Secondary use of clinical data for biomedical research purposes has great potential for various types of noninterventional, data-driven studies. We define secondary use of clinical data as the collection and reuse of clinical data in data gathering, noninterventional biomedical research, or learning activities; clinical data are collected during and for the purpose of patient care [1]. Research using clinical data has the ethical and efficiency advantages of not requiring additional physical interventions or collection of additional data. Although secondary use aims at improving biomedical knowledge and, in turn, medical care, it does not imply a direct benefit for the patient who has released their data.

The blurring of the boundaries between research and care, as envisaged in concepts of learning health care systems, is currently visible only in few areas [1,2]. The endeavor to merge these different system logics is faced with emerging challenges such as limited utility of specific consent models for research or false expectations regarding their benefits on the part of patients [3]. The goal of this paper was to contribute the patients’ perspective to the debate and potential solutions to the current challenges of secondary use of clinical data in the context of learning health care systems.

Previous studies with citizens and patients have already shown that certain aspects seem to be crucial for supporting secondary use, such as who conducts the research (eg, academic or commercial), whether data are transferred to other countries, and what consent model is applied [2-9]. However, owing to varying research designs, for example, by examining different study units, applying different survey instruments, and being conducted in diverse health care systems, these studies, taken together, have heterogeneous results.

Consent is a crucial component of respecting patient autonomy and building trust in health research. However, the specific consent paradigm of clinical trials cannot easily be applied to the secondary use of clinical data because most scientific questions are unknown at the time consent is obtained, that is, when the patient receives care. Newly applied models for secondary use of clinical data, such as broad consent or data use by default (with the option to opt out any time), facilitate research with clinical data, but are criticized from an informational self-determination perspective for offering patients insufficient control over their clinical data. However, previous studies have identified patients’ and citizens’ openness toward these new models [5,10-12]. Other empirical studies show that, to increase research benefits, participants seem willing to accept the loss of control over their data [13-16]. However, no studies have yet been conducted to assess the acceptance of consent models in light of the trade-off between the control of clinical data and research utility.

The objective of this study was to assess (1) patients’ general willingness and relevant requirements to share pseudonymized clinical data for research purposes, (2) acceptance of different consent models including characteristics of data control and research utility, (3) preferences regarding the setting to provide consent, and (4) general expectations toward data use and other stakeholders.

The questionnaire (Multimedia Appendix 1) was based on a review of the relevant scientific literature and a preparatory expert interview study among stakeholder groups engaged or affected by the planned secondary use of clinical data in Germany [17]. In total, 2 representatives for patient interests were included in the expert sample. The questionnaire was developed through several discussion and feedback rounds by the international and interdisciplinary project team, consisting of social scientists; ethicists; legal scholars; and clinicians with expertise in social, ethical, legal, or practical aspects of secondary use of clinical data. To ensure comprehensibility and technical functionality of the questionnaire, cognitive interviews (n=5) with patients with cancer and survivors of cancer who had provided consent were conducted in the pretest phase, resulting in minor adaptions.

To allow participants to develop an informed opinion, the survey included background information about risks and benefits associated with the secondary use of clinical data. The survey consisted of 33 items on the following topics: sociodemographic and disease-related information, expectations and risk perception toward secondary use, willingness to provide clinical data under certain requirements, and acceptability of consent models and procedures. Attitudinal questions were designed as 5-point Likert scale. The survey was approved by the data protection officer of the Heidelberg University Hospital.

In total, 3 vignettes were developed to measure the acceptability of 3 consent scenarios: specific consent, broad consent, and data use by default (with the option to opt out at any time). Acceptance was measured using a 4-point Likert scale (Figure 1).

Previous studies have reported that participants made a trade-off between research utility and data control [13]. Hence, 3 consent scenarios were designed with information about research utility and control over data. In the process of operationalization, we further reduced the complexity of the theoretical concept to ensure good comprehensibility of the survey material:

The Cancer Registry of the German federal state of Baden-Württemberg sent postal invitations to a random sample of patients with cancer and survivors of cancer, proportionally stratified by age and gender, requesting study participation (n=4219). The sample frame consisted of all registered patients in Baden-Württemberg, Germany, with a diagnosed tumor disease who were aged ≥18 years. Participants had the option of either completing an anonymous and self-administered web-based survey (the hyperlink was provided in the cover letter) or returning an envelope by mail, consenting that their address may be forwarded to the research group to subsequently receive a paper-and-pencil questionnaire. Survey instruments were adapted to the requirements of a mixed-mode survey [20].

