Adolescence and early adulthood are critical developmental periods characterised by significant physical, emotional and cognitive changes. Reference Larsen and Luna1 At the same time, stressors related to family, health, social, economic and environmental factors can detrimentally impact a young person’s emotional well-being. Reference Reiss, Meyrose, Otto, Lampert, Klasen and Ravens-Sieberer2 Accordingly, between 10 and 25 years of age represents the peak risk period for onset of mental disorders. Reference Solmi, Radua, Olivola, Croce, Soardo and Salazar de Pablo3 Challenges during this life stage are often complex, comorbid and multifaceted, reflecting the diversity of young people’s experiences. Reference Gao, Telford, Filia, Menssink, Albrecht and McGorry4
Globally, the prevalence of mental disorders for young people aged 14–24 years is approximately 14%, Reference Kieling, Buchweitz, Caye, Silvani, Ameis and Brunoni5 although some data suggest even higher rates. For example, according to the Australian Bureau of Statistics (ABS) National Study of Mental Health and Wellbeing, 2020–2022, 39% of individuals aged 16–24 years experienced a mental disorder in the previous 12 months, the highest rate across all age groups. 6
In high-sociodemographic index countries, these mental disorders account for 25% of the total burden of disease among individuals aged 10–24 years. 7 This burden continues to increase as young people face growing uncertainty in all aspects of daily life, from social relationships to education and employment, a trend exacerbated by the coronavirus 19 (COVID-19) pandemic and concerns about the impacts of climate change. Reference McGorry, Mei, Dalal, Alvarez-Jimenez, Blakemore and Browne8
Australia has been a global leader in youth mental health, pioneering early intervention models and service innovations that have set international standards. Reference Tapp, Scheurer, Burgess, Le, Mihalopoulos and Currier9 This includes services such as the Early Psychosis Prevention and Intervention Centre (EPPIC, Victoria) and the national headspace network of enhanced primary care, Reference Tapp, Scheurer, Burgess, Le, Mihalopoulos and Currier9 alongside schemes such as Better Access, which provides Medicare-subsidised mental healthcare. Reference Tapp, Scheurer, Burgess, Le, Mihalopoulos and Currier9 Models such as headspace have contributed to improved access to youth mental healthcare, reduced stigma and the provision of more holistic, youth-friendly services. Reference Rickwood, Paraskakis, Quin, Hobbs, Ryall and Trethowan10 At the same time, as with any large-scale service reform, several challenges have been identified, including variability in effectiveness, retention and service integration, as well as questions regarding the capacity to meet the needs of more complex presentations and considerations of cost. Reference Kisely, Bastiampillai, Allison and Looi11 These ongoing areas of evaluation are important for informing continued refinement and optimisation of service models.
Despite the increasing prevalence and burden of mental disorders in young people and the availability of services such as headspace, individuals and their families still struggle to navigate systems and access developmentally appropriate care. Reference Hodgins, McHugh, Hu, Georgiou, Paul and Eapen12 Integrated service delivery across primary and tertiary care settings is essential, Reference Hodgins, McHugh, Hu, Georgiou, Paul and Eapen12 but we lack a clear understanding of care pathways and the services young people access for their mental health and broader health and social support needs. Other relevant sectors, such as alcohol and other drug, housing, child protection, youth justice and social support, are poorly integrated, resulting in fragmented care and missed opportunities for holistic support.
We also do not fully understand how care pathways and cross-sector service access can be restricted by geographical boundaries, age limits and rigid eligibility criteria, and services may be perceived as inaccessible or culturally unsafe by some population groups. These barriers might disproportionately affect specific subgroups, including young people in disadvantaged areas, First Nations youth, ethnic minorities and those from lesbian, gay, bisexual, transgender, queer or intersex (LGBTQI+) communities. Reference Filia, Teo, Brennan, Freeburn, Baker and Browne13,Reference Gao, Filia, Bedi, Menssink, Brown and Rickwood14 Services also tend to prioritise acute presentations and crisis response, rather than adopting preventive or long-term recovery strategies. Reference Barratt, Rojas-García, Clarke, Moore, Whittington and Stockton15
Calls for reform across both mental health and related service sectors have been made at state and federal levels. 16,17 However, implementation remains challenging due to a lack of clarity of service availability and limited understanding of who receives care and where gaps exist. These issues hinder efforts to improve the quality, efficiency and equity of mental healthcare.
