The value of healthcare data: to nudge, or not?

ABSTRACT

The processes of datafication, digitization and automation of healthcare and medicine are making new types and data available for analysis, and at greater volume. While the newly available data is often hailed as a solution to various problems in healthcare, there is only little discussion about who the use of such data empowers and who bears the costs. The use of healthcare data for “nudging”–e.g. to get patients to adopt healthier lifestyles–is a case in point: While such interventions are presumed to be cheap and effective, I argue that their value is a priori unclear. Both because of its assumed value-freeness, and because of its focus on individual behaviour, nudging draws attention away from the societal, political and economic factors that shape human practice. I conclude with a call upon policy makers to facilitate the use of healthcare data to build better institutions and to address social determinants of health before they seek to “fix” individual behaviour through nudging.

KEYWORDS:

• Data-rich medicine
• healthcare data
• nudging
• behavioural public policy
• big data

1. The use of healthcare data as a new challenge for policy makers

It has become a trope to speak of the “digital revolution” in healthcare (e.g. Flores et al. 2013; Hollis et al. 2015; Topol 2015). Even those who prefer not to use this term agree that digital practices have transformed many features of healthcare, ranging from organizational aspects to diagnosis to treatment (e.g. Lupton 2013; Lupton et al. 2018). What we commonly subsume under the label of digitization arguably consists of three overlapping processes that signify specific developments: Digitization in the strict sense of the word refers to moving things and practices from the analogue to the digital domain, such as the digital capture of doctor’s notes that were previously kept on paper. Automation means the transfer of tasks from humans to machines–such as the automatic filling of repeat prescriptions instead of patients having to order them over the phone or by visiting their doctor. The term datafication refers to the phenomenon that ever wider aspects of our bodies and lives that used to remain private are now recorded (typically, but not necessarily, by digital means).¹

Taken together, automation, digitization and datafication are converging to change healthcare in profound ways. They change the roles and practices of clinicians. They reconfigure what is expected of, and experienced by, patients, and they change the ways in which patients are represented and–literally–seen. In the context of Personalized and Precision Medicine (NAS 2011; Juengst et al. 2016), the emphasis has shifted from narrative and experiential information forming the lens through which patients are seen, towards structured, digital, quantified and computable data (see also de Mul 1999; Webster 2002; Nettleton and Burrows 2003; Nettleton 2004; Hartzband and Groopman 2016).

While these developments could be seen as processes that concern policy makers in the field of healthcare only, the implications of these transformations are much wider. At the same time as they reconfigure the work and experiences of health professionals and patients, digital practices also give rise to new business models, new actors, and new stakes. Not only are technology companies such as Amazon and Apple establishing themselves as powerful actors in the healthcare domain, and pose new challenges for government (Sharon 2016; Prainsack 2017; see also Introduction to this Special Issue), but healthcare data are becoming increasingly valuable assets (Birch 2017; see also Birch, Chiappetta, and Artyushina, 2020). For example, a patient using a digital insulin pump may control it with a smartphone application; the data stored in the smartphone app creates new opportunities for the company who built the pump or the app, and who legally owns the patient data,² to extract value from that data. If the data from the app can be linked with other information about that specific patient, such as her geolocation and her physical activity levels, for example, and if a company has enough data from a large enough number of patients, the data stored in the app can be used to discern patterns that could help to predict the behaviours and needs of other patients.³ Also social media content is now implicated in these processes as a source of “real world health data” (McDonald et al. 2012). The ability to predict the behaviour of patients and other citizens, in turn, lies at the heart of value creation in contemporary capitalist societies: It is the stuff that is traded in markets of “behavioral futures” (Zuboff 2019, 8; see also Prainsack and Van Hoyweghen in press).

