Use of Medicare-subsidised treatment services among people prescribed opioids for chronic non-cancer pain

Design

The POINT study protocol has been published.¹² Briefly, eligible participants were aged ≥18 years, self-reported living with CNCP for ≥3 months, and had been prescribed restricted opioids (Australian ‘Schedule 8’) for >6 weeks. Individuals currently prescribed opioid agonist treatment were excluded. Participants were recruited from community pharmacies between 28 July 2012 and 14 April 2014.

Data collection

Participants completed interviews, questionnaires, and medication diaries over 5 years. At cohort entry, consent for MBS data linkage was received from 1214 participants (80.2%, Supplementary Fig. S1). The study period was 1 January 2012–31 December 2018. Data were provided by the Australian Institute of Health and Welfare (AIHW) in two extracts (1 January 2012–31 December 2016 and 1 January 2017–31 December 2018), which were analysed together. For individuals known to have died, observation ceased at their last POINT follow-up date; eight individuals died before the first 3-month follow-up and were excluded. Fifty-four individuals withdrew consent and were excluded from the second MBS extract; observation for these participants ceased on 31 December 2016. For all other participants, observation ceased 3 years after their cohort entry date.

Outcomes

MBS item numbers were used to identify claims for general practitioner attendances, CDM items, allied health services, and specialist attendances (Supplementary Table S1). Allied health services were categorised as physical (physiotherapy, exercise physiology, chiropractic, osteopathy), mental (clinical psychology, other psychology, other mental health), and other (podiatry, dietetics, occupational therapy, speech pathology, diabetes education, audiology, other). Categories conformed to those used in AIHW reports;⁸ an additional category was created for pain medicine physician attendances (MBS subgroup A24.1–A24.2). Diagnostic, pathology, and dental services claims were excluded.

Covariates

Baseline data included self-reported demographics (age, gender, income, private health insurance, Indigenous status). Low income was categorised as ≤AUD400/week, representing the social security eligibility threshold. Index of Relative Socioeconomic Advantage and Disadvantage (IRSAD) quintile, and whether participants lived in a major city per the Accessibility/Remoteness Index of Australia, were derived from postcodes.^15,16 Pain characteristics included self-reported pain conditions, years living with pain, and duration of opioid use, collected at baseline. Baseline medication diaries were used to calculate mean daily opioid doses as oral morphine equivalent milligrams (OME mg), categorised as 1–49, 50–89, 90–199, and ≥200 mg on the basis of guideline thresholds and recommendations.^17,18 The Brief Pain Inventory (BPI) was used to assess pain severity and interference (the extent to which pain interferes with activities), each scored out of 10 where a higher score indicates more pain severity/interference.¹⁹ Other health-related characteristics were collected. Participants completed the 12-Item Short-Form Health Survey physical component (SF-12-PCS); poor physical health was dichotomised as <30 out of 100 (two standard deviations below the normative population mean).²⁰ Participants completed the Patient Health Questionnaire 9-item depression assessment (PHQ-9) and Generalized Anxiety Disorder 7-item scale (GAD-7); symptoms of moderate-to-severe depression or anxiety were dichotomised as ≥10 on the PHQ-9 or GAD-7, respectively.²¹

Statistical analysis

Descriptive statistics are presented, including the number of participants with at least one claim for services. Rates were calculated by dividing the number of claims by contributed person-time, expressed per 100 person-years (PY).

Descriptive comparisons were made between service use for both the study cohort and the Australian adult population in the 2014–15 financial year, chosen as POINT enrolment ended in April 2014 to minimise the impact of attrition. Population service use was derived from AIHW data,⁸ dividing the number of people receiving a service by the estimated resident population, restricted to those aged ≥15 years. As the POINT cohort skews older (>75% ≥45 years at baseline), a comparison was also made with the general population aged ≥45 years.

Logistic regression was used to measure associations between covariates and odds of having a claim for service categories of interest (CDM, physical/mental/other allied health, specialist/pain medicine attendances), adjusted for contributed person-time. Poisson regression, adjusted for overdispersion and offset by log of person-time, was used to measure associations between covariates and rates. Covariates included in models included demographics and socioeconomic characteristics on the basis of prior research about economic or logistic barriers to health service access;^9–11 pain characteristics to determine whether pain intensity was associated with service utilisation; and health measures to determine whether people with greater comorbidity used more services. Polychoric correlations were assessed (Supplementary Table S2); PHQ-9 (depression scores) and GAD-7 (generalised anxiety scores) demonstrated a high correlation coefficient (0.77), leading to the exclusion of GAD-7 from models. Because of multiple comparisons being made, a higher threshold for statistical significance was chosen (P ≤ 0.005). As IRSAD quintile and opioid dose had ≥2 categorical levels, differences across these variables were tested using Wald Chi-square and likelihood ratios. Analyses were undertaken using SAS v9.4 (SAS Institute, Cary NC).

