2.1 Eligibility

The review included original, peer-reviewed articles. Participants were residents of Australian RACFs. Residential aged care facilities are synonymous with ‘nursing homes’ and ‘long-term care facilities’ in other countries. In Australia, analgesics are predominantly prescribed by general medical practitioners (GPs), dispensed by off-site community pharmacies and administered by qualified nurses.¹⁷ Included studies reported prescribing, dispensing or administration of overall regular or overall PRN opioid use and/or specific opioid medications. Studies utilising the same study sample were included where different characteristics of opioid use were reported. Specific opioids were defined as buprenorphine, codeine, fentanyl, hydromorphone, methadone, morphine, oxycodone, oxycodone and naloxone, pethidine, tapentadol and tramadol. There were no restrictions on study design. Studies that reported opioid initiation only were excluded. Studies that focused explicitly on cancer treatment and/or palliative care were excluded.

2.2 Search strategy

MEDLINE, Embase, CINAHL, AgeLine, Web of Science Core Collection, InformIT and International Pharmaceutical Abstracts were searched for studies published from inception to September 2021. Google, the Royal Commission into Aged Care Quality and Safety and Analysis & Policy Observatory, and reference lists of included studies were searched. The complete search strategy is shown in Appendix 1. When relevant conference abstracts were identified, attempts were made to identify potential full-text articles.

2.3 Study selection and data extraction

The title and abstract screen, followed by a full-text screen, were performed independently by two review authors (author 1 and author 2 or author 3). Any discrepancies were discussed until consensus was reached. A standardised data extraction tool was developed and pre-piloted based on the ‘Joanna Briggs Institute (JBI) guidance for systematic reviews of prevalence and incidence studies’.¹⁸ If prevalence studies also reported elements of opioid appropriateness, this information was extracted. Criteria for assessing appropriateness were not prespecified but included analgesic co-prescription, concomitant use of other CNS-active drugs, ADEs and opioid dose.^{5, 19, 20} Data extraction was performed independently by a team of review contributors, then validated independently by two review authors (author 1 and author 2). Attempts were made to contact authors for additional information when needed.

2.4 Data synthesis

Results were synthesised for regular or PRN use, data source (i.e. prescribing, dispensing or administration records) and data collection period. For the purpose of the review, opioid prescribing was considered synonymous with charting. For studies reporting opioid prevalence based on Residential Medication Management Review (RMMR) records, this was considered under prescribing. Opioid use was calculated as the number of residents charted, dispensed or administered regular or PRN opioids as the numerator and the study sample as the denominator. Percentages were adjusted to whole numbers. For randomised controlled trials (RCTs), baseline data were used in the synthesis.

2.5 Quality assessment

Three review authors (author 1, author 2 and author 5) independently assessed the risk of bias using an adapted version of the ‘JBI Checklist for Prevalence Studies’ (Appendix 2, Table S3).²¹ The checklist was adapted to appraise the quality of opioid assessment in each study (i.e. not the overall quality of the study), irrespective of whether the studies were opioid-related. Potential biases were assessed across seven criteria relating to sampling and opioid prevalence measurement. Sampling was considered appropriate if the study reported characteristics and demographics representative of the Australian RACF population. The measurement of opioid prevalence was considered to have a low risk of bias if the authors differentiated between regular and PRN use, clearly described how opioid use was measured (e.g. training or experience of those extracting medication data) and clearly reported the numerator and denominator.

