Mental health,big data and research ethics: Parity of esteem in mental health research from a UK perspective

Inequalities in research funding

Inequalities in service provision are replicated in research; as despite mental health conditions affecting approximately 23% of the UK population, it only acquires 6% of the UK health research expenditure. ⁹ Notably, less than 30% of that mental health research is focused on young people, despite this being the most common age for onset, and its consequences lasting through into adulthood. ¹⁰ Indeed, the MQ report demonstrated that while cancer research was awarded £402 m, mental health research only acquired £112 m despite figures showing that the cost of cancer to the economy was £12,853 m compared to the £51,600 m that mental health conditions cost. In other words, cancer attracted a research spend of £178 per person affected and yet mental health only attracted £8 per person affected. This points to serious inequalities in research spending and something that must be addressed. However, the British Research Councils are slowly beginning to address the gap between mental and physical health research, arguing that mental health is a new priority. For example, the Medical Research Council ¹¹ recently noted that the government has published a ‘Framework for Mental Health Research’ document that is shaping research funding strategies setting out core recommendations to improve the impact of mental health over the next decade. The focus for this is on prevention and early intervention, as well as increased engagements from patients and the public which will hopefully be reflected in the kinds of research undertaken.

Research on mental health data needs to be appraised in the context of the political discourse of parity of esteem and cannot be viewed as separate from it. This has implications for the ethical parameters and processes that are imbued in research. It is essential to undertake more research to better understand the nuances between the manifestation of symptoms, the aetiology of conditions, and any genetic contribution to illness. Furthermore, it is crucial to understand the barriers to help-seeking. However, this kind of research needs to account for the ethical issues that such sensitive work involves. Research areas could include understanding stigma, the narratives and views of those awaiting and receiving diagnosis, the lived experience of mental health conditions and the links to the social and psychological. It is only through this integration of genetic, quantitative and qualitative research that we can truly begin to strive towards parity of esteem, so that policy and practice can be established based on best evidence, and work that is person-centred, accounting for the positions, views and experiences of key stakeholders. With a rather disconcerting picture of disparity of esteem in mind, we now turn to more specific tensions presented when using BD in mental health research, developing a critique related to some of the taken-for-granted assumptions about its efficacy. We highlight some challenges and dilemmas before turning attention to the ethical challenges implicit in using BD to pursue knowledge in the field of mental ill-health.

The place of data in mental health research

BD and use of the digital media are ever more relevant in social life, with a growth of automated ‘production, use and analysis of transactional and other big data by private corporations and governments’ (Frade, ¹² p.864). A central concern in our argument regarding parity of esteem, and the ethics of mental health research, is the place of BD and the dilemmas created by its use with this population. In the digital age, increasing volumes of personal data on individuals’ health are available to be used for research purposes and can be ‘mined’ and analysed with benign intentions. NHS Digital in the UK holds patient data centrally for the benefit of patients, services and researchers. Their mental health section captures data which are routinely collected at the interface of patient care in specific categories. In addition, there are specific psychiatric databases being developed for researchers with some attention paid to the sensitivity of such data and the development of highly sophisticated systems to protect anonymity while also allowing researcher access. ¹³

There is however no consistency in the kinds of psychiatric databases which currently exist which is arguably ethically and practically problematic. In a ‘snapshot’ overview of mental health databases available globally, there were no definable boundaries in BD sets in mental health with some generic health databases containing mental health information and with others more specialised and related solely to mental health conditions. ¹⁴ The types and range of specific and non-specific data and the lack of integration of existing sources all present challenges for research. Even if these operational problems can be overcome, we argue that there is a need for a questioning approach to its use in mental health research and a reflective consideration of ethics.

The characteristics of data collected also limit the potential for knowledge generation. An example of this is the lack of completeness and comprehensiveness of data relating to BME populations and mental health where categories are largely defined through reference to the UK’s colonial past with the risk of omitting experiences of people who would define their ethnicity differently. ¹⁵ A critical approach to the use of BD would seek out the gaps highlighting how social contexts shape data collection, precision and comprehensiveness. Furthermore, social inquiry might raise epistemological issues, asking what kinds of discovery and inquiry are missing. While all data (not just BD) collected for research purposes require this kind of critical approach, the ethical dilemmas and ethical tensions raised in BD research are still emerging, especially when we consider the relationship of this with the rhetoric of parity of esteem. This means that researchers need to reflexively engage with the nature and purpose of their inquiry. We now turn to look at some specific ethical challenges presented.

Who is the research ‘subject’ in data?

