INTRODUCTION
Uses of patient data have become increasingly varied as precision medicine and artificial intelligence (AI) expand across the US healthcare system.1 This raises questions about how patient concerns and preferences regarding their data may shift, with implications for willingness to disclose information and the desire for communication around data sharing. Building on previous literature,2,3 we conducted a longitudinal survey of US adults to understand how their perceptions of data use are changing.
METHODS
The survey was fielded by the National Opinion Research Center (NORC) using the AmeriSpeak Panel, with a nationally representative sample of US adults aged 18 and older. The first wave of the survey was conducted in May 2019 (n=2157) and the second wave was conducted in October and November 2021 (n=1541). The weighted retention rate was 78.4%. NORC calculated post-stratification longitudinal survey weights based on the Current Population Survey and non-response to produce national estimates. Only respondents that answered both survey waves and had nonmissing values for our measures of patient preferences were included in the analytic sample (n=1517). Participants viewed a brief video describing data use in healthcare to ensure a shared understanding of the topic.4
Participants reported their agreement with statements about their (1) desire for notification, (2) comfort with the use of their deidentified data, and (3) comfort with the use of their identifiable data on a 4-point Likert scale (1=not true, 2=somewhat true, 3=fairly true, 4=very true). We conducted weighted multivariable linear regressions with covariates (demographic variables, perceptions of integrity and trust,3) and respondent fixed effects to examine changes in these outcomes over time as well as their associations with self-reported demographic variables and trust in the healthcare system.
RESULTS
In 2019, 60.7% of respondents reported desire for notification of data use; this increased to 74.2% in 2021. Between 2019 and 2021, comfort with deidentified data use increased with 45.2% of respondents reporting comfort in 2019 compared to 47.2% in 2021. In contrast, comfort with identifiable data use decreased between waves with 14.8% reporting comfort in 2019 and 14.5% reporting comfort in 2021.
In our primary analyses (Table 1), the desire for notification of data use increased from 2019 to 2021 by 0.23 points, while comfort with data use, both de-identified and identifiable decreased by 0.09 and 0.21 points, respectively. However, the decline in comfort with deidentified data use is not statistically significant.
Table 1.
Estimates from the Weighted Multivariate Model with Covariates
| Notification of data use (p) |
Comfort w/ deidentified data (p) |
Comfort w/ identifiable data (p) |
|
|---|---|---|---|
| Wave 2 | 0.232 | −0.0852 | −0.207 |
| (<0.001***) | (0.180) | (0.013*) | |
| Age | −0.0159 | −0.0253 | −0.00557 |
| (0.149) | (0.197) | (0.847) | |
| Male | ref. | ref. | ref. |
| Female | 0.676 | 2.008 | 0.618 |
| (0.260) | (0.003**) | (0.354) | |
| White | ref. | ref. | ref. |
| Black | 1.478 | 0.487 | −0.573 |
| (0.009*) | (0.529) | (0.476) | |
| Hispanic | −0.267 | 0.0713 | 0.778 |
| (0.750) | (0.934) | (0.435) | |
| Other race | −0.232 | 0.757 | 0.190 |
| (0.748) | (0.293) | (0.775) | |
| Less than HS | −0.208 | 0.132 | −0.215 |
| (0.612) | (0.697) | (0.506) | |
| High school | ref. | ref. | ref. |
| Some college | −0.741 | 0.0820 | −0.0376 |
| (<0.001***) | (0.245) | (0.640) | |
| College or higher | 0.0899 | 0.779 | −0.337 |
| (0.763) | (0.025*) | (0.269) | |
| Married | 0.0126 | −0.103 | −0.0961 |
| (0.953) | (0.763) | (0.848) | |
| Employed | −1.402 | −1.028 | −0.250 |
| (<0.001***) | (<0.001***) | (<0.001***) | |
| Income: <$30,000 | ref. | ref. | ref. |
| Income: $30,000–$59,999 | −0.267 | −0.0376 | 0.114 |