To inform meaningful reform, alternative research approaches are needed. While epidemiological studies provide valuable insights into the prevalence and risk factors for mental illness among young people, Reference Gustavson, Knudsen, Nesvåg, Knudsen, Stein and Reichborn-Kjennerud18 traditional cohort designs often face challenges with participant engagement and retention, particularly among historically marginalised groups. Reference Brownell and Jutte19 These limitations can compromise data completeness and generalisability of findings. Reference Bonevski, Randell, Paul, Chapman, Twyman and Bryant20 In parallel, there is a wealth of administrative data collected across services and sectors. 16 Data linkage offers a powerful complementary approach by integrating these datasets across multiple domains, Reference Andrew, Sundararajan, Thrift, Kilkenny, Katzenellenbogen and Flack21,Reference Cotton, Menssink, Hamilton, Filia, Teo and Wang22 enabling a more comprehensive picture of service use patterns, pathways through care and associated outcomes than any single dataset can provide. Nordic countries including Denmark, Finland, Iceland, Norway and Sweden, have led this field, supported by nationwide registries and unique personal identifiers that facilitate large-scale, high-quality linkage across health, population, education and social data. Reference Laugesen, Ludvigsson, Schmidt, Gissler, Valdimarsdottir and Lunde23 In contrast, Australia lacks a universal population identifier across sectors, limiting routine linkage. As a result, linkage studies require careful planning, bespoke approvals and coordinated infrastructure to enable cross-sector integration. Reference Smith and Flack24 Although data linkage is well established in other areas of health research, their uptake in mental health has been comparatively slower and has often focused on outcomes such as suicide or hospitalisation. Reference Cotton, Menssink, Hamilton, Filia, Teo and Wang22
There have been some instances of linkage studies involving young people and mental health service use; Reference Gubi, Sjöqvist, Viksten-Assel, Bäärnhielm, Dalman and Hollander25,Reference Dalsgaard, McGrath, Østergaard, Wray, Pedersen and Mortensen26 however, these have largely focused on associations between early mental health and later outcomes, rather than explicitly modelling cross-sector service utilisation pathways or continuity of care. By linking data across health, mental health and human services, we can better understand young people’s care needs, preferences and experiences, and gaps in service provision. Reference Cotton, Menssink, Hamilton, Filia, Teo and Wang22 This approach has the potential to inform more coordinated, responsive systems of care and mitigate the wide-ranging and long-term impacts of mental ill health during adolescence and young adulthood. Reference McGorry, Mei, Dalal, Alvarez-Jimenez, Blakemore and Browne8
‘The who, what, where, when and why of primary youth mental health care: The 5W research program’ is a National Health and Medical Research Centre (NHMRC) Partnership grant (#1198696). The programme comprises three data linkage studies with the overarching aim of better understanding the mental health, health and human services needs of Australian young people experiencing mental ill -health. Specifically, it seeks to identify who needs care, why they need it, what care they require, where it should be provided and when it is most effective. In doing so, the programme aims to inform youth mental health policy and service reform.
Although not the primary focus of this paper, the 5W programme includes a complementary linkage study integrating the headspace minimum dataset with Victorian mental health, health and human services data (5W-hL). A further study, the 5W-National Health Data Hub (5W-NHDH), will leverage the NHDH to examine national patterns of health service utilisation among young people, including Federal data assets such as Medicare Benefits Schedule and Pharmaceutical Benefits Scheme data. In this paper, we describe the protocol and methodologies for the first 5W study, the 5W-Population Linkage (5W-PL).
5W-PL is a comprehensive data linkage study integrating Victorian service use and outcome data across mental health, health and human services, with a focus on the holistic care needs of young people with mental disorders. Focusing on Victorian young people provides a unique opportunity to examine mental health use within a rapidly reforming and highly fragmented system. Unlike Nordic settings, where integrated register systems are well established, Victoria’s multi-sector landscape allows investigation of real-world service navigation, discontinuities in care and the ‘ missing middle ’ . The study will also enable examination of the impacts of major policy reforms, including the Victorian Royal Commission into Mental Health. 17 In addition, 5W-PL will examine the impacts of the COVID-19 pandemic on young people’s mental health and service use, leveraging Victoria’s distinctive context, including the longest cumulative lockdown period globally (∼262 days between 2020 and 2021).