With the increasing inroads of data-valorizing practices in the medical and healthcare field, the profound power asymmetries that characterize the relationship between data subjects and data users in the digital economy are now also a feature of healthcare (Andrejevic 2014; Sharon 2016). Besides the question of what types of data should be collected in healthcare contexts in the first place, and where we should refrain from datafication, another question of fundamental importance for policy makers is thus what uses of healthcare data–for what purposes, for whose benefit, and at what costs–should be allowed, and which ones deserve to be facilitated. Ultimately, this is a question about the public value of data, if public value is understood as the plausibly justified assumption that data use will have clear benefits for specific or groups of people, for society as a whole, or for future generations, and no person or group will experience significant and undue harm (see also Prainsack and Buyx 2016, 497).⁴

In this paper I will discuss the example of using healthcare data for nudging. I will do so to illustrate the intricacies of determining the public value of data use, and to argue that, despite of the presumed benefits of nudging,⁵ the public value of using healthcare data for this purpose is a priori unclear. Even if nudging improved the situation of individual people or groups targeted by the intervention, the implications of nudging on the institutional polity in a specific policy field could be so problematic that overall the public value of the nudge is to be considered negative. Moreover, the choice of the instrument of nudging over other policy instruments is a value judgment that is, in itself, in need of scrutiny. I conclude the paper by positing that, besides improving clinical outcomes, we should use patient data first and foremost to create better institutions, and not to tackle “behaviour” at the individual level.

2. The values of nudging

The concept of nudging was popularized by Richard Thaler and Cass Sunstein’s book Nudge (2008). These authors famously defined nudges as “any aspect of the choice architecture that alters people’s behavior in a predictable way without forbidding any options or significantly changing their economic incentives. To count as a mere nudge, the intervention must be easy and cheap to avoid” (Thaler and Sunstein 2008, 6). This definition– and the rhetoric of individual freedom and choice within which it is embedded–gave nudging the label of libertarian paternalism: according to the proponents of nudging, people are “helped” to do what is good for them while their freedom of choice is preserved.

Nudging is now an established instrument within the toolbox of behavioural public policy, which comprises tools that were “tested, informed or at least aligned to evidence from behavioural research” (Lourenço et al. 2016; see also Straßheim and Beck 2019, 2). The more datafied our societies become, the more tempting it is for public policy to utilize the available data to steer the conduct of citizens in a “personalised” way (Straßheim and Beck 2019, 18; see also Prainsack and Van Hoyweghen in press). Policy makers in countries such as the United Kingdom, the Netherlands or Australia have established Behavioural Insights Units to support them with evidence and tools from behavioural research and practice (John 2014; Feitsma 2019). A range of other countries, despite not having such dedicated “nudge units”, also use behavioural instruments to try to get citizens behave more healthily, submit their taxes on time, or behave in ways that incurs less costs for the collective (Whitehead et al. 2014).

One of the great attractions of nudging for public policy is the idea that through nudging, it is possible to fix the “demand side” of problems, rather than trying to address the supply side (see also Williams and Fullagar 2019). Moreover, proponents argue that nudging is a value-free vehicle to support people in reaching their own goals and living up to their own values, without policy makers imposing their own views and values top-down (e.g. Piniewski, Codagnone, and Osimo 2011). While this explains, to an extent, the popularity of nudges in societies where the legitimate role of the state is considered to be minimal, this assumption of the value-freeness of nudging is highly problematic (e.g. Pykett 2012; Jones, Pykett, and Whitehead 2013; Bubb and Pildes 2014; John 2014; McMahon 2015). Nudges articulate value judgments in at least two respects: First, they typically seek to make people's behaviour more “rational” in a sense that complies with a dominant sense of reason and morality. A smartphone app counting calories and rewarding people for caloric restriction is a nudge because it helps the person to avoid overeating. But a person who overeats is also seen as incurring costs for society if she ends up in poor health. In other words, every nudge is part of a larger system of values and valuation where an action or omission is assessed as sub-optimal according to societal values or standards that are seen as politically and morally acceptable. Second, nudges express judgments also about the appropriateness of certain forms and formats of policy interventions (e.g. McCrudden and King 2016). While this is, to some extent, the case for every policy instrument, what sets nudging apart is that it treats “the human subject as a target of correctional re-rationalisation” (Whitehead et al. 2012, 305). The following quote from a report of one of the Joint Research Centres of the European Commission illustrates this dynamic:

… aggressive attempts at improving health care delivery (supply side) has [sic] left us remarkably inept at transforming the health as well as the health costs of crowds. […] Hence an urgent and paradigmatic shift in public policy making is proposed. Communities and individuals must play a key role in co-creating the knowledge engines that support evidence-based investment of public health funds. […] Information technologies must be integrated with the expressed purpose of optimizing human performance and lifting our collective health talents. (Piniewski, Codagnone, and Osimo 2011, 9)

