Public Attitudes Towards Consent When Research is Integrated Into Care—Any “Ought” from All the “Is”?

Abstract

Research approaches that integrate learning into ongoing clinical activities offer the potential to accelerate the generation of knowledge to improve the health of individuals and populations. Yet, integrating research into clinical care raises difficult ethical and regulatory questions. Multiple empirical studies have explored patients’ and public attitudes toward approaches to consent for pragmatic research. Open questions remain, however, regarding the appropriate role of the resulting empirical data in resolving normative debates. We propose that data on acceptability or informed preferences should be prioritized; suggest caution regarding the weight given to majority viewpoints; and identify circumstances when empirical data can be overridden. We argue that empirical data bolster normative arguments that alterations of consent should be the default in pragmatic research; waivers are appropriate only when the pragmatic research would otherwise be impracticable and also has sufficiently high social value. We conclude with suggestions for additional work to guide policy decisions.

Serious, well-documented evidentiary shortcomings afflict the U.S. healthcare system. Though believing medical care should be informed by high-quality evidence, patients, clinicians, policymakers, and other key stakeholders often lack such evidence to inform and guide their decisions.¹ A distinct but closely related concern is that our existing research enterprise is too expensive, too slow, and too rarely responsive to key stakeholders’ needs.² There is widespread belief that novel approaches that deliberately embed research into the delivery of clinical care can address these two problems by answering important questions more quickly and more efficiently than traditional research methods.³

Take, for example, “pragmatic” clinical trials (PCTs), which are intended to remedy the concern that results from traditional “explanatory” trials often aren’t applicable in practice. PCTs are designed to evaluate the effectiveness of interventions under real-world conditions and to produce results that can be broadly generalized and used to inform both delivery of care and development of policy.⁴ To do this, they commonly take place where patients are already receiving their care and are designed to be flexible along dimensions such as eligibility and adherence.⁵ As an indicator of the enthusiasm for PCTs, the National Institutes of Health (NIH), with support from the Common Fund, has invested extensive resources into creating an infrastructure to support the design, conduct, and analysis of PCTs—first through the NIH Collaboratory and more recently through the National Institute on Aging (NIA) Imbedded Pragmatic Alzheimer’s Disease and AD-Related Dementias Clinical Trials (IMPACT) Collaboratory, which fund pilot (IMPACT) or Demonstration (Collaboratory) PCTs, and aim to develop data, tools, and resources to “propel a transformation in how clinical research is conducted.”⁶ Similarly, the Agency for Healthcare Research and Quality and the Patient-Centered Outcomes Research Institute have invested over $40 million in the Learning Health System Centers of Excellence initiative, aimed at developing learning health systems (LHS), an aspirational model in which research is embedded into ongoing clinical care to develop new knowledge, which is then systematically translated into altered clinical care processes to improve the health of individuals and communities.⁷

Research-care integration approaches such as PCTs and LHS hold great potential, but they also present ethical and regulatory challenges.⁸ Research and care have long been understood to be distinct activities entailing distinct normative commitments.⁹ Therefore, integration challenges the norms of research and of care.¹⁰ This raises a number of critical questions.

One question is whether consent for PCTs and other activities deliberately integrating research into care should more closely resemble consent for care or consent for research. In both the clinical and research contexts, informed consent is the culmination of a process of dyadic communication—in care, between a clinician and patient and, in research, between an investigator and prospective participant. Consent for research, however, typically requires higher levels of information-sharing, formality, and oversight than consent for care.¹¹ A key reason for this is that research, unlike care, aims to produce generalizable knowledge for the benefit of future patients.¹² As a result, research often exposes participants to risks and burdens that are justified not by any prospect of direct benefit but rather by their potential to generate socially valuable knowledge. The traditional view of research ethics grounds the requirement to obtain participants’ consent in the principle of respect for persons. As assumption of research-related risks is not generally considered obligatory, prospective participants should have the opportunity to determine for themselves whether enrolling in a particular study is consistent with their values and interests. Waivers to and alterations of consent requirements are, however, permissible under the Federal Policy for the Protection of Human Subjects or “Common Rule” when (among other factors) a study poses no more than minimal risk to participants, the research could not practicably be carried out without a waiver or alteration, and the waiver or alteration does not adversely affect the rights and welfare of participants.¹³ Arguing that these conditions often apply when research and care are integrated, some ethicists have proposed that consent for pragmatic research can be truncated, integrated into clinical consent, or forgone altogether.¹⁴