Specific aims of 5W-PL are to: (a) identify subgroups of young people based on their pathways through care and health service use behaviours (e.g. varied care intensities, use of services across sectors), including those who encounter treatment delays, treatment discontinuation or challenges in securing access to care; (b) understand geographical patterns of service needs and access by subgroup to inform enhanced regional planning of youth mental health services; and (c) develop predictive models of type and intensity of health services required based on clinical, socioeconomic and environmental factors.
Method
Study design
This study forms part of a larger NHMRC funded research programme (#1198696, 5W), examining health service utilisation among young people who have accessed a health service for their mental health problems. This research programme was formed through partnerships with Orygen, headspace National, Ambulance Victoria and the Victorian Department of Health, and includes academics from The University of Melbourne, University of Canberra, University of Sydney, Monash University, Deakin University, Victoria University and the University of Auckland. The research programme encompasses three data linkage studies. The current study, 5W-PL, involves data linkage with Victorian mental health, health and human services.
Setting
Here, we focus on Victorian population data across multiple sectors because of our existing strong partnerships with Victorian government departments and Ambulance Victoria; however, this protocol could be adapted and applied to other Australian states. We further note the difficulties in conducting cross-jurisdictional research in Australia due to infrastructure, legal and regulatory challenges. Reference Olver27
The Victorian Department of Health governs the Centre for Victorian Data Linkage (CVDL; https://vahi.vic.gov.au/ourwork/data-linkage). The CVDL manages a record linkage system for health and human services databases, which can be accessed for research purposes. The CVDL will identify a cohort of individuals from an established data linkage framework, as outlined below.
Expert consultations
During the planning stages of this study, consultative activities were conducted with young people, professionals involved in their care (clinicians and service providers) and technical advisors from the CVDL. Orygen’s Youth Research Council provided input on the acceptability of the study design and data management procedures, ensuring that the study aligns with youth perspectives. A dedicated 5W youth advisory group will assist in ensuring that translation activities are meaningful and acceptable to young people, including those who access mental health services. Other experts, such as policy makers, have and will be engaged throughout the study.
Cohort
Our study cohort will comprise individuals born between 1 January 1970 and 31 December 2010 who have been in contact with any of the included services (see Table 1). The primary group of interest is young people aged 12–25 years as of January 2018, encompassing the critical developmental period marking the transition to adulthood. This period represents the highest risk for the onset of mental disorders and aligns with the target age range for youth-specific primary and specialist mental health services. Reference McGorry, Mei, Dalal, Alvarez-Jimenez, Blakemore and Browne8 To allow for comparative analyses, we will also include younger and older age groups. Including a younger cohort enables examination of factors preceding the onset of mental health difficulties in adolescence and young adulthood, while the older cohort allows investigation of longer-term outcomes and the effects of treatment after transition from youth to adult services. Health and human service use data will be sourced from 1 January 2015 to 31 December 2025 (noting that the final date may vary across datasets due to differences in service-level data management and reporting, ranging from June 2019 to December 2025). This time frame enables examination of service use patterns across pre-, during- and post-COVID periods.
Datasets to be sourced for the 5W-PL study
Table 1 Long description
The table contains information about various datasets used in a study. It has six columns: Name, Abbreviation, Description, Available data, and Date data collection started. The table includes 15 rows of data, each representing a different dataset. Row 1: Mortality, COD URF, ABS codes cause of death data from the Registry of Births Deaths and Marriages using ICD-10., Age at death; gender; date of death; state/territory of usual residence; underlying cause of death; coroner case flag., January 1937 (however, time lags in data access exists). Row 2: Victorian Death index, VDI, The Victorian Deaths Index (VDI) data are sourced from the Registry of Births, Deaths and Marriages., Demographics; date of death; age at death; cause of death (ICD codes are not available)., January 1994. Row 3: Clinical public mental health services, CMI/ODS, Clinical specialist mental health services focus on assessment and treatment of people with a mental health condition. This database comprises two systems: the Client Management Interface (CMI) is the local client information system used by each public mental health service and the Operational Data Store (ODS) manages a set of selected data items from each CMI., Mental health client demographics; mental health service admission details; indicators for alcohol and drugs or justice services; mental health programme and referral; mental health diagnosis and treatments; mental health and well-being details., January 1960. Row 4: Mental Health Community