Individuals who completed the survey were not compensated.

Data collection occurred from May 2021 to July 2021.

Descriptive statistics were used to express the categorical variables as counts and percentages. Differences in proportions were assessed for statistical significance (P<.05) using chi-square tests. The 2-tailed Pearson correlation coefficients were computed. All analyses were performed using SPSS (version 28; IBM Corp).

The study obtained ethics approval from the University of Heidelberg’s research ethics committee (reference number S-361/2018). Informed consent was obtained from the individuals who participated in the study pretest measurement and the written survey.

Of the 4155 patients with cancer approached by the Cancer Registry Baden-Württemberg, 838 (20.17%) participants completed the survey. Approximately half of the participants who answered the respective question were women (389/820, 47.4%; Table 1). Of 832 participants, 390 (46.9%) participants were aged between 60 and 74 years, and of 826 participants, 541 (65.5%) participants were retired. In total, 29.8% (247/830) of the participants had a university degree. The most common types of cancer were breast cancer (204/826, 24.7%), prostate cancer (187/826, 22.6%), and gastrointestinal cancer (79/826, 9.6%). The distribution of age, gender, and cancer entity mirrored that of the general distribution of patients with cancer in the Cancer Registry Baden-Württemberg, with minor deviation.

Most participants indicated that they are generally willing to make their clinical data available either without restrictions (527/838, 62.9%) or under certain conditions (283/838, 33.8%). Only 0.7% (6/838) of the participants generally refused to provide clinical data.

Then, the participants who indicated general willingness were asked about certain requirements under which they would provide their clinical data. When asked about the general requirements they deemed relevant, most participants stated the highest possible data security standards (482/838, 57.5%), followed by use of their data for as many research projects as possible (254/838, 30.3%), and being informed about the most important research results (208/838, 24.8%; Multimedia Appendix 2).

Most participants (591/832, 70.5%) stated that they would support research with their data in countries with high level of data protection comparable with German standards; 17.9% (149/832) of the participants stated that they would restrict data use to domestic research projects; and 8.8% (73/832) of the participants agreed to support international projects, independent of the level of data protection (Multimedia Appendix 3).

When asked how long their initial consent should be valid, 38.5% (320/832) of the participants set no time limit and approximately half of the participants demanded to renew consent either after 3 years (181/832, 21.8%), 10 years (227/832, 27.3%), or 30 years (10/832, 1.2%), respectively. In total, 10.2% (85/832) of the participants favored renewal of consent each time their data are used for specific research projects (Multimedia Appendix 4).

A large proportion of participants (532/832, 63.4%) said that they would grant access to researchers, independent of their affiliation; however, 22.7% (189/832) of them did not want to share their data with researchers at for-profit companies that conduct medical research (Multimedia Appendix 5). Only a small proportion opposed the secondary use of their clinical data by their physicians (56/832, 6.7%) or researchers at universities and university hospitals (48/832, 5.8%).

The questionnaire provided information about 3 consent models that correspond to specific consent, broad consent, and data use by default (with the option to opt out at any time), including the trade-offs of each model between control over clinical data and the facilitation of medical research (Table 2). For each consent model, the participants rated the level of acceptance on a 4-point Likert scale. Each of the 3 consent models showed a medium degree of acceptance with significant mean differences. Of the 838 participants, 491 (58.6%) accepted the broad consent model, 421 (50.2%) accepted data use by default (with the option to opt out at any time), and 323 (38.5%) accepted the specific consent model. Of the 323 participants accepting the specific consent model, 102 (31.6%) did not accept any other model (102/838, 12.2% of the total sample). Sociodemographic characteristics were not significant, except for older participants being more likely to accept data use by default (Pearson coefficient, 2-tailed: r=0.138; P<.001).

Participants were asked about the most appropriate setting for providing consent for the secondary use of their clinical data for research purposes. Most of them preferred to decide at their general practitioner’s practice (528/838, 63%), and a small proportion of participants preferred to decide during the admission to a hospital (174/838, 20.8%; Multimedia Appendix 6).

When asked about preferred information formats, most participants selected a brief written summary of key points in easy-to-understand language to learn more about secondary use (616/838, 73.5%), followed by face-to-face consultation with physicians (347/838, 41.4%; Multimedia Appendix 7). Participants were asked about who should decide about data access and use by individual research projects: most participants (393/838, 46.9%) favored committees with experts in which the opinion of patients is represented, for example, by patient representatives, whereas a small proportion of participants preferred to leave the decision to an expert committee (without patient representation; 185/838, 22.1%) or to decide for themselves (200/838, 23.9%; Multimedia Appendix 8).

