Ethical issues in biomedical research using electronic health records: a systematic review

The first distinguished category is related to the intrinsic and instrumental values of EHR-based research that are mentioned in the included papers. The first reason for implementing EHR-based research is public interest (A1). EHR-based research is deemed instrumental in improvement of public health as well as in increasing efficacy and efficiency of healthcare. The second reason is efficiency and efficacy of EHRs research (A2): EHR-based research is relatively inexpensive and this kind of research can circumvent shortcomings of randomized clinical trials by providing researchers with data about all groups of patients and in that way contributing to generalizable knowledge. The included papers also discuss the fact that research with EHRs can give us a comprehensive picture of a healthcare system, and therefore allow for a more equitable allocation of resources (A3). Finally, the included papers discussed the fact that EHR-based research also promises benefits to the private sector: private hospitals, as well as insurance and technological companies (A4).

Even though EHR-based research is valuable, its implementation is not always easy. The papers we analyzed discussed various factors that can either facilitate this kind of research or obstruct it. The most frequent factor that impedes EHR-based research is regulations that put a burden of administrative work on researchers (B5). This could be especially cumbersome in multicenter international research, where there are unharmonized regulations and multiple research ethics committees. Moreover, a research project can fall within the blurred lines of the distinction between research and quality improvement, making it even more difficult for researchers and research ethics committees to decide which regulatory framework should be used.

Policy makers, however, are well aware that regulations can put too heavy a burden on researchers' shoulders, and the included papers also discuss regulatory facilitations and instrumental support of EHR-based research (B6). The support spans from administrative instruments such as abbreviated ethics review and waivers of informed consent requirement to financial investments into technological infrastructure and political support for embedded research. Financial and infrastructure investments are necessary because EHR-based research could encounter technical difficulties with implementation with regards to amount and versatile character of data, data quality, as well as other organizational problems (B7). A separate issue that we distinguished in the included papers is factors that hinder informing participants and obtaining their informed consent. These factors are related to ineffective outreach to participants, insufficient communication with vulnerable and underprivileged groups, problems with processing and documenting consent, especially in regards to tailoring consent to individual preferences (B8).

There are also two context-dependent factors of implementation of embedded research: public awareness (B9) and researchers’ attitude (B10). The public can favor EHR-based research or oppose it. Public awareness is also associated with previous experience with EHRs and healthcare, with information policies and practices, as well as multiple individual factors such as confidence in one’s computer skills, perception of sensitivity of health data and level of risk. Also, researchers’ attitudes (B10) is also a factor that could influence the conduct of EHR-based research. Researchers recognize ethical and legal challenges of embedded research and designing research protocols and they try to balance the imperatives to benefit society and to develop science. However, sometimes they can come to a conclusion that administrative and regulatory barriers make their research projects unworthy of pursuit.

We found several aspects of ethical data management in the literature. First of all, the included papers discuss an issue of safe and secure data storage. There is a variety of specific procedural and technical security measures (C11) such as firewalls, data safe havens and secure data access that can be deployed. Their main goal is to control access to data and protect patients’ confidentiality while at the same time streamlining legitimate research. Data used in research has different levels of identifiability (pseudonymized, de-identified, minimized data sets, aggregated data sets) at different stages of research (C12). Some elements of EHRs can contain sensitive information about a patient (C13), such as information about mental illness, fertility, face images, free text with references to third person. Also, quality, quantity and integrity of data (C14) has ethical importance, since researchers can draw conclusions that are meaningful and beneficial to wide social groups only when the data are representative and of good quality. Poor data quality can be also considered as a waste of resources. A separate problem that we discerned in the literature, is data ownership, management and curation (C15). This category encompasses an issue of data control and maintenance, and the possibility of selling data. Meaningful sharing of data (C16) means that data is a valuable resource that can be wasted, when it is not properly used, or when it is collected in such a data format that prevents sharing. We also distinguished in the literature a separate issue concerning data extraction and transfer (C17). From an ethical point of view it is important who has access to health data before it is extracted for the purpose of research, which elements of an EHRs are extracted. If the data contain sensitive information, who is responsible for storage and extraction of data: healthcare workers, researchers or any third entities, as well as who is responsible for decisions about extracting or using data of particular people or groups. This problem is linked to legitimacy of uses and users (C18). Medical professionals, researchers, private companies, healthcare providers, data institutes, disease foundations, governments and patients themselves can use EHR for various legitimate and (potentially) illegitimate purposes, such as research and development, marketing, and education. Empirical research shows (Table 2) that patients and participants sometimes insist on higher ethical scrutiny for certain users and uses. For instance, research use of EHR data by private companies can be considered less legitimate than use by university researchers.