Ethics

The study was approved by the UNSW Human Research Ethics Committee (HC12149, HC16916).

Baseline characteristics

Of 1206 participants included in this analysis, 57.2% were female, and the mean age at cohort entry was 58.4 years (Table 1). Half lived in a major city (49.5%), and one-third held private health insurance (37.7%). Health measure scores indicated poor overall physical health, and almost half met the criteria for moderate-severe depression. Pain scores were suggestive of moderate pain severity and interference, with 55.5% reporting a daily opioid dose of ≥50 OME mg at baseline. Individuals included in this analysis were comparable to the overall POINT cohort (Supplementary Table A3).

Medicare-subsidised service use

Over a 3-year period, 76.9% of individuals had at least one claim for a CDM item (198.7/100 PY) (Table 2); 68.4% received at least one new or reviewed care plan, and 32.2% received a general practitioner mental health service. Claims for health assessments and multidisciplinary case conferencing were identified for 175 and 14 participants, respectively.

Allied health service claims were identified for 86.7% of participants (234.3/100 PY). The most common services were not directly pain related: optometry (74.0%) and podiatry (30.4%). One-third of participants attended a physical allied health service (32.8%, 55.5/100 PY), predominately physiotherapy (26.0%). Most participants had ≥1 specialist attendance (82.0%, 439.2/100 PY), with 15.0% attending a pain medicine physician (17.9/100 PY).

Comparison with the general population

When compared with the general population (n = 19,035,964)⁸, use of most services over a 12-month period was higher in the study cohort (Supplementary Table S4). The study cohort was observed to have ≥2-times higher use of CDM items (study cohort 58.8% vs general population 21.6%), care plans (50.7% vs 13.5%), physical allied health services (16.9% vs 3.6%), and specialist attendances (64.6% vs 32.8%).

Restricting to the general population aged ≥45 years (n = 9,224,292), similar patterns were observed. Use of CDM items remained higher (study cohort 58.8% vs general population ≥45 years 30.0%), as did use of care plans (50.7% vs 22.8%), physical allied health services (16.9% vs 5.9%), and specialist attendances (64.6% vs 46.2%).

Factors associated with service use

After adjusting for time, sociodemographic, health, and pain characteristics, being older was associated with higher use of specialist and other allied health services; conversely, mental allied health use reduced with age (per 10 years, adjusted odds ratio [aOR] 0.65, 99.5% confidence interval [CI] 0.53–0.79) (Table 3). Living in a major city was not associated with odds of use for any service category; however, it was associated with increased rates of pain medicine attendances (adjusted incident rate ratio, [aIRR] 1.70, 99.5%CI 1.12–2.56) (Table 4). Holding private insurance was associated with both increased odds and rates of specialist (aOR 4.29, 99.5%CI 2.32–7.91; aIRR 3.13, 99.5%CI 2.52–3.89) and pain medicine attendances (aOR 3.08, 99.5%CI 1.81–5.25; aIRR 3.11, 99.5%CI 2.11–4.59).

Mental health scores indicative of moderate-to-severe depression were associated with use of mental allied health services (aIRR 2.96, 99.5%CI 1.93–4.53) and CDM items (aIRR 1.39, 99.5%CI 1.12–1.73); no associations were observed for physical health or pain severity measures. A small inverse relationship was observed between longer pre-study opioid use and odds of CDM item use (per year, aIRR 0.99, 99.5%CI 0.97–0.99).

Limitations

Combining interview and administrative data had several benefits, allowing collection of information often lacking in administrative datasets such as health measures, and limiting biases from self-reported service use.²⁷ Both methods have limitations. Sociodemographic characteristics and health measures were self-reported and cross-sectional. MBS data do not include diagnoses, and individual conditions were not examined, limiting the ability to examine the impact of specific pain conditions or comorbid conditions on service use. Medicare-subsidised service use is under-representative of true use: services accessed without a Medicare rebate (e.g. privately) and those provided by the Department of Veterans Affairs were not captured, representing areas for future research.

More recent follow-up data were unavailable. On the basis of reports of population-level service use from 2013 to 2019,⁸ and in the absence of major changes to eligibility for services examined in this study, significant changes to service use over this period are not anticipated. From 2020, the impact of COVID-19 requires exploration, including potential disruptions to service access. Policy changes enacted during the pandemic, including expansions to Medicare-subsidised telehealth services, will have implications for use and should be a focus of future work.