In recent years, multiple empirical studies have explored the perspectives of diverse stakeholders, particularly patients and the public, on approaches to consent for pragmatic research. These studies share an assumption that the views of those involved in and affected by research-care integration are relevant to ethics and policy deliberations. This seems intuitively right as a matter of respect for persons. Yet, while there is apparent agreement that empirical data are relevant to debates regarding appropriate approaches to consent for pragmatic research, to date there has been no specification of the ways in which empirical data can and should inform these deliberations. Are they determinative? Are they simply inputs to be weighed? And if so, what relative weight should be allotted to empirical data as compared to other relevant inputs? Answering these questions is of great import: without consensus on when consent is required “our health system will never learn to learn.”¹⁵

Our purpose here is three-fold. First, we seek to make sense of the existing empirical data on patients’ and the public’s views on approaches to consent for pragmatic research. Second, having marshalled the existing empirical data, we consider the normative implications, focusing on the conditions under which alterations and waivers of consent are appropriate. Finally, we outline an agenda for further empirical research.

Empirical Data on Approaches to Consent for Pragmatic Research

Here, we examine empirical data from the United States regarding approaches to consent for “pragmatic research”—an umbrella term we use to cover PCTs, comparative effectiveness research (CER), research on medical practices (ROMP), and other approaches that deliberately integrate research into the delivery of care. This empirical research has been motivated by an explicit recognition that research-care integration challenges traditional ethical and regulatory approaches to consent and by an implicit belief that addressing this challenge requires assessing the views of patients and the public, those whose care is the locus of learning.

Survey Studies

To date, five nationally representative surveys have examined public and/or patient views on approaches to consent for pragmatic research.¹⁶ In all, they represent the views of more than 10,000 American adults. Notably, each of these surveys employed hypothetical scenarios, admitting for the possibility that views would differ in real-world situations. Survey respondents have been asked about four general approaches to consent for pragmatic research: (1) traditional written consent; (2) streamlined consent with either (a) verbal permission or (b) brief written permission; (3) general notification, in which all patients within a health system are notified (e.g., via posters, brochures, or letters) that a health system regularly conducts learning activities, that patients will be automatically enrolled, and that patients will not be explicitly asked whether or not they would like to participate; and (4) post-enrollment notification, in which patients are first enrolled in a study and later asked to consent for data use.

When the findings from these five surveys are considered together, three cross-cutting themes emerge. First, people prefer prospective, study-specific consent—a category that encompasses traditional written consent and streamlined consent—to other approaches. Second, as between the two methods of prospective, study-specific consent, people generally express a preference for traditional written consent. Third, many individuals who prefer traditional written consent nevertheless find streamlined consent acceptable, but general notification is generally unacceptable. We will consider the data supporting each of these themes in turn.

A consistent finding across these surveys is that most respondents prefer prospective, study-specific consent to other approaches. Surveys by both Rahul Nayak and colleagues and Mildred Cho and colleagues examined respondents’ preferred approaches to consent for hypothetical randomized CER studies comparing two medications for hypertension.¹⁷ Nayak et al found that 71.6% of respondents preferred written informed consent to general notification for the hypothetical study.¹⁸ Similarly, respondents to the survey by Cho et al overwhelmingly favored streamlined consent—a discussion followed by either brief written (51.0%) or verbal (24.2%) permission—over general notification (14.8%).¹⁹ Limited evidence suggests that the preference for prospective, study-specific consent may, however, be tempered when research is conducted in the midst of acute, potentially life-threatening illness. Neal Dickert and colleagues conducted an experimental survey of preferences for consent in the context of a pragmatic trial for myocardial infarction. Only a bare majority of respondents (51.5%) preferred written consent to post-enrollment notification.²⁰ This may reflect situational constraints on decision-making specific to the acute-care context.