Digitalization of healthcare systems and implementation of embedded research impacts healthcare providers and alters their relationship with patients. This impact of digitalization has also an ethical dimension that is discussed in the literature. First of all, EHR-based research changes professional relationships within institutions and between institutions (D19) in a way that blurs the distinction between practice, research and public health activities. Moreover, healthcare workers have to face new responsibilities and adapt to new practices in regard to data processing. As reported in the literature, digitization of health records and embedded research poses yet another challenge to communication with patients (D20), who have to be informed about data policies, including policies on data storage, sharing, and roles of a physician and provider as intermediaries between researchers. As is indicated in the literature, this process must foster trust between healthcare professionals and patients. Trust is crucial because patients should be still willing to share with their physicians all sensitive information that is important for their health. Another category that we distinguished in the included papers is directly connected with already discussed alterations in healthcare practice: digitization of health records brings new ethical responsibilities of the medical staff (D21). Healthcare professionals have to understand ethical challenges associated with EHRs and embedded research, especially if they are involved in data processing. Involvement into data processing, notably, when it directly concerns cooperation with researchers, can also be a source of moral distress and unwillingness to share patients’ data outside the context of their current care. Moreover, as reported in the analyzed papers, embedded research also requires additional work from staff and patients (D 22). Staff is involved in data curation and informing patients; patients have to respond to research invitations and fill out additional documents. In both cases, all these activities are discrete forms of resource allocation. Another aspect of embedded research, discussed in the literature, is the pivotal role of patients (D23). LHSs are sustainable in the long run only when there is broad patient participation, public acceptance and support. Some included papers addressing this issue underline the necessity of patient involvement into policy-making and implementation. Finally, a few papers included in our sample envision and discuss an idea that we called “digital patient-citizenship” (D24). Digital patient-citizenship is a proposal to encourage patient participation in research activities, research oversight, policy making and all the responsibilities associated with curation of data. In that vision, a personal EHR is a tool of patient empowerment. It is also an attempt to take a bit broader look at EHRs, not merely as an element of the healthcare system, but an element of contemporary digital culture.

Although usually it is believed that embedded research does not pose more than minimal risk, due to the fact that research use of EHRs is not associated with additional risk other than everyday medical practice, we found fifteen different categories of risk posed by EHR-based research. The most obvious seems to be risk to privacy (E25), then risk to patient autonomy (E26), when one loses control over one’s data. Because data use can violate a patient’s beliefs and values, it is associated with a risk of dignitary harm (E27). Risk for patients encompasses also harmful use of data (E28), e.g., when one’s data is used by a misleading pharmaceutics marketer, or when one loses a job after a leak of health information. In the papers we also encountered a risk of legal (E29) and psychological harm (E30). Some authors discuss information risk (E31) that is associated with the fact that patients could be insufficiently informed about research or that they can misunderstand provided information. Information risk category encompasses also a common situation of social science: it is difficult to study patients’ opinions and preferences, since people usually form opinions during research itself. Another category of risk that we encountered in the literature is a risk of exploitation (E32), when one’s data are used without consent for commercial purposes. In the included papers we also identify undue pressure to participate in research (E33). This kind of risk is closely associated with the risk to the whole healthcare system: when patients realize that their data are being used for illegitimate purposes or by illegitimate users that could undermine trust between the healthcare system and patients (E34) and lead to a massive drop out from research activities. Undermined trust can also result in other risks to patients. For example compromised care (E35) because patients withhold certain information from healthcare professionals; compromised care can be also a result of burdens that are associated with processing information and deflection of attention from a patient. We also distinguished a category of group mediated risk (E36). Research using EHRs provides detailed health related knowledge about individual patients, as well as about certain groups (e.g., ethnic minorities) and populations. Based on this knowledge, individuals and groups can be stigmatized and discriminated against by, for example, refusal of certain services. We also found a risk of financial conflict of interest (E37): researchers can take advantage of access to data and sell it. This kind of risk is more often mentioned by respondents in empirical research. Finally, their EHR-based research also poses risk to healthcare providers. First, because they can face additional work and administrative burden (E38). Second, disclosure of information about providers can sometimes undermine their reputation and commercial interests (E39).