BD research requires ethical reflection possibly requiring new regulatory frameworks to be devised or a consideration of how such research fits into existing frameworks. ¹⁶ Whatever such reflections might be, the use of BD for the purposes of mental health research challenges our understandings and implementation of research ethics in profound ways. Even the foundational question of identification, of who the human research ‘subject’ is, becomes problematic because humans are largely invisible in data. An ethically reflexive approach would nevertheless seek identification of a ‘data subject’ and what they are owed in ethical terms. ¹⁷ The conventional and theoretical foundations for ethical conduct in research (typically summarised as beneficence, non-maleficence, autonomy and justice) ¹⁸ are potentially problematic in BD research. This is because the ‘subject’ or ‘participant’ becomes integrated and embedded within the broader data corpus and individual contribution as well as individual consent, may be problematic, obscured or lost. In other words, it may not always be clear to individuals how or when their data may be used, in what context it may be embedded or integrated, or what contribution they may be making at some potentially unidentified point in the future. This is then important in terms of ethical obligations, as BD seeks out trends rather than more focused individual contributions, raising questions as to how we might differently construct the individual hidden in the data to incorporate what is owed ethically.

Harm/benefit ratio

Research on large sets of data is different. First, research is on groups or populations often seeking to generalise broad trends. Consent is altered because the individual is consenting to use of their data as part of a larger data set. The harms/benefits ratio requirement needs to be reconceptualised as it cannot be described in relation to an individual. Protection from risk may in this context be protection from what has been referred to as ‘data dredging’ (Corrigan, ¹⁹ p.41) where the temptation to investigate associations in treatments is irresistible but potentially flawed when data have been collected without the primary purpose of an investigation in mind. Perhaps then protection from harm is protection from poor research practice. Nevertheless, it is important to be mindful that susceptibility to risk is not universal ²⁰ and neither does exposure to risk always result in harm ²¹ at an individual or group level.

Ethics regulation ensures that researchers are mindful of individual harms, but they are not required to consider or account for harms that may affect communities. ¹⁷ The implications for mental health research are that researchers interact with data investigating particular questions, but they are not explicitly required to consider the consequences and problems of their research and the potential of such inquiry to lead to (unintentional) further stigma. It might be argued that the ethical concern with communities of participants could be incorporated into regulatory frameworks and made a requirement of review. However, at this stage of emerging ethical and epistemological challenges in using BD there is a prior consideration. Researchers need to reflexively consider whether they are undertaking research simply because certain kinds of data exist or are easily extracted. For example, in psychiatric research, the ease of access to data on medication has resulted in extensive medication-orientated research questions. ¹⁴ Researchers are working at a distance from the communities most concerned with the research and who are unlikely to have the opportunity to shape research questions. This raises new questions about who research ‘subjects’ are and what trust means in these emerging research contexts. These ethical concerns may be resolved but require researchers to engage ‘with a rapidly changing set of research dynamics that should be addressed in context’. ¹⁷

Vulnerability and protection from risk

The inclusion of participants with mental health conditions in research frequently raises concerns for ethics committees, and a common rationale for this is the protectionist position taken as these participants are usually constructed as a vulnerable group who are more likely to come to harm through participation. While we critically question treating those with mental health conditions as a homogenous group, we also consider the tensions that the notion of vulnerability raises in the context of BD, arguing that the homogenising of a population constructs and treats that population as more vulnerable than may be the case in practice.

The argued risk for this population is that those with mental health conditions may experience cognitive impairment or may find it difficult to make fully informed decisions about participation in research, and furthermore are subject to stigma and discrimination, which advances their vulnerable status. Such status is central to ethical decision-making, which has been heavily influenced by historical exploitation of these groups. However, to construct a group or population as vulnerable is complex, and there has been limited consensus in the evidence regarding how this construct should be defined. ²² Somewhat problematically, it has been used interchangeably with other concepts such as ‘hard-to-reach’, ‘sensitive groups’ or ‘hidden people’. ²³

The notion of vulnerability is indexical, not static, as it is tied to contexts ²⁴ which need to be accounted for when positioning groups. Vulnerability is not dispositional to the person but fluid because mental health conditions affect different people in different ways. In other words, it is not simply the characteristics of individuals or groups that define the population as vulnerable, but the context of the health research and how that may be associated with vulnerabilities. Those with health conditions may experience consent-based vulnerability, in the sense that they may lack the capacity to express their autonomy, or may experience fairness-based vulnerability, as mental health diagnosis renders them as members of a group who lack opportunity or freedoms leaving them susceptible to coercion. ²⁵

To summarise, in considering the ethics of BD research, vulnerability and its association with increased risk cannot be merely assumed or attributed to this population. This would underestimate the agency of these individuals to make their own decisions and deny the potential to influence research. Vulnerability is complex as a construct and it manifests in different ways. Indeed, it has been argued that all humans experience inherent vulnerability by the general affective and social nature of the human species that renders people at risk of psychological and physical harms. ²⁶

We argue that it is problematic to treat a group, (like those with mental health conditions) as homogenous. It is inappropriate to make sweeping conceptualisations of whole groups just because they loosely share the characteristics of a ‘mental health condition’. Labelling groups as ‘vulnerable’ risks an unjustifiable exclusion from research. ²⁷ A plausible solution to this challenge therefore is a view of vulnerability as etic or emic. ²⁸ Using this perspective, etic vulnerability focuses on the individual’s susceptibility to harm and risk assessed externally by others, that is, the person is either vulnerable or they are not. In contrast, an emic position would describe a position of being under threat, positioning vulnerability from the perspective of the participant. This latter would counter a protectionist ethical discourse by placing the voice of participants as central to ethical considerations.