3.1 Characteristics

The 23 included studies involved 286,141 residents (Table 1, Table S1). The mean/median age of residents ranged from 81 to 91 years, and the proportion of female residents ranged from 58% to 78%. Study sample sizes ranged from 43 to 203,894.^{22, 23} Five^{8, 24-27} and two studies^{28, 29} reported opioid prevalence utilising the same study sample. One RCT was included,²² and the remaining studies were observational. Of 22 observational studies, 18 reported the cross-sectional prevalence of opioid use across various assessment periods.^{8, 23-39} One study combined a baseline cross-sectional and 12-month longitudinal audit.⁴⁰ Three studies investigated period prevalence using a cohort design over 12 months,⁴¹ 24 months⁴² and 41 months.⁴³ The majority of the included studies (18/23) were published in 2015 or later.^{8, 22-27, 29-32, 34, 35, 38-41, 43} Data were sourced from South Australia,^{8, 24-27} New South Wales,^{33, 37, 42} Victoria,^{31, 40} Queensland,²² Tasmania³⁹ and Western Australia.³⁰ Of 23 studies, 16 reported overall opioid prevalence (Table 1)^{8, 22-25, 27, 28, 30, 31, 33, 34, 37-41} and 14 reported prevalence of specific opioids (‘specific opioid classes’, Table S2).^{8, 24, 26-30, 32, 36, 39-43} Definitions of opioid prevalence are provided in Table 1 and Table S1.

3.2 Prevalence of pain

One study reported that 28% of residents self-reported pain, and of these residents reporting pain, 32% were prescribed opioids.³³ Four studies reported the prevalence of pain and opioid use separately. In terms of clinician-observed pain, one study reported 42% of residents were estimated to have moderate pain on the nurse-estimated pain scale (0–10), with a mean score of 3.2 (±2.2).²² This study reported 28% of residents and 42% of residents were prescribed regular and/or PRN opioids at baseline, respectively. Veal et al.³⁹ reported 26% of residents were observed to experience pain or reported pain during the preceding seven days. This study reported 29% of residents and 35% of residents were prescribed regular and PRN opioids at baseline, respectively. Two studies (utilising the same study sample) reported 30% of residents were observed to be in pain on the Pain Assessment in Advanced Dementia Scale (PAINAD), while 60% of residents self-reported pain.^{8, 27} These studies reported 29% of residents were administered an opioid over a 24-hour (24 h) period.

3.3 Prescribing

Five studies reported prescribing of regular opioids as per medical or RMMR records, ranging from 28% to 34% across various data collection periods. Two studies reported 29%³⁹ and 30%²⁵ of residents had a regular opioid charted over 24 h. Three studies reported 28% over 1 week,²² 28% over 30 months²⁸ and 34% over 1 month had a regular opioid charted.³¹ No studies analysed GP prescribing records to determine opioid prevalence.

Five studies reported prescribing of PRN opioids, ranging from 11% to 42% across various data collection periods. Of 7309 residents indicated for a RMMR, 11% had a PRN opioid charted as recorded in the Medscope™ database.²⁸ The dose and frequency of PRN opioid administration was not specified in the database. Three studies reported 30%,⁴⁰ 35%³⁹ and 37%²⁵ of residents were charted a PRN opioid over a 24-h data collection period. A pilot RCT reported 42% of 43 residents were charted a PRN opioid at baseline.²² One study did not specify whether opioids were charted for regular or PRN use. In this study, 32% were charted an opioid over a seven-day data collection period.³³

3.4 Dispensing

Three studies reported opioid prevalence based on pharmacy dispensing data, ranging from 27% to 50% over 12 months.^{23, 38, 41} It was not possible to determine whether dispensed opioids were intended for regular or PRN administration in these studies. The largest nationwide study (n = 203,894 long-term residents) reported the prevalence of chronic opioid use was 27% in 2016. This study analysed Pharmaceutical Benefits Scheme (PBS) dispensing data from the Registry of Senior Australians (ROSA).²³ Taxis et al.³⁸ reported 43% of 1560 residents were dispensed an opioid in 2009. This study did not specify what type of database was used for data collection. The highest overall annual opioid prevalence was 50% among residents in the Australian Government Department of Veterans’ Affairs (DVA) dispensing claims database from July 2015 to June 2016.⁴¹

3.5 Administration

The lowest reported prevalence was 4% of 3054 residents administered regular opioids over two weeks from data collected in 2003.³⁷ Five studies reported the administration of regular and PRN opioids over 24 h. Two of which extracted data from an electronic medication management system and reported 21% of residents were administered opioids.^{30, 34} Three studies that analysed administration data for 383 residents of six RACFs reported 29% were administered an opioid.^{8, 24, 27}

