Exploring patient motivations and preferences for medical data sharing with researchers: a simulation study using the iAgree platform

Abstract

Objective

This study explores patient motivations and preferences for sharing medical data with researchers using the iAgree platform. We examine how study characteristics, including data type requested and data-sharing arrangements, influence consent decisions, and assess the role of demographic factors, privacy concerns, and perceived benefits in shaping data-sharing behavior.

Materials and Methods

We conducted a mixed-methods study with 527 US adults (≥18 years) recruited via advisory boards, social media, clinics, and newsletters. Participants completed 3 of 4 simulated studies on iAgree, each varying by data elements requested and data-sharing scope. Participants provided consent and data-sharing decisions and completed a post-simulation survey capturing demographics, data-sharing motivations, privacy concerns, and patient activation. We used logistic regressions to examine associations between demographics, privacy concerns, and patient activation and: (1) consent status and (2) willingness to share particular data elements. Finally, we applied thematic analysis to open-ended responses.

Results

Consent status did not significantly vary by data type or study design. However, participants citing altruism, personal benefit, and patient solidarity were more likely to share data. Higher privacy concerns were linked to lower willingness to share family health and mental health information. Participants with higher patient activation were also less likely to share data.

Discussion

Demographic factors were not significantly associated with consent or willingness to share data, countering common assumptions about disparities in sharing preferences.

Conclusion

Altruism and perceived benefit drive willingness to share health data, while privacy concerns and patient activation may reduce it, emphasizing the need for patient-centered, transparent consent models.

Background and significance

Medical record data play a pivotal role in advancing biomedical research, enabling discoveries that improve patient care, inform public health strategies, and drive medical innovation.^1–4 Access to comprehensive clinical data allows researchers to identify patterns, develop predictive models, recruit patients for clinical trials, and tailor interventions to diverse populations.^5–8 However, the process of data sharing is complex, requiring careful consideration of patient privacy, ethical standards, and trust in the research enterprise.^9–16 While data sharing accelerates scientific progress, respecting patients’ rights to control their personal health information remains an essential ethical consideration.¹⁴

Prior research has highlighted patients’ desire for greater transparency in how their clinical data are shared with researchers.^12,16–19 Patients have expressed a preference for understanding not only the purpose of data use but also the types of institutions involved, particularly distinguishing between for-profit and nonprofit entities.^14,17,19 For example, a study of 1246 participants who shared their data sharing preferences found that 28.3% declined to share their medical records data with outside non-profit institutions, while 47.4% declined to share their data with outside for-profit institutions.¹⁹ Additionally, patients have voiced a strong interest in selecting which categories of their data are shared, rather than being subject to blanket data-sharing policies.^10,17 Patient stakeholders have also voiced a desire for global research results when their data are accessed or used in secondary data analyses.¹⁶ These findings underscore the importance of developing patient-centered data governance models that respect individual preferences while supporting the collective benefits of biomedical research.¹⁸

Understanding patients’ motivations for sharing health data is essential. In general, studies find high levels of public trust in the ability of healthcare institutions to protect patients’ sensitive data,²⁰ although there are differences in levels of trust among different populations. A critical review identified key factors influencing willingness to share data: trust in providers and institutions, consent and transparency, perceived societal or personal benefits, and data sensitivity.²¹ It also found that older age (>65), younger age (<24), lower education, female gender, non-White race/ethnicity, better health, and lower internet use were linked to lower willingness to share data.²¹ Individuals with low levels of trust perceive the risks of data breaches to be higher, perceive the consequences of data breaches to be worse, and are more likely to request that permission be asked for sharing their data.²⁰ Investigating factors that lead individuals to have low levels of trust in data sharing for research can help develop more inclusive and responsive data-sharing policies.

Many healthcare systems use an opt-out consent model for medical record data sharing, meaning patients must actively decline to prevent their data from being used in secondary research.²² While this approach increases data availability, it often lacks transparency and may not align with patient expectations for autonomy and informed consent, especially since many are unaware their data is being used.^21,23 Moreover, some have argued that even with de-identified data, re-identification is possible.²⁴ One systematic review of patients’ perspectives on opt-out consent found that patients perceived that sharing secondary data with opt-out consent processes generally benefitted medicine and future generations, but had concerns about privacy, sharing data with other parties, and about the overall low levels of knowledge about opt-out consent.²²

To explore an alternative to the opt-out consent process, we developed iAgree, an interactive platform that engages patients in the data-sharing process. This study focused on de-identified data, which is generally seen as lower risk by IRBs, regulators, and patients. In prior work, we found that patients support data sharing for societal benefit but want more transparency and control, while researchers prioritize data completeness and avoiding bias if patients opt out.¹⁸ Although de-identified data can legally be shared without patient permission, we intentionally designed our simulation around de-identified information to test how participants would respond when given an affirmative choice in a setting where consent is not required. This approach allowed us to focus on the central aim of iAgree: evaluating whether a more patient-centered, opt-in consent process can enhance transparency and trust, even in contexts traditionally viewed as low risk. iAgree enables patients to review secondary data use requests, selectively consent to specific studies, and choose which data elements to share. By shifting toward an opt-in model, this initiative explores how health systems might enhance patient agency while preserving access to valuable clinical data for public-interest research.

Unlike prior studies of dynamic consent—such as biobank research, where participants actively opt in and can later revisit their consent decisions—the iAgree study examined dynamic consent in the context of secondary data that were originally collected for clinical rather than research purposes.^25,26 This is an important distinction, because patients who enroll in biobanks do so with the understanding that their data may be used for research, whereas patients whose secondary medical record data are repurposed for observational studies may not even be aware that such research is taking place.²² Moreover, with the increasing use of administrative and clinical data being used for studies using Artificial Intelligence, we wanted to better understand whether participants would adopt such a granular data sharing platform. Methodologically, the iAgree study is novel because it extends the application of dynamic consent to observational research, specifically testing its feasibility in the context of administrative and clinical data that were originally collected for non-research purposes.