Second, three of the surveys elicited respondents’ preferences as between the two prospective, opt-in approaches to consent. Here, the data are mixed. Both Nayak et al and Cho et al found that respondents preferred traditional written consent to streamlined consent.²¹ Dickert et al, however, found that more individuals preferred streamlined consent with verbal permission to traditional written consent for studies of acute events like myocardial infarction (60% vs. 35.5%).²² Again, situational constraints on decision-making specific to the acute-care context may be at play.

Finally, the survey data suggest that, while respondents generally prefer traditional written consent, a majority considers streamlined consent acceptable. Cho et al asked individuals who initially preferred traditional written consent: “If getting written permission or consent would make this research too difficult to carry out, how would you prefer to be notified about this research?” Confronted with the trade-off between forgoing traditional written consent or forgoing the research, nearly two-thirds of respondents (65%) favored streamlined consent, while only 17% preferred that the research not be conducted.²³ This finding is supported by findings from a survey by Kristina Hallez and colleagues (a companion study embedded in a larger survey study by Nancy Kass and colleagues, described below), which asked respondents if they would find a streamlined approach to consent—which was illustrated via animated videos—acceptable for a CER study comparing two antihypertensives. Nearly all respondents (92%) deemed streamlined consent with verbal permission acceptable.²⁴ Respondents were asked to provide free-text explaining why they did or did not find streamlined consent acceptable. The most commonly offered reason for why it was unacceptable was that respondents valued documentation of their agreement. Two survey experiments have found similar results. In the first, Kevin Weinfurt and colleagues found no difference in the perceived acceptability of traditional written consent as compared to streamlined approaches.²⁵ In the second, Nancy Kass and colleagues compared attitudes on traditional written consent to six variations on streamlined consent and found no differences in perceived respectfulness.²⁶ General notification does not achieve comparable levels of acceptability. In fact, Weinfurt et al found that general notification approaches were unacceptable to the majority of the public.²⁷ Some respondents expressed concerns related to the lack of transparency associated with general notification; they worried patients might not see or read notices about research and might therefore be enrolled in a study without their knowledge or authorization.

Deliberative Engagement Studies

The survey data just described are generally consistent with findings from two deliberative engagement studies (DESs) with patient-participants that have looked at attitudes towards consent in learning health systems. In brief, DESs are structured—through a mix of didactic and discussion-based sessions—both to inform and to encourage reflection on different views when selecting among potential approaches to a policy decision.²⁸

Nancy Kass and colleagues had patient-participants engage in structured, iterative plenaries and small-group discussions about consent in the context of a learning health system.²⁹ While patient-participants initially preferred traditional written consent for a hypothetical CER study of two antihypertensives, there were no differences in the proportion preferring traditional written consent to streamlined consent after a day-long deliberation session. Patient-participants’ comments provide insight into this shift. Many described streamlined consent as a “good middle of the road approach” that preserved patient choice while minimizing impact on patient care and promoting quantity and quality of research.³⁰ Even after deliberative engagement, general notification was viewed far less favorably than either traditional written consent or streamlined consent. Some patient-participants expressed concern that a general notification approach would provide insufficient access to study information and found it inconsistent with their desire to make their own decisions regarding research and treatment.

Reshma Jagsi and colleagues conducted a DES with a focus on consent models for oncology learning health systems.³¹ In contrast to the aforementioned studies, Jagsi et al examined only attitudes towards research involving secondary use of data, and not those toward prospective randomized studies to compare two or more treatments. After day-long deliberation sessions consisting of educational presentations and facilitated small-group discussions, a majority of patient-participants expressed a desire for “at least some level of informed consent beyond notification” for secondary use of their data. A third (32.7%) indicated that it was either critically or very important for doctors to get a patient’s permission each time a medical record is used for research “even if it means a great deal of research will not be done.” This was down only slightly from 38.4% at baseline.