Such calls to move attention from the “supply side” to the “demand side” of policy shift the responsibility for outcomes to individual citizens–despite rhetorical allusions to “collective health talents” and sometimes even to solidarity. In this manner, behavioural interventions such as nudging are not merely an option for policy makers in a context of austerity to “do more with less” (see Redden, Dencik, and Warne, 2020; see also Briant and Harkins 2017), but they also have the effect to absolve public policy from producing good outcomes. Moreover, the use of terms such as “demand” and “supply” imports into policy discussions the assumptions of market economics where an equilibrium is possible and the state should only interfere when markets fail. In an ironic twist, by such use of market terms in public policy, the latter actively purports its own de-legitimization.

Nudging articulates value-preferences also in terms of its institutional effects. As Lepenies and Małecka (2015, 428) point out, “[w]e should not look only at what a nudge does or does not do to a nudgee […] but also whether the adoption of nudging as a policy impacts institutional structures” (ibid). By using healthcare data to nudge people to adopt healthier, more effective, or otherwise “better lifestyles”, we are also changing how we do things in our society, and gradually also changing understandings of how things should be done.

2.1. Nudging as an impediment to solidarity?

In order to explicate the values that underpin the very instrument of nudging, it may be helpful to take a short look at the history of nudging.⁶ The first use of the term nudging the title, keywords, or abstract of a scientific paper (outside of the field of biology⁷) seems to have occurred in 1972. In that year, the medical sociologist Irving Zola published a paper “on the social-political consequences of medical influence” (Zola 1972). He argued that medicine is not merely the application of neutral and objective scientific knowledge into the clinic. Through its expanding jurisdiction over areas such as ageing, addiction, or pregnancy, as well as through other processes such as the medicalization of ever wider aspects of “normal” (i.e. non-pathological) daily life, medicine became an authority that controls and steers the conduct of people–among other things, via the instrument of risk. In Zola’s words,

The belief in the omnipresence of disorder is further enhanced by a reading of the scientific, pharmacological, and medical literature. For there one finds a growing litany of indictments of “unhealthy” life activity. From sex to food, from aspirins to clothes, from driving your car to riding the surf, it seems that under certain conditions, or in combination with certain other substances or activities or if done too much or too little, virtually anything can lead to certain medical problems. [… E]very aspect of our daily life has in it elements of risk to health. (Zola 1972, 498)

Through these mechanisms, medicine locates the source of the problem, and thus the key to improving it, in the individual person, while neglecting the larger, structural, collective issues. What Zola was thus foreshadowing in his 1972 article was the importance of upstream interventions and addressing social determinants in solving societal problems. In other words, if policy makers do not tackle social ills first, and only then try to change people’s behaviour, they are buying into what Slater called “The toilet assumption”, namely:

the notion that unwanted matter, unwanted difficulties, unwanted complexities and obstacles will disappear if they are removed from our immediate field of vision … Our approach to social problems is to decrease their visibility: out of sight, out of mind … The result of our social efforts has been to remove the underlying problems of our society farther and farther from daily experience and daily consciousness, and hence to decrease the mass of the population, the knowledge, skill, resources, and motivations necessary to deal with them. (Slater 1990, 19)

The problem that policy makers focus on isolated aspects of a problem that can, seemingly, be fixed, at the cost of addressing the more inert, structural dimensions of a problem, remain highly acute in the new millennium. Despite growing evidence for the immense impact that structural, socio-economic factors have on health and wellbeing, austerity politics in many countries have led to divestments from these areas in favour of “quick fix” approaches addressing individual behaviour. This is the case despite the fact that the impact of social determinants on population health and the health of individuals is well known (e.g. Marmot and Wilkinson 2005; Schrecker and Bambra 2015; Marmot 2018; Williams and Fullagar 2019). Also outside of the health domain, work of scholars such as Richard Wilkinson and Kate Pickett (2010), who famously argued that greater social equality benefits everyone and not only the worst off, are endorsements of reforming social, political, and economic institutions and policies to reduce social inequalities, instead of focusing on individual behaviour. To speak with David Brady, who, in his 2009 book Rich Democracies, Poor People, analyzed data from 18 high-income countries across the world, the best predictor of poverty is not the family, but the country that someone is born into. This applies not only to the comparison between poor and rich countries, but it also applies among rich countries, such as in comparison between Sweden and the United States, for example.