Objectives

The objectives of this study were to: (1) analyze consenting preferences for various types of research studies requesting different types of data and data-sharing scenarios (eg, sharing with outside institutions) by demographics, healthcare engagement, privacy concerns, and motivations for sharing data for research, and (2) examine data-sharing preferences among individuals who consented to a study based on the same characteristics.

Materials and methods

iAgree platform

The iAgree platform is a proof-of-concept tool that enables patients to manage their health data sharing preferences. After linking their account to participating institutions via an authentication module, patients can (1) set default data sharing preferences, (2) view study-specific data requests, and (3) choose whether to share their data and customize what information is shared for each study.

After providing e-consent, participants accessed the iAgree website via an internet browser. For this study, to reduce respondent burden, we modified the platform by removing login requirements, disabling the authentication-authorization module, and hiding the navigation menu. Each participant was asked to complete only 3 of 4 simulated consent studies. For each, they reviewed a brief scenario and instructions, with the understanding that no real medical data would be shared, they were to imagine themselves as patients.

Participants viewed a simulated study webpage on iAgree, which included the study purpose, requested data elements, and a consent form (Figures 1 and 2). They then indicated whether they would hypothetically consent to participate. If they consented, they selected which data elements they would be willing to share (Figure 3). If they declined to share specific elements, they were asked to briefly explain why (Figure 4).

Figure 1.

Example of information related to a hypothetical simulated study displayed on the iAgree platform.

Figure 2.

Example of a consent form for a hypothetical simulated study displayed on the iAgree platform.

Figure 3.

Example of a webpage on the iAgree platform that allows participants to set their data sharing preferences for a simulated research study.

Figure 4.

Example webpage on the iAgree platform for participants to explain why they are not willing to share specific data elements for research.

Simulated studies

We created 4 simulated studies which were taken from real-world examples of research study protocols that had previously been approved by the Institutional Review Board (IRB) at a academic health care system. In addition to randomizing the order in which simulated studies are presented, each participant will be randomized to receive a specific version of each simulated study. We developed 4 simulated studies, each with multiple versions that varied by up to 2 “altered facts,” that is, sensitive data elements requested for sharing. These alterations were designed to test how socially sensitive information might influence participants’ willingness to consent, or whether whom the data was shared with may influence consent preferences.

The 4 simulated studies included:

1. Depression and Readmissions: Versions varied by whether family health history or social/economic data were requested, with one control version requesting no sensitive data.

2. Meningioma Registry: Versions varied by inclusion of genetic data and whether data would be shared with a for-profit vs. nonprofit institution.

3. Peripheral Nerve Injuries Registry: Versions varied by single site vs. multi-site design and whether mental health data was requested.

4. IVF Outcomes: One version included a request for pregnancy-related information; another served as a control with no altered facts.

Table S1 in File S1 summarizes each study’s structure, including sample sizes, versions, and the altered facts. Some versions included only one sensitive data element, others combined 2, and control versions included none. This factorial design enabled a systematic evaluation of how different data requests affect willingness to share medical information, both across and within study topics. We include a total of 13 variations. Consent forms for all study versions are available in File S2.

Post-simulation survey

After completing the simulated studies, participants completed a survey assessing satisfaction with the iAgree platform, demographics, data privacy attitudes, and motivations for sharing health data (Table 1). We included numerous sub-scales developed by the US National Cancer Institute as part of the Health Information National Trends Survey (HINTS), which regularly collects information about the American public’s knowledge about health-related information, health-information technology, privacy, patient activation (defined as: “having the knowledge, skills, and confidence to manage one’s health”²⁷), communication trends and practices, and relationships with health care providers.^28,29 The sub-scales used are available in Table 1. The full survey is available in File S3.

Table 1.

Key measures used in the survey following the simulated studies.

Measure category	Instrument/description
Platform satisfaction	Net promoter score (0-10 scale; higher = greater satisfaction)
Motivations for sharing	Custom items assessing reasons for sharing medical data for research
Trust in providers	HINTS patient-physician relationship sub-scale28
Patient activation	HINTS patient activation sub-scale28
Privacy attitudes	HINTS privacy concerns sub-scale28
Chronic disease status	HINTS issue involvement sub-scale28
Health literacy	TOFHLA health literacy instrument25
Data breach concern	5-point Likert scale (“Not at all concerned” to “Extremely concerned”) + open-ended response
Demographics	Prior research participation, education, employment, income, race, ethnicity, gender identity, sexual orientation, age

We also included free-text questions, including, “If iAgree was available to me where I receive health care, I would personally use it. (Briefly explain your answer)”; “How likely is it that you would recommend iAgree to a friend or colleague?”; “Please share any other reasons that would motivate you to share your medical record data for research.”

The survey also included 2 attention check questions to ensure that participants were not entering data at random. The survey was administered online using REDCap (Vanderbilt University, Nashville, TN, United States).

Data collection

We pilot tested the iAgree platform and survey with 19 participants from February to May 2024, leading to minor revisions for clarity. The full survey launched in June 2024. Participants first completed a screening form to confirm eligibility: (1) age 18 or older, (2) currently receiving healthcare in the United States, and (3) English proficiency. Eligible participants were then randomized to complete 3 of 4 simulated studies on the iAgree platform.

We recruited participants through multiple channels, including emails to Cedars-Sinai’s Patient and Family Advisory Board, snowball sampling, social media (LinkedIn, X, Facebook), and flyers distributed at other iAgree sites (University of Colorado, UCSF, UC Irvine, UC San Diego, and Yale). Participants received a $10 gift card for participation.

Quantitative analysis

We summarized demographic and clinical characteristics using univariate analyses. Continuous variables were summarized by median and interquartile range (IQR) and categorical variables were summarized by frequencies.

We conducted bivariate analyses to examine associations between study design and consent status, using Pearson’s Chi-square tests when assumptions were met and Fisher’s exact tests when expected cell counts were below 5.