Other Relevant Evidence

Existing data suggest that streamlined consent may be at least as good as traditional written consent in promoting understanding, voluntariness, and trust—all of which are important functions of informed consent.³² Weinfurt et al and Kass et al used their survey experiments, described above, to examine understanding of study-related information following traditional written consent and streamlined consent approaches. They assessed respondents’ knowledge of factors such as whether treatment would be assigned by randomization, whether the study was placebo-controlled, and whether participation would require “extra things” beyond what would be expected in usual care. Neither Weinfurt et al nor Kass et al found significant differences in levels of understanding between respondents randomized to traditional informed consent and those randomized to streamlined consent.³³ Only Kass et al examined perceived voluntariness, defined as whether it would be easy or hard for a prospective participant who did not want to join the study to say no to participation. Again, there were no differences between traditional written consent and streamlined consent.³⁴ Weinfurt et al found that levels of reported trust in both the doctor and in the health care institution were similar following traditional written consent and streamlined consent. Notably, reported trust was significantly lower among respondents randomized to approaches forgoing prospective, study-specific consent.³⁵

In addition, several recent studies indicate that members of the public do not distinguish between observational or interventional research when assessing approaches to consent for pragmatic research. The surveys conducted by Cho et al, Nayak et al, and Kraft et al, as well as the DES conducted by Kass et al, explicitly sought to determine whether respondents’ attitudes differed if the research did or did not affect clinical treatment decisions. None of these teams found a statistically significant difference in respondents’ views on approaches to consent for research methods involving randomization as compared to research methods involving medical record review.³⁶

Notably, at least two related studies suggest that other stakeholders do in fact consider whether research affects clinical treatment decisions when assessing approaches to consent. Kraft et al surveyed institutional review board (IRB) professionals, including IRB staff and administrators. They found that, unlike patients, IRB professionals find streamlined consent far more acceptable in observational research than in randomized trials.³⁷ Similarly, in a companion study to the Kass et al DES study with patients, Stephanie Morain and colleagues reported that a multistakeholder group—including clinicians, those responsible for ethical and regulatory research oversight, CER researchers, health system administrators, payers, and research funders—voiced strong support (80%) for a general notification approach to an observational study of antihypertensives. Support for general notification did not, however, persist among these same stakeholders in the context of a randomized interventional study.³⁸ Collectively, the findings of Kraft et al and Morain et al suggest that the views of the public may not accord with those of other stakeholders. While further work is merited to explore the nature of these differences, Kraft et al point to related research in the biobanking context to suggest that these differences may reflect that patients place less emphasis on “methodology and regulations [and] instead prioritize trust and transparency.”³⁹

Additional research suggests that there may also be differences between stakeholders on the effect of the consent approach on willingness to participate in PCTs. In a discrete choice experiment assessing the attitudes of nephrologists and patients receiving outpatient dialysis, Katherine Courtright and colleagues found that the consent approach influenced willingness to participate for both patients and physicians, with both groups finding no notification less acceptable than either opt-in or opt-out approaches.⁴⁰ However, though a statistical comparison was not performed, “the effect of consent approach on willingness to participate appeared to be stronger for physicians than for patients.”⁴¹ This finding merits further exploration given physicians’ roles as gatekeepers for PCTs.⁴²

The overall number of studies exploring patient and public views on approaches to consent for pragmatic research remains limited. As noted previously, a shared limitation of these studies is that they elicited responses to hypothetical scenarios; it is possible that views formed in response to real-world situations views would differ from the results reported herein. Nevertheless, it is reassuring that the findings are generally consistent in both direction and magnitude across studies.

What to do with these data?

So often in bioethics, we bandy about the phrase, “That’s an empirical question!” It is an empirical question, for instance, whether people will accept ethicists’ arguments that consent for pragmatic research can be truncated, integrated into clinical consent, or forgone altogether. Now, with the empirical data on patients’ and the public’s views on approaches to consent for pragmatic research arrayed before us, we must ask and answer a threshold (and decidedly unempirical) question: Should empirical data inform ongoing ethics and policy debates regarding approaches to consent for pragmatic research?

As a starting point, we can rule out the extremes. Surely there is consensus that empirical evidence alone—even robust empirical evidence—cannot yield definitive ethical guidance. Reaching normative conclusions from descriptive data would be a failure to attend to the is-ought distinction.⁴³ The evidence summarized above cannot—without something more—conclusively tell us how to approach consent for pragmatic research. But, of course, accepting that we cannot divine moral truth from empirical evidence alone in no way implies that empirical evidence is irrelevant to ethics and policy debates. At times it may be: we don’t, for instance, need to conduct surveys to determine whether murder is morally permissible. Yet, in the context of deliberations about human subjects research ethics, it seems implausible that we could uphold the foundational principle of respect for persons—one application of which is consent—without some effort to identify how consent for pragmatic research is construed and negotiated by potential human subjects. Bolstering our intuition that empirical data are relevant, it’s unclear why so many prominent ethicists would gauge patient and public views on approaches to consent for pragmatic research if they didn’t think these views could inform ongoing normative debates. Thus, while possession of empirical data is not a sufficient condition for reaching normative conclusions, in this particular context, it appears to be a necessary one.

