Association between the location of Opioid Agonist Treatment (OAT) providers and heroin-related ambulance attendances

Abstract

Introduction

Opioid agonist treatment (OAT) is an evidence-based intervention that reduces harms associated with heroin use. Ambulance services often serve as the first point of contact for people experiencing these acute harms. OAT access relies on dosing points and prescribers, which may be unevenly distributed geographically. This study examines the geographical distribution of heroin-related ambulance attendances across Victoria and assesses whether the presence, availability, and number of OAT service providers are associated with these harms.

Method

We merged 2023/24 Victorian heroin-related ambulance attendance from the National Ambulance Surveillance System with Victorian OAT service availability data from a statewide helpline. Three negative binomial regression models tested associations between OAT availability and heroin-related ambulance attendances at the local government area (LGA) level.

Results

Fourteen LGAs had higher than average, population-adjusted, heroin-related ambulance attendances. In Model 2, LGAs with more pharmacies offering OAT vacancies than those without had a 50% lower risk of heroin-related harms (IRR = 0.52, p = 0.014). Model 3 found that every 10% increase in the proportion of OAT doctors with current availability was associated with a 9% reduction in heroin harms (IRR = 0.91, p = 0.04). A similar trend was observed for pharmacists (IRR = 0.90, p = 0.08). All models found heroin harms to be approximately 70% lower in regional versus metropolitan LGAs.

Discussion and conclusions

The availability and geographic distribution of OAT providers—especially those with current capacity—are linked to reduced heroin-related ambulance attendances. Strategic placement and resourcing of OAT services could better align with population need and target hotspots. The metropolitan-regional differences warrant further investigation.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13011-026-00706-9.

Introduction

Opioid agonist treatment (OAT) is a well-established harm minimisation intervention for opioid use disorder, significantly reducing illicit heroin and synthetic opioid use, fatal and non-fatal overdose [1–4], and emergency health service utilisation [5, 6]. Despite its effectiveness, OAT uptake is far from universal. A recent meta-analysis estimated that only one in five people with heroin dependence had sought treatment in the previous 12 months [7].

People who use opioids report multiple logistical barriers to OAT, including limited numbers of prescribers and dosing points, long travel distances, costs and restrictive service hours [8, 9]. In Australia, OAT is predominantly prescribed by general practitioners (GPs) accredited through an OAT training course [10], yet in 2022, only 7% of doctors prescribed OAT [11]. Furthermore between 2001 and 2017, the number of doctors ceasing prescribing OAT increased fourfold [12]. Reported reasons for non-prescribing include inadequate support and training, perceived complexity of the client group, safety and disruption concerns, and low remuneration [13, 14].

Dispensing occurs through public or private clinics, community pharmacies, and correctional facilities [15], with pharmacies providing approximately 75% of all OAT doses [15]. In 2023, 3,082 sites dispensed OAT across Australia [15], an increase of 50% since 2012 [15]. In Victoria, dosing points increased by 40% between 2006 and 2023 [15]. Nonetheless, fewer than half (44%) of pharmacies supply OAT [15, 16], often citing stigma, fear, the opt-in nature of provision, lack of awareness or training, and poor remuneration [17]. This contributes to geographical inequalities [15, 18]; in 2024, over half (53%) of dosing points were in major cities, with fewer in regional and remote areas [15].

Geographic location affects not only access, but also patterns of healthcare use. A New South Wales study found OAT reduced urgent emergency department (ED) attendances in both metropolitan and regional areas [6], but non-urgent visits decreased only in metropolitan settings, likely due to some regional clients receiving OAT dosing in EDs [6]. Aboriginal and Torres Strait Islander Australians and younger populations are more likely to live further from dosing points [18], and greater travel distances are linked to lower OAT uptake and retention [19, 20]. Conversely, one US study found proximity to large methadone clinics was associated with increased heroin use [21], underscoring the complexity of location-harm relationships.

Ambulance services are often the first healthcare contact for acute heroin harms [22–24], which can include fatal and non-fatal overdose, intoxication adverse effects, injection-related complications, associated trauma or mental health and withdrawal related problems. However only around half of such heroin related ambulance attendances result in hospital transport [25]. As such, ambulance data capture more acute heroin-related events than ED or mortality datasets [26, 27]. However, standard ambulance coding can underestimate AOD harms due to broad diagnostic categories [28, 29]. The National Ambulance Surveillance System (NASS), addresses this limitation through manual coding of paramedic clinical records, producing highly detailed, geocoded datasets on alcohol and drug harms, including co-occurring substances, comorbid self-harm, violence and situational stressors [25, 30, 31].

