The study received ethics approval from the Sydney Local Health District (Camperdown, New South Wales, Australia; Reference #, X20-0401). This manuscript was written in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist.11

We conducted a prospective observational cohort study conducted on opioid-naïve patients undergoing cancer-related surgery who were prescribed opioids at the time of discharge for acute pain management. Specifically, patients with three types of cancer (breast, gynecologic, or head and neck) were studied. The recruitment period was 18 months, from June 2021 to January 2023.

The literature reported a broad, divided incidence of overprescribing ranging from 42% to 71% in one to four weeks after discharge.12 The definition of overprescribing is controversial; however, it is commonly accepted as a prescription that exceeds the anticipated need after discharge.13 For this study, we defined overprescribing as prescriptions with 50% or more unused opioids within the first seven days after discharge. The period of seven days aligned with the initial opioid analgesic prescription duration in primary care settings.14 To reflect the current standard in the estimation of opioid prescription for discharge, the equivalent days of opioid supply were calculated as below.15 This was based on the total amount of oMEDD prescribed divided by the patients’ 24-hr opioid use immediately before discharge:

$$Equivalent \,days \,of\, opioid\, supplied= \frac\, hr\, prior\, to \,discharge\, (mg/day)}$$

The study site is a 140-bed quaternary cancer referral hospital, specializing in cancer surgery, oncological and radiation treatments, and cancer research. The Chris O’Brien Lifehouse Hospital (COBLH) is situated in Sydney, Australia. The staff are providing cancer-related surgeries for patients residing both locally and in regional and rural areas of New South Wales. The COBLH Pharmacy Department employs a part-time opioid stewardship pharmacist, whose responsibility includes staff education, reviewing hospital opioid use policy, and monitoring opioid prescriptions and dispensing patterns. Using these data, we identified the top three surgical specialties that prescribed the most opioids at the time of discharge for their patients. These were the gynecology teams, breast surgery teams, and the head and neck surgical teams.

Participants were eligible to be enrolled in the study if they were admitted to COBLH for more than 24 hr and admitted under the care of the three targeted surgical specialities. Clinical nurse specialists used a convenience sampling method and informed the admitted patients about our study, after which the patients were recruited by the research staff. Written informed consent was obtained from all participants. The participant inclusion criteria were: 1) proficient in English with age ≥ 18 yr, 2) opioid-naïve, defined as having history of opioid use ≤ 30 mg oMEDD for less than seven days in the three months before surgery, 3) prescribed opioids at the time of discharge, 4) the intention of surgery was cancer-related and curative, 5) minimum postoperative stay of 24 hr, 6) admission for surgery for gynecologic, breast, or head and neck cancer, and 7) reachable via telephone for follow-up. The three specialties were selected based on an internal audit using the Merlin® pharmacy dispensing software (Pharmhos Inc., Melbourne, VIC, Australia), which identified them as having the highest average monthly opioid discharge prescriptions in the hospital. Patients were excluded if they were opioid-dependent with daily use > 30 mg oMEDD for seven days or more before surgery or if the surgery was for palliative intent.

Before data were accessed, a data analysis and statistical plan was written and filed with the COBLH research governance board. Data collected included participant demographic data (age, gender, body mass index [BMI]), total number of psychological comorbidities (including attention-deficit/hyperactivity disorder, obsessive-compulsive disorder, bipolar, major depression, anxiety, schizophrenia, and pre-existing chronic pain), cancer information (cancer type, clinical staging, cancer recurrence, other cancer treatments), surgical information (types of surgery, surgical intent, length of stay, regional anesthetic blocks), previous and current opioid use (amount of 24-hr inpatient opioid use before discharge, time of last opioid dose, presence of opioid-related adverse events), discharge opioid prescriptions (name, strength, and quantity), and patient-reported pain scores using the Brief Pain Inventory Short Form (BPI).16 The BPI assesses the severity of pain and the interference to daily functionality due to pain. The instrument is reported on a 0–10 numerical rating scale, with each scale of measure reported as the average of different conditions, thereby providing a more comprehensive measure of pain than would be expected if using just one scale of measure. The pain interference scale determines the impact that pain has on the patient’s daily activities and has been recommended by international consensus as a core outcome measure for overall quality of life.17

The equivalent days of opioids supplied were calculated using the total amount of opioids supplied and the amount of opioids consumed 24 hr before discharge. Patients were also asked about their understanding of any surgery-specific advice on pain management for self-care at home. Patient demographics, cancer information, surgical information perioperative anesthetic blocks, and discharge opioid prescription information were collected through the Meditech® electronic health record (Medical Information Technology Inc., Canton, MA, USA). Bedside medication charts were used to verify the details of medication use, including nonopioid analgesics and supplements to ease the opioid-related adverse effects. All initial patient-reported outcomes were measured at the time of immediate discharge.

After discharge, participant follow-up was conducted on day 7 via telephone using a structured proforma. Participants were asked to identify the last date and amount of opioid dose taken. All opioids were converted to the oMEDD using the opioid conversion calculator developed by the Faculty of Pain Medicine of the Australian and New Zealand College of Anaesthetists.18 Secondary outcome measures included self-reported pain score on the numerical rating scale, use of other pharmacologic analgesics, presence of opioid-related adverse events, and number of urgent medical visits for prescription.19 We used Research Electronic Data Capture (REDCap®, Vanderbilt University, Nashville, TN, USA) to capture and store all data electronically.

To reduce potential researcher bias, the interview transcripts were created for each session to ensure consistency between the researchers in the team. Validated questionnaires were chosen to assess pain and daily interferences at the time of discharge. All participants completed an expression of interest form before being approached for full consent. While surgical teams were aware of the study, they were not informed of the specific variables collected. Participants were informed of the observational nature of the study to minimize the subconscious influences of their reports of opioid use after discharge.

We calculated the sample size to identify a prevalence of overprescribing of opioids by 53%, with a power of 0.80 and an alpha of 0.05. Accounting for a 20% attrition rate, a sample size of 102 was required. This sample size was informed by a previous study that showed 53% of patients receiving short-stay breast surgeries were overprescribed opioids before discharge.20 Normally distributed continuous variables were compared using independent t tests and nonnormally distributed data were analyzed using Mann–Whitney U test. Categorical variables were compared using the Chi square test. Multivariable analyses of predictors of opioid overprescribing were performed using Poisson regression, with opioid use in the 24 hr before discharge and the equivalent days of opioid supplied as the independent variables and the clinically significant variables such as age, BMI, and length of stay as covariates. A P value < 0.05 was considered statistically significant. Data were analyzed using IBM SPSS Statistics for Windows version 26 software (IBM Corp., Armonk, NY, USA).