How can this situation be explained? In Brady’s words, the decisive factor is a society’s willingness or failure to “collectively take responsibility for ensuring the economic security of its citizens” (Brady 2009, 181). European welfare states have structures in place that ensure the economic security of their citizens, instead of assuming that poverty is ultimately due to bad choices of individuals. They do not, to paraphrase Slater, blind themselves to the underlying and structural roots of phenomena that are undesirable. Instead, they accept a collective responsibility to address these root causes, even if it hurts–in the sense that it costs money, and that it may mean, in some cases, that people benefit who we may think have not deserved our support. What matters is that collectively, as humans, people are seen to deserve support in times of need. To take this collective responsibility seriously means to have institutions organized according to solidaristic principles, underpinned by the commitment to support also those who cannot contribute as much as they “owe”.

This principle of solidarity, which has been characteristic of continental European welfare states but has had impact on the design of policies and institutions far beyond, is challenged by the ideology of nudging. In contrast so solidarity, which foregrounds people’s connections and interdependencies with others (Prainsack and Buyx 2017), nudging is underpinned by an individualistic ontology that treats people as ideally bounded, independent, and “rational” (in a strategic sense) entities. In this sense it cements the important role that individualism has played in Western political and social thought even more firmly in the domain of policy making.

2.2. Atomistic individualism as an enabler of nudging

That the autonomous individual is the central bearer of agency is so deeply engrained in Western social and political institutions that it may sound trivial (e.g. Taylor 1985; Nedelsky 2011; Siedentop 2014). But ascribing agency to autonomously acting individuals who are ideally independent from others in the exercise of their freedoms is not the only way to think of human agency. Agency could also be conceived as an outcome of relations: Scholars such as Catriona Mackenzie and Natalie Stoljar (2000), Baylis, Kenny, and Sherwin (2008), and Jennifer Nedelsky (2011), have, over the last decades, put forward a different conception of autonomy, namely relational autonomy. Relational autonomy treats human autonomy as an outcome of people’s connections and interdependencies with their human, natural, and artefactual environments. Not only are people’s needs and interests influenced by their connections to these environments, but these connections make us to who we are. Within this view, the person is not conceptualized as a bounded, independent individual, and certainly not as somebody who only behaves, or should only behave, rationally in a strategic sense. The relational autonomy approach sees people as strong and autonomous because of their connections to, and interdependencies with, others.

In most of the work on and in nudging to date, however, people are seen as bounded individuals who should decide freely and act rationally–and if they cannot do this on their own, they should be helped. Many nudging initiatives are designed in this spirit. While proponents of nudging say they try to help people do what they “really” want to do, e.g. by helping them to overcome their biases, the assumption of such neutrality of nudging is an illusion. Decisions on what is good for people are generally made not only with reference to a dominant sense of morality, but also according to an understanding of instrumental rationality moulded after the rational actor paradigm. The rational actor, of course, is the very homo oeconomicus of neoclassical economics that behavioural science claims to get rid of: “an atomistic individual who has stable, coherent and well-defined preferences rooted in self-interest and utility maximization that are revealed through their choices” (McMahon 2015, 141). Rather than undermining this axiom, hovewer, behavioural interventions arguably reinforce it. They do so by “helping” people and societies to compensate for the rationality and consistency deficits that real human beings display in their behaviour. By overcoming cognitive and other biases (Kahneman 2011), people can strive to behave more rationally.

This attempt to “fix the rationality” of people, to help us be more strategically rational, is another reason that nudging has been so popular in recent decades (see also McMahon 2015). At a time when empirical evidence from a number of disciplines–anthropology, psychology, behavioural economics–shows that people do typically not, in fact, act rationally, and that the rational actor paradigm, may be not only empirically wrong but also conceptually misguided, nudging seems to give hope: people can be helped to be more rational. At the same time, nudging upholds the illusion of free choice, which is a particularly powerful move at a time when so many people feel increasing pressures and constraints from economic and political factors such as rising costs of living. This combination of (seemingly) enabling “free choice” at the individual level, and absolving government from their responsibility for policy outcomes at the same time, makes nudging an almost irresistible package for policy makers.⁸

In summary, when policy makers or other actors choose the instrument of nudging, they are implying that the most effective way of addressing social problems is to change the behaviour of individuals. When healthcare data is used for this purpose, this implies that such use of healthcare data is a good and valuable use. It is often only a small step from assuming the benefits of nudging and absolving nudgers from scrutinizing the impact of the intervention on marginalized groups and on the system as a whole.