While OAT has been shown to reduce heroin overdose and there is evidence that distance to services affects engagement, no prior research has examined whether the location and availability of OAT providers, both prescribers and dosing points, are associated with the location of heroin-related harms. Understanding this relationship could inform more effective, needs-based placement of OAT services.

Therefore, this study aims to (i) map ambulance attendances for acute heroin-related harms across Victoria at the local government area (LGA) level; (ii) assess whether the number, type, and availability of OAT services in an LGA are associated with heroin-related harms, hypothesising an inverse relationship where greater OAT service availability is associated with fewer heroin-related ambulance attendances; and (iii) explore how these relationships vary by metropolitan/regional classification and socioeconomic status.

Methods

Study design and sample

This study employed negative binomial cross-sectional regression to examine the association between the location of heroin-related ambulance attendances and the location of OAT services over a four-month period (1 November 2023 to 29 February 2024). The 4-month time frame was used because thrice per year updating is the industry standard based on the experience and expertise of the statewide helpline team. Therefore, the OAT availability data was assumed to stay constant from November 2023 and February 2024, whereas if we used a greater duration (e.g., 12 months), vacancy status is likely to change, potentially reducing the accuracy of availability data.

Ethics

The Eastern Health Research Ethics Committee provided ethical approval (E122-0809-14552) for data collection via the NASS.

Measures

Heroin-related presentations between November 2023 to February 2024 were obtained from the Victorian arm of the NASS [25, 30]. The NASS is a population level surveillance system for alcohol- and drug-related harms, mental health harms and suicide and self-harm. The NASS data comes from electronic paramedic clinic records (ePCRs), which are coded by a specialist team at Turning Point in order to identify specific substances involved in the presentation, as well as self-harm or suicidal behaviours. Further details of the NASS methodology can be found in the NASS methodology papers [25, 30]. OAT services (including prescriber and dosing points), and whether the services were currently accepting (or not accepting) referrals were obtained from a Victorian 24/7 alcohol and other drug statewide helpline [32], which provides advice, referral and support to the Victorian community. The statewide helpline operates an opioid prescriber and dispensing directory for callers, and maintains a regularly updated record of OAT prescribers and OAT dosing point locations across Victoria for practitioners and pharmacies who have agreed to be part of the referral program and who have at some point, since 2020, provided OAT services [32]. In Victoria OAT is predominantly (> 97%) delivered by a general practitioner (prescribed) and pharmacy (dispensed) model, which is partly funded by Medicare [15]. However, other Australian jurisdictions differ in the delivery of OAT, for example, New South Wales has 75% Medicare funded general practitioner and pharmacist model with the remaining 25% provided via public community clinics, non-government organisations, hospitals or doctors surgeries [15].

Population description variables

Population level variables describing the heroin-related ambulance cohort included age, gender, transportation to hospital and metropolitan or regional location. Metropolitan Melbourne areas were local government areas (LGAs) surrounding Melbourne as per the State of Victoria Department of Environment, Land, Water and Planning 2017 [33]. Regional areas were defined as all other LGAs. In Victoria, there are 79 LGAs, of which 31 are metropolitan LGAs and 48 are regional LGAs [33]. Descriptive variables of LGA regions with heroin harms were reported and included location (metropolitan or regional), population and socioeconomic status via the socioeconomic index for areas (SEIFA) [34, 35].

Outcome variables

The outcome variable was the number of heroin-related attendances that occurred in the 4-month time period in each LGA. Heroin-related presentations were classified as any ambulance attendances that had been coded as related to heroin including overdose, withdrawal, or complications from heroin use disorder.