Support Services, MHCSS, MHCSS provide non-clinical mental health services, for example, supporting people with psychiatric disability to achieve a broader quality of life. Services include Individualised Client Support Packages, Youth Residential Rehabilitation Services, Adult Residential Rehabilitation Services and Supported Accommodation Services., Service user demographics; disability group; carer details; client support units and service types; referral type; recovery plan details; mental health assessments., July 2008. Row 5: Community Health Minimum Data Set, CHMDS, Each year, the Community Health Program provides over one million hours of allied health, counselling and nursing services to the Victorian community, including services to particular groups including children, vulnerable pregnant women, refugee and asylum seekers and those with chronic disease., Client demographics; health condition or diagnosis; mental health diagnosis; occasions of service; reasons for attendance; service type., Month 2000. Row 6: Victorian Admitted Episode Dataset, VAED, The VAED provides a comprehensive dataset of illness details and immediate outcomes following treatment during a patient’s admission to a Victorian hospital service., Demographics include residential location (SA2); cause, effect and nature of illness; type of diagnosis; procedures; source of funding (public v. private); and admission and discharge information., January 2000. Row 7: Non-Admitted Episode, VINAH, The VINAH dataset comprises data from the Family Choice Program (FCP), Specialist Clinics (OP), Post-Acute Care (PAC), Subacute Ambulatory Care Services (SACS), Transition Care Program (TCP) and Victorian HIV Service (VHS)., Client demographics; presence of a carer; usual accommodation; purpose of contact; episode characteristics; service delivery setting; type of service and provider; and factors affecting health (e.g. family violence, housing instability, mental health conditions)., July 2007. Row 8: Victorian Ambulance Clinical Information System, VACIS/AV, In Victoria, ambulance paramedics are required to complete an electronic patient care record for every incident they attend or for which they provide a service. Unlike other datasets included, ambulance data comprise both structured and free-text data., Operational details (e.g. location of the incident; dispatch time; resources used); clinical assessments (e.g. vital signs, patient condition); primary diagnoses; treatments provided and transport outcomes., July 2008. Row 9: Victorian Emergency Minimum Dataset, VBMD, The Victorian Emergency Minimum Dataset (VEMD) comprises demographic, administrative and clinical data regarding presentations at public hospitals with emergency departments., Demographics include residential location; diagnosis and procedures; arrival and departure details., July 1999. Row 10: Child Protection, CP, The Child Protection Data Set (CPDS) is an annual collection of information on child protection services provided by the Department of Health and Human Services (DHHS)., Child protection case demographics; child protection notifications, investigations and substantiations; care and protection orders; living arrangements; sibling relationships; intake and placement., January 1992. Row 11: Disability Services, DISAB, Disability support services provided under the National Disability Insurance Scheme (NDIS) and DHHS, including short- and long-term supports, behavioural supports, case management, therapy and supported accommodation., Client demographics; main source of income; type of disability carer availability; mental health needs; functioning; supports., July 2009. Row 12: Sexual Assault Services, SASS, Sexual assault services respond to individuals who experience sexual assault. Services include crisis care, counselling, group work, advocacy and telephone crisis services., Client demographics; referral details; client issues and service activities; case outcome; closure reasons., July 2008. Row 13: Specialist Homelessness Services Collection/Homelessness Data Collection, SHSC/HD, Homelessness services collection includes long-term housing programmes as well as temporary accommodation and homelessness support., Demographics; information about homelessness services provided; homelessness issues including crisis, family violence and homelessness risk., June 2015. Row 14: Victorian Alcohol and Drug Collection (formerly Alcohol and Drug Services), VADC/ADIS, The VADC comprises data from alcohol and drug treatment services funded by the Victorian Government. These agencies fund providers that include alcohol and other drug treatment, drink-drive assessment and education and Local Drug Strategy Primary Health Services., Demographics; alcohol and drugs; family violence; homeless, justice and mental health indicators; drug use and treatments; physical and mental health assessments., July 2000.
5W-PL, 5W-Population Linkage; ABS, Australian Bureau of Statistics.
Outcomes and risk factors
We are interested in a broad range of outcomes, including mortality (rates and causes of death), diagnoses (health and mental health), frequency and type of services used (e.g. emergency, mental health, health and other human services) and cross-sectoral involvement (e.g. child protection, disability, sexual assault, housing support, drug and alcohol) (see Table 1 for data sources). We will explore the influence of various risk factors on outcomes (e.g. sociodemographic characteristics, residential location, housing stability, culturally and linguistically diverse background, Indigenous status, comorbidities, forensic history, family violence exposure and child protection involvement).