It should also be noted that the idea of changing the “choice architecture” of people hinges on the assumption that individuals can freely choose what they do. Whoever speaks of behaviour “choice” thus draws of attention away from the larger structural elements that shape people’s practices-the ones that go beyond the design of a supermarket or cafeteria display. References to behaviour choice may fall victim to Slater’s toilet assumption as they move out of sight the unequal distribution not only of power and resources but also of duties, entitlements, and other factors that enable people to do certain things (see also Goodwin 2012; Hansen 2018; MacKay and Quigley 2018). This is, unfortunately, not only a philosophical problem, but one that has tangible economic and policy implications. In the United Kingdom, it has been argued that nudging “diverts government from its responsibility to use other, more effective, instruments” (McCrudden and King 2016, 84; see also House of Lords 2011⁹).

3. Healthcare data for nudging – or to build better institutions?

Does this mean that nudging is problematic in principle, and we should not use healthcare data for the purpose of nudging at all? While I will not go as far as rejecting nudging in principle (Prainsack and Buyx 2014), if we use healthcare data for nudging, we need to avoid two fallacies: First, that we treat people’s practices as “choices” that individuals are making freely and unconstrained from their relations to their human, natural, and artefactual environments. Those who devise and use nudges need to be conscious of the social, economic, and political factors that shape human practice, even if the intervention itself tackles the practice itself. Second, we need to avoid that nudging depoliticizes discussions about what we want to achieve by it. Nudging does not only align policy making more closely with the values and goals of individual citizens, as nudging proponents argue, but also the other way round: It aligns people’s decision making more closely with the values and goals of authorities. Nudging must not suspend political deliberation about these values.

An implication from trying to avoid the first fallacy, namely that nudging treats people practices as resulting from free, individual “choice”, is a shift in focus of policy making. Once we look at individual practice in the context of the social, economic and political factors that create the conditions of possibility for this practice in the first place, it may appear that the best way to address the problem may not be through targeting individual “behaviour”: It may be through changing institutions. Instead of using real-time data about people’s attendance of hospital emergency departments in order to “nudge” those that did not need emergency care to seek help elsewhere next time (e.g. Mustafee et al. 2017), for example, the same data could be used to learn where the need for faster, more easily accessible, or better healthcare is particularly needed, and expand healthcare services in that region.

Placing our focus on conditions of possibility for human practice, and not on the practice of individual people itself, would mean to go against a paradigm that has characterized policy making for the last decades. In the words of Elinor Ostrom, in a lecture she gave upon receiving the Sveriges Riskbank Prize in Economic Sciences in Memory of Alfred Nobel in 2009:

Designing institutions to force (or nudge) entirely self-interested individuals to achieve better outcomes has been the major goal posited by policy analysts for governments to accomplish for much of the past half-century. Extensive empirical research leads me to argue that instead, a core goal of public policy should be to facilitate the development of institutions that bring out the best in humans. We need to ask how diverse polycentric institutions help or hinder the innovativeness, learning, adapting, trustworthiness, levels of cooperation of participants, and the achievement of more effective, equitable, and sustainable outcomes at multiple scales. (Ostrom 2010, 9)

Also here, Ostrom’s argument is not that nudging, in itself, is wrong. But the implication is that nudging policies that do not take into consideration the larger context of the institutional setting, and the larger political economy, are counterproductive. Moreover, in many instances, focusing on the structural factors may be the more effective way of changing people’s “behaviour” than tackling it directly.

