Mental health inequalities

Inequalities in mental health disorders and mental healthcare are well established. Those who are disadvantaged, such as older individuals, people from minoritised ethnic backgrounds and those with lower income and education levels, are more likely to suffer from mental health disorders, but are also less likely to access or benefit from treatment. The growing incidence of mental health disorders worldwide, partly linked to the COVID-19 pandemic, geopolitical events and cost-of-living crises, is set to exacerbate a vicious cycle of inequity. While digital mental health treatments, such as evidence-based intervention programmes and applications providing assessment and monitoring, offer effective ways to overcome some traditional barriers to accessing care, this shift risks leaving behind certain groups in society. This has the potential to widen the very inequalities that digital care was meant to address and perpetuate a situation where those most in need of mental health support are least able to access it. Better understanding the drivers of inequalities in mental healthcare, and what we can do about them, requires a focus on the groups who have been underrepresented in these settings. The generation of such data can then support the development of new insights that have the potential to inform clinical practice.

Identifying and reducing healthcare inequalities through data science

Data science is a diverse field bringing together informatics, data analytics and statistics to derive insights from data. In recent decades there has been a revolutionary expansion in the collection and use of vast amounts of data of different types across a multitude of sectors. Using big data (data-sets characterised by complexity and variety) and advanced analytics, it is possible to move beyond describing inequalities to predicting risk and identifying equitable interventions. It enables a shift away from binary frameworks to a more nuanced consideration of the intersection of multiple individual characteristics like age, gender, ethnicity, disability and geographic location, which create complex patterns of inequality. Data science is also fundamental to how we interpret emerging data types, such as those collected through digital sources, and to harness such information to support clinical decision-making. This holds the potential to provide practitioners with information about clinical needs, individual response to treatment and longer-term outcomes from care. Data science is therefore an essential tool to further aid the identification and reduction of mental health inequalities.

Through the analysis of routinely collected healthcare records, it might be observed that some patient groups are less likely to be assessed or offered care by services following referral, or that there are specific groups for whom certain treatments do not yield the same benefits as others. Understanding such causes of inequalities is essential to reducing them, and data science can help design better targeted interventions and personalised approaches. Such work has been conducted in the context of inequalities in cancer care where linked administrative data in Wales has identified characteristics of people less likely to participate in screening ^{Reference Chiovoloni, Bedston, Abbasizanjani, Antoniou, Gordon and Akbari1} and used to help identify racial and geographical disparities that increase the risk of poor maternal outcomes. ²

What about the data in psychiatry?

A primary obstacle to applying data science in psychiatry is the limited availability of consistent and actionable clinical data. Historically, there have been significant challenges regarding both the populations and the variables measured within mental health services, making inequalities difficult to recognise. Despite national guidance in the UK emphasising the need for routine outcome monitoring, ³ this practice is not consistently implemented across clinicians in community mental health services. Consequently, the data that are collected focus on administrative and service-planning metrics, such as admissions, discharges and referrals, and often these can’t be followed between services. While useful for logistical purposes, this information provides insufficient insight into how services can improve, or how health inequalities can be reduced for individual patients, teams or services. Notable exceptions exist, particularly within specialised service delivery models, such as Early Intervention for Psychosis services, where a tool (PsyMaptic) has been developed that used epidemiological data to forecast expected service need in different areas of England. ^{Reference McDonald, Ding, Ker, Dliwayo, Osborn and Wohland4} Such tools can model service user characteristics to help identify inequalities in access. However, the historical lack of standardised and routinely collected outcome measures makes it difficult to examine corresponding inequalities in treatment outcomes, limiting our ability to assess the effectiveness and equity of the care provided.

Overcoming the challenge of inconsistent data requires a whole system approach to measurement, a feat achieved in several healthcare models, for example the National Health Service Talking Therapies for anxiety and depression (NHS TTad) programme in England. The mandated use of sessional outcome measurement has been instrumental in improving care. This has been achieved through a standardised data collection specification, allowing the evaluation of treatment outcomes for the hundreds of thousands of patients who receive treatment each year. Equally valuable is the collection of standardised patient data on sociodemographic and clinical characteristics. This data-set has been crucial for identifying significant inequalities in both service access and clinical outcomes, and provide the ability to track how these inequalities change over time to model how services might be improved to mitigate them. ⁵ Services now have the ability to monitor how their initiatives improve patient outcomes. ^{Reference James, Saxon and Barkham6}

Such a large, standardised data-set is unfortunately rare in psychiatry. The first major barrier is a historical and cultural resistance among some clinicians to routine collection of outcome data. This often stems from concerns over time burden, ethical considerations and perceived lack of value. For many, the introduction of such measures can appear at odds with the empathic, trust-based and highly individualised nature of psychiatric care. Despite demonstrated benefits in services where outcome measurement has been effectively implemented, this remains a significant hurdle.

For clinicians who are using outcome measurement to improve their practice, a further issue is heterogeneity in outcome measures used, as it becomes difficult to compare between individuals, let alone groups, to examine inequalities. A previous requirement to use the Health of the National Outcome Scales was viewed by many as serving bureaucratic purposes and less about patient needs. The agreement of a set of patient-reported outcome measures (PROMs) for people with severe mental illness ³ might begin to address existing issues, but do not resolve concerns about the time required to complete them routinely.

