Data Resource Profile Update: The Opioid Agonist Treatment and Safety II (OATS II) Study, 2001–22, New South Wales, Australia

Data resource basics

Although medicines such as methadone and buprenorphine are established as the most effective treatment for opioid dependence, there are risks associated with opioid agonist treatment (OAT), especially at the start [1–3]. Previous studies have illustrated that the benefits have outweighed the risks. Whilst on treatment, evidence of reductions in the risk of death, opioid use, HIV, and hepatitis C infection has been documented [4–6]. The use and costs associated with the health service and the criminal justice system are also reduced while people are retained in OAT [7–11]. There is less evidence of the benefits of OAT in subpopulations, such as pregnant women and older people.

This data resource (OATS II) has been constructed for a population-based, retrospective cohort study through a single data linkage (2001–22). It is an update to the Opioid Agonist Treatment and Safety (OATS) study (2001–18) [12], which was established to determine the incidence of adverse clinical events. Medicines that were available during this updated study period included methadone, buprenorphine, buprenorphine–naloxone, and long-acting injectable buprenorphine (LAIB). The resource includes all people who were approved to receive OAT for opioid dependence in New South Wales (NSW), Australia, through the NSW Opioid Treatment Program between 1 January 2001 and 30 November 2022 (N = 55 160). These approvals are recorded in the Controlled Drugs Data Collection (CoDDaC), which is a revised and updated version of the Electronic Reporting and Recording of Controlled Drugs database (used in OATS originally).

The CoDDaC includes basic demographic data, drugs of concern, and dates of applications to prescribe OAT. The prescriber, prescribing setting (e.g. primary care, public clinic, private clinic, correctional center, hospital), and dispensing setting (e.g. clinic, community pharmacy, correctional center) are provided. An episode of OAT commences once approval has been granted and continues through multiple dosing occasions and changes in medication, prescriber or dosing location, and concludes once the prescriber has submitted the treatment exit form or is cancelled at the last known dosing date. OAT approvals that never resulted in the dispersal of medication may be recorded with an episode duration of 1 day. New treatment episodes that commence within 6 days of a treatment cessation of daily dosing of methadone and buprenorphine or within 36 days for LAIB initiated by Justice Health after 2018 are considered a continuous treatment episode [13].

Overall, 160 889 OAT episodes were recorded, with durations that ranged from 1 day to a maximum of 14 611 days, and a median duration of 235 days (mean duration 1095 days). Participants had a median of 2 episodes each, with a range of 1 to 53 episodes.

This new cohort has been linked to 10 NSW-wide administrative datasets (Figure 1) and content data for all datasets have been extended by ≤4 years (to 2022). The extension includes the period in which two formulations of LAIB were registered in Australia with the Therapeutic Goods Association, allowing the opportunity to measure the impact that an extended-release depot OAT formulation has on health and social outcomes and treatment retention. The addition of cost weight variables from the hospital admission data collection will aid in studies that estimate the economic evaluations of changes in health utilization costs from interventions of changes such as the introduction of LAIB.

Figure 1.

Datasets linked in the Opioid Agonist Treatment and Safety (OATS II) Study.

Open in new tabDownload slide

With the inclusion of perinatal data, we will examine the effects of OAT on perinatal health using real-world data; pregnant women are typically excluded from clinical trials. The addition of ambulance data provides a more complete picture of overdose and gaps in the current literature can be studied with the addition of notifications of infectious diseases (including hepatitis C and sexually transmitted infections), perinatal, and cancer registry data collections.

The linked data were received in 2023 and are stored in secure file storage at UNSW Sydney.

Data collected

Linkage of data was performed by the Centre for Health Record Linkage, which is a specialized data linkage unit within the New South Wales Ministry of Health [1, 14]. The probabilistic linkage process used information such as full name, aliases, dates of birth, sex, and dates of last known contact. Adherence to the data separation principle [15] ensured that those who were working with identifying information did not also have access to content data.