We recoded variables to address small sample sizes. We dichotomized age at the median. For education, we combined “High school/GED,” “9-11 grades,” and “No schooling” as “High school or less”, and “Bachelor’s” and “Graduate degree” as “bachelor’s degree or above.” For income, we grouped responses as “$49 999 or less” and “$100 000 or more.”

We combined cisgender and transgender individuals by gender identity (male and female, respectively). Due to small numbers, we grouped “American Indian or Alaska Native,” “Native Hawaiian or Pacific Islander,” and “Middle Eastern or North African” as “Other” in race. We treated responses indicating “more than one race” (n = 9) as missing. We excluded ethnicity from models due to the very small number of Hispanic participants.

We used unadjusted logistic regression models to assess associations between consent status and (1) the type of information requested and data-sharing arrangements, and (2) participant characteristics, including demographics, satisfaction with iAgree, motivations for data sharing, HINTS sub-scale scores, data breach concerns, health literacy (TOFHLA), and prior research participation. Separate models were estimated for willingness to share family health, genetic, pregnancy, mental health, and social/economic information.

We then estimated adjusted multivariable logistic regression models to further assess these associations. Covariates were selected based on prior studies of patient data-sharing preferences, which emphasize the role of demographic characteristics, privacy attitudes, and patient engagement in shaping consent decisions. Variables were retained in multivariable models if their inclusion changed the odds ratio of the primary predictors by >10%, consistent with established confounder selection practices. The multi-variate models controlled for: age, race, gender, education, income, HINTS Patient Activation Sub-Scale,²⁸ HINTS Privacy Concerns Sub-Scale,²⁸ a question with concerns about data breaches (“if your medical record data were used for research, how concerned would you be about the potential for data breaches?” Not at all concerned, slightly concerned, moderately concerned, very concerned, and extremely concerned), and the TOFHLA Health Literacy score.³⁰

Model assumptions were assessed by examining residuals, leverage, and deviance statistics to ensure adequate fit. For contingency tables with expected counts <5, Fisher’s exact test was used in place of chi-square. To mitigate bias related to small sample sizes, we applied Firth’s penalized likelihood logistic regression, which reduces small-sample bias in maximum likelihood estimation. P-values were adjusted using the Benjamini-Hochberg false discovery rate procedure.

Qualitative analysis

We used a thematic analysis approach to analyze the free-text questions, which included reading through all the responses for each free-text question, identifying common sub-themes, and categorizing the sub-themes into larger themes.^31–33 We used an inductive and deductive iterative approach. All of the free-text question responses were pasted into an Excel document. We did not use an a priori codebook or framework, but rather created codes from the data to create a preliminary codebook, and then coded the data using this codebook.³² One researcher with experience in qualitative coding [CL] coded all the data, created the sub-themes and larger themes, and a second senior researcher [MSK] reviewed the codes and analysis. Nonsensical responses were tagged, and the entire observation was removed if there was a nonsensical response.

Ethics approval

This study was reviewed and approved by the Cedars-Sinai Medical Center Institutional Review Board.

Results

Characteristics of study participants

After excluding 62 participants who failed attention checks or provided nonsensical responses, the final sample included 527 participants (Table 2). The median age was 35 years (IQR: 31.0-39.0). Most participants identified as White (69.6%), followed by Black or African American (20.8%), Asian (4.6%), and Other (5.0%). The majority were non-Hispanic (96.4%), held at least a bachelor’s degree (44.6%), and identified as male (54.7%). Most reported annual incomes between $50 000 and $99 999 (55.0%).

Table 2.

Overall demographics and other characteristics of the study sample.

	Overall (n = 527)
Demographics
Age median (IQR)	35.0 [32.0, 39.0]
Gender, n (%)
Femalea	238 (45.16)
Maleb	288 (54.65)
Non-binary	1 (0.19)
Race, n (%)
Asian	24 (4.62)
Black	108 (20.77)
White	362 (69.62)
Other	26 (5.00)
Missing	7
Ethnicity, n (%)
Hispanic	19 (3.61)
Non-Hispanic	508 (96.39)
Education, n (%)
Bachelor’s degree or above	235 (44.59)
Some college	138 (26.19)
High school or less	37 (7.02)
Technical or trade school	117 (21.31)
Income, n (%)
$100 000 or more	55 (10.44)
$50 000 to $99 999	290 (55.03)
$49 999 or less	176 (33.40)
Prefer not to say	6
Reasons for sharing medical record data
a. I will personally benefit from the research	7.0 [5.0, 8.0]
b. Family or friends will benefit from the research	7.0 [5.0, 8.0]
c. Support medical research in general	7.0 [5.0, 8.0]
d. Help other patients with the same health conditions as me	7.0 [6.0, 8.0]
e. Like to help other patients, no matter health conditions	7.0 [5.0, 8.0]
f. Take advantage of medical research, should also make contribution	7.0 [4.0, 8.0]
g. Being patient myself, I feel connected to future patients	7.0 [6.0, 8.0]
h. Know other patients who have shared their data for research	7.0 [6.0, 8.0]
i. Grateful to my doctors	7.0 [6.0, 8.0]
HINTS scores
Patient-Physician Relationship Sub-Scale median (IQR)	15.0 [13.0, 18.0]
Patient Activation Sub-Scale, median (IQR)	8.0 [7.0, 10.0]
Issue Involvement Sub-Scale, median (IQR)	6.0 [5.0, 7.0]
Privacy Concerns Sub-Scale, median (IQR)	6.0 [5.0, 7.0]
Other characteristics
Concerns about data breaches, median (IQR)	3.0 [2.0, 4.0]
TOFHLA Health Literacy Score, median (IQR)	8.0 [6.0, 9.0]
Prior participation in research study, n (%)
Yes	80 (15.18)
No	447 (84.82)

^aIncludes individuals who identified as trans-female.