The included papers also discuss a variety of protective measures against negative consequences of participation in EHR-based research: independent review by a research ethics committee or privacy board (F40), requirement for informed consent or authorization (F41), legal privacy regulations and ethical guidelines (F42), risk assessment procedures (F43), primary care provider consent (F44), as well as community or patients’ panel consent (F45). When one considers a healthcare institution as a participant of embedded research, then some authors also propose providers’ consent mechanism (F46) that would be devised for protection of healthcare providers’ interests, such as control of access to data.

The papers that we analyzed discuss several different types of consent in the context of EHR-based research. The first problem associated with research is approaching a patient and initiation of contact (G47). Embedded research might require different levels of patient involvement and in the literature a broad variety of options spanning from no need for consent, through opt-out to fully-informed and document consent (G48) are discussed. A separate issue is documentation of consent for EHR-based research and whether it should be a written document or just a verbal consent that is then marked in a patient’s electronic documentation. The included papers discuss also a seemingly opposing approaches to consent: broad consent (G50), where a patient agrees in advance for a whole spectrum of different and unknown research belonging to a certain category, and interactive and granular consent (G51), when a patient can actively select not only studies that she wants to participate in, but also pieces of information from her EHR that she is willing to share with researchers. The papers we analyzed also discuss such issues as when should researchers be able to obtain retrospective consent (D52) or what conditions have to be met to waive this requirement. Finally, papers also mention proxy consent (G53) and assent (G54). The latter two issues, however, are not thoroughly discussed in the literature.

As a separate topic we distinguished a category of content of consent (H). This category encompasses all pieces of information that a patient could or should be provided with in EHR-based research. Several of the included papers discuss issues such as the issue of data management, purpose of research, possible future use of data, storage and sharing details (H55). Some authors include additional items to the list of items that should be discussed in a consent form: security measures (H56), benefits, risks and burdens (H57) associated with research, as well as commercial application of data (H58). Only 4 papers discuss the issue of communication of study results (H59), and 2 mention that informed consent should contain information about research funding sources (H60).

Empirical research on patient attitudes towards EHR-based research is summarized in Table 2. Generally, patients or respondents representing the general public express their willingness to make one’s EHR available for the sake of research, as well as express other motives, such as altruism and solidarity, support for science, and health of future generations (I61, I62). Decisions to participate in research depend on trust of involved institutions (I63) and personal and sociodemographic factors such as race, education, income, living in a city, and employment status (I64).

Mainly empirical papers allowed us to distinguish also the emotional component of attitude toward EHRs and EHR-based research. Some patients and participants express their positive emotional attitude towards research encompassing a sense of comfort, trust, and even excitement with new vistas for biomedical research (J65). Others responded with emotions indicating negative attitudes toward research, such as discomfort, wariness, lack of commitment, anxiety and confusion (J66). Finally, there were also patients and participants whose attitude was inconsistent and they expressed enthusiastic support for research and great concern for privacy, felt uncomfortable about the fact that facilitation of research decreases the level of personal control over data (J67).