Predictor variables

Predictor LGA variables included in the model were geographical area (binary variable with metropolitan = 1 or regional = 0) and SEIFA quintile (categorical variable with 1 being most disadvantaged and 5 being most advantaged). The OAT dispensers/dosing points and OAT prescribers and their vacancy status variables were determined via the statewide helpline report [32]. The vacancy status groups that were used, which were linked to the OAT prescribers and dosing points were current vacancies and limited vacancies. Current vacancies were defined as when the OAT dosing point, or OAT prescriber had been confirmed to be currently accepting referrals for new OAT clients within the last four months of the statewide helpline report. Limited vacancy status was when the OAT dosing point or OAT prescriber service was unavailable, had no current vacancies or the vacancy status was not reported within four months of the statewide helpline report. Three negative binomial models were analysed with the OAT prescriber and dosing point variables differing between the three models. The models are described below:

Model 1 OAT service provider variables

There was one variable for OAT dosing point and one variable for OAT prescriber. The OAT dosing point and OAT prescriber variables in this model were binary variables that were equal to 1 if there was at least one OAT prescriber/OAT dosing point accepting referrals in an LGA and was equal to zero if there were no OAT prescriber/OAT dosing point with spots available in an LGA.

Model 2 OAT service provider variables

There was one variable for OAT dosing point and one variable for OAT prescriber. The OAT dosing point and OAT prescriber variables in this model were binary variables that were equal to 1 if the number of OAT prescriber/OAT dosing points currently accepting referrals in an LGA was greater than the number of OAT prescriber/OAT dosing point spots not currently accepting referrals.

Model 3 OAT service provider variables

The OAT dosing point and OAT prescriber variables in this model were continuous percentage variables that were equal to the number of doctors and pharmacists with current vacancy over the total number of pharmacies or doctors (who were included in the statewide helpline report and had at some time since 2020 supplied OAT) in an LGA multiplied by 100 to provide the value as a percentage (number of pharmacies accepting referrals in an LGA/total number of pharmacies in the statewide helpline report in an LGA*100).

Statistical analysis

To construct one dataset, the NASS heroin-related ambulance attendances were merged with the statewide helpline report data by LGA, which provided the number of OAT dosing points (and their vacancy status) and OAT prescribers (and their vacancy status) per LGA. Demographic descriptive statistics were presented for individuals who experienced a heroin-related ambulance attendance.

The mean number of heroin ambulance harms per 100,000 individuals, adjusted for LGA population, was identified for the whole sample and then LGAs with higher than the Victorian average number of harms were reported. The demographics including socio-economic disadvantage, population and rurality of the LGAs with above average heroin harms were also reported. The socio-economic disadvantage and population were reported as per the Australian Bureau of Statistics Socio-Economic Indexes for Areas (SEIFA) data [34] and the location status was reported as either metropolitan or regional [35].

Negative binomial mixed-effects model was used to model overdispersed count data, with random effects to account for clustering by LGA. Three models were used to test the hypothesis as to whether the location, number and vacancy status of OAT prescribers and dosing points had an impact on heroin-related ambulance harms; LGA was included as a random variable. Variables were organised into two hierarchical levels. Level one variables include the OAT prescriber and OAT dosing point variables (model 1: any dosing point/prescriber vacancy in an LGA, model 2: number of dosing point/prescriber current vacancy was greater than number of dosing point/prescriber no current vacancy in an LGA, model 3: proportion of doctors and pharmacists with current or limited vacancy); level two variables included population density, SEIFA level and location of the ambulance presentation (metropolitan or regional).

The negative binomial models were developed using a three-stage analysis strategy [36]. The null model, which had only the random intercept, was stage one and was analysed to determined geographical variance associated with heroin-related ambulance harms. Second, the level one variables (OAT prescriber and OAT dosing point variables) were added to examine the impact of random effects. Thirdly, the level two variables including location (metropolitan and regional) and SEIFA quintile were added into the model. The model development and model diagnostics (Bayesian Information Criteria (BIC), Aikake Information Criteria (AIC) and log-likelihood) are provided in Appendix 1. The models were adjusted for LGA population as an exposure variable. As per the West et al. analytic approach, potential interaction effects were examined between variables and non-significant interaction variables were removed. As heroin is used at different rates in metropolitan [37, 38] and regional areas [39, 40], an interaction term between OAT prescriber/dosing point variables and location (metropolitan or regional) was examined, as well as between OAT prescriber and dispenser variables. Incident rate ratios (IRR) were reported at significance levels of p < 0.05 and p < 0.1. For Model 3, IRRs for the OAT service provider percentage variables were expressed per 10-percentage-point increase rather than per 1-percentage-point increase. This was achieved using the lincom command to rescale the two predictor variables—current pharmacy vacancy proportion and current doctor vacancy proportion—by a factor of 10. This approach allowed IRRs to be interpreted as the effect of a 10% increase, improving interpretability. The lincom IRR value was then converted to a proportional change using the following formula: (IRR-1)*100% [41].