^bIncludes individuals who identified as trans-male.

Reasons for sharing medical record data

Across all motivations for sharing medical record data, the mean score was 7 (the scale was from 1 to 8), which indicates high agreement with each motivation (Table 2).

Patient-physician relationship, patient activation, issue involvement, and privacy concerns

Table 2 presents descriptive statistics for the overall sample. Participants reported a median score of 15.0 (IQR: 13.0-18.0) on the HINTS Patient-Physician Relationship sub-scale, out of a maximum score of 28. The median HINTS Patient Activation score was 8.0 (IQR: 7.0-10.0) out of a maximum score of 16. HINTS Privacy Concern sub-scale scores had a median of 6.0 (IQR: 5.0-7.0) out of a maximum score of 10. Median scores also reflected low chronic disease burden (Issue Involvement: a median of 6.0 out of a maximum score of 16, IQR: 5.0-7.0). The median on data breach concerns was 3 out of a maximum score of 5 (3.0, IQR: 2.0-4.0). Health literacy, as measured by the TOFHLA, had a median of 8.0 out of a maximum score of 15 (IQR: 6.0-9.0), a range typically classified as “adequate” literacy in prior validation studies.³⁰

Table S2 explores how study design, types of information requested, and data-sharing arrangements were associated with consent decisions across 4 study types using chi-square tests. Unadjusted logistic regressions (Table 3) showed no statistically significant associations between consent status and either the type of information requested or the data-sharing scenario.

Table 3.

Unadjusted logistic regression analysis of consent status by information type and data-sharing scenario across 4 studies.

	Odds ratio (OR)	95% CI		P-value
Depression and readmissions (Study 1)
Social and economic information	1.70	0.43	6.74	0.22
Family health history information	0.64	0.21	1.95	0.16
None (reference)	–	–	–	–
Meningioma registry (Study 2)
For-profit institution	0.91	0.34	2.48	0.43
For-profit institution (genetic information)	1.21	0.42	3.52	0.96
Non-profit institution (genetic information)	1.79	0.55	5.85	0.30
Non-profit institution (reference)	–	–	–	–
Peripheral nerve injuries registry (Study 3)
Multiple site study	1.91	0.594	6.12	0.89
Single site study (mental health information)	6.06	1.047	35.09	0.09
Multiple site study (mental health information)	1.45	0.51	4.11	0.40
Single site study (reference)	–	–	–	–
IVF outcomes (Study 4)
Pregnancy-related information	0.75	0.328	1.715	0.27
None (reference)	–	–	–	–

In Study 1 (Depression and Readmissions, n = 308), participants were asked for social/economic information (35.3%), family health history (32.2%), or no personal information (32.5%). Consent status did not significantly differ based on the type of requested information; the chi-square test was not significant (P = .257). Participants asked to provide social and economic information showed a non-significant trend toward greater consent compared to the reference group, who was not asked for any personal information (OR = 1.70, 95% CI: 0.43-6.74, P = .22, Table 3), while those asked for family health history information showed a non-significant trend toward lower consent (OR = 0.64, 95% CI: 0.21-1.95, P = .16).

In Study 2, a chi-square test found no significant difference in consent rates across the 4 organization and information type combinations (P = .69). Neither for-profit institution sponsorship alone (OR = 0.91, 95% CI: 0.34-2.48, P = .43, Table 3), for-profit institution sponsorship with genetic information requested (OR = 1.21, 95% CI: 0.42-3.52, P = .96), nor non-profit institution sponsorship with genetic information requested (OR = 1.79, 95% CI: 0.55-5.85, P = .30) was significantly associated with consent compared to the non-profit reference group.

In Study 3 (Peripheral Nerve Injuries Registry, n = 374), design varied by single- vs multi-site data sharing and whether mental health information was requested. A chi-square test found no significant difference in consent rates across site and information type combinations (P = .09). Neither a multiple-site design alone (OR = 1.91, 95% CI: 0.59-6.12, P = .89, Table 3), a single-site design requesting mental health information (OR = 6.06, 95% CI: 1.05-35.09, P = .09), nor a multiple-site design requesting mental health information (OR = 1.45, 95% CI: 0.51-4.11, P = .40) was significantly associated with consent compared to the single-site reference group, though the single-site mental health condition showed a trend approaching significance.

Study 4 (Adverse Outcomes of In Vitro Fertilization, n = 239) involved requests for pregnancy-related information. A chi-square test found no significant difference in consent rates between those asked for pregnancy-related information and those asked for nothing (P = .108). Overall consent was 97.14%, though those asked for pregnancy-related information were somewhat overrepresented among non-consenters (83.33% of non-consenters vs 48.12% of the total sample). In Study 4, participants asked to provide pregnancy-related information showed a non-significant trend toward lower consent compared to the reference group (OR = 0.75, 95% CI: 0.33-1.72, P = .27, Table 3).

Across all studies, consent decisions were not significantly influenced by information type, data-sharing scope, or organizational affiliation.

Demographics/other characteristics and willingness to share medical record data

Table 4 presents unadjusted logistic regression results across all 4 studies, examining factors associated with participants’ willingness to share various types of sensitive health information. No demographic variables were significantly associated with sharing in any study. While several motivational and psychosocial factors showed significance in unadjusted models, none remained significant in multivariate analyses (Tables S3-S7).

Table 4.

Unadjusted logistic regression analysis results regarding demographic and other characteristics and willingness to share different types of medical record data.

