Predictors of Long-term Opioid Use After Elective Primary Total Shoulder Arthroplasty

Materials and Methods

Demographic Variables and Comorbidities

Demographic variables are recorded in the MarketScan database according to information from insurance enrollment. The location of residence and the location where the surgery was performed are classified by regions as defined by the US Census Bureau.⁸ For the purposes of this study, the region of residence was recorded as the primary region for analysis of long-term postoperative opioid use.

Comorbidities were defined with ICD-9 and/or Current Procedural Terminology codes (Table A, available in the online version of the article). The authors identified patients who had 1 or more outpatient visits during the 3 months before surgery for any of these comorbidities or a diagnosis of any of these comorbidities during the hospital stay after TSA.

Table A: Codes Used to Identify Comorbidities

Opioid Use

Outpatient pharmaceutical claims were identified for drugs that were classified as “opioid agonists/partial agonists” in the MarketScan database. National Drug Codes assigned by the Food and Drug Administration⁹ were used to identify individual drugs, formulations, and dosages. With methods that were similar to those used in previous studies, the authors classified patients who had filled opioid prescriptions 7 days or more before surgery as “using opioids preoperatively.”^10,11 Prescriptions that were filled within 7 days before or 7 days after surgery were assumed to be postoperative opioid prescriptions. The MMED was calculated for each individual prescription according to the formula:

Conversion to MMED enables comparison of opioid prescription strength across different formulations and daily doses. Long-term postoperative opioid use was defined as filling prescriptions totaling a 120-day or greater supply during the 3- to 12-month postoperative period.¹² Patients who had MMED greater than the 99th percentile or less than the 1st percentile of the mean MMED were excluded because these values were attributed to data entry errors.

Results

Study Population

A total of 5676 patients were included in the final cohort (Figure 1). Mean patient age was 57±5.6 years, and 42% of patients were women. Of the patients, 17% underwent reverse TSA. Mean duration of hospital stay was 1.7±1.2 days, and 81% of patients were discharged home after surgery. Additionally, 36% of patients had been taking opioids before surgery, and 79% had a prescription filled postoperatively. Patients were represented across all regions of the United States (Table 1).

Figure 1: Patient selection flowchart. Abbreviations: CPT, Current Procedural Terminology; ICD-9, International Classification of Diseases, Ninth Revision; TSA, total shoulder arthroplasty.

Table 1: Characteristics of Patients Who Underwent Elective Primary Total Shoulder Arthroplasty in the United States, 2010–2015 (N=5676)

Initial Postoperative Opioid Prescriptions

Among the opioid prescriptions that were filled postoperatively, 2005 (45%) were for short-acting hydrocodone and 1878 (42%) were for short-acting oxycodone. These two short-acting medications accounted for most opioid prescriptions. Long-acting oxycodone, short-acting hydromorphone, codeine, and long-acting morphine were prescribed less commonly (Table 2).

Table 2: Initial Postoperative Opioid Prescriptions for Patients Who Underwent Primary Total Shoulder Arthroplasty (N=4485)

Short-acting oxycodone was prescribed at a mean strength of 113±166 MMED. Short-acting hydrocodone was prescribed at a mean strength of 95±138 MMED for initial postoperative prescriptions.

Predictors of Long-term Postoperative Opioid Use

Univariate analysis of factors associated with long-term opioid use after TSA is shown in Table 3. The multivariate model was significant (P<.001), with an area under the receiver operating characteristic curve of 0.78, indicating fair predictive value (Figure 2).⁶

Table 3: Univariate Odds of Long-term Postoperative Opioid Use After Total Shoulder Arthroplasty

Figure 2: Receiver operating characteristic (ROC) curve of the logistic regression model used to predict postoperative opioid use after total shoulder arthroplasty.

After controlling for comorbidities, factors that were independently associated with long-term postoperative opioid use on multivariate analysis were preoperative opioid use (odds ratio [OR], 4.7; 95% CI, 4.0–5.5), history of drug abuse (OR, 2.5; 95% CI, 1.3–4.9), depression (OR, 1.9; 95% CI, 1.6–2.3), anxiety (OR, 1.4; 95% CI, 1.2–1.7), surgery performed in the Western United States (OR, 1.8; 95% CI, 1.4–2.3), and reverse TSA (OR, 1.5; 95% CI, 1.2–1.8). Older age was associated with lower odds of long-term opioid use (45–54 years: OR, 0.39, P=.02; 55–64 years: OR, 0.26, P<.01) (Figure 3).

Figure 3: Adjusted odds of long-term opioid use after total shoulder arthroplasty (TSA). Odds ratios (OR) control for the comorbidities shown in Table 3.

Discussion

After undergoing elective primary TSA, 16% of patients met the criteria for long-term opioid use. Patients who had been taking opioids preoperatively, had a history of drug abuse, had depression or anxiety, underwent reverse TSA, or underwent TSA in the Western United States had the highest odds of long-term postoperative opioid use.