Avoiding the second fallacy–namely that nudging blacks out and thus depoliticizes discussions about why we want to nudge, and for what purpose–is connected to the first insofar as our lack of attention to structural and institutional factors blinds us also to the values embedded in them–even if these values are implicated in creating the problem in the first place. Take, for example, a mobile app that collects data about the diet, exercise levels, and alcohol and nicotine intake of people who eat, drink, or smoke too much and provides positive reinforcement and rewards for “good” behaviour. Even if, with the help of the app, the target group makes small positive changes in their lifestyle, this does not change the reasons why people overeat and smoke too much–especially if these reasons include high levels of stress, worries about money, and a lack of hope for a more positive future. Similarly, to tackle the problem of opioid overuse, researchers have suggested that physicians should be nudged to prescribe fewer opioids by lowering the default number of pills prescribed to outpatients after a surgery (Zhang, Dossa, and Baxter 2018). This suggestion assumes, first, that individual prescribing practices by physicians are an important contributor to the problem; second, that physicians do not already think carefully about the number of tablets that they prescribe; and third, that lowering the default number of tablets from 30 to 12 will not lead to a worsening of the problem by physicians prescribing two packs instead of one. At best, these assumptions are correct, and the nudge will have the desired effect. At worst, it will not be effective, and it will have channelled our energies into addressing a problem at an individual level that we should best address by changing structural factors: That companies profit from overprescribing, and that the systems in place to review and reward the performance of doctors and hospitals incentivize practices that satisfy patients on the short term, rather than addressing problems at their roots (not to mention the deeper causes lying within our current system of drug development and pricing).

How can we translate the importance of looking at the values that underpin the nudges that we devise, and our decision to use nudging in the first place, into specific guidance for policy and practice? Let us assume that a nudging intervention seeks to convince pregnant women to quit smoking; it does so by creating a smartphone app that visualizes the development of healthy foetuses of non-smoking women compared to the foetal development of smoking women. Women who participate in the nudging programme are required to log the number of cigarettes they smoke every day in the app; the same app shows women the projected impact of smoking on foetal development. Every woman who completes her cigarette log throughout her pregnancy receives a moderate cash payment at the end of the study; women who decide to quit smoking during the study receive cessation counselling and a higher cash payment at the end of the pregnancy. Smoking status is established by urine tests.

While such a nudging intervention could be seen as entirely beneficial at first sight, it is problematic in at least two ways: First of all, it targets a group that is already stigmatized. Smoking during pregnancy is associated with social and economic deprivation (Drope et al. 2018), and devising an instrument that focuses on such a group specifically runs the risk of increasing stigma. Second, as noted above, the reasons for smoking during pregnancy are complex and go far beyond the “choice” of an individual to engage in harmful behaviour (e.g. Glover and Kira 2011). These factors–poverty, poor mental health, or smoking partners and families–require policy interventions that address the root causes of smoking. Measures such as poverty reduction, better mental health services, better healthcare and social services for deprived and vulnerable populations could help people to lead happier and healthier lives. Healthcare data has a place in devising such measures; for example, using data from healthcare providers or social workers on the types of problems that people report can be used to tailor services to the needs of people in a region. These solutions would all focus on the “supply side” of service provision instead of tackling individual behaviour.

For those cases where addressing individual behaviour–tackling the “demand side”–may still seem most effective and desirable, Table 1 contains a list of questions aimed to help to ensure that: (a) nudging does not distract from attending to structural factors; (b) that the values that nudges articulate and promote are not excluded from deliberation; (c) that there is critical reflection of the choice of policy instruments to reach certain goals. This list is not meant to be a tick-box exercise but to invite systematic and critical reflection.

Table 1. Questions for a structured exploration of the value and values of nudging.

1	Problem definition	What problem does the nudge set out to address?
2		Whose problem definition is this? How else could the problem be defined?
3		What assumptions are inherent in the problem description?
4		What are the values underpinning the problem description?
5		Who benefits from the status quo? Who is harmed by the status quo?
6		What is the ideal outcome of the nudge?
7		Who would benefit from the ideal outcome? Who would be harmed by it?
8		What are the likely institutional consequences of the ideal outcome, if any? Who would benefit? Who would be disempowered?
9	Structural factors	What structural factors contribute to the problem? How could they be addressed effectively?
10	Target group	Who does the nudge target?
11		What characteristics do the people who are targeted by the nudge have in common? What are the most important differences within this group?
12		Does the nudge target a stigmatized or marginalized group?
13		Is there a risk that the nudge could stigmatize (or increase stigma) of the group that is targeted?
14		Is the group that is targeted the main group that causes the problem that the nudge seeks to address? If not, why is the nudge targeted at this group and not at the group(s) that cause(s) the problem?
15		Would it be beneficial to change/broaden/limit the target group?
16	Goals and format of nudging	Who defined the goal of the nudge?
17		What values underpin the nudge and its goals? Who has had a say in determining its goals? Whose voices and views have not been heard although they should have?
18		Would other policy instruments (infrastructure investments, binding law, etc.) be more appropriate to solve the problem that the nudge seeks to address?
19		What happens to people who are targeted by a nudge but do not follow it?