| (0.276) | (0.854) | (0.437) | |
| Income: $60,000–$99,999 | −0.441 | −0.403 | −0.0670 |
| (0.279) | (0.090) | (0.745) | |
| Income: ≥$100,000 | −0.614 | −0.251 | −0.205 |
| (0.124) | (0.399) | (0.373) | |
| Region: New England | ref. | ref. | ref. |
| Region: Mid-Atlantic | 1.636 | −1.524 | 0.316 |
| (<0.001***) | (0.027*) | (0.689) | |
| Region: East North Central | 1.279 | −2.073 | 0.0278 |
| (0.001**) | (<0.001***) | (0.953) | |
| Region: West North Central | 1.441 | −1.750 | −0.536 |
| (<0.001***) | (<0.001***) | (0.150) | |
| Region: South Atlantic | 1.455 | −1.712 | −0.227 |
| (<0.001***) | (<0.001***) | (0.693) | |
| Region: East South Central | 1.038 | −2.034 | 0.528 |
| (0.012*) | (0.002**) | (0.426) | |
| Region: West South Central | 1.143 | −2.214 | −0.00833 |
| (0.040*) | (<0.001***) | (0.988) | |
| Region: Mountain | 1.220 | −2.392 | −0.411 |
| (0.001**) | (<0.001***) | (0.296) | |
| Region: Pacific | 1.431 | −2.050 | −0.386 |
| (<0.001***) | (<0.001***) | (0.368) | |
| Insured | −0.0647 | 0.00972 | −0.102 |
| (0.456) | (0.932) | (0.357) | |
| Integrity | 0.0667 | 0.144 | 0.152 |
| (0.359) | (0.024*) | (0.046*) | |
| Global trust | 0.00520 | 0.283 | 0.150 |
| (0.915) | (<0.001***) | (0.025*) | |
| Constant | 4.499 | 4.045 | 2.098 |
| (<0.001***) | (<0.001***) | (0.176) | |
| Wave 1 mean | 3.11 | 2.71 | 1.86 |
| n | 3034 | 3034 | 3034 |
p-values in parentheses. The estimates are weighted using the post-stratification longitudinal survey weights. Other race is an indicator variable equal to 1 if the respondent identifies as Asian, multi-racial, or as a race or ethnicity other than White, Black, or Hispanic. Additional control covariates include binary indicators for marital status, employment, living in a metropolitan area, and having health insurance. Categorical variables for income and region are also included, along with composite measures of integrity and global trust. Respondent fixed effects are included in the regression
***p <0.001; **p<0.01; *p<0.05
Looking at the perceptions of data use over time by race (Table 2), we observe significant increases in the desire for notification among Black and White respondents and significant decreases in comfort with identified data use among White respondents. There was no significant change in comfort with data use among Black respondents between 2019 and 2021.
Table 2.
Estimates from the Weighted Multivariate Model with Covariates by Respondent Race
| Notification of data use | Comfort w/ deidentified data | Comfort w/ identifiable data | ||||
|---|---|---|---|---|---|---|
| White | Black | White | Black | White | Black | |
| Wave 2 | 0.202 | 0.235 | −0.0989 | −0.204 | −0.322 | −0.142 |
| (0.025*) | (0.007**) | (0.366) | (0.276) | (0.002**) | (0.364) | |
| Age | −0.0173 | −0.0173 | −0.0179 | −0.0233 | 0.0329 | −0.0305 |
| (0.524) | (0.122) | (0.628) | (0.389) | (0.320) | (0.196) | |
| Male | ref. | ref. | ref. | ref. | ref. | ref. |
| Female | −1.072 | −0.730 | −2.713 | 0.378 | −0.103 | 0.0465 |
| (0.121) | (0.362) | (0.006**) | (0.268) | (0.897) | (0.943) | |
| Less than HS | 2.126 | 0.00859 | 1.653 | −0.514 | 1.071 | 1.062 |
| (0.013*) | (0.994) | (0.182) | (0.518) | (0.187) | (0.195) | |
| High school | ref. | ref. | ref. | ref. | ref. | ref. |
| Some college | −0.795 | −1.346 | 0.133 | 0.105 | −0.0665 | −0.551 |
| (<0.001***) | (0.097) | (0.134) | (0.829) | (0.514) | (0.567) | |
| College or higher | 2.248 | −1.355 | 2.230 | −2.165 | 1.012 | −2.255 |
| (0.008**) | (0.289) | (0.068) | (<0.001***) | (0.195) | (0.020*) | |
| Married | 0.0496 | −0.250 | −0.0495 | −0.116 | −0.944 | −1.017 |
| (0.825) | (0.742) | (0.872) | (0.785) | (0.002**) | (0.022*) | |
| Employed | −2.325 | 1.996 | 2.101 | −1.499 | −0.270 | −0.656 |
| (0.005**) | (0.006**) | (0.023*) | (0.010*) | (0.700) | (0.358) | |
| Income: <$30,000 | ref. | ref. | ref. | ref. | ref. | ref. |