Furthermore, to accurately identify inequalities by patient characteristics, it is necessary to achieve consistent recording of inequality markers. This often proves a challenge; missing information on demographic information beyond age and gender is common, and individuals from disadvantaged backgrounds can be more reluctant to provide detailed sociodemographic information. However, programmes such as NHS TTad show that these barriers can be overcome in many scenarios, offering a blueprint for a systemic shift in how data and measurement are approached in psychiatry.

How can we improve data in psychiatry?

Improving data collection in psychiatry requires a multi-faceted approach, integrating cultural shifts, data linkage and technological innovation. Key strategies include securing clinician buy-in, enriching data-sets through linkage and leveraging new technologies for data capture and analysis.

The first step is convincing clinicians and services of the value of standardised data collection, by demonstrating how collecting data routinely can both inform the care of individual patients but also shape service improvement. Showcasing evidence of tools and methods across different disciplines will be vital to these efforts, alongside methods to address concerns around the burden of this work.

The scope of psychiatric data can be significantly expanded through data linkage, incorporating information about physical and environmental factors. Linking to broader clinical and epidemiological data-sets (e.g. those which identify inequalities in cancer screening ^{Reference Chiovoloni, Bedston, Abbasizanjani, Antoniou, Gordon and Akbari1} ), can provide a greater understanding of the interaction between mental health, physical health and the social determinants of health.

New technologies provide a further vehicle for increasing the availability of data in routine care. New tools such as natural language processing have the potential to ‘create’ analysable data from either free-text (e.g. in case records), or from transcripts of clinical appointments, generating data that might uncover inequalities in how care is provided. Such methods have been used to develop models from electronic health record data that demonstrate accuracy in predicting psychosis risk. ^{Reference Irving, Patel, Oliver, Colling, Pritchard and Broadbent7} Although such approaches have are still be tested in the real world, they have clear potential application to identify inequalities, and how these might be addressed. The emergence of advanced Large Language Models may enhance these approaches, for example, by accurately extracting sociodemographic features alongside clinical characteristics from patient records. Implementation of these tools poses ethical and logistical challenges, including data protection, patient privacy, the risk of model inaccuracies and the practicalities of integration into existing clinical workflows, which need to be overcome.

The importance of patient and public involvement and engagement

Patient and public involvement and engagement (PPIE) will be essential for improving the quality and relevance of data in psychiatry and achieving equity in mental healthcare. As the ultimate goal of psychiatry is to support and serve patients, incorporating patient voices in how data is collected, used and interpreted is indispensable. PPIE members bring unique insights into what data should be collected and why. They are uniquely able to help identify outcomes that matter most to patients and highlight the most meaningful markers of inequality. For example, PPIE directly contributed to the development of the EQ Health and Wellbeing measure by identifying domains that were previously overlooked. ^{Reference Brazier, Peasgood, Mukuria, Marten, Kreimeier and Luo8} PPIE members can also help understand the burden of collecting PROMS, limitations to PROMs in capturing individual experiences of mental health, uncover biases in data, address concerns around data security and the use of sensitive information in predictive tools and raise standards of transparency and accountability in data practices. In the collection of psychiatric data, it will be extremely important to co-produce ethical frameworks with PPIE members to address privacy concerns and build trust, which is key to improving inclusion and equity. An open discussion about data practices with PPIE stakeholders will provide additional perspectives to uncover any hidden biases and inequalities across different levels of mental healthcare systems.

A major challenge in employing data science to reduce inequalities is the underrepresentation of minoritised groups who might be particularly hesitant to engage due to concerns around data security. Levels of stigma around mental health vary across different cultures and communities, and those who are most underserved by mental health services are often the ones in greatest need. It is therefore crucial to actively recruit and involve individuals from seldom-heard backgrounds in PPIE early on, ensuring that this is a safe avenue for their voices to be heard. Both researchers and PPIE members have noted that involvement in data science research is often difficult due to technical jargon and unfamiliar statistical concepts. Recent initiatives in statistical PPIE (e.g. PPI-SMART) ^{Reference Worboys, Broomfield, Smith, Stannard, Tyrer and Vounzoulaki9} aim to address these challenges by making statistical concepts more accessible. With appropriate support, PPIE members can meaningfully contribute to all stages of data science, and can serve as ambassadors of research quality, making sure that data science is not only ethical and socially relevant, but also technically robust. Sustained investment in PPIE within psychiatric data science will ensure that data science is conducted with and for patients, rather than about or to them.

Future of data science in psychiatry

Improving the availability of structured, consistent data is critical to accelerating research into the drivers of mental health inequalities, identifying characteristics increasing the risks that people are unable to access care, or are less likely to benefit from available treatments. Such advances will begin with mandating and incentivising better data adoption. These projects should not be developed in isolation. Progress requires partnerships with clinicians, service users (through PPIE) and community members. Knowledge exchange and the awareness of data science best practices are fundamental. Promoting the value of technology and measurement as tools to support clinical practice, not supplant it, will help develop further insights. Investing in data science training for clinicians, service users and the community can empower these groups to become effective advocates for data quality. This integrated approach is the most promising path towards a future where personalised medicine can effectively reduce mental health inequalities.