	Study 1—Willingness to share family health information (n = 102)			Study 1—Willingness to share social and economic information (n = 106)			Study 2—Willingness to share genetic health information (n = 180)			Study 3—Willingness to share mental health information (n = 186)			Study 4—Willingness to share pregnancy information (n = 115)
Demographics	OR	95% Cis	P-value	OR	95% CIs	P-value	OR	95% CIs	P-value	OR	95% CIs	P-value	OR	95% CIs	P-value
Age
20-35	1.03	(0.97, 1.1)	0.32	0.97	(0.39, 2.45)	0.95	1.22	(0.59, 2.53)	0.60	1.09	(0.55, 2.15)	0.81	2.23	(0.86, 5.82)	0.10
35+ (Reference)
Gender
Malea (Reference)
Femaleb	0.88	(0.38, 2.06)	0.77	3.00	(1.11, 8.15)	0.46	1.39	(0.66, 2.91)	0.38	1.14	(0.57, 2.28)	0.32	NAc
Nonbinaryd	NA			1.43	(0.02, 134.04)	0.93	NA			NA		NA	NA
Race
White (Reference)
Asian	0.62	(0.06, 6.8)	1.00	0.66	(0.11, 3.77)	0.46	0.90	(0.12, 6.6)	0.99	1.17	(0.25, 5.4)	0.93	0.83	(0.11, 6.3)	0.82
Black	0.72	(0.25, 2.04)	0.80	1.72	(0.47, 6.26)	0.44	0.76	(0.32, 1.79)	0.74	1.23	(0.52, 2.93)	0.80	1.67	(0.39, 7.23)	0.15
Other	0.37	(0.07, 2.00)	0.37	1.34	(0.18, 10.06)	0.81	0.90	(0.12, 6.6)	0.99	1.03	(0.13, 8.1)	0.93	0.16		0.26
Education
Bachelor’s degree or above (Reference)
High school or less	1.90	(0.64, 5.7)	0.67	0.82	(0.24, 2.82)	0.45	1.50	(0.62, 3.59)	0.81	1.71	(0.7, 4.15)	0.41	1.90	(0.57, 6.33)	0.99
Some college	2.40	(0.29, 20.04)	0.60	4.09	(0.16, 104.5)	0.30	2.64	(0.39, 17.89)	0.49	2.02	(0.42, 9.67)	0.45	5.97	(0.25, 140.64)	0.33
Technical or trade school	1.47	(0.48, 4.47)	0.82	0.59	(0.15, 2.29)	0.19	1.75	(0.63, 4.84)	0.84	0.92	(0.37, 2.27)	0.23	1.18	(0.36, 3.9)	0.34
Income
$100 000 or more	4.09	(0.58, 28.59)	0.46	5.44	(0.25, 118.47)	0.48	3.25	(0.53, 20)	0.19	3.07	(0.72, 13.14)	0.31	0.73	(0.08, 6.98)	0.56
$50 000 to $99 999	2.11	(0.86, 5.17)	1.00	0.54	(0.19, 1.52)	0.40	0.98	(0.46, 2.12)	0.25	1.06	(0.51, 2.21)	1.00	1.58	(0.6, 4.2)	0.69
Prefer not to say	2.21	(0.02, 220.2)	0.97	0.24	(0.01, 4.31)	0.23	1.00	-	-	0.15	(0.02, 1.4)	0.16	2.18	(0.07, 72.17)	0.68
$49 999 or less (Reference)
Satisfaction with iAgree
PSSUQ total	0.96	(0.92, 1)	0.11	0.97	(0.94, 1.01)	0.14	0.98	(0.96, 1.01)	0.33	0.98	(0.96, 1.02)	0.30	0.95	(0.92, 0.99)	0.07
Net Promoter Score	1.23	(1.01, 1.5)	0.09	1.10	(0.89, 1.35)	0.37	1.26	(1.07, 1.48)	0.06	1.15	(0.99, 1.34)	0.09	1.36	(1.09, 1.7)	0.03
Reasons for sharing medical record data
a. I will personally benefit from the research	1.35	(1.09, 1.67)	0.01	1.05	(0.86, 1.28)	0.64	1.17	(0.99, 1.37)	0.11	1.05	(0.91, 1.22)	0.62	1.39	(1.11, 1.73)	0.01
b. Family or friends will benefit from the research	1.04	(0.84, 1.28)	0.72	1.11	(0.91, 1.34)	0.30	1.09	(0.93, 1.28)	0.38	1.14	(0.98, 1.32)	0.38	1.17	(0.94, 1.46)	0.38
c. Support medical research in general	1.24	(1.03, 1.48)	0.12	1.00	(0.84, 1.21)	0.97	1.00	(0.85, 1.19)	0.96	1.05	(0.91, 1.21)	0.53	1.13	(0.93, 1.37)	0.21
d. Help other patients with the same health conditions as me	1.23	(1.01, 1.5)	0.10	1.02	(0.83, 1.25)	0.85	1.13	(0.97, 1.33)	0.21	1.04	(0.89, 1.21)	0.81	1.37	(1.12, 1.69)	0.01
e. Like to help other patients, no matter health conditions	1.58	(1.26, 1.97)	0.00	1.13	(0.92, 1.38)	0.33	1.03	(0.87, 1.21)	0.73	1.24	(1.06, 1.45)	0.02	1.21	(0.99, 1.48)	0.11
f. Take advantage of medical research, should also make contribution	1.29	(1.06, 1.57)	0.03	1.27	(1.05, 1.54)	0.03	1.11	(0.95, 1.28)	0.19	1.15	(0.99, 1.32)	0.08	1.23	(1.03, 1.48)	0.04
g. Being patient myself, feel connected to future patient	1.42	(1.16, 1.74)	0.00	1.37	(1.08, 1.72)	0.01	1.14	(0.98, 1.34)	0.09	1.26	(1.08, 1.47)	0.01	1.41	(1.16, 1.72)	0.00
h. Know other patients who have shared their data for research	1.14	(0.94, 1.39)	0.30	1.13	(0.9, 1.41)	0.30	1.11	(0.93, 1.32)	0.30	1.13	(0.97, 1.32)	0.30	1.19	(0.97, 1.45)	0.30
i. Grateful to my doctors	1.05	(0.85, 1.31)	0.80	1.09	(0.89, 1.33)	0.67	0.99	(0.72, 1.37)	0.97	1.11	(0.95, 1.3)	0.49	1.32	(1.06, 1.66)	0.07
HINTS scores
Patient-Physician Relationship Sub-Scale	0.83	(0.72, 0.96)	0.06	0.92	(0.82, 1.05)	0.21	0.87	(0.78, 0.98)	0.05	0.91	(0.83, 1)	0.05	0.91	(0.79, 1.04)	0.18
Patient Activation Sub-Scale	0.76	(0.62, 0.92)	0.02	0.77	(0.62, 0.96)	0.02	0.80	(0.67, 0.95)	0.02	0.87	(0.75, 1)	0.05	0.74	(0.59, 0.93)	0.02
Issue Involvement Sub-Scale	1.17	(0.92, 1.49)	0.21	1.00	(0.81, 1.24)	1.00	1.04	(0.85, 1.27)	0.69	1.09	(0.91, 1.29)	0.35	0.89	(0.72, 1.1)	0.28
Privacy Concerns Sub-Scale	0.59	(0.41, 0.86)	0.02	0.94	(0.69, 1.27)	0.67	0.91	(0.69, 1.19)	0.59	0.64	(0.49, 0.85)	0.01	0.71	(0.48, 1.03)	0.12
Other characteristics
Concerns about data breaches	1.11	(0.76, 1.61)	0.59	1.16	(0.82, 1.65)	0.40	0.96	(0.71, 1.3)	0.81	1.20	(0.9, 1.61)	0.22	0.82	(0.55, 1.22)	0.32
TOFHLA Health Literacy Score	0.90	(0.75, 1.09)	0.29	0.94	(0.78, 1.14)	0.52	0.90	(0.76, 1.06)	0.22	1.05	(0.91, 1.2)	0.51	0.90	(0.72, 1.13)	0.36
Prior participation in research study
No (Reference)
Yes	5.30	(0.89, 31.57)	0.33	2.19	(0.51, 9.37)	0.51	1.67	(0.49, 5.62)	0.41	1.23	(0.47, 3.17)	0.68	2.46	(0.41, 14.94)	0.52