Results

Demographics of the NASS opioid sample

There were 760 heroin-related ambulance attendances over the four-month period, with just over half transported to hospital (57%). The mean age was 42 years, and the majority of attendances were male (73%) and located in a metropolitan area (82%).

LGA areas with the highest heroin-related attendances

The mean number of heroin-related harms adjusted for LGA population for the four-month period was 8.95 per 100,000 individuals (standard error = 1.74, 95% confidence interval = 5.49, 12.40). Figure 1 provides details of LGAs with higher than average heroin-related harms (when 95% CI is considered). The LGAs with the highest heroin-related harms, adjusted for population, were Yarra (102 per 100,000 individuals), Melbourne (63 per 100,000 individuals) and Northern Grampians (42 per 100,000 individuals).

Fig. 1.

LGAs with higher than average heroin-related ambulance attendances

Demographic details of the 14 LGAs with higher than average heroin-related attendances, including SEIFA quintile, location (metropolitan or regional) and population are reported in Table 1. One third of the LGAs with the highest heroin-related attendances were in the most advantaged SEIFA quintile, whilst another third of LGAs were in the second most disadvantaged SEIFA quintile. Of the LGAs that experienced the highest heroin-related attendances, half were large LGAs (population greater than 100,000), 29% were medium sized LGAs (population between 50,000 and 100,000) and 21% were small LGAs (less than 50,000).

Table 1.

Demographics of areas with higher than average heroin-related ambulance attendances

	Frequency, n = 14	%
SEIFA quintile
1 (most disadvantaged)	1	7.1
2	5	35.7
3	1	7.1
4	2	14.3
5 (most advantaged)	5	35.7
	Frequency, n = 14	%
Location
Metropolitan	7	50
Regional	7	50
	Frequency, n = 14	%
Population
Small (< 50,000 people)	3	21
Medium (between 50,000 and 100,000 people)	4	29
Large (> 100,000 people)	7	50

Negative binomial regression model results

Table 2 shows the adjusted regression results for Model 1, which analysed the association between heroin-related ambulance attendances in an LGA and whether an LGA had at least one OAT dosing point and one OAT prescriber with a current vacancy. The results found that heroin-related harms were 70% less in regional LGAs compared to metropolitan LGAs. The interaction terms were non-significant.

Table 2.

Adjusted negative binomial regression for heroin-related ambulance attendances per LGA (any OAT service in an LGA)

Variable	Heroin IRR (95% CI)
At least one dosing point with a current vacancy	1.57 (0.70, 3.50)
At least one prescriber with a current vacancy	0.91 (0.48, 1.73)
Location (regional)	0.30 (0.17, 0.56)**
SEIFA quintile	0.88 (0.69, 1.14)
Constant	0.0006
Pseudo R2	0.05
alpha	0.83 (0.54, 1.27)
Log likelihood	-199.32
Aikake Information Criteria (AIC)	410.64
Bayesian Information Criteria (BIC)	424.78
N (LGA)	78

*Significant at p < 0.1

**Significant at p < 0.05

Table 3 presents the adjusted regression results for Model 2, which examined the association between heroin-related ambulance attendances at the LGA level and whether an LGA had more OAT services (prescriber and dosing points analysed separately) accepting referrals than not accepting referrals, as reported in the statewide helpline dataset. Model 2 found that LGAs with a greater number of pharmacists offering OAT vacancies compared to those without vacancies had a 48% lower risk of heroin-related harms. Additionally, heroin-related harms were 74% lower in regional LGAs compared with metropolitan LGAs. Interaction terms were non-significant.

Table 3.

Adjusted negative binomial regression of heroin-related ambulance attendances per LGA, comparing LGAs with a greater number of OAT services with current vacancies to those with more OAT services without vacancies

Variable	Heroin IRR (95% CI)
Number of dosing points accepting OAT referrals is larger than number of dosing points not accepting referrals	0.52 (0.28, 0.92)**
Number of prescribers accepting OAT referrals is larger than number of dosing points not accepting referrals	0.60 (0.25, 1.44)
Location (regional)	0.26 (0.15, 0.48)**
SEIFA quintile	0.94 (0.74, 1.19)
Constant	0.001
Pseudo R2	0.06
alpha	0.73 (0.47, 1.14)
Log likelihood	-196.30
Aikake Information Criteria (AIC)	404.59
Bayesian Information Criteria (BIC)	418.73
N (LGA)	78