Now, accepting that empirical data on attitudes towards consent for pragmatic research are neither wholly irrelevant nor wholly determinative of the issue, we must do the spadework of defending the middle position. This requires that we address two questions. First, given the variety of empirical data that have been collected about approaches to consent for pragmatic research, which data should we focus on? Second, what degree of consensus among patients or members of the public is sufficient to influence normative debate? Once those questions are answered, we can set to work integrating empirical data into our normative analyses.

Which data should we focus on?

When using empirical data to inform normative debate about approaches to consent for pragmatic research, it’s important that we justify our focus rather than cherry-pick data that conform with our normative priors or advance our favored outcomes.

First, we wish to draw a distinction between initial and informed preferences. We contend that informed preferences—the preferences people endorse if they have the relevant information and make use of it—are more relevant than initial preferences to ethics and policy debates. It is reasonable to assume that initial preferences will often, though not always, be uninformed. For instance, people might not initially understand the expected outcomes associated with different policy options. As a result, ethics and policy decisions made in light of initial preferences may lead to sub-optimal outcomes. It is also important to identify the reasons and values that underlie preferences, whether initial or informed. Understanding why people prefer what they do can help us understand whether their preferences reflect reasons and values we should respect in our normative analyses.

Next, we distinguish what is preferable (i.e., what is more desirable) from what is acceptable (i.e., what can be tolerated or allowed). It is not our intention to imply that there is a rigid dichotomy between preferability and acceptability, only that it is more instructive to know what people will tolerate than what they view as optimal. There are three key reasons for this. First, prioritizing acceptability in the context of consent for pragmatic research is consistent with our expectations for public policy more broadly. As Jonathan Ives and Heather Draper have observed, it is certainly possible that a given policy might be mutually preferred by a group of like-minded individuals.⁴⁴ Yet, most public policies—including those governing the conduct of human subjects research at a national or even an institutional level—will rarely achieve this level of consensus.⁴⁵ Thus, policy work often strives for acceptability, which it has been argued, is more likely to achieve broad levels of agreement across individuals than are preferences.⁴⁶ A second reason to prioritize acceptability can be found within the logic of research ethics. The primary goal of research is to generate socially valuable knowledge for the benefit of future patients, rather than to benefit individual research participants—though, of course, risks to participants should be minimized and reasonable in relation to the expected benefits.⁴⁷ Reasoning by analogy, approaches to consent for pragmatic research do not necessarily need to satisfy the preferences of individual research subjects, but instead to help society while being acceptable to participants. Third, acceptability more closely resembles an “all things considered” judgment than preferability. Laura Beskow has made a similar point in the context of biospecimens: “[w]hat individuals might prefer…is not the same as what they might find acceptable, once they are aware of the risks, benefits, costs, and trade-offs at stake for the array of interests they would like to see advanced.”^48,28 (This quote suggests there may be some overlap between informed preferences and acceptability.)

Of course, there are limits to any reliance on acceptability. If patients or the public would accept something that was ethically impermissible, we would accord little or no weight to that view. Yet, in this context, there is no reason to think that the public is endorsing as acceptable anything impermissible. As recently outlined by a group of research ethicists led by Dickert, informed consent serves at least seven functions: (1) providing transparency, (2) allowing control and authorization, (3) promoting concordance with participants’ values, (4) protecting participants’ welfare interests, (5) promoting trust, (6) satisfying regulatory requirements, and (7) promoting the integrity of research and researchers.^49,23 Many respondents in the empirical studies reviewed above seemed prepared to accept streamlined consent for pragmatic research as a compromise position that accounts for a range of competing interests. Ethical analyses as well as existing (albeit limited) empirical data suggest that streamlined approaches to consent are capable of satisfactorily performing many of the functions of informed consent.