The CoDDaC is a dataset that has been revised and renamed on multiple occasions and is the primary administrative dataset on which the OATS II Study data resource is built. The revised dataset has been cleaned and missing data and errors have been corrected and documented, particularly concerning dates of treatment entry and exit. However, treatment exit date errors may remain, as these data rely on timely reporting by prescribers, which may not always occur. Once linked, data validation across datasets is conducted, which includes logic checks, correction of obvious typographical errors, and variable comparisons across datasets to determine erroneous values. For instance, Indigenous status is often poorly recorded in many health datasets [16]. We have chosen to classify an individual as Aboriginal and/or Torres Strait Islander if they are identified as such in any of the included datasets. Exclusions from the cleaning process are provided in Supplementary Figure S1 (see online supplementary material for a color version of this figure).

The original OATS Study data resource included linked data from the following NSW administrative datasets: Admitted Patient Data Collection, Emergency Department Data Collection, Mental Health Ambulatory Data Collection, Re-offending Database (ROD) Court and Custody data, Registry of Births, Deaths and Marriages, and Cause of Death Unit Record File (COD URF). Data on OAT prior to 2001 are also included for cohort members. The updated OATS II Study data resource extends the collection of data for all linked datasets (COD URF to 2020, ROD to 2021, all others to 2022) and adds linkage to the following NSW administrative datasets:

• Perinatal Data Collection (1994–2021): records of all live births and stillbirths of ≥20 weeks’ gestation or birthweight of ≥400 g delivered in public and private hospitals as well as homebirths.

• Cancer Registry (1972–2020): all cancer diagnoses including type, site of origin, morphology codes, and degree of spread.

• Notifiable Conditions Information Management System (1993–2022): diagnoses of certain infectious diseases and infections (i.e. hepatitis C, sexually transmitted infections) and adverse events following immunization, notified to the NSW Department of Health by laboratories, hospitals, medical practitioners, schools, and childcare centers.

• Ambulance data (2009–22): for all ambulance attendances, operational information from the computer-aided dispatch system, and data documented by clinicians in the paper-based patient healthcare record and electronic medical record. Data includes patient vital signs and emergency and urgent episodes of care for patients who were transported to a hospital, left at a scene following clinical assessment, or who died at the scene.

Further details on all linked datasets in this updated data resource are available via https://www.cherel.org.au/datasets.

Data resource use

The cohort size increased by 7012 participants to 55 160 after linkage validation; the number of participants on methadone each year has remained stable; the increase in buprenorphine prescribing between 2018 and 2022 has resulted in a nearly 50:50 ratio across medicines, methadone to buprenorphine (Figure 2). This updated cohort has similar demographics to the earlier cohort; Table 1 summarizes new data and full demographic characteristics, and hospitalizations are provided as Supplementary Tables S1 and S2 (see online supplementary material for color versions of these tables). The proportion of participants who identify as Aboriginal and/or of Torres Strait Islander descent has increased from one in five to just over one in four (26.8%). The median age of entry into the cohort has remained unchanged, although the age distribution of people in treatment year on year has been steadily increasing: on 1 January 2022, almost half of the people in NSW on OAT were >45 years old.

Figure 2.

OATS II: number of people receiving OAT on 1 January annually, 2001–22, by medicine prescribed.

Open in new tabDownload slide

It has been shown that, for older people with chronic conditions such as circulatory and respiratory disease, mortality risk was lower in those prescribed buprenorphine compared with methadone [17]. Further investigation into treatment management for an older population is important to ensure continued safety. The number of male participants <45 years old in treatment have remained constant each year but there has been a decrease in females in the same age group year on year (Figure 3). It is important to establish whether this is due to a lack of need or because females of childbearing age are not seeking treatment. Among the cohort, 4656 women gave birth across the follow-up period (n = 8620 births).

Figure 3.

OATS II: number of people receiving OAT on 1 January annually, 2001–22, by age and sex.