^aIncludes individuals who identified as Trans-Male.

^bIncludes individuals who identified as Trans-Female.

^cOnly individuals who identified as Female were included in this study.

^dOnly Study 1 with the altered fact “willingness to share socioeconomic information” had a sample size large enough to included individuals who identified as Nonbinary.

In Study 1 (Depression and Readmissions), among 102 participants asked to share family health information, the strongest predictor of willingness to share was altruism, particularly endorsing the motivation “Like to help other patients” (OR: 1.58, 95% CI: 1.26-1.97, P < .001). Other significant motivations included personal benefit (OR: 1.35, 95% CI: 1.09-1.67, P < .05), a sense of moral obligation (“Take advantage of medical research, should also make contribution”; OR: 1.29, 95% CI: 1.06-1.57, P = .03), and patient solidarity (“Being a patient myself, feel connected to future patients”; OR: 1.42, 95% CI: 1.16-1.74, P < .01). Higher scores on Patient Activation (OR: 0.76, 95% CI: 0.62-0.92, P = .02) and Privacy Concerns (OR: 0.59, 95% CI: 0.41-0.86, P = .02) were negatively associated with sharing. Among 106 participants asked for social and economic information, sharing was significantly associated with feeling connected to future patients (OR: 1.37, 95% CI: 1.08-1.72, P = .011) and a belief in contributing to medical research (“Take advantage of medical research, should also make contribution”; OR: 1.27, 95% CI: 1.05-1.54, P = .032). Again, higher Patient Activation scores predicted lower willingness to share (OR: 0.77, 95% CI: 0.62-0.96, P = .02).

In Study 2 (Meningioma Registry), among 180 participants asked to share genetic health information, no motivations for sharing reached statistical significance. However, higher Patient Activation scores were negatively associated with willingness to share (OR: 0.80, 95% CI: 0.67-0.95, P = .02), and the Patient-Physician Relationship Sub-Scale showed a borderline negative association (OR: 0.87, 95% CI: 0.78-0.98, P = .05).

In Study 3 (Peripheral Nerve Injuries Registry), among 186 participants asked to share mental health information, significant motivations included altruism (“Like to help other patients”; OR: 1.24, 95% CI: 1.06-1.45, P = .02) and patient solidarity (“Being a patient myself, feel connected to future patients”; OR: 1.26, 95% CI: 1.08-1.47, P = .01). Higher Privacy Concerns scores were negatively associated with willingness to share (OR: 0.64, 95% CI: 0.49-0.85, P = .01), and the Patient-Physician Relationship Sub-Scale again showed a borderline negative association (OR: 0.91, 95% CI: 0.83-1.00, P = .05).

In Study 4 (IVF Outcomes), among 115 participants asked to share pregnancy-related information, several motivations were significantly associated with willingness to share, including personal benefit (OR: 1.39, 95% CI: 1.11-1.73, P = .01), helping other patients with the same condition (OR: 1.37, 95% CI: 1.12-1.69, P = .01), moral obligation (“Take advantage of medical research, should also make contribution”; OR: 1.23, 95% CI: 1.03-1.48, P = .04), and patient solidarity (“Being a patient myself, feel connected to future patients”; OR: 1.41, 95% CI: 1.16-1.72, P < .001). Higher Patient Activation scores were again negatively associated with willingness to share (OR: 0.74, 95% CI: 0.59-0.93, P = .02). Additionally, a higher Net Promoter Score, reflecting greater satisfaction with the iAgree platform, was positively associated with willingness to share (OR: 1.36, 95% CI: 1.09-1.70, P = .03).

Across all studies, while unadjusted models highlighted the influence of altruistic and identity-based motivations, these associations did not persist after adjustment.