Finally, we identified 7 explicit ethical topics discussed in the context of EHR-based research. 49 papers discuss the traditional patient rights to information and to autonomous choice, for instance to donate one’s EHR to be used after the donor's death (K68). Next most frequently discussed issue is information, public education and public engagement (K69) that is also associated with such values as transparency, empowering of patients and communities. In the context of embedded research, the ethical principle of beneficence (K70) is also discussed. In addition, as some authors suggest, not only researchers and healthcare workers are obliged to conduct research for the public benefit, but patients are also under similar obligation to participate in low risk embedded research. Another quite frequently invoked ethical principle is the principle of justice (K71), which is translated in the context of EHR-based research into fair benefit sharing, fair recruitment and protection of vulnerable groups and individuals, and involving local communities in the process of research. A few papers discuss the problem of research integrity and respect for intellectual property (K73). Less frequently, a right to optimal healthcare and right to clinical judgment is discussed, as well as a right to compensation (K74).

Legal regulations in many countries (e.g. US and EU) already allow for modification or even a complete waiver of informed consent in the case of EHR-based research. Nevertheless, a waiver of informed consent disables individual control over data and can undermine public trust in research and healthcare institutions. The idea of empowering patients by giving them access to their data through IT tools is in line with the empirical studies that show patients want to know what is happening with their EHRs and that information policies play an important role in preserving trust towards healthcare institutions (Tale 2). Technology opens an opportunity to a more granular approach to informed consent (G51). It means that a patient, logging through a patient interface, is able to choose which elements of her records can be accessed by researchers and for how long. Such digital tools are already applied in research and clinical practice (Shelton 2011, Wallace and Miola 2021). This is in congruence with Neal Dickert’s and his colleagues analysis that distinguishes seven different functions of informed consent. Informed consent 1. Makes the process of research transparent; 2. Allows to control and authorize research; 3. Gives a patient opportunity to participate only in those research projects which conform to her values; 4. Protects and promotes welfare; 5. Promotes public trust, 6. Is required by regulations, and researchers who follow regulations are protected, and 7. Promotes research integrity (Dickert et al. 2017). Technological advances therefore give us an opportunity to balance individual control and public responsibility, because it seems reasonable to think that not all functions of informed consent have to be performed in all circumstances. There are some tradeoffs that a society or community could negotiate. For instance, one can agree that public health goals might require obligatory accessibility of health information concerning infectious diseases, especially in the time of pandemic, but other elements of an EHR could be under patient’s control.

The empirical studies that we analyzed also reveal that some patients hold a belief that data ownership is an appropriate instrument to control their data (Table 2). The idea of health data ownership has been discussed thoroughly, especially in the US context (Evans 2011; Haislmaier 2006; Hall and Schulman 2009; Hall 2010; Kish and Topol 2015; Mirchev et al. 2020; Purtova 2015, 2017; Rodwin 2009). It was argued that every EHR as a byproduct of therapeutic encounters belongs to a patient and medical facility (Haislmaier 2006; Hall and Schulman 2009; Hall 2010). An independent intermediary—EHR bank—can manage collection, exchange and access to databases for researchers, and other parties. The revenue would then be shared among the patient, the medical facility, and the EHR bank. Thus, it was thought, an invisible hand of the market will give a spur to the economy and research enterprise at once.

Nonetheless, this idea has been criticized for several reasons. The proponents of public interests argued that privatization of EHRs would increase the cost of public health and epidemiological research, as well as result in biased, not representative, research samples (Rodwin 2009; Evans 2011). Barbara Evans argues that in the US legal context property right is not an absolute one, and in reality it does not give an individual stronger privacy protection instruments (Evans 2011).

Generally, two different regimes of data protection with some variations can be distinguished (Painsack 2019). In the first regime an individual has control over data. This control can be exercised by property rights (the US) or data can be considered, as in the EU, an inalienable individual possession protected by civil rights (Painsack 2019). The second regime of data protection introduces an element of collective control over data, and as Prainsack argues, this element of collective control can be reconciled with the concept of data as inalienable possession. A similar idea was also discussed in one article included in our sample (Grande et al. 2014). Public data stewardship is an element of digital citizenship, where the community as a whole can balance individual rights and common good through deliberation and decisions can be made in a legitimate political process. Thus the problem of who can use people’s data and how is left neither to paternalistic protection of public health nor to purely economic forces. Data stewardship, nonetheless, requires additional education efforts, and can be probably implemented in societies with a high level of public trust and solidarity (K70-71). It seems that as societies, we need digital education that would explain the benefits and risks of new technologies.