Open in new tabDownload slide

From 2012, data on participants’ primary opioid of concern and other drugs of concern were collected (see Supplementary Figures S2 and S3; see online supplementary material for color versions of these figures). Since 2018, the number of patients who reported pharmaceutical opioids as the primary drug of concern has dropped below those clients for whom the primary opioid of concern was heroin, and we see a steady increase in methamphetamine recorded as another drug of concern. However, these data should be interpreted with care due to potential variation between the ways in which prescribers collect the data (e.g. asking clients about specific drugs of concern versus asking clients whether they have any other drugs of concern).

In this cohort, around half of the participants had had at least one contact with an ambulatory mental health service (48%; Table 2). Around three-quarters had had at least one emergency department (ED) presentation (76%) and hospitalization (69%) during the study period. Nearly 60% of the cohort had been the recipient of an ambulance callout and 24% had had a callout for a suspected drug overdose. One-quarter of the cohort had had a notification for hepatitis C, which contributed to nearly three-quarters of the notifications of all infectious diseases among the cohort (Table 2).

Expanded outcome variables include all pregnancy and neonatal outcomes, cancer, and infectious disease notifications. More than half of overdose presentations are handled by first responders and are not transported to a hospital; with the addition of the ambulance callout data, we have more visibility of acute outcomes of substance use [18].

This data resource will enable detailed exploration of OAT trajectories, individual and treatment setting variables, and how these may influence outcomes. We expect findings from these analyses to inform OAT clinical guidelines and the expansion of diverse treatment models that increase access to OAT. Among individuals who gave birth, we quantified the pattern of OAT use for opioid dependence during pregnancy and following childbirth. Between 2005 and 2021, among >5200 women who gave birth and had OAT within the past 4 years, we found that nearly 4100 (79%) had had OAT during pregnancy [19]. Of those, three-quarters were on OAT prior to treatment and >80% retained treatment until childbirth [20]. Among 3900 pregnant women who were on OAT at childbirth, the majority (88%) continued treatment within the first 6 months following childbirth [21].

Among people who were released from prison, we compared treatment retention between methadone versus LAIB and found that 46.6% of people who were released while prescribed LAIB were still retained 1 year later compared with 58.8% of people who were released while prescribed methadone. Re-incarceration within 1 year of release was lower for people who were released on methadone: 50.8% were re-incarcerated compared with 58.5% released on LAIB [22].

As previously noted, this study is an update of an existing cohort of people who were prescribed OAT in NSW: the OATS Study. The OATS Study investigated the critical periods of elevated risk for fatal and non-fatal overdose, suicide, and injecting-related infections [3, 23–26]. Fatal overdose, suicide, and self-harm hospitalizations were at their lowest while in treatment (post the first month after treatment entry) and >12 and >17 times higher in the first 4 weeks out of treatment for fatal overdose and suicide, respectively, compared with the treatment period following the first 4 weeks whereas the highest overdose rate for non-fatal opioid overdose was in the first 4 weeks of treatment, suggesting either an inadequate dosing schedule or a need for education on poly-substance use management [3, 26]. Retention on OAT was associated with a reduced rate of injecting-related infections. However, the first 4 weeks of treatment were associated with an increased rate, possibly due to referral pathways between hospitals and community OAT services [23]. The first 2 weeks following release from prison were also found to have elevated risk for injecting-related bacterial infections, as well as the early periods of treatment initiation and cessation [24].

People who are dependent on opioids have a 15-year gap in life expectancy compared with the general population [27]. Between 2001 and 2020, it was estimated that OAT provision reduced overdose and other cause mortality among the cohort by 53% and 27%, respectively. It was estimated that 1.2 deaths were averted and 9.7 life years gained per 100 person-years on OAT. The community and prison OAT program in NSW has substantially reduced population-level overdose and all-cause mortality in the past 20 years, partially due to high retention, which should be a key consideration of national programs [28]. An assessment of the characteristics of prescribers of OAT in NSW unveiled a service at risk. OAT prescribing is concentrated in a small group of mature prescribers and new prescribers have low retention. There is a need to identify and respond to the barriers of prescribing if we are to continue to avert deaths, increase life years gained from OAT, and increase accessibility from 89% of the NSW population [29, 30].