A small proportion of the participants (99/838, 11.8%) showed major general concerns regarding their clinical data being used for research purposes (Figure 2). Then, all participants were asked about more specific concerns: the largest proportion of participants were worried about the data being misused in countries other than Germany (246/838, 29.4%), data being misused by criminals (244/838, 29.1%), and data being used by companies for something other than medical research (235/838, 28%). Concerns about participants being discriminated against because of cancer were very low (32/838, 3.8%).

Approximately all participants (788/838, 94%) expected a benefit for other patients from making their clinical data available for research purposes (Figure 3). More than half of the participants (482/838, 57.5%) mistakenly expected a personal benefit, even though the explanatory text explicitly stated the opposite. Of the 838 participants, 676 (80.7%) participants supported the claim that all patients should voluntarily make their clinical data available for research purposes. In total, 68.3% (572/838) of the participants expected their physicians to protect the participants’ clinical data in all circumstances, and approximately all participants (758/838, 90.5%) expected their physicians to support research, if consent was provided, by making their patients’ clinical data available for research.

Information about the requirements under which patients would make their treatment data available for research is important for any form of policy that regulates the secondary use of such data. This paper provides the results of a representative sample of German patients with cancer on general willingness and decisive requirements for sharing their data for research purposes and on the acceptance of consent models and expectations toward relevant stakeholders. The following are the main findings. First, we found an unprecedentedly high general willingness (810/838, 96.7%) to make clinical data available even after being informed about the potential risks of secondary use; however, relevant requirements included the following: ensuring a high level of data security, comparable data protection standards for data use abroad, and renewed consent at regular time intervals. Second, in contrast to previous studies, three-fourths of respondents (620/838, 73.9%) supported data use also by researchers in commercial companies. Third, the highest acceptance rate was found for a broad consent model (494/838, 58.9%), followed by data use by default (419/838, 50%); and specific consent for every study (327/838, 39%). Fourth, high expectations for physicians and fellow patients to support data sharing for research purposes were found.

To the best of our knowledge, this is the first representative study on attitudes toward the secondary use of clinical data and acceptance of consent models in combination with characteristics of data control and research utility.

The recruited sample is largely representative of the population of patients with cancer in the federal state of Baden-Württemberg in age, gender, and cancer entity. However, we found that the educational level in our sample was higher than that of the corresponding age cohorts of the German population [36]. The educational level of the German population presumably applies to the group of patients with cancer and survivors of cancer. Owing to the topic of the survey, we suspect a self-selection bias correlating with high educational level. According to a study in the context of genetic research and biobanking, high educational level positively correlates with willingness to provide data [24]; consequently, our results may overestimate willingness to provide clinical data. In addition, false expectations of personal benefits from providing data for secondary use may have increased the participants’ willingness to share clinical data.

A considerable proportion of participants who had previously agreed to hypothetically make their clinical data available without restrictions, favored restricted use of their clinical data when asked about specific requirements such as data user, duration, and data use in other countries (Multimedia Appendix 9). We assume that the participants have not yet formed a strong opinion about sharing their clinical data. Hence, the general willingness to provide clinical data seems to measure an overall attitude toward secondary use, rather than the actual willingness to provide clinical data without restrictions for research purposes.

Our study shows very high general willingness of patients with cancer to make their clinical data available for biomedical research purposes. However, the willingness to provide clinical data may be overstated owing to the above-average educational level of the respondents. For a considerable proportion of patients with cancer, willingness depends on certain requirements. In addition to the basic prerequisite of high level of data security and transparency in the use of the data, most patients shared the view that the data must not be used in countries with low data protection standards and that they should have the possibility to renew consent. In contrast to previous studies, the exclusion of use of data for private sector studies is not a requirement for most participants.

High willingness on the part of patients to accept loss of control over clinical data in favor of research benefits and request to maximize accessibility (and usability) of data for research were found. This is consistent with the acceptance of more research-friendly and low-control models, namely the broad consent model, followed by data use by default (with the option to opt out at any time). The striving for maximizing data use is also reflected by patients’ expectations toward physicians and other patients to support secondary use.

Policy makers are well advised to account for patients’ views when designing and implementing secondary use, with the aim to contribute to a socially legitimized culture of data sharing.