Three studies reported the administration of PRN opioids and compared this with prescribing at baseline. In two studies, 37% of 242 residents and 30% of 392 were prescribed an opioid at baseline.^{25, 40} In these studies, 22% of residents were administered an opioid in the previous 7 days and 42% of residents were administered an opioid in the following 12 months. A pilot RCT investigating the effects of an interactive robotic companion seal (PARO) in people with dementia reported 42% of 43 residents were prescribed a PRN opioid at baseline. Over the 6-week intervention, one resident in the intervention group and four residents in the control group were administered a PRN opioid.²² Three studies reported opioids were among the most frequently administered PRN medication class.^{25, 34, 40} One reported that opioids accounted for 26% of all PRN administrations.³⁴

3.6 Opioid appropriateness

Of 23 studies reporting opioid prevalence, nine studies reported elements of opioid appropriateness.^{24, 27-30, 33, 41, 43} Six studies reported analgesic co-prescription.^{27-30, 33, 43} Co-administration of opioids and paracetamol ranged from 24%²⁷ to 80%.³⁰ One study reported 50% of residents charted regular opioids were co-prescribed paracetamol at 3-4 g/day and 48% were co-prescribed anxiolytics/hypnotics.²⁸ Kalisch Ellett et al.⁴¹ reported that the prolonged use of opioids was common, despite guideline recommendations for short-term use. One study reported analgesic ADEs and found daytime sleepiness was not associated with opioid use in the previous 24 h.²⁷

Three studies reported morphine milligram equivalents (MEQ)/day.^{24, 28, 29} One study reported the median MEQ/day was 21 mg, and 6% of residents administered opioids were administered doses exceeding the Australian maximum recommended dose of 100 mg MEQ/day for non-cancer pain at the time of the study.²⁴ Veal et al.²⁸ reported the mean and median charted dose of regular opioids was 58 and 30 mg MEQ/day, respectively. In a second study by Veal et al.,²⁹ 13% of residents charted opioids were charted doses exceeding 100 mg MEQ/day. Other studies reported analgesic load,^{8, 27} defined daily doses (DDD)⁴³ and specific opioid doses.³⁰ Of 23 studies, nine reported inabilities to assess or validate clinical indications, diagnoses or doses as a limitation.^{8, 26, 27, 30, 31, 34, 36, 40, 41}

3.7 Specific opioid classes

Fourteen studies reported the prevalence of specific opioid classes (Table S2).^{8, 24, 26-30, 32, 36, 39-43} Oxycodone was the most prevalent opioid in seven studies and was dispensed in up to 28% of the study sample.^{25, 26, 29, 30, 32, 41, 43} Buprenorphine was the most prevalent opioid in four studies and was dispensed in up to 15% of the study sample.^{8, 24, 27, 39} The prevalence of opioids with active metabolites or which can accumulate in renal impairment (e.g. codeine, tramadol, morphine or hydromorphone) was <8% in all studies. The least prevalent opioids were methadone and pethidine.⁴³

3.8 Quality assessment

Of 23 studies, three studies focused specifically on opioids,^{24, 29, 30} seven studies investigated pain, pain management or analgesics,^{8, 22, 27, 28, 33, 39, 43} while 13 studies reported opioid prevalence as part of a larger study.^{23, 25, 26, 31, 32, 34-38, 40-42} Sampling was considered appropriate in all but one study, which did not provide adequate information on how sampling was performed (Appendix 2, Table S3).³³ Evidence of adequate sample sizes was provided in 11 studies.^{8, 23-29, 36, 41, 43} Reporting of PRN and/or regular opioid use was differentiated in 13 studies.^{8, 22, 24-31, 37, 39, 40} Opioid administration to residents was reported in nine studies.^{8, 22, 24, 25, 27, 30, 34, 37, 40} Nine studies did not provide adequate information on how opioid prevalence was measured; therefore, reliability of methods could not be ascertained.^{22, 32-34, 36, 38, 41-43}