*Significant at p < 0.1

**Significant at p < 0.05

Table 4 presents the adjusted regression results for Model 3, which examined the association between heroin-related ambulance attendances per LGA and the proportion of doctors and pharmacists with current vacancies, relative to the total number of doctors and pharmacists in that LGA who have provided OAT since 2020 (as listed in the statewide helpline report). Model 3 found that for a 10% increase in the proportion of OAT doctors with current capacity to accept referrals in an LGA, as a proportion of overall number of doctors in the statewide helpline report, is associated with an IRR of 0.91, indicating a 9% reduction in the risk of heroin-related harms. For example, in an LGA with 10 doctors listed (4 accepting OAT clients and 6 not accepting), an additional doctor with a vacancy (which would be increasing the doctor total proportion by 10%) would be associated with a 9% reduction in heroin-related harms. Similarly, a 10% increase in the proportion of OAT pharmacists with current vacancies was associated with IRR equal to 0.90, which is a 10% reduction in heroin-related harms. Consistent with Models 1 and 2, Model 3 also found that heroin-related harms were 72% lower in regional LGAs compared with metropolitan LGAs. Interaction terms were non-significant.

Table 4.

Adjusted negative binomial regression for heroin-related ambulance attendances per LGA (proportion of Doctors and pharmacists with current or limited vacancy over the total number of pharmacies or Doctors in an LGA)

Variable	Heroin IRR (95% CI)
Current OAT vacancy pharmacy proportion	0.90 (0.80, 1.01)*
Current OAT vacancy doctor proportion	0.91 (0.83, 0.99)**
Location (regional)	0.28 (0.14, 0.56)**
SEIFA quintile	1.00 (0.76, 1.33)
Constant	0.001
Pseudo R2	0.06
alpha	0.72
Log likelihood	-150.49
Aikake Information Criteria (AIC)	312.98
Bayesian Information Criteria (BIC)	324.80
N (LGA)	53

*Significant at p < 0.1

**Significant at p < 0.05

Discussion

This study identified the 14 Victorian LGAs with the highest rates of heroin-related ambulance attendances and examined their demographic characteristics. Heroin harms were observed across LGAs of all population sizes, socioeconomic quintiles, and in both regional and metropolitan areas. Notably, LGAs with a higher number of OAT prescribers and dosing points with current vacancies experienced a reduced risk of heroin-related ambulance attendances. Differences were also observed by geographic classification. Overall, regional LGAs had a lower risk of heroin-related ambulance attendances compared with metropolitan LGAs.

OAT is known to reduce heroin overdose harms [1–4], which in turn decreases heroin-related ambulance attendances [5]. Our study suggests that elevated harms in certain areas may, in part, reflect limited OAT availability, leaving individuals who could benefit from treatment unable to access it. Targeted planning to identify priority areas and increase the number of prescribers and dosing points with current vacancies could ensure timely, efficient, and locally accessible OAT provision. Population-based service planning has the potential to not only reduce costly and resource-intensive ambulance attendances, but also to strengthen preparedness for future synthetic opioid outbreaks in Australia.

Model 1, which assessed whether there was any prescriber or dosing point with any current vacancy in an LGA, was not associated with changes in heroin-related harms. In contrast, both Model 2 (more dosing points accepting referrals than not accepting referrals) and Model 3 (higher proportions of prescribers and dosing points with current vacancies) were associated with reduced harm. This suggests a dose-response relationship. A single available OAT service in an LGA may be insufficient to influence harm levels, whereas increasing the number and proportion of services with vacancies is associated with measurable reductions.

These findings highlight the need to incentivise and expand OAT service provision in high-need areas, particularly increasing the number of pharmacists operating as dosing points. Potential supply-side strategies include revising funding models, providing targeted incentives for establishing services in priority LGAs, streamlining prescriber and pharmacist training, and expanding the scope of pharmacists to include prescribing under collaborative care models. Such pharmacist-prescriber collaborations have been shown to reduce prescriber workload and waiting lists, while improving client outcomes [42], addressing a key barrier to OAT access [8].

While the lack of current availability is a critical supply-side barrier, demand-side factors also contribute to low OAT uptake [8]. These include stigma, negative perceptions of OAT, lack of flexibility in treatment models, and travel distance [8]. Addressing these barriers may involve improving public transport links, extending service opening hours, and tailoring service models to increase accessibility, particularly in rural and regional areas, where travel time, costs, and physical access are significant obstacles [43].