Data sources
Fourteen datasets have been selected to cover the domains of mortality, mental health, health, emergency services and human services. Mortality will largely be determined through the National Cause of Death Unit Record File and the Victorian Death Index, but may also be identified in other datasets. Key mental health datasets include the Client Management Interface (CMI)/Operational Data Set (ODS), which comprises details of any contacts with Victorian public mental health services and Mental Health Community Support Services (MHCSS) and is the clinical outcome registry for all publicly funded in-patient and out-patient mental health services in Victoria.
Health service datasets include the Community Health Program Minimum Dataset and both the Victorian Admitted Episode and Non-Admitted Episode databases. Attendance by an emergency ambulance will be identified through the VACIS® database, Ambulance Victoria’s electronic patient care record system. Emergency department presentations will be ascertained from the Victorian Emergency Minimum Dataset (VEMD). Human service datasets include Child Protection, Disability Support Services, Sexual Assault Services, Specialist Homelessness Services collection and Victorian Alcohol and Drug Collection (formerly Alcohol and Drug Services). Most of these datasets can be accessed via the CVDL, except for VACIS. Each dataset is described in Table 1.
Data linkage
The CVDL has developed the Victorian Linkage Map (VLM), comprising a collection of personal identifiers (names and date of birth) of all users of Victorian health and human services. This map is then used to link individuals’ data across multiple datasets using a combination of deterministic and probabilistic linkage methodologies. The CVDL extracts population-wide data (both individuals and episodes) for the cohort of interest from each dataset within the VLM and Integrated Data Resource (IDR; i.e. linked datasets; https://vahi.vic.gov.au/ourwork/data-linkage).
Ambulance Victoria data are not held by the CVDL. Ambulance Victoria’s paramedics record clinical information within the electronic Victorian Ambulance Clinical Information System. The data contain both structure fields and free-text case notes capturing contextual details such as event location, situation at the scene and patient transportation. Ambulance Victoria provides its de-identified dataset directly to the research team for processing and extraction of relevant variables. Data to be extracted include instances of mental health- or alcohol- and substance use-related attendances. Additional text mining will be conducted based on ambulance clinical notes to map more specific attendance information (e.g. suicidal ideation). After approval by Ambulance Victoria, these data will be securely transferred to the CVDL to be merged with existing VLM datasets to create the final integrated dataset with the rest of the linked data using the probabilistic linkage method.
The final linked data, comprehensive, pseudo-anonymised longitudinal records of treatment and service engagement episodes with mapped individuals across all datasets, will be securely stored and maintained by the CVDL in a restricted cloud environment. Access will be tightly controlled, with no direct internet connection and a regulated file import/export approval process.
Data processing
The CVDL will provide our team with individual datasets comprising the project identification number. A systematic approach will be used by the research team to evaluate the quality, completeness, accuracy and consistency of each dataset. Cohort demographic data, such as age, gender at birth, First Nations status, level of education and gender identity, will be modified to harmonise variables across datasets. Recognising that there will be inconsistencies in both coding and recording of some variables (e.g. gender), the research team will implement a systematic approach to data harmonisation and report this for transparency. We acknowledge that key variables relating to marginalisation are often incomplete or inconsistently captured in administrative datasets. Where available, we will incorporate these variables; however, analyses will explicitly account for limitations in data completeness and quality. A validation pipeline for linked data integration will be established to ensure linkage accuracy (e.g. mapping linkage errors such as service use recorded after death).
Diagnoses will be extracted based on available data for different datasets; for example, ICD-10 codes in hospital records and final primary assessments in ambulance records. Diagnostic categories of mental disorders and alcohol- and substance use-related conditions will be established following the Global Burden of Disease framework for ICD-10. Other diagnostic categories (e.g. neurological and physical) to be used will be based on the Australian Institute of Health and Welfare (AIHW) ICD character definition. 28 Residential areas will be harmonised or approximated from multiple datasets using residential postcodes or other geographical information (e.g. event postcode). Additional area-based social, demographic and environmental factors will be extracted from other available sources (e.g. population distribution, socioeconomic status markers and unemployment data from the ABS) and integrated into the linked data using geospatial information.
Data analysis
Detailed statistical analysis plans will be developed for each aim outlined above and other related areas of interest. The cohort under investigation may vary as a function of the area of interest relating to specific research questions. Broadly, we will incorporate a range of traditional and contemporary analytical techniques to address these research questions.