| Income: $30,000–$59,999 | 0.0144 | 0.498 | −0.308 | 0.762 | 0.434 | 0.482 |
| (0.932) | (0.366) | (0.352) | (0.002**) | (0.109) | (0.191) | |
| Income: $60,000–$99,999 | 0.0487 | −0.464 | −0.614 | 0.704 | 0.223 | 0.0110 |
| (0.848) | (0.608) | (0.103) | (0.176) | (0.451) | (0.988) | |
| Income: ≥$100,000 | −0.163 | −0.511 | −0.106 | −0.559 | 0.478 | −0.719 |
| (0.540) | (0.410) | (0.822) | (0.151) | (0.171) | (0.138) | |
| Region: New England | ref. | ref. | ref. | ref. | ref. | ref. |
| Region: Mid-Atlantic | 2.133 | 1.898 | −0.674 | 0.0783 | 0.649 | 0.514 |
| (0.002**) | (0.042) | (0.419) | (0.851) | (0.444) | (0.484) | |
| Region: East North Central | 1.613 | 1.666 | −2.426 | −0.916 | −0.418 | 0.454 |
| (0.011*) | (<0.001***) | (0.001**) | (0.043) | (0.444) | (0.328) | |
| Region: West North Central | 1.456 | 0.818 | −1.722 | 0.259 | −0.936 | −0.00819 |
| (<0.001***) | (0.364) | (<0.001***) | (0.684) | (0.027*) | (0.993) | |
| Region: South Atlantic | 1.661 | 1.828 | −2.206 | −1.614 | −1.089 | −0.431 |
| (0.014*) | (0.018*) | (0.001**) | (0.002**) | (0.155) | (0.528) | |
| Region: East South Central | 1.382 | 1.311 | −2.371 | −1.646 | 0.0336 | −0.0777 |
| (0.026*) | (0.092) | (0.005**) | (0.004**) | (0.963) | (0.909) | |
| Region: West South Central | 2.060 | 1.466 | −1.609 | −1.211 | 0.312 | −0.0757 |
| (0.001**) | (0.003**) | (0.012*) | (0.002**) | (0.588) | (0.923) | |
| Region: Mountain | 1.566 | 1.584 | −2.660 | −1.489 | −0.830 | 0.377 |
| (0.007**) | (0.024*) | (<0.001***) | (0.015*) | (0.111) | (0.668) | |
| Region: Pacific | 1.692 | 1.637 | −2.165 | −0.832 | −0.960 | 0.485 |
| (0.002**) | (0.041*) | (<0.001***) | (0.071) | (0.045*) | (0.523) | |
| Metropolitan area | −0.0525 | −0.914 | 0.478 | −0.803 | −0.146 | −0.146 |
| (0.822) | (<0.001***) | (0.043*) | (<0.001***) | (0.689) | (0.823) | |
| Insured | −0.0660 | −0.0112 | 0.0374 | 0.0182 | −0.0802 | −0.365 |
| (0.623) | (0.966) | (0.749) | (0.960) | (0.540) | (0.394) | |
| Integrity | 0.0682 | 0.203 | 0.220 | 0.352 | 0.189 | 0.282 |
| (0.455) | (0.092) | (0.009**) | (0.167) | (0.115) | (0.152) | |
| Global trust | −0.0627 | 0.103 | 0.239 | 0.193 | 0.0823 | 0.210 |
| (0.229) | (0.413) | (<0.001***) | (0.429) | (0.174) | (0.437) | |
| Constant | 4.525 | 3.235 | 3.533 | 4.943 | 0.320 | 4.503 |
| (0.003**) | (0.002**) | (0.077) | (0.005**) | (0.877) | (0.025*) | |
| Wave 1 mean | 3.08 | 3.23 | 2.82 | 2.53 | 1.80 | 2.01 |
| n | 1852 | 426 | 1852 | 426 | 1852 | 426 |
p-values in parentheses. The estimates are weighted using the post-stratification longitudinal survey weights. Other race is an indicator variable equal to 1 if the respondent identifies as Asian, multi-racial, or as a race or ethnicity other than White, Black, or Hispanic. Additional control covariates include binary indicators for marital status, employment, living in a metropolitan area, and having health insurance. Categorical variables for income and region are also included as well as composite measures of integrity and global trust. Respondent fixed effects are included in the regression, but not shown
***p <0.001; **p<0.01; *p<0.05
DISCUSSION
We observe an overall increase in desire for notification and decrease in comfort with data use among the US adult population between 2019 and 2021. The associated decrease in comfort with data use is driven by White respondents’ opinions. As use of patient data for AI and precision medicine increase, there are implications for patient trust and comfort across demographic groups.4,5 It is possible that general awareness of the myriad uses of patient data within and external to the healthcare system is increasing,6 leading to public concerns about data sharing. Given patients’ reliance on electronic communication and trust in their providers,7 the stakes for communication around data sharing are high.