For further publications, please see https://ndarc.med.unsw.edu.au/project/opioid-agonist-treatment-and-safety-oats-study and https://ndarc.med.unsw.edu.au/project/opioid-agonist-treatment-and-safety-ii-oats-ii-study.

Strengths and weaknesses

A key strength of this cohort is the comprehensive identification of all individuals who have been undergoing OAT for opioid dependence in NSW over the last two decades. The extensive size and long duration of the cohort allow the analysis of uncommon adverse outcomes and temporal changes. Additionally, we have linked multiple health and criminal justice datasets, offering a diverse array of covariates for our predictive models.

The administrative purpose of the study data collections introduces some limitations. The CoDDaC records authorizations to treat with opioid agonists (methadone and buprenorphine) and does not include daily dose data. It does not distinguish between mono-buprenorphine and buprenorphine–naloxone, or LAIB, although we do know that long-acting buprenorphine (monthly injectable) was prescribed in NSW prisons from 2018. There is no record of whether people have access to unsupervised (take-home) dosing. It should not be assumed that buprenorphine prescriptions are for unsupervised dosing, as the NSW Opioid Treatment Program clinical guidelines recommend that all methadone and buprenorphine doses should be supervised during treatment induction and stabilization [31]. When a participant achieves stabilization, a level of unsupervised dosing may be considered (e.g. a client who is considered at “moderate risk” may receive up to two take-home doses of methadone per week) [31, 32].

Some important factors, such as HIV or homelessness, can only be inferred from hospitalization data, which almost certainly underestimate true prevalence. Socioeconomic status is inferred indirectly by applying the Socio-Economic Indexes for Areas to the residential postcode that is listed for each participant [33].

Data resource access

The data underlying this article cannot be shared publicly due to the guidelines from the data custodians. To ensure privacy and confidentiality, the approval process for linking health data in NSW is governed by strict conditions concerning data storage, retention, and usage. Currently, data can be stored at a single location—UNSW Sydney—for a maximum of 7 years after publication of the results.

We welcome inquiries from interested parties regarding potential secondary data analyses. It is important to note that legislation mandates that data must be stored and analysed exclusively within NSW. Requests for data access should be directed to Professor Louisa Degenhardt ([email protected]) and they will be reviewed by the OATS II investigator team. Collaborators will need to obtain approval for data access and any specific secondary analyses from the NSW Population and Health Services Research Ethics Committee. Furthermore, those who are aiming to address research questions related to Aboriginal peoples must also seek approval from the Aboriginal Health and Medical Research Council.

Acknowledgements

We acknowledge the Australian Institute of Health and Welfare, NSW Ministry of Health, Centre for Health Record Linkage, and the Bureau of Crime Statistics and Research for their support in data provision and linkage. A draft of this paper has been reviewed by the Chair of the Aboriginal Reference Group prior to publication.

Author contributions

Data curation was conducted by N.J., F.N.G., and D.T.T. Analysis was conducted by N.J. and F.N.G. N.J. and A.G. wrote the first draft of the manuscript and all authors critically reviewed the manuscript.

Supplementary data

Supplementary data is available at IJE online.

Use of artificial intelligence (AI) tools

No AI tools were used in the collecting and/or analysing data, producing images or graphical elements, or in writing the paper.

Conflict of interest

None declared.

Funding

The linkage for the original data resource was funded by fellowship grants to L.D. This revised data resource was supported by the National Institutes of Health (R01 DA144740 to L.D.). N.J. and L.D. are supported by an Australian National Health and Medical Research Council Research Investigator grant (2016825). L.D. is supported by an Australian National Health and Medical Research Council Research Fellowship (1135991). The National Drug and Alcohol Research Centre is supported by funding from the Australian Government Department of Health under the Drug and Alcohol Program.

References