Diverse potential outcomes and risk factors will be harmonised across multiple datasets including the extraction of rich data from ambulance paramedic clinical notes. The large amount of unstructured text can help to understand the underlying causes and trajectories of acute mental health service use, which is otherwise not feasible with other structured datasets. Natural language processing and artificial intelligence methods (e.g. Bidirectional Encoder Representations from Transformers) Reference Devlin, Chang, Lee and Toutanova29 will be used to extract key features that can be incorporated into further analysis.
To delineate subgroups of young people according to pathways through care, we will use contemporary machine-learning techniques, including sequential pattern mining, deep learning, latent Markov model, ensemble clustering models and multi-view clustering models. Reference Gao, Dwyer, Zhu, Smith, Du and Filia30 Initially, these models will be applied to individual data sources to understand service use patterns within each. The models will then be extended to linked datasets to capture the interactions between different services and provide a more comprehensive understanding of care pathways. Strategies for dealing with missing data will be developed, incorporating imputation models seamlessly into cross-validation frameworks for model development. Reference Gao, Dwyer, Zhu, Smith, Du and Filia30 These analyses will allow us to address research questions such as: (a) What distinct patterns of service utilisation can be identified among young people? (b) Are there subgroups that differ in timing, intensity and sequencing of service use across sectors? (c) Are there identifiable profiles of ‘high-need’ or ‘high-risk’ trajectories (e.g. repeated acute care)?
We will also explore geographical patterns of service use, outcomes (e.g. mortality) and distribution of different subgroups identified through clustering methods. Spatial time-series models will be applied to reveal geographical and temporal risk factors associated with heavy acute mental health service use and to identify whether a relationship exists between subgroups with unmet needs and geosocial factors such as health service accessibility, provider availability and social inequity. Geospatial analyses will help inform future service planning for providers. A network of health services will also be developed to facilitate an understanding of care pathways within and between different services (e.g. transferring from youth psychiatric in-patient units to adult services). This will enable an evaluation of service availability in different areas, identify areas with unmet needs and inform future service planning. Questions that could be addressed by these analyses include: (a) How does service use and outcomes vary geographically across regions? (b) Are areas of high unmet need associated with service accessibility, workforce availability or socioeconomic disadvantage? (c) Do geographic patterns predict transitions into acute or crisis care?
To investigate the interplay between clinical, socioeconomic and environmental factors in relation to service use (i.e. subgroups identified by clustering models) and outcomes (e.g. mortality and the need for acute services), we will employ statistical and interpretable machine-learning models, such as multinomial logistic regression, generalised additive models, random forest and XGBoost causal forest. Reference Shatte, Hutchinson and Teague31 Building on these findings, we will apply a causal inference framework Reference Hernán, Wang and Leaf32,Reference Prosperi, Guo, Sperrin, Koopman, Min and He33 to evaluate counterfactual scenarios of service configurations. This approach will allow assessment of the impact of various care models on clinical outcomes, tailored to specific youth subgroups and at-risk profiles, providing critical insights to improve service design and delivery. Example research questions include: (a) Which clinical, socioeconomic and environmental factors are associated with poor engagement, high service use or adverse outcomes? (b) Which intervention points in the system yield the greatest benefit for high-risk groups?
As noted, our primary focus is on young people aged 12–25 years; however, other age groups (<12 and 26–53 years) will be used as comparators to determine differences in latent profiles, associated risk factors and temporal and spatial distribution.
Ethics
Ethics approval has been obtained from The University of Melbourne Human Research Ethics Committee (ID: 21041), with consideration of the appropriate ethical and design guidelines (e.g.). 34 Additionally, we will seek institutional approvals (e.g. Ambulance Victoria Research Committee) where required.
Confidentiality
Confidentialisation (the process of modifying data to protect individuals’ identities); 35 and a quality evaluation will be undertaken before the research team is provided with access to the linked data. Confidentialisation ensures that, even though there are no individually identifying data, a person cannot be re-identified based on recognisable patterns within the combined data. With personal identifiers removed, each case is assigned a new unique identifier, preventing re-identification. When reporting, we will adhere to the ABS data suppression rules 35 to ensure confidentiality (e.g. suppressing cells with small counts or dominated by a few contributors).
Data security
All datasets (with the exception of case note text from Ambulance Victoria data) are stored on the CVDL’s high-security, cloud-based technical platform, which utilises Information Security Registered Assessors Program Certification and multi-factor authentication. Access to this platform is restricted to specific team members (e.g. analysts) for a set period.