Rates of long-term postoperative opioid use vary for different surgical procedures.¹³ Sun et al¹⁴ analyzed more than 640,000 opioid-naïve patients who underwent various surgical procedures and showed that the highest rate of long-term postoperative opioid use was 1.4%, occurring after total knee arthroplasty. They also found that patients undergoing total hip arthroplasty, open or laparoscopic cholecystectomy, open appendectomy, mastectomy, or cesarean delivery had greater odds of long-term opioid use compared with patients undergoing other procedures; however, they did not analyze the risk of long-term opioid use after TSA.¹⁴ Although the authors used the same definition for prolonged opioid use as in the current study, the incidence found (16%) in this study is much higher. One reason for this difference is that Sun et al¹⁴ analyzed only new cases of long-term use in opioid-naïve patients, whereas the current study included all patients with long-term postoperative use, including those who were taking opioids before surgery. In an analysis of more than 79,000 patients who were treated by different surgical specialists at 1 institution, Jiang et al¹⁵ found a mean rate of long-term (at least 3 months after surgery) postoperative opioid use of 9.2%, regardless of whether patients had been taking opioids preoperatively. The authors reported a range of rates by surgical specialty, from 4.4% for cardiac surgery to 23.8% for orthopedic surgery, but did not analyze specific procedures. Gil et al¹⁶ reported long-term opioid use in 8.3% of patients after arthroscopic shoulder surgery, which is lower than the rate of 16% reported in the current study, but this finding is not unexpected, given the focus on an arthroplasty population.

Patient-specific factors associated with prolonged postoperative opioid use are important to identify because many can be modified or treated before surgery. Factors such as preoperative opioid use are associated with long-term opioid use after bariatric surgery, organ transplant, spine surgery, and total hip and knee arthroplasty.^17–21 Zarling et al²¹ reported that, of 367 patients who underwent total hip or knee arthroplasty, those who had been taking opioids preoperatively had higher odds of continued opioid use 1 year after surgery compared with opioid-naïve patients. Other patient factors, such as pre-operative substance abuse, also have been associated with long-term postoperative opioid use.¹⁴ Brummett et al²² stratified more than 36,000 patients according to whether they underwent “major” or “minor” surgery. They found that rates of long-term postoperative opioid use did not differ significantly by surgery category but rather that long-term use was associated with patient-specific factors, such as smoking, drug and alcohol abuse, anxiety and mood disorders, and preoperative pain disorders.²² The authors suggested that long-term opioid use was not caused by surgical pain but rather by modifiable patient-level factors.²² This finding is consistent with the current findings, which showed that drug abuse, depression, and anxiety were associated with increased odds of long-term opioid use after primary elective TSA.

Patients who underwent reverse TSA had greater odds of long-term postoperative opioid use than those who underwent anatomic TSA. One explanation for this finding may be the different indications for anatomic and reverse TSA. Because reverse TSA is indicated for patients who have rotator cuff insufficiency or cuff tear arthropathy, many of these patients may have been taking opioids for chronic rotator cuff-related pain before surgery and therefore may be at higher risk for long-term postoperative opioid use. However, logistic regression analysis controlled for confounders when making this comparison, so the higher rate of long-term use after reverse TSA shown in this study should prompt further investigation.

Rates of opioid use and opioid-related deaths vary by US region but have increased at higher rates in areas with greater poverty and unemployment.²³ Regions such as Appalachia, New England, parts of the Midwest, and the West, including counties in northern California and southwestern Oregon, have higher rates of opioid use than other regions.²³ Geographic differences in postoperative opioid prescribing practices also have been shown. Vail et al²⁴ analyzed more than 73,000 patients who underwent surgery for lumbar spondylolisthesis and showed substantial geographic variation in postoperative opioid use, even after controlling for patient and procedure characteristics.²⁴ In their study, Arizona, Idaho, Minnesota, Nevada, New York, Oklahoma, Oregon, South Carolina, Utah, and Washington showed the highest rates of postoperative opioid use.²⁴ This finding agrees with the current study, which suggests that having surgery in the Western United States is associated with long-term opioid use after elective primary TSA and warrants further study of variations in geographic opioid prescribing patterns.

This study had several limitations. The MarketScan database includes information only for privately insured patients and does not include data for Medicare or Medicaid beneficiaries. Although the exclusion of these 2 groups limits the ability to study opioid use patterns in elderly patients and examine differences across socioeconomic classes, it produces a more homogeneous patient population for analysis. The MarketScan database does not include data on opioids that patients obtain without a prescription (eg, from friends, family, or previous supply), nor does it provide data on opioid prescriptions that were filled but not taken by the patient. The database cannot account for over-the-counter pain medications that may affect postoperative pain and the amount of opioid medication used. Additionally, the MarketScan database relies on hospital coding of diagnoses and procedures, which can be subject to coding errors. However, any coding errors that occur should be distributed evenly among patients, regardless of whether they use opioids on a long-term basis after surgery. Finally, the MarketScan database does not contain orthopedic-specific outcome measures, such as return to previous activity level or assessment of postoperative functional outcomes, but this limitation should not affect the findings on opioid use.

Conclusion

After undergoing elective primary TSA, 16% of patients had long-term opioid use. Factors associated with increased odds of long-term postoperative opioid use were preoperative opioid use, history of drug abuse, depression, anxiety, undergoing reverse TSA, or having surgery in the Western United States. Understanding risk factors for long-term opioid use after elective TSA can help in identifying modifiable factors before surgery and aid in preoperative patient counseling and postoperative pain management planning.