What level of consensus is sufficient to move decision-making?

When we survey or otherwise engage the public about approaches to consent for pragmatic research, we must determine when there is sufficient consensus behind any given position for that position to be persuasive. Recognizing that unanimity will rarely be achieved in a heterogeneous society, does a bare majority suffice? Does a two-thirds majority? Of course, any line we draw will be arbitrary. Yet, given that research ethics is grounded in respect for persons, we think that the larger the portion of the population that endorses a particular view, the more persuasive this view is or should be in our normative debates.

We acknowledge, however, that there is an inherent weakness to any such majority rule. It is possible for the majority to pursue its own interests at the expense of those in the minority, a phenomenon known as “the tyranny of the majority.” A question then looms over this discussion: if research ethicists roughly follow public opinion, are they capable of protecting minority rights against the majority? In answering this, it may be particularly important to know who constitutes the minority. Courts worry about “discrete and insular minorities” and look to see if petitioners come from groups that have been historically disadvantaged or lacked effective representation in the political process.⁵⁰ At present, we don’t have adequate empirical data to definitively answer questions about, for instance, racial/ethnic minorities’ attitudes regarding approaches to consent for pragmatic research. Even in the absence of such data, however, we would suggest that ethicists, like courts, have an institutional obligation to be mindful of minority interests as well as the historical events and social forces that shape them and—when needed—to act decisively to protect them. Minority views cannot automatically be discounted even in the face of a supermajority. Drawing the analogy to courts out further, some legal scholars worry about what is known as “the counter-majoritarian difficulty.” There is a fear that unelected—and ostensibly unaccountable—judges will impose their preferences on an unwilling citizenry. That argument, however influential it is in the law, lacks force in the present context. Research ethics is not—nor does it claim to be—democratic. When there is evidence that a consensus is emerging among patients or the public, there is an obligation to consider it, but it can be rebutted with robust normative argumentation.

Elsewhere, Rosalie Clawson and Zoe Oxley contend that public preferences are an important consideration for policy decision-making in a democratic society but are not the sole relevant criterion.⁵¹ The same, we argue, is true in research ethics. While there are good reasons to give weight to patients’ and the public’s informed preferences and reported acceptability—including for ethical reasons such as demonstrating respect for persons and also for practical reasons such as maintaining social support of and trust in the research enterprise—such preferences are not determinative. Overriding the public may, in some circumstances, be justified by other considerations.

How do we integrate our empirical and normative perspectives?

Let’s return to the original challenge that motivated this article. There are open questions about how best to approach consent for pragmatic research. Conducting pragmatic research is desirable, as is obtaining traditional written consent. However, it may not, in all circumstances, be possible to have both. The options to resolve such a dilemma are either to not conduct the research or to revise our approach to consent. Unpacking the latter option, it has variously been argued that consent for pragmatic research can be truncated, integrated into clinical consent, or forgone altogether. Deciding what to do is, we suggest, ultimately a normative task—but one that can and should be informed by empirical data. As we’ll show, ‘is’ can inform ‘ought’ in various ways.⁵²

First, ‘is’ can enrich our process of moral justification. Scott Kim and Franklin Miller have persuasively argued that waivers and alterations of consent are “two ethically different departures from regulatory consent” (i.e., traditional written consent).⁵³ Oddly, while these two approaches are ethically distinct, no comparable distinction is found in the Common Rule. Pragmatic research must meet the same regulatory criteria to qualify for either a waiver or an alteration of informed consent (45 CFR §46.116). Noting this gap between ethics and regulations, Kim and Miller have asserted that determining whether an alteration or waiver of consent is appropriate requires us to appeal to an ethical framework that takes into account respect for persons. Their emphasis on respect for persons is consistent with our central premise that empirical data have a role in normative debates because of the privileged place of respect for persons in research ethics. Applying their framework, Kim and Miller suggest that departures from traditional written consent “should be the minimum necessary.” Because a waiver is more extreme than an alteration, waivers should be viewed as a last resort even though satisfaction of the regulatory criteria outlined in the Common Rule can justify either.