Qualitative analysis of free-text questions

We identified 3 themes from 214 responses to the free-text question, “Please share any other reasons that would motivate you to share your medical record data for research”: (1) Social responsibility and empowerment in research participation, (2) Recognition of research’s importance for disease prevention and treatment, and (3) Trust, confidentiality, and transparency. About half of respondents fell into the first theme, emphasizing a sense of civic duty and personal fulfillment. One participant described “fulfillment in contributing to the greater good and helping others through their participation in research.” Others who had benefited from research felt compelled to give back:

“Knowing that my data could help researchers understand the disease better and potentially lead to improved therapies or even cures feels incredibly empowering. It’s like being part of something bigger than myself and actively contributing to progress.”

“I have benefited from medical research myself, and I am willing to give back to society and support more related projects.”

Several participants also expressed a desire for more communication about how their data was used. As one wrote:

“I am interested in receiving updates or feedback about the research findings. Being informed about how my data has contributed to new knowledge can provide a sense of fulfillment and engagement in the research process.”

Another 40% of participants had responses that were categorized under the theme, (2) Recognize importance of research for prevention and treatment of disease. Participants wrote about the importance of sharing data to “contribute to the development of personalized medicine approaches, where treatments are tailored to individual genetic, environmental, and lifestyle factors,” and aiding “in the development of new medical technologies and diagnostic tools.” For some participants, there was a recognition of the importance of data sharing in cases where there were few with a particular condition. As one participant wrote:

“Rare diseases, each piece of data is vital. My contribution can greatly support research efforts in these fields,” explained one participant.

Approximately 6% of participants wrote about the Importance of trust, confidentiality, and transparency and noted that while they were willing to share data, their reasons for sharing were predicated on trust in the institution and the importance of transparency. Wrote one participant:

“I rarely am not open to sharing data. However, it is important to me that the consent is explicit each and every time and that I also have an opportunity to receive the study results.”

One participant noted that they both understood how important it was to share data given their profession, but also had concerns about how the data could be used:

“Just this is who i am and i work in healthcare and know how beneficial data is. But I have my fears of using them against me as well (lack of trust maybe).” A few participants wrote that they would not share their medical record data “for any reason.”

We analyzed responses to the free-text question, “If your medical record data were used for research, how concerned would you be about the potential for data breaches?” and identified 2 main themes: (1) Data security and privacy concerns and (2) Trust, awareness, and perceived importance. The first theme captured fears about unauthorized access and misuse of personal information. Participants expressed concern over “misuse or unethical purposes,” potential “stigmatization or social repercussions,” and “violation of privacy” involving identifiable data. The second theme reflected varying levels of institutional trust and perceptions of legal safeguards. Some participants voiced skepticism, citing concerns about future legal changes and the durability of privacy protections. Others expressed confidence in existing laws and institutional practices, noting their belief that “data security is guaranteed under the legal framework” and that research institutions would “keep my health records data as secure as possible.”

Discussion

In this mixed-methods study, we used the iAgree simulation platform to examine what motivates people to share their medical record data. Participants interacted with simulated consent scenarios and completed a comprehensive survey. We found that consent status did not significantly differ based on study design or the type of information requested. However, participants were more willing to share their data when motivated by altruism, perceived personal benefit, and a sense of solidarity with other patients. Those with higher privacy concerns were less likely to share family health and mental health information. Participants with higher patient activation scores, who tend to be more engaged in managing their health, also showed greater caution in sharing data. One possible explanation is that more activated patients, who are generally more informed and engaged in managing their health, may also be more cautious about risks and more deliberate in exercising control over their information. They may also be more aware of potential privacy concerns or more skeptical of data use that does not clearly align with their own goals. These findings should be further explored in future research. Our findings suggests that patient engagement strategies around data sharing may need to be tailored: for highly activated patients, emphasizing transparency, granular consent options, and alignment with patient priorities may help address concerns, while for less activated patients, highlighting the societal benefits of research and reassuring patients about safeguards may be more effective.

In free-text responses, participants described three main motivations for data sharing: a sense of social responsibility and empowerment; recognition of research as essential for disease prevention and treatment; and the need for trust, confidentiality, and transparency. Many expressed a desire to contribute to the greater good, while 40% emphasized the value of data sharing for advancing personalized medicine and rare disease research. A smaller group said their willingness to share depended on receiving clear, explicit consent and trusting the institutions involved. When asked about data breaches, participants raised two key concerns: the risk of unauthorized access or misuse, and their level of trust in institutional safeguards and legal protections.

Our study purposefully framed all simulated scenarios around the use of de-identified data. While federal law and IRB regulations generally permit secondary use of de-identified information without explicit consent,³⁴ the central goal of iAgree is to examine the feasibility of an affirmative, opt-in consent process that exceeds legal requirements. By using de-identified data as the test case, we explored how participants would respond when invited to exercise greater agency in data-sharing decisions that are typically made on their behalf. This framing may have influenced participant perceptions of risk—some may have felt reassured that data could not be re-identified, reducing concerns that otherwise deter sharing. It is also important to recognize that iAgree introduces a layered “double opt-in” architecture: individuals must first self-select into the platform itself and then decide whether to consent to specific study requests. This architecture likely differs in meaningful ways from both traditional opt-out consent models and from other opt-in contexts, and may help explain why we did not observe demographic differences in consent rates, contrary to some prior studies.^20,35–37

Prior studies have shown that patients value transparency and autonomy in data-sharing decisions, often preferring opt-in models that provide granular control over which data elements are shared and with whom.^{17,19,35,38,39} Our findings align with this literature, reinforcing that patients’ willingness to share data is influenced by trust in data governance and perceived benefits to themselves and others. We specifically examined how demographic factors such as race, ethnicity, sex, and age influenced study participation. According to a systematic review on barriers and facilitators of research participation by racial and ethnic minority groups, minoritized groups’ mistrust of medical research stems from fears of purposeful mistreatment, unfamiliar procedures, and unintended consequences.⁴⁰ One prior study found that African American participants in the United States preferred greater control over their biospecimen sample in a biobank.³⁶ Unlike some earlier studies that found demographic factors significantly influenced data-sharing preferences,^37,41,42 our analysis did not reveal significant associations with age, race, education, or income. This discrepancy underscores the fact that when faced with making real-world choices about consent or data sharing about specific studies, as opposed to hypothetical questions about data sharing without additional context, participants may be fine with sharing their data for medical research. One potential explanation for our findings is that we included more transparency about the research, including the organization conducting the study (which was a trusted institution), the name of the PI, and data about the specific elements being used in the secondary data analysis. Participants were also given greater control over which data elements they wanted to share. Prior research has found that participants increase their trust in research when there is greater transparency.⁴³ The need for our study was driven by limited empirical data on how providing greater transparency, greater control over one’s data and how varying study characteristics and types of data influence consent decisions, which our simulation-based approach directly addressed.