4. Conclusion

This article started out with a discussion of how the interwoven processes of datafication, digitization, and automation give rise to new practices and business models that render patient and other healthcare data an important asset in healthcare. If we see data as a key driver of practices and innovations in medicine and healthcare, then policies that shape whose data can be used, for what purposes, and who will benefit from such use, are moving to the foreground. Currently, however, discussions on the use of healthcare data still focus mostly on data protection and ownership, and leave aside the larger political questions of how different types of data use affect the distribution of power and resources within our societies.

Nudging is an increasingly popular scenario of putting patient and other healthcare data to use. Following the (by now) classical definition by Thaler and Sunstein (2008), nudging means to alter the choice architectures to make people behave in ways that are assumed to be good for them (without making other choices difficult or costly). If judged according to the rhetoric of nudging programmes and proponents, which emphasize the benefits of nudging in terms of getting people to behave more rationally and more healthily, most instances of data-driven nudging would be seen as empowering patients. But this conclusion would be rash: Nudges are never merely neutral tools that help people to do what is good for them; despite the focus on individual freedom, nudgers normally only consider a nudge necessary when the “behaviour” of individuals goes against the desired social and political order. Some authors explicitly acknowledge this when they distinguish between “pro-self nudges” and “pro-social nudges”, for example (Barton and Grüne-Yanoff 2015, 344). But even those nudges that claim (or are considered to be) pro-self nudges take reference to socially accepted norms that are, in turn, closely linked to dominant understandings of the greater good.

In addition, the very instrument of nudging contains value judgements: It assumes that addressing the practices of people directly is better than changing structural factors. It has been shown, however, that a focus on individual practices directs attention and resources away from tackling the more structural, systemic characteristics that shape the problem in the first place (e.g. Shove, Pantzar, and Watson 2012; MacKay and Quigley 2018). For example, when nudging takes the form of sending text-messages to people who “needlessly” visited emergency rooms, telling them that they could have saved public resources by visiting their general practitioner, then this measure distracts from the larger problem of underfunded emergency and hospital care. I have argued that, in order to establish whether or not data-driven nudging is a good use of healthcare data, we need to make explicit the values that are articulated by the nudge–and by the choice of nudging as an instrument. The goals and purposes for which nudging instruments are used need to be a matter of political deliberation.

This also means that we cannot assume that the use of patient data for nudging necessarily has public value (understood as creating benefits for specific or groups of people, for society as a whole, or for future generations, and no person or group will experience significant and undue harm; Prainsack and Buyx 2016, 497). The harms incurred by nudging could outweigh the benefits. Against a backdrop of a dominant rhetoric where the value and benefits of nudging are assumed to be evident, and the only problems with it pertain to its potential conflict with individual autonomy, this stance may be counterintuitive. To aide a systematic consideration of aspects that, taken together, shape the value of nudging, I have developed a list of questions to be asked of nudging programmes and plans (Table 1).

This paper addresses and problematizes the use of patient and other healthcare data in the form of nudging. In the context of a political economy that is characterized by profound power asymmetries between data subjects and data users, other questions about the role of patient data in the political economy still need to be addressed. For example, in the era of data-driven personalization within virtually any domain of life, where people’s data can be used against themand against others, we are in urgent need of better harm mitigation instruments (McMahon, Buyx, and Prainsack 2019). Moreover, there is an equally pressing need to ensure that, insofar as their fundamental needs of people are concerned–such as healthcare, housing, education, people do not need to satisfy them on the “free market”, where the risks of surveillance, profiling, and discrimination are particularly severe. Exposing people to data-driven practices such as personalized marketing is less problematic when they are buying a new hairdryer, or a holiday; it is highly problematic when it could interfere with the satisfaction of people’s fundamental needs. Ideally, people’s most important needs and interests should be met by publicly funded and provided services. Also here, the use of patient data has an important role to play: Patient data–besides helping to improve individual and population health outcomes–should be used not for risk stratification or pricing, but to target services and improve infrastructures where they are especially needed. In other words, in the much-hailed era of big data, it is more important than ever that we use data to build better institutions, instead of trying to solve problems through tackling individual behaviour.