Kim and Miller’s ethical arguments and, perhaps more strikingly, their policy prescriptions, are bolstered by empirical evidence showing that patients and members of the public clearly distinguish streamlined consent approaches—or, in the language of the Common Rule, alterations of consent—from approaches such as general and post-enrollment notification that are more akin to waivers. Moreover, a supermajority of patients and members of the public deem streamlined consent acceptable, whereas a supermajority deem general and post-enrollment notification unacceptable. Integrating our empirical and normative perspectives, it is clear that there needs to be a greater distinction between waiver and alteration. Given that no such distinction can be found in the Common Rule, the distinction must emerge in how investigators understand and IRBs apply the regulations. Stated otherwise, in practice, streamlined consent should be the default when traditional written consent is not feasible.

There will be times, however, when pragmatic research can only be conducted with a waiver of consent. Kim and Miller argue that waivers may be appropriate if there are no feasible alternatives; we share this view. Yet this view is hard to reconcile with the empirical evidence showing that general and post-enrollment notification are generally unacceptable to patients and members of the public, and that a sizeable minority would rather forgo socially valuable research than conduct it in the absence of prospective, study-specific consent. Furthermore, when asked about the reasons underlying their preferences, respondents identified reasons that demand respect—namely, the importance of transparency and preserving patient choice. How, then, do we bridge this normative-empirical divide?

As explained above, even a persuasive level of public consensus can be thoughtfully overridden by other considerations. Here, we draw upon recent arguments related to the social value of PCTs. Kirstin Borgerson has proposed that deviations from standard research practices will generally find stronger justification in PCTs than in explanatory trials given that PCTs “have direct social value, and so, are more likely to be immediately applicable to practice.”⁵⁴ Whereas explanatory trials typically seek to elucidate if and how an intervention works under idealized conditions, PCTs aim to determine whether an intervention works under real-world circumstances. Therefore, PCTs are directly relevant to key stakeholders because they are designed to answer questions central to clinical decision-making and policy. Similarly, Shona Kalkman and colleagues have argued that social value—in addition to being a threshold requirement for ethical research—may serve a “leveraging” function in PCTs.⁵⁵ It can provide additional justification for deviations from standard research ethics practices when other relevant conditions hold, such as that the PCT impose no more than minimal risk and that the research would otherwise be impracticable. It follows from these social value arguments, then, that there might be reasons to impose waivers of consent, despite the fact that many patients and members of the public find waivers unacceptable.

Who should get to determine whether the social value is sufficiently high (and that other conditions are appropriately met) such that it is permissible to override reported unacceptability and allow a waiver of consent? This is an instance where the ‘is’ of empirical data illuminates a new problem and highlights the need to consider new normative prescriptions. We suggest there is a role for novel patient-engaged governance mechanisms for pragmatic research. For example, one of us (EL) proposed elsewhere that research-care integration may benefit from alternative governance models, such as the use of a governing body or council responsible for the overall management of data used in a learning health system.⁵⁶ As outlined above, consent serves many functions, some of which can be achieved by means other than consent. Further normative and empirical work is needed to explore how governance models might fulfill the ethical commitment to demonstrate respect for persons while facilitating pragmatic research with waivers of consent.

What additional data do we need?

Considerable empirical work remains to be done on approaches to consent for pragmatic research if we are to continue moving normative work forward. We have identified six pressing areas for future research.

First, do the Common Rule criteria for alterations of informed consent adequately capture the conditions under which patients and the public find streamlined consent acceptable? General criteria for waivers and alterations are outlined in the Common Rule, and additional work has been done to further specify how these criteria might be applied in the context of research-care integration.⁵⁷ However, it is unclear whether patients and members of the public view these criteria as the appropriate means for adjudicating when streamlined consent is appropriate. Empirical work is needed to explore how regulatory criteria might be effectively translated into specific, action-guiding directions to support investigators and IRBs in determining when regulatory criteria are satisfied, as this remains a persistent challenge. Specifically, the ‘is’ of empirical research here can help us understand the context that mediates moral action, and can provide data to stimulate institutional accountability. That IRBs and others continue to struggle with the “blurred lines” encountered when evaluating pragmatic research suggests far more work is needed to guide their decision-making.⁵⁸

Second, how can the implementation of streamlined consent approaches be refined to further increase their acceptability? Data from Kraft et al and Hallez et al suggest, for example, that the acceptability of streamlined consent may be enhanced by taking relatively simple steps, such as including a signature requirement.⁵⁹ In normative work, Nancy Kass and Ruth Faden have proposed more extensive structural features that might enhance the acceptability of streamlined consent, including engaging patients in decision-making about what types of research activities health systems undertake, transparency about the types of research underway in the health system, and accountability to use knowledge gained from research conducted within the health system to improve the quality of care delivered to those patients who contributed to that knowledge development.⁶⁰ Such proposals merit further empirical exploration.