Assessing risk/benefit ratio: The risk of ‘more of the same’

Harms and benefits might be less obvious or transparent than the convention of weighing up risk and benefit when using data. Harm may not be direct. Nevertheless, potential difficulties arise if populations are treated as homogenous and knowledge about conditions continues to be driven by narrow conceptualisations of the nature of mental ill-health. To date, inquiry appears to reflect a continued medical approach. The potential for change is clear but a search conducted by the authors of the associated Clinical Record Interactive Search ²⁹ (publications between 2009 and 2018) revealed a strong focus on treatment, onset and the co-morbidity of substance misuse or physical conditions and mental disorder. While this is worthwhile research, the nature of studies should introduce a note of caution. Generally, it is difficult to clearly identify direct benefits for those whose data have formed part of research. Arguably, the potential for making notable discoveries using data that has not been collected for (specific) research purposes is overrated. Research is shaped by the kinds of data which exist and the interests of researchers. In addition, if data are not collected for specific research purposes, we need some assurances that the research is worthwhile. Put bluntly, better research requires better research data – not ‘spurious analyses with junk’ (Ioannidis, p.41). ¹⁹

Research is not necessarily seeking answers to questions that participants want answers for. A good example of this is in autism research where 56% of research studies focus on brain, biology and cognition, and 15% on aetiology, with very little in funds spent elsewhere. ³⁰ For some, this is argued to be insufficient as parents of autistic children want more research in areas that help them support their children and address problem behaviours. ³¹ Others have taken a stronger position arguing that genetic testing and identifying the genetics associated with autism may put mothers under pressure to terminate pregnancies ³² which reflects a new wave of eugenics. ³³ Continued research for a schizophrenia gene presents similar disturbing ethical questions. ³⁴ Social and economic aspects of mental health are less frequently the subject of inquiry. People living with mental health conditions which have affected their quality of life may view social barriers, stigma, isolation, exclusion and the associated limiting consequences of these, as at least as important as symptom control. There is a risk of losing the human behind the data and the real risk may be research that would be considered iatrogenic to those living with mental health conditions.

With parity of esteem in mind, researchers need to consider epistemological and methodological elements of investigations to ensure that empirical findings can challenge inequalities in mental health. Ethically reflexive questions for researchers might relate to how the existence and content of data shapes research questions and, to what extent inquiry is influenced by populations experiencing everyday challenges. There is a risk that BD research elides important questions of disparity and the complex interconnections of mental ill-health with psychological and social factors thus constraining our understanding of mental health.

Consenting to research

Finally, we turn to specific consent issues. Mental health research which directly involves people experiencing mental health conditions is limited. This can in part be explained by a focus on biological, genetics and outcomes-driven research calls and as we have shown the barriers presented by ethical regulation of research involving vulnerable groups. The obstacle presented by consent requirements is particularly problematic. A significant difficulty is that informed consent as a principle potentially excludes large groups of people, including for example those who are viewed as mentally unstable or disabled. ³⁵ This exclusion of those deemed incompetent to consent may lead to their lives being concealed and limit the extension of knowledge which may be of benefit. In research ethics, the a priori assumptions of vulnerability in the operationalisation of informed consent can result in silencing voices. ³⁶ A way forward would be to enable a reasonable assessment of the level of protection needed. Current ethical regulation arguably privileges consent above all other ethical considerations and has been described as an ‘ethical panacea counteracting the potential danger of paternalistic and autocratic practices’ (Corrigan, p. 769). ³⁷ We are not arguing that consent should be dispensed with but that regulatory processes permit a reasonable debate about consent in context. In other words, viewing consent as fluid and iterative rather than a one-off event. ³⁸

There are important and alternative ways that consent might be considered in relation to BD. In his discussion of consent in the use of data which have not been collected for research purposes (often referred to as ‘routinely available’), Ioannidis argues that dispensing with consent constitutes an oxymoron in that research is claimed to simultaneously be non-interventionist and ‘routine’, presenting little impact on patients while resulting research and analyses of data are viewed as highly important and influential. Mental health data are potentially sensitive because stigma and its consequences create an uncertain and often unsafe world for prospective participants who understandably value their privacy. Consequently, we regard consent as important albeit different in BD research. This is because participants are consenting to research which they may have no agency to influence or shape. For this reason, the continuation of assertive opt-in rather than the necessity of an assertion to opt-out remains important. We acknowledge the efforts made to include service-users/patients in how data should be accessed but this does not always amount to influencing inquiry. Involving service-users and carers in setting priorities for research offers a useful way forward. ³⁹