The problem of data stewardship is even more complicated, because a community has to balance not only individual and societal interests, but also interests of corporate entities. Gregory Simon realizes that a healthcare provider cannot be unconditionally included into embedded research, because this kind of research also entails some risks for the whole institution (Simon et al. 2017, Piasecki & Dranseika 2021). We discussed the problem thoroughly in a separate publication, and we proposed three different strategies of finding a balance between a provider's professional obligation to contribute to the development of healthcare and duty to protect important interests of the institution. The first approach is the self-regulating model, probably the most suitable for free market driven healthcare systems, like in the US, where balancing is managed by healthcare providers themselves. The centralized model is more suitable for a centralized public healthcare system. In this model, the process of managing patients’ data is governed by a state body. The most democratic and patient- empowering approach is possible in the mediating model, where providers, the state and the citizens can negotiate what kind of data can be made accessible to researchers (Piasecki & Dranseika 2021).

EHRs are a key element of collecting data in a systematic manner during medical practice and then using these data to develop generalizable knowledge. However, in everyday life researchers usually do not have direct access to EHRs and practitioners’ access to EHRs is strictly regulated, as well. Thus, the use of EHRs presents a set of practical and ethical challenges: who can access an actual patient’s record and to what extent? How are the data from the record extracted and stored? (Evans 2011) How does this process of data extraction for the purpose of research influence everyday healthcare provider operations? (D19–22) How to contact patients and how to provide them with information about the research projects? Moreover, the empirical research with healthcare staff that is involved in processing and sharing the data reveals that healthcare professionals are not comfortable with sharing data outside the context of healthcare (Stevenson 2015) (D21). They are trained to keep patients’ data confidential. They build their trust relationship with patients on the basis of this commitment to professional confidentiality. As it has already been mentioned, building an LHS requires, then, not only technological tweaks and contact with patients, but also changing the organizational culture of healthcare providers (Foley and Fairmichael 2015).

In a LHS healthcare professionals: physicians, nurses, administrative clerks are assigned new roles (F29, K69–72). They are becoming a part of the LHS, obliged to provide additional information to researchers, overview and assess collected data and handle contacts with researchers. These new assignments transform ethical responsibilities of healthcare workers. The healthcare professionals have to accept and internalize these new responsibilities in order to make the LHS run in an efficient and ethical way. Similarly, patients have to recognize that quality of care, effectiveness and safety of provided therapies depend on their contribution and cooperation.

But it seems that these relatively miniscule changes in the healthcare systems constitute a more general ethical problem. Namely, what is the role of healthcare and healthcare systems in modern societies? Is healthcare a human right and a response to individual health needs? Is healthcare just one of the different goods that are available in the free market economy? (Daniels 2001) This question is not always dealt with directly. It seems that when one discusses a new ethical framework for LHSs, this implicitly answers it (Faden et al. 2013a; b). A raw fact of altered accessibility of EHR, accumulation of data and rising computing powers change our approach to clinical ethics, research ethics and public health ethics (Piasecki & Dranseika 2019a). We are facing a stark choice: what values our regulatory and technological environment should espouse: give precedence to individual interests or promote the public good? Currently, in both clinical and research ethics, the binding principle is still that of precedence of the individual interest. The new approach seems to take a different turn and underscores the moral importance of public good. In public health ethics the main goal is good of the community as a whole. But in the pursuit of that public good we cannot entirely discard the value of individuals (Kass 2001, Piasecki & Dranseika 2019b). And this new question “What should be the organizing principle of EHRs use in LHS?” also emerges from the results of our study. This issue has its consequences on all levels of the healthcare system and it affects not only the framing regulations, but also professional roles of healthcare workers, and patients’ attitudes and expectations.