De-identified Ambulance Victoria data will be processed (e.g. for text mining) by the research team within a highly secure, isolated computing environment, the Secure Research Environment, at The University of Melbourne.
Dissemination
Processes and findings will be reported according to reporting guidelines for observational studies Reference von Elm, Altman, Egger, Pocock, Gøtzsche and Vandenbroucke36 and data linkage. Reference Benchimol, Smeeth, Guttmann, Harron, Moher and Petersen37 We will use an ongoing consultation process involving young people and stakeholder engagement to seek recommendations on the interpretation and dissemination of findings. Dissemination will include reports, peer-reviewed academic publications, presentations at national and international conferences and to government agencies overseeing various services, and policy recommendations. Both expert and youth advisory reference groups will play a significant role in translation activities to ensure that outputs are meaningful, impactful and are acceptable to a youth audience and consumers.
Discussion
There is a youth mental health crisis characterised by increasing rates of depression and anxiety, eating disorders and suicide. Reference McGorry, Mei, Dalal, Alvarez-Jimenez, Blakemore and Browne8 Despite this, many young people do not seek professional help, Reference Filia, Teo, Brennan, Freeburn, Baker and Browne13 face inequities in accessing services Reference Filia, Teo, Brennan, Freeburn, Baker and Browne13,Reference Gao, McDonald, Hamilton, Simons, Menssink and Filia38 and frequently find that existing services do not adequately address their needs. Reference Menssink, Gao, Zbukvic, Prober, Kakkos and Watson39 Our understanding of the additional health and human services that young people depend on to support their mental well-being remains limited, making it essential to identify gaps both within and across sectors.
This research programme represents the largest of its kind in Australia aiming to better understand the outcomes and care needs of young people with mental disorders. Responding to the Productivity Commission’s call for improved use of routinely collected administrative data, 16 we have adopted a holistic approach by examining young people’s needs beyond mental health and healthcare. Using data linkage methodologies, we will create a unique multi-sectoral data asset that can be used to improve our understanding of mental health, health and human service utilisation among young people experiencing mental health challenges in a way that could not otherwise be determined from individual datasets. This approach allows for the identification of subgroups who may benefit from either lower-intensity interventions or may require more intensive, tailored support to address complex needs, as well as for improving pathways through and between services and sectors. We will address specific systemic challenges such as the ‘missing middle ’ service gap, where poor service coordination, rigid eligibility criteria and inadequate care leave individuals in the system without appropriate support. Reference Menssink, Gao, Zbukvic, Prober, Kakkos and Watson39 Our findings will be shared with a diverse group of stakeholders, including individuals with lived experience, healthcare professionals and those involved in service planning, provision and evaluation.
The study has several additional strengths. Youth participation has been integrated from the outset, alongside engagement with key stakeholders across services, sectors and governments. These partnerships are essential for ensuring the relevance and impact of data linkage research. Reference Jewell, Pritchard, Barrett, Green, Markham and McKenzie40 The richness of the data will allow us to address a wide array of research questions, such as mapping the impact of COVID-19 on mental health outcomes and service access and enabling preparedness for future crises. Additionally, the protocol developed for this study is reproducible across other Australian states and territories, enhancing its scalability. The outcomes of this project will foster multidisciplinary collaborations, facilitating diverse perspectives to interpret findings and inform strategies for effective translation and policy reform.
Some limitations should be acknowledged. First, we note the broad barriers associated with conducting data linkage research, including high linkage and data storage costs, lengthy ethics approvals, extended data linkage and processing times, complex coordination with multiple data custodians and the challenging structure of linked data for statistical analysis. 16
Second, there can be issues with the data content and quality. Services and sectors often operate with different systems and administrative platformsand use diverse identifiers and routine outcome measures. Reference Cotton, Menssink, Hamilton, Filia, Teo and Wang22 There can also be a lack of information regarding clinical operational models, care quality and risk factors. Specific subgroups may be underrepresented, including those with First Nations or culturally and linguistically diverse (CALD) backgrounds, because of greater challenges for these groups in accessing and navigating healthcare and social services. 41 Following rigorous methodologies and appropriate guidelines, we will document any issues with data quality.