Third, the relationship between demographic factors and views on approaches to consent for pragmatic research remains under-studied. As attitudes towards research generally vary across the population, it is reasonable to expect that attitudes towards consent will also vary.⁶¹ To date, however, only Dickert et al and Jagsi et al have systematically examined the association between demographics and perspectives on consent in the context of pragmatic research. Dickert et al found that respondents of all races examined (white, Black, Hispanic, and “Other”) favored verbal consent in the acute-care context and noted that the transparency afforded by such an approach “may be especially important for communities who are less trustful of research.”⁶² Race was a significant predictor of lower odds of being willing to enroll. Jagsi et al also found significant differences by race/ethnicity.⁶³ Whereas only a quarter (23.2%) of non-Hispanic whites expressed a desire for consent to be obtained for each secondary use of de-identified health information, half of minority participants did (51.6%). Greater understanding of the relationship between demographics and attitudes towards consent for pragmatic research is critical for ensuring that selected approaches do not exacerbate existing disparities in perceptions of, access to, participation in, and benefit from biomedical research.

Fourth, there are open questions about who should seek consent when research and care are integrated. Classic statements of research ethics advise against “dual role consent,” the practice of physician-investigators obtaining consent from patients with whom they have preexisting treatment relationship. Only Kraft et al and Cho et al have directly examined patient or public attitudes regarding who should solicit consent for pragmatic research. In both studies, respondents overwhelmingly favored having their physician, rather than a third-party not involved with their care, discuss study participation with them and ask them whether they would like to participate.⁶⁴ Notably, Kraft et al compared the preferences of patients to those of IRB members and found their views sharply diverged. While approximately 85% of patients preferred that their physician obtain consent, more than half of IRB members stated that the patient’s physician was the least preferred individual to fill this role. Here, the ‘is’ could illuminate gaps between espoused ethics and lived practice, gaps that can be remediated through education or reassessment of our ideals. Additional empirical work would inform an emerging body of conceptual work challenging the conventional view of dual-role consent.⁶⁵

Fifth, preliminary data noted throughout this article suggest that attitudes of other stakeholders, including those charged with research oversight, diverge from those of patients and the public. Further work is needed to explore where stakeholder views diverge and to identify factors driving these differences. For example, some divergence in the views of patients as compared to other stakeholders may reflect differential weighting of reasons; several empirical studies indicate patients, in contrast with IRB members, may place relatively greater emphasis on trust and transparency than on methodological approaches.⁶⁶ Understanding these respective weights can help ensure that decisions made in the interests of patients do, in fact, reflect patients’ dominant interests. Stated otherwise, ‘is’ can help us modify the weight given to ethical principles in light of their relevance to key stakeholders.

Finally, research to date has not exhausted possible approaches to consent. There are more options than the four explored herein—traditional written consent, streamlined consent, general notification, and post-enrollment notification. Above, we called for a patient-engaged governance system. Additional empirical and conceptual research is needed to identify novel approaches to consent—or robust alternatives to consent—and to test their acceptability.

Conclusion

The integration of research and care demands reflection on the appropriate approaches to consent for pragmatic research. Empirical data on informed preferences and acceptability are not the only factors that should inform research ethics and policy decisions regarding consent for pragmatic research. Nevertheless, such data provide critical insights. Empirical data support the normative argument that alterations and waivers of traditional written consent are ethically distinct. In pragmatic research, alterations of consent should be the default and waivers a last resort justified by high social value. Additional empirical research is needed to further inform normative debates, and particular attention should be paid to developing patient-engaged governance models that achieve the functions of consent by other means.