Preferences for artificial intelligence clinicians before and during the COVID-19 pandemic : discrete choice experiment and propensity score matching study

Abstract

Background: Artificial intelligence (AI) methods can potentially be used to relieve the pressure that the COVID-19 pandemic has exerted on public health. In cases of medical resource shortages caused by the pandemic, changes in people's preferences for AI clinicians and traditional clinicians are worth exploring. Objective: We aimed to quantify and compare people's preferences for AI clinicians and traditional clinicians before and during the COVID-19 pandemic, and to assess whether people's preferences were affected by the pressure of pandemic. Methods: We used the propensity score matching method to match two different groups of respondents with similar demographic characteristics. Respondents were recruited in 2017 and 2020. A total of 2048 respondents (2017: n=1520; 2020: n=528) completed the questionnaire and were included in the analysis. Multinomial logit models and latent class models were used to assess people's preferences for different diagnosis methods. Results: In total, 84.7% (1115/1317) of respondents in the 2017 group and 91.3% (482/528) of respondents in the 2020 group were confident that AI diagnosis methods would outperform human clinician diagnosis methods in the future. Both groups of matched respondents believed that the most important attribute of diagnosis was accuracy, and they preferred to receive combined diagnoses from both AI and human clinicians (2017: Odds ratio [OR] 1.645, 95% CI 1.535-1.763; P<.001; 2020: OR 1.513, 95% CI 1.413-1.621; P<.001; reference: Clinician diagnoses). The latent class model identified three classes with different attribute priorities. In class 1, preferences for combined diagnoses and accuracy remained constant in 2017 and 2020, and high accuracy (eg, 100% accuracy in 2017: OR 1.357, 95% CI 1.164-1.581) was preferred. In class 2, the matched data from 2017 were similar to those from 2020; combined diagnoses from both AI and human clinicians (2017: OR 1.204, 95% CI 1.039-1.394; P=.011; 2020: OR 2.009, 95% CI 1.826-2.211; P<.001; reference: Clinician diagnoses) and an outpatient waiting time of 20 minutes (2017: OR 1.349, 95% CI 1.065-1.708; P<.001; 2020: OR 1.488, 95% CI 1.287-1.721; P<.001; reference: 0 minutes) were consistently preferred. In class 3, the respondents in the 2017 and 2020 groups preferred different diagnosis methods; respondents in the 2017 group preferred clinician diagnoses, whereas respondents in the 2020 group preferred AI diagnoses. In the latent class, which was stratified according to sex, all male and female respondents in the 2017 and 2020 groups believed that accuracy was the most important attribute of diagnosis. Conclusions: Individuals' preferences for receiving clinical diagnoses from AI and human clinicians were generally unaffected by the pandemic. Respondents believed that accuracy and expense were the most important attributes of diagnosis. These findings can be used to guide policies that are relevant to the development of AI-based health care.