Should I Choose Artificial Intelligence or Clinicians' Diagnosis? a Discrete Choice Experiment of Patients' Preference UNDER COVID-19 Pandemic in China.
Author(s)
Liu T*1;Tsang W2;Huang F2, Ming WK2
1University of Groningen, Liyang, 32, China, 2Jinan University, Guangzhou, China
OBJECTIVES: This study aims to quantify the strength of patients’ heterogeneous preferences for various aspects of Artificial intelligence (AI) diagnosis versus clinicians under the epidemic of COVID-19 in china. Moreover, to illustrate the different decision-making factors of the latent class of discrete choice experiment (DCE) and prospects for application of AI techniques in diagnosis treatment in the pandemic of SARS-CoV-2 and future.
METHODS: A DCE approach was used. Attributes from different dimensions which have been hypothesized were diagnostic method; waiting periods; diagnosis time; the precision of diagnosis; Follow-up support service, and diagnostic expenses. With data from the DCE component, a restricted latent class model was estimated with fixed size of each class to determine discrete ‘classes’ of diagnosis preferences. For statistical analysis, we construct generalized logit and mixed logit models with the 428 datasets.
RESULTS: 55.74 per cent of the respondents are opted for AI diagnosis regardless of the description of the clinicians. Logit models presented the evident patients' preference of 'AI+clinician' diagnosis method. And clearly, the higher accuracy is, the more patients would prefer. In addition, the most acceptable latent class model is consisting of three latent classes of respondents, defined by different preferences for the diagