Abstract
This survey study reports the perspectives and preferences of US adults regarding use of photographs of their skin in medical research, education, and development of image-based artificial intelligence (AI).
Increasing use of imaging for research, education, and development of image-based artificial intelligence (AI) is parallel to increasing concerns about confidentiality and autonomy. Regulatory requirements for collecting and processing personal information vary geographically, but even the most stringent legal guidelines do not require informed consent for sharing of deidentified data.1,2,3,4 Despite widespread use of imaging in dermatology and portability of digital image formats that enable both rapid intentional and inadvertent image sharing, information about attitudes and preferences on consent and identifiability are limited. Consequently, processes for obtaining informed consent are not standardized across clinical practices and research journals.4 To inform practices for protecting privacy, we performed a survey study to elucidate perspectives on image use, consent, and identifiability.
Methods
In December 2022, we recruited 1316 English-speaking US adults using an online survey recruitment platform (Prolific). The sample size was divided into subgroups of gender identity, and race and ethnicity with the same proportions as the US population (eg, the sample has a similar proportion of Black individuals as the US). Respondents completed a 29-item cross-sectional survey (eAppendix in Supplement 1) and were compensated $19.43 per hour. Race and ethnicity, along with other sociodemographic characteristics, were self-reported. We calculated descriptive statistics using R 4.2.2 (R Core Team). The Memorial Sloan Kettering Institutional Review Board approved this study. Respondents provided oral consent.
Results
Demographics of 1316 respondents (625 [47.6%] men, 673 [51.0%] women, 18 [1.4%] other or nonbinary gender identity; median [IQR] age, 43 [30-58] years) are depicted in Table 1. Most respondents (1048 [79.6%]) agreed that images are necessary for skin-related medical care (Table 2). Most respondents (843 [64.1%]) reported no change in trust in the quality of their care if their images were used for research or education. If they were asked for permission to use their images, over half of respondents (741 [56.3%]) reported an increase in trust and 533 (40.5%) reported no change in trust. Respondents reported comfort with use of their clinical images for publication in a journal (596 [45.3%]), training AI models (718 [54.6%]), presentation at an academic conference (596 [45.3%]), and education of students and medical trainees (816 [62.0%]).
Table 1. Demographics of Survey Respondents.
| Characteristic | Survey respondents, No. (%) (N = 1316) | Proportion in 2020 US population, %d |
|---|---|---|
| Age, median (IQR), y | 43 (30-58) | NA |
| Gender identity | ||
| Man | 625 (47.5) | 49.6 |
| Woman | 673 (51.1) | 50.4 |
| Other and nonbinarya | 18 (1.4) | NA |
| Race and ethnicityb | ||
| Asian | 78 (5.9) | 6.1 |
| Black or African American | 168 (12.8) | 13.5 |
| Native American or Alaska Native | 6 (0.5) | 1.3 |
| Native Hawaiian or Pacific Islander | 1 (0.1) | 0.3 |
| White | 1040 (79.0) | 75.9 |
| Otherc | 20 (1.5) | NA |
| Unknown | 3 (0.2) | NA |
Abbreviation: NA, not applicable.
Other gender identities were not specified.
Race and ethnicity were self-reported in the survey.
Other race and ethnicity groups were not specified.
Source: US Census Bureau.
Table 2. Survey Questions and Responses.
| Question | Response, No. (%) | |||||
|---|---|---|---|---|---|---|
| “Indicate how much you agree or disagree with the following statement: Photographs are necessary for my skin-related medical care.” | Strongly agree: 471 (35.8) | Somewhat agree: 577 (43.8) | Undecided: 179 (13.6) | Somewhat disagree: 64 (4.9) | Strongly disagree: 20 (1.5) | No response: 5 (0.4) |
| “How much do you trust your medical provider?” | A lot: 488 (37.1) | A fair amount: 692 (52.6) | Not very much: 77 (5.9) | Not at all: 15 (1.1) | I don’t have a medical provider: 44 (3.3) | No response: 0 |
| “Would it impact your trust in the quality of care you receive” | ||||||
| “If your medical provider shares medical photographs of your skin with your other care providers?” | Strongly increase trust: 29 (2.2) | Increase trust: 86 (6.5) | No change in trust: 974 (7.4) | Decrease trust: 188 (14.3) | Strongly decrease trust: 34 (2.6) | No response: 5 (0.4) |
| “If your medical provider uses medical photographs of your skin for research or education purposes?” | Strongly increase trust: 34 (2.6) | Increase trust: 104 (7.9) | No change in trust: 843 (64.1) | Decrease trust: 263 (20.0) | Strongly decrease trust: 71 (5.4) | No response: 1 (0.1) |
| “If your medical provider asked you whether they could use medical photographs of your skin for research or education purposes?” | Strongly increase trust: 222 (16.9) | Increase trust: 519 (39.4) | No change in trust: 533 (40.5) | Decrease trust: 36 (2.7) | Strongly decrease trust: 6 (0.5) | No response: 0 |
| “How comfortable would you be with medical photographs of your skin being used in the follow[ing] ways?” | ||||||
| “To help my medical provider diagnose conditions” | Very comfortable: 862 (65.5) | Somewhat comfortable: 294 (22.3) | Undecided: 43 (3.3) | Somewhat uncomfortable: 34 (2.6) | Very uncomfortable: 34 (2.6) | No response: 49 (3.7) |
| “To communicate with other medical providers within my care team about my health” | Very comfortable: 667 (50.7) | Somewhat comfortable: 399 (30.3) | Undecided: 110 (8.4) | Somewhat uncomfortable: 59 (4.5) | Very uncomfortable: 30 (2.3) | No response: 51 (3.9) |
| “To communicate with other medical providers outside of my care team about my health” | Very comfortable: 398 (30.2) | Somewhat comfortable: 358 (27.2) | Undecided: 219 (16.6) | Somewhat uncomfortable: 228 (17.3) | Very uncomfortable: 87 (6.6) | No response: 26 (2.0) |
| “To identify skin changes to bring up with my medical provider” | Very comfortable: 772 (58.7) | Somewhat comfortable: 317 (24.1) | Undecided: 127 (9.7) | Somewhat uncomfortable: 62 (4.7) | Very uncomfortable: 28 (2.1) | No response: 10 (0.8) |
| “How comfortable would you be with the use of medical photographs of your skin for the following research or education purposes?” | ||||||
| “For a publication in an academic journal” | Very comfortable: 240 (18.2) | Somewhat comfortable: 356 (27.1) | Undecided: 243 (18.5) | Somewhat uncomfortable: 258 (19.6) | Very uncomfortable: 217 (16.5) | No response: 2 (0.2) |
| “For training artificial intelligence models to help improve diagnosis” | Very comfortable: 333 (25.3) | Somewhat comfortable: 385 (29.3) | Undecided: 235 (17.9) | Somewhat uncomfortable: 201 (15.3) | Very uncomfortable: 160 (12.2) | No response: 2 (0.2) |
| “For a presentation at an academic conference” | Very comfortable: 271 (20.6) | Somewhat comfortable: 325 (24.7) | Undecided: 239 (18.2) | Somewhat uncomfortable: 256 (19.5) | Very uncomfortable: 222 (16.9) | No response: 3 (0.2) |
| “For the education of students and medical trainees” | Very comfortable: 379 (28.8) | Somewhat comfortable: 437 (33.2) | Undecided: 190 (14.4) | Somewhat uncomfortable: 183 (13.9) | Very uncomfortable: 124 (9.4) | No response: 3 (0.2) |
| “How comfortable would you be with the use of the following medical photographs of your skin for research or education purposes?” | ||||||
| “Photos that do not contain identifiable visual features” | Very comfortable: 770 (58.5) | Somewhat comfortable: 347 (26.4) | Undecided: 100 (7.6) | Somewhat uncomfortable: 62 (4.7) | Very uncomfortable: 32 (2.4) | No response: 5 (0.4) |
| “Photos that contain identifiable visual features” | Very comfortable: 84 (6.4) | Somewhat comfortable: 194 (14.7) | Undecided: 201 (15.3) | Somewhat uncomfortable: 428 (32.5) | Very uncomfortable: 399 (30.6) | No response: 10 (0.8) |
| “If you have had or were to have medical photographs taken of your skin, would you be concerned that certain visual features of the photographs are identifiable?” | Not concerned: 271 (20.6) | A little concerned: 408 (31.0) | Undecided: 44 (3.3) | Somewhat concerned: 406 (30.9) | Very uncomfortable: 185 (14.1) | No response: 2 (0.2) |
| “Would you want to be asked before the use of” | ||||||
| “Identifiable photographs of your skin for research or education?” | Yes: 1286 (97.7) | NA | NA | NA | NA | No response: 2 (0.2) |
| “Nonidentifiable photographs of your skin for research or education?” | Yes: 1166 (88.6) | NA | NA | NA | NA | No response: 0 |
Abbreviation: NA, not applicable.
Five hundred ninety-one respondents (45.0%) reported concern over identifiability; 679 reported (51.6%) little or no concern. Most respondents (1117 [84.9%]) were comfortable with use of nonidentifiable images; only 278 (21.1%) were comfortable with use of identifiable images. Most respondents preferred to be asked before use of both nonidentifiable (1166 [88.6%]) and identifiable (1286 [97.7%]) images.
Discussion
Sharing images for clinical research, education, and development of image-based AI has associated ethical and legal concerns. Prior studies suggest patients might share their images, especially for clinical purposes,2 but consent preferences have not been published. Informed consent is not required for using deidentified data for clinical research and education, abiding by US and European laws, and practices protecting privacy remain unstandardized.1,2,3 The CLEAR (Checklist for Evaluation of Image-Based AI Reports in Dermatology) guidelines provide a framework for addressing the challenges of image use for AI in dermatology, but guidelines on consent based on patient preferences are needed.5
Findings of this study highlight the need for transparent and standardized consent and deidentification processes. Most respondents reported that their trust in the quality of care they receive would increase if they were asked for consent to use their medical images, regardless of identifiability. This response aligns with studies that conclude health information transparency plays a role in engendering trust.6
The study is limited in generalizability because we used a convenience sample of technology-literate English-speaking individuals, broadly defined clinical images, and did not ask baseline health care experience or discuss the range of potentially identifiable features. We also did not investigate attitudes toward data collection for for-profit AI companies, an important area for future study. Larger studies that are powered to examine historically excluded subgroups are needed for generalizable findings.
eAppendix. Survey Questions
Data Sharing Statement
References
- 1.Shinkai K, Bruckner AL, Robinson JK. Best practices for sharing images in clinical care, research, and education-protecting patient privacy. JAMA Dermatol. 2023;159(7):695-697. doi: 10.1001/jamadermatol.2023.1080 [DOI] [PubMed] [Google Scholar]
- 2.Ly S, Reyes-Hadsall S, Drake L, et al. Public perceptions, factors, and incentives influencing patient willingness to share clinical images for artificial intelligence-based healthcare tools. Dermatol Ther (Heidelb). 2023;13(11):2895-2902. doi: 10.1007/s13555-023-01031-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Kovarik CL, Sanabria B, Stoff BK. Image consent and the development of image-based artificial intelligence. JAMA Dermatol. 2022;158(5):589. doi: 10.1001/jamadermatol.2022.0689 [DOI] [PubMed] [Google Scholar]
- 4.Nanda JK, Marchetti MA. Consent and deidentification of patient images in dermatology journals: observational study. JMIR Dermatol. 2022;5(3):e37398. doi: 10.2196/37398 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Daneshjou R, Barata C, Betz-Stablein B, et al. Checklist for evaluation of image-based artificial intelligence reports in dermatology: CLEAR Derm consensus guidelines from the International Skin Imaging Collaboration Artificial Intelligence Working Group. JAMA Dermatol. 2022;158(1):90-96. doi: 10.1001/jamadermatol.2021.4915 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mangal S, Park L, Reading Turchioe M, et al. Building trust in research through information and intent transparency with health information: representative cross-sectional survey of 502 US adults. J Am Med Inform Assoc. 2022;29(9):1535-1545. doi: 10.1093/jamia/ocac084 [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.
Supplementary Materials
eAppendix. Survey Questions
Data Sharing Statement