Participants’ decisions to share data were influenced by complex psychological and social mechanisms, including altruism, perceived reciprocity, and personal identification with future patients, which has been found in other studies.^39,44,45 These findings suggest that fostering patient solidarity and emphasizing communal benefits in consent communications may enhance data-sharing rates. Practically, healthcare institutions and researchers can leverage these insights to design patient-centered consent platforms that highlight collective benefits without undermining individual autonomy. An actionable implication of this work is that consent platforms should provide patients with greater flexibility in their decisions. For example, dynamic consent platforms could allow participants to consent on a study-by-study basis and to select which categories of data they are comfortable sharing. While our quantitative analyses did not reveal significant demographic differences in willingness to share specific data types, our qualitative findings highlighted that participants valued the ability to exercise this control, underscoring its importance for fostering trust and transparency.⁴⁵

Participants in our study shared significant concerns about trust in institutions, specifically trust in the stability of federal laws to protect patients from being denied insurance coverage or concerns that their data could affect their healthcare. Participants in other studies have expressed similar concerns.^35,46,47 In one study examining genomic research and wide data sharing, one participant noted, “This is the privacy issue: that there’s no failsafe, as far as I’m concerned. And I would trust researchers, but I don’t trust the insurance industry, and I don’t trust the government.”⁴⁶ In another study using focus groups to understanding genetic research participation, another participant noted, “The government does shady things, so you never know where it’s going. So you can trust the government like only so far.”⁴⁷ These studies point to the need for researchers to work with policymakers to ensure there are strong legal protections for medical records data. While simulated consent scenarios provide valuable insights, real-world decisions may differ when patients consider institutional trust, provider recommendations, or potential risks of data misuse. For example, patients who have high trust in their health system or who receive encouragement from their clinicians may be more willing to share data when tested in real-world conditions, while those with prior negative experiences or concerns about insurers and commercial use may be more reluctant. Similarly, provider interactions could play a pivotal role: a recommendation from a trusted clinician may increase patients’ comfort with data sharing, whereas a lack of provider engagement may reduce participation. Future work should therefore pilot test iAgree within clinical workflows to evaluate whether the patterns observed in our study hold when patients make choices about their own records, and to better understand how trust, transparency, and provider relationships influence consent behaviors.

This study has several limitations. External validity may be limited by the sample’s demographics—primarily White, non-Hispanic, and highly educated—reducing generalizability. The sample composition may have led to increased levels of willingness to consent to the studies and higher willingness to share data compared to populations with lower levels of education and historically lower levels of institutional trust.^36,40 Moreover, our sample skewed relatively young (mean age: 35), which reduced our ability to understand how some of these questions may be answered by older individuals, who may have lower levels of trust with regards to data sharing.²¹ We also combined responses from cisgender and transgender participants due to the small number of transgender respondents; future research should explore this group specifically. While self-reported data may introduce social desirability bias, the use of anonymous surveys helps mitigate this concern. The design of iAgree required participants to actively opt in twice, once to participate in the platform and again to share data for each simulated study. This “double opt-in” design may have introduced selection effects related to health literacy, digital literacy, or institutional trust that we did not measure. It is possible that participants with certain sociodemographic characteristics did not participate in the study at all, and thus we lack data from non-responders, a limitation of all survey research. Third, by focusing on de-identified data, our findings may not fully reflect decisions participants would make if identifiable information were requested. These contextual features could help explain why we did not observe demographic differences in consent and highlight the importance of comparing opt-in versus opt-out architectures in future research. Despite these limitations, one of the strengths of the study is in using real-world examples of studies which have used clinical data for observational studies, which increases its external validity.

Conclusion

This study found that motivations for medical data sharing were driven mainly by altruism, perceived personal benefit, and patient solidarity, while privacy concerns and higher patient activation were linked to lower willingness to share. Notably, demographic factors did not significantly influence willingness to share, countering common concerns about disparities in data sharing across patient groups. These findings support the value of patient-centered consent models that prioritize transparency, data security, and clearly communicate the benefits of data sharing. Future research should assess how these patterns hold in more diverse populations and real-world clinical settings.

Author contributions

Michelle Sophie Keller (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), An Thi Nguyen (Data curation, Formal analysis, Investigation, Methodology, Project administration, Writing—review & editing), Chloe Leder (Data curation, Formal analysis, Investigation, Project administration, Writing—review & editing), Yunan Chen (Investigation, Methodology, Writing—review & editing), Brad Morse (Investigation, Methodology, Writing—review & editing), Lisa Schilling (Funding acquisition, Investigation, Methodology, Writing—review & editing), Di Hu (Methodology, Writing—review & editing), Spencer Soohoo (Funding acquisition, Investigation, Methodology, Writing—review & editing), and Lucila Ohno-Machado (Funding acquisition, Writing—review & editing)

Supplementary material

Supplementary material is available at Journal of the American Medical Informatics Association online.

Funding

This work was supported by the National Human Genome Research Institute of the National Institutes of Health under Award Number R01HG011066 (PI: Lucila Ohno-Machado).

Conflicts of interest

None declared.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.