Patterns of heroin and synthetic opioid use and harms also differ geographically. Our study identified notable metropolitan-regional differences in the relationship between heroin-related harms, with regional areas having approximately 70% lower heroin attendances compared to metropolitan areas. Metropolitan and large regional hub areas often experience higher rates of heroin harms [37], due to higher levels of use and availability in metropolitan areas. Higher drug use in metropolitan areas likely occurs because illicit opioids typically enter Australia via major cities, which therefore means reduced access and use of illicit drugs in regional areas [44]. This is reflected in our finding that of the fourteen LGAs with higher than average heroin-related ambulance attendances, six out of the top seven LGAs were metropolitan and only one was regional, this is despite regional Victoria covering a much larger land mass. Furthermore, we note that the LGA with the highest heroin harm in Victoria (Yarra) has a medically supervised injecting room (MSIR) located within in. Medically supervised injecting rooms have been found to reduce heroin-related overdoses [45, 46] and therefore ambulance attendances related to heroin. This indicates that without the MSIR the mean number of heroin ambulance attendances in this LGA would likely be larger.

Overall, our findings underscore the importance of strategically locating OAT services to align with population need. Careful data-driven planning of OAT services based on population need has the potential to improve equity of access, reduce heroin-related harms, and strengthen preparedness for future synthetic opioid outbreaks. Future research should identify optimal service placement to ensure equitable treatment access for all Victorians, and evaluate whether expanding OAT availability in high-need areas leads to measurable reductions in opioid-related harms.

Limitations

This study has several limitations. First, the statewide helpline report, while regularly updated and containing most Victorian OAT services and their availability status, is primarily a clinical referral tool and does not capture all prescribers and dosing points. This may have resulted in an underestimation of service availability in some locations. Second, the statewide helpline is not a real time data set and therefore vacancy status may have changed within the 4-month period. Third, the analysis was limited to Victorian data, and findings may not be generalisable to other Australian jurisdictions. Fourth, the NASS ambulance dataset uses conservative coding protocols, which may lead to under-reporting of heroin-related attendances [30]. Finally, only two datasets were combined, limiting the ability to incorporate individual-level variables such as OAT use, engagement with other services, and hospital admissions. Future research linking additional datasets, including primary care (Medicare Benefits Schedule data), prescribing (Pharmaceutical Benefits Schedule data), emergency department (ED) and hospital records, would provide a more comprehensive understanding of heroin-related harms and support more precise OAT service planning and mapping.

Conclusion

An increased number of OAT prescribers and dispensers with current vacancies in an area was associated with a dose-responsive reduction in heroin-related ambulance presentations. This suggests that limited access to OAT, despite its proven ability to reduce premature death, overdose, and illicit opioid use, may be contributing to elevated harm rates. Expanding the number and capacity of OAT providers through targeted incentives, streamlined training, and enhanced service models, could reduce heroin-related ambulance attendances and improve preparedness for future synthetic opioid threats. While certain areas exhibited higher rates of heroin harms, these patterns were not driven by specific demographic characteristics, underscoring the need for OAT service planning to be guided by population need rather than geography alone.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

NH conceived the idea, wrote the manuscript, completed the data analysis and prepared all the figures and tables. DL aided with idea development, reviewed and provided significant feedback for the manuscript. RL provided data and overall project advice and ideas. RO reviewed the manuscript and aided with data analysis. BR reviewed the manuscript and provided significant guidance with the data analysis. ZN reviewed the manuscript and provided significant content feedback.

Funding

No funding or financial support was received for this study.

Data availability

Data is confidential and is therefore not available for dissemination.

Declarations

Ethical approval

This study received ethical approval from the Eastern Health Research Ethics Committee (ethics approval: E122-0809-14552). The Eastern Health Human Research Ethics Committee acts in accordance with the by-laws of Eastern Health, and operates within the guidelines of the Australian National Health and Medical Research Council (NHMRC).

Consent to participate

Not applicable.

Consent to publication

All manuscript authors, Natasha Hall, Bosco Rowland, Rowan Ogeil, Rick Loos, Ziad Nehme and Dan Lubman, provide consent for their work to be published.

Competing interests

The authors declare no competing interests.