Acknowledgements

I am grateful to Francesco Camboni, Alexander Degelsegger-Marquez, Carrie Friese, Robert Lepenies, Hanna Kienzler, Thomas King, Claire Marris, Stefan Neumann, Christoph Novak, Katharina Paul, Anna Pichelstorfer, Mirjam Pot, Joanna Redden, Lukas Schlögl, Wanda Spahl, Janneke Toussaint, and Hendrik Wagenaar for very helpful comments on this manuscript. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes on contributor

Barbara Prainsack is a Professor at the Department of Political Science at the University of Vienna, and at the Department of Global Health & Social Medicine at King’s College London. Her work explores the social, regulatory and ethical dimensions of biomedicine and bioscience, with a focus on personalized and “precision” medicine, citizen participation, and the role of solidarity in medicine and healthcare.

Correction Statement

This article has been republished with minor changes. These changes do not impact the academic content of the article.

Notes

1 Using the example of a person taking a walk in a park 30 years ago, only the person herself and the occasional passers-by would have known of this walk. Today, if the person carried her mobile phone, her exact route and the number of her steps would be captured; if she took a picture and shared it on social media then even her visual impressions would be datafied. Her personal devices may also have been recognized by wireless networks. Even without carrying her mobile phone, her image may have been captured in somebody else’s photo or a surveillance camera. Moreover, in contrast to an acquaintance who may have recognized her walking in the park, but who may have forgotten the details a few days later, the devices that collected this data would not “forget” any of the coordinates and details of her walk, which would remain stored and searchable for many years to come.

2 I use the term of patient data very widely to refer to personal data from patients, whether identified or de-identified (e.g. health records, insurance claims data, molecular tests), as well as other, non-personal data that is created by patients journeying through care pathways (e.g. aggregate data on the use of pain medication in a hospital ward). My use of the term patient data thus includes both individual-level and aggregate data coming from the (self-)observation of patients in the widest sense of the word.

3 This is done by looking for correlations between different characteristics and practices of people in the past, and to use them to infer the likely actions of specific people in the future: If it has turned out that a set of specific men who are undergoing a divorce and have more than one child under 18 have a high likelihood of defaulting on their mortgage, for example, this information could be applied also to other men with children who are going through a divorce: They would marked as at risk for defaulting, even if they have not done so.

4 Alena Buyx and I argued in another place that public value is more pronounced if “the benefits are likely to materialise for underprivileged groups than for privileged people, due to the overall lower baseline and potential size of impact” (Prainsack and Buyx 2016, 497).

5 On the side of proponents and users of nudging, the beneficial effect of nudges is typically presumed because nudges are claimed to help people to do what they should be wanting to do anyhow, and what is good for them. In his 2016 book on The Ethics of Influence, Cass Sunstein also claims that nudges should foster the welfare and wellbeing of people.

6 For an overview of the use of behavioural instruments in public policy, see Straßheim and Beck 2019.

7 Prior to the 1970s, occurrences of the term “nudging” in published academic articles typically referred to the sexual behaviour of animals.

8 While some authors note that part of the attraction of nudging is its resonance with processes of deregulation and privatization (Madi 2017; see also Jones, Pykett, and Whitehead 2011; Pykett 2012; Leggett 2014; McMahon 2015), it should be noted that many nudging proponents now recommend nudging to complement, not to replace, traditional regulation (e.g. Bhargava and Loewenstein 2015).

9 5.15:

We therefore urge ministers to ensure that policy makers are made aware of the evidence that non-regulatory measures are often not likely to be effective if used in isolation and that evidence regarding the whole range of policy interventions should be considered before they commit to using non-regulatory measures alone.