Third, we focused on health and human service use within a single state because of longstanding challenges with cross-jurisdictional data linkage in Australia. Reference Andrew, Sundararajan, Thrift, Kilkenny, Katzenellenbogen and Flack21,Reference Olver42 In addition to practical challenges, key data gaps may affect the study’s ability to capture full service use, particularly the limited linkage to national datasets (e.g. Commonwealth-funded services). To address this, analyses will explicitly account for data coverage, interpret findings within defined boundaries and use sensitivity analyses where possible. Findings will be reported transparently, with careful consideration of whether observed patterns reflect true service use or limitations in data availability.
However, recent advances in data linkage methodologies and infrastructure are increasingly enabling integration of data across services, sectors and jurisdictions. Reference Cotton, Menssink, Hamilton, Filia, Teo and Wang22 These limitations will be progressively addressed through the broader 5W research programme, including the 5W-NHDH study, which covers linkage across jurisdictions, including with Commonwealth datasets (e.g. Medicare Benefits Scheme, Pharmaceutical Benefits Scheme). We will also integrate Victorian population-level data with additional sources, including headspace National clinical and demographic data (5W-hL), to gain a more thorough understanding of, and enable prediction of, service utilisation and outcomes in Australia’s evolving mental health system. This approach will inform practice and drive policy reform in Australia and internationally.
Establishing this 5W-PL linked data asset is a complex, multi-stage process requiring data access approvals, governance negotiations, data cleaning and linkage, which can take several years to complete. In the current context, timelines of approximately 2–4 years are typical, particularly when linking across multiple sectors with differing custodians and legislative requirements. Despite these challenges, the resulting asset provides a uniquely powerful platform to examine longitudinal, cross-sector service utilisation and outcomes at scale. While situated within the Australian context, this work has clear international relevance. Many countries are seeking to better integrate health and human services data to address system fragmentation and unmet need among young people. As such, this study offers a model for developing and leveraging linked administrative data to inform policy, service planning and early intervention strategies in comparable health systems globally.
Data availability
Data sharing is not applicable to this article as no datasets were yet generated or analysed for the purpose of the current study.
Acknowledgements
We would like to acknowledge Amelia Callaghan, Nic Telford, Jason Trethowan and Lan Du for their ongoing role in the 5W Investigator team and their significant contributions to the development of the 5W research programme.
Author contributions
S.M.C., C.X.G., K.M.F., M.P.H., J.M.M., D.J.R. and C.I.M. contributed to the conceptualisation of the study. All authors contributed to the methodology. A.E.W. and J.M.M. contributed to the writing of the original draft. All authors contributed to the review and editing of the manuscript. S.M.C. and C.X.G. provided supervision. S.M.C. and C.X.G. contributed to project administration. S.M.C., C.X.G., K.M.F., D.J.R., P.D.M., I.B.H., S.E.H., A.P., A.R.Y., K.S. and C.I.M. contributed to funding acquisition.
Funding
The project is funded by the NHMRC (GNT#1198696) and is a collaborative partnership between Orygen, headspace, the University of Melbourne, the Department of Health and Ambulance Victoria. Individual authors report the following sources of support: K.W. – NHMRC Emerging Leader Investigator Grant (GNT#1177787) and a Dame Kate Campbell Fellowship from the Faculty of Medicine, Dentistry, and Health Sciences at The University of Melbourne. Z.N. is supported by fellowships from the National Heart Foundation (GNT#105690) and the NHMRC (GNT#2034615).
Declaration of interest
I.B.H. is the Co-Director, Health and Policy, at the Brain and Mind Centre (BMC), University of Sydney, Australia. The BMC operates an early-intervention youth service, headspace Camperdown, under contract to headspace. I.B.H. has previously led community-based and pharmaceutical industry-supported (Wyeth, Eli Lilly, Servier, Pfizer, AstraZeneca, Janssen Cilag) projects focused on the identification and better management of anxiety and depression. He is the Chief Scientific Advisor to, and a 3.2% equity shareholder in, InnoWell Pty Ltd. which aims to transform mental health services using innovative technologies. K.W. is a member of the BJPsych Open Editorial Board. She did not take part in the review or decision-making process for this manuscript. All other authors have nothing to disclose.
Ethical standards
This study was approved by the University of Melbourne Human Research Ethics Committee (HREC approval number 2023-21041-44208-4). The project involves analysis of routinely collected administrative data. As no direct contact with participants will occur and data will be de-identified prior to analysis, a waiver of consent was granted in accordance with the National Statement on Ethical Conduct in Human Research (2007, updated 2018).
Open access
eLetters
No eLetters have been published for this article.