Study design and participants
The study was nested within a prospective cohort study capturing pre-operative, acute-care and longer-term data from patients undergoing primary TKA or THA for osteoarthritis with full methods described elsewhere [20]. The larger study involved a part-random, part-convenience sample of 19 high-volume sites (defined as performing over 275 knee or hip arthroplasties annually) from five Australian States. The inclusion criteria for participants in the study were: consenting adults (over 18 years) with a primary diagnosis of osteoarthritis undergoing primary TKR or THR; sufficient English to comprehend the protocol; and available to participate in follow-up for 12 months.
Ethical approval and consent to participate
The study was approved by the following Human Research Ethics Committees (HREC) prior to study commencement: Hunter New England HREC (NSW); St Vincent’s Health and Aged Care HREC (Queensland); Austin Health HREC (Victoria); Barwon Health HREC (Victoria); Epworth HREC (Victoria); Calvary Health Care Clinical and Research Ethics Committee (Tasmania) and; Calvary Healthcare Adelaide HREC (South Australia). Prior to commencement the study protocol was registered on the 15 July 2013 (NCT01899443) [21]. All participants provided written, informed consent and all methods were performed in accordance with the relevant guidelines and regulations.
Baseline data and post-operative outcomes
A consecutive series of eligible people who provided informed consent to participate in this prospective observational study provided basic demographic, sociodemographic, comorbid and analgesia usage data during a pre-admission visit or telephone call 2–6 weeks prior to surgery. Pre-operative opioid use for any reason was also obtained from the medical record. Participants also completed patient-reported outcomes measures (PROMs). The ASA classification of physical health was used to grade comorbid pre-operative health, the ASA score is a subjective assessment of overall health based on five classes (I. healthy fit patient; II. mild systemic disease; III. severe systemic disease that is not incapacitating; IV incapacitating disease with constant threat to life; V. moribund, not expected to live 24 h with or without surgery) [22]. Acute-care outcomes such as length of stay and inpatient rehabilitation were provided by the sites using a standardised pro-forma and verified by site audit at a later time. Inpatient rehabilitation includes physiotherapy, occupational therapy, and speech/language therapy, delivered under the supervision of professionally qualified allied health staff. Inpatients receive a minimum of three hours per day for who have the capacity to tolerate this amount of therapy which should occur on a minimum of five days per week [23]. Participant ineligibility included revision surgery, documented dementia, rheumatoid arthritis in the joint being replaced and another arthroplasty planned within three months of the index surgery.
Thirty-five- and ninety-day post-operative opioid use and PROM data were obtained via telephone follow-up by trained study personnel not involved in care delivery. The Oxford Knee or Hip Score (OKS, OHS) [24] and the EuroQol survey (EQ-5D-5L) [25] were used to measure joint-specific impairment and health-related quality of life, respectively. The OKS and OHS comprise 12 joint-specific Likert-style questions each concerning pain and functional impairment over the last four weeks. Each item is scored from zero (maximal discomfort/pain or maximal impairment) to four (no pain/discomfort or impairment), providing a total score out of 48 with higher scores indicating better joint status.
The EuroQol (EQ-5D-5L) provides a simple descriptive profile in five health dimensions (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression) and a single index value for health status. Each question is scored as a single number 1 (no problems), 2 (some problems) 3 (moderate problems), 4 (severe problems), 5 (extreme problems). A 0–100 VAS scale for perceived overall health is also part of the tool: 0 indicating worst health, and 100 indicating best health imaginable.
Data definitions
An opioid analgesic was defined as codeine 15 mg or over, buprenorphine, dextropropoxyphene, fentanyl, hydromorphone, methadone, oxycodone, or tramadol. A comorbidity was defined as the presence of one or more additional disorders including heart disease, stroke, high blood pressure, high cholesterol, cardiac disease, kidney disease, liver disease, diabetes, cancer, lung disease, depression, anxiety or mental health disease, bleeding disorders, gastroesophageal reflux disease, sleep apnoea, neurological disease, parkinson’s disease, co-occurring with the primary disease (OA), diagnosed by a doctor and requiring daily medications. The presence of any comorbid condition was treated as binary (Yes/No).
A major complication was defined as experiencing at least one of any of the following events: pulmonary emboli, deep venous thrombosis, deep surgical site infection, revision, reoperation, joint fracture, dislocation, manipulation under anaesthesia, stroke, acute myocardial infarction and/or re-admission. Major complications were grouped, based on time of occurrence, as acute (occurring during hospital admission) or post-discharge (occurring after discharge up until 90 days post-surgery).
Statistical analysis
Data were analysed using SAS Enterprise Guide version 8.2. Baseline characteristics were compared for those taking opioids at 90 days post-operatively (the opioid group) and those not taking opioids (the non-opioid group). The relationship between collected or measured variables and opioid grouping were assessed in univariate logistic regression. The exposure of interest was inpatient rehabilitation or not, while all other variables were chosen as being known or suspected predictors of ongoing opioid use. A backward-elimination model was utilised which started with all explanatory variables and then discarded the least statistically significant variables, one by one. The final model was chosen when each variable remaining in the equation was statistically significant. Some variables were collapsed prior to their addition into the model, specifically, education and smoking status. Interactions between inpatient rehabilitation with acute- major complications, and inpatient rehabilitation with post-major complications were considered in the models. A sensitivity analysis was performed for the opioid naïve patients (non-opioid users pre-operatively). Sample size was dictated by the original study hence no a priori calculation was performed. Odds ratios (ORs) were calculated with the 95% confidence interval (95% CI). A p-value < 0.05 was considered significant. Patients with any missing predictor data were removed from analysis.