This study has limitations. We do not identify causal relationships. Rather, we identify a significant decrease in comfort and increase in desire for notification over time. There may be a relationship between comfort and effects of Covid-19, for which we cannot empirically account. However, these results indicate a general shift over time and additional longitudinal measurement will be important for understanding how patient perspectives evolve. Additionally, future work should continue to explore the important questions raised about demographic differences in comfort with data sharing.
The shift in patient preferences regarding use of their data and desire for notification indicate the importance of transparency and patient engagement as the data landscape in healthcare changes. Health systems and clinicians should be aware of this shift and endeavor to communicate about data use to prevent negative impacts on their relationships with patients.
Author Contribution
Drs. Gangaram and Nong contributed equally to this work.
Funding
The project was supported by the National Institute of Biomedical Imaging and Bioengineering (5R01EB030492-04).
Data Availability
The datasets analyzed in this work are available from the corresponding author on reasonable request.
Declarations
Ethical Approval
NORC received IRB approval for this work.
Consent to Participate
NORC obtained written consent from all survey participants.
Consent to Publish
Drs. Gangaram and Nong consent to the publishing of this work.
Conflict of Interest
Drs. Gangaram and Nong have nothing to disclose.
Footnotes
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
References
- 1.Wu K, Wu E, Theodorou B, et al. Characterizing the Clinical Adoption of Medical AI Devices through U.S. Insurance Claims. NEJM AI. 2023;1(1):AIoa2300030. 10.1056/AIoa2300030 [Google Scholar]
- 2.Spector-Bagdady K, Trinidad G, Kardia S, et al. Reported Interest in Notification Regarding Use of Health Information and Biospecimens. JAMA. 2022;328(5):474-476. 10.1001/jama.2022.9740 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Platt JE, Jacobson PD, Kardia SLR. Public Trust in Health Information Sharing: A Measure of System Trust. Health Serv Res. 2018;53(2):824-845. 10.1111/1475-6773.12654 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Nong P, Adler-Milstein J, Kardia S, Platt J. Public Perspectives On the Use of Different Data Types for Prediction in Healthcare. J Am Med Inf Assoc. Published online February 1, 2024:ocae009. 10.1093/jamia/ocae009 [DOI] [PMC free article] [PubMed]
- 5.Quinn TP, Senadeera M, Jacobs S, Coghlan S, Le V. Trust and Medical AI: the Challenges We Face and the Expertise Needed to Overcome them. J Am Med Inform Assoc. 2020;28(4):890-894. 10.1093/jamia/ocaa268 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Shabani M, Bezuidenhout L, Borry P. Attitudes of Research Participants and the General Public Towards Genomic Data Sharing: a Systematic Literature Review. Expert Rev Mol Diagnos. 2014;14(8):1053-1065. 10.1586/14737159.2014.961917 [DOI] [PubMed] [Google Scholar]
- 7.Moore A, Chavez C, Fisher MP. Factors Enhancing Trust in Electronic Communication Among Patients from an Internal Medicine Clinic: Qualitative Results of the RECEPT Study. J Gen Intern Med. 2022;37(12):3121-3127. 10.1007/s11606-021-07345-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The datasets analyzed in this work are available from the corresponding author on reasonable request.