing risks through a descriptive model as the number of deaths within 30 days was too small for a Fine-Gray calculation.

Finally, we performed a subanalysis of each step above after removing the antispastics, as the Beers criterion does not apply to muscle relaxants typically used for the management of spasticity (i.e., baclofen and tizanidine) instead of pain, although these drugs are in fact also used for postoperative pain and can also cause substantial adverse effects [2]. Therefore, we aimed to assess whether there was a difference between muscle relaxant use as a whole with skeletal muscle relaxants specifically.

We conducted analyses using SAS 9.4 and Stata 17, and plots were generated with R. Data were last extracted on 2/26/2024. This manuscript complies with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies (e Appendix2) [7]. The study was approved by the University of California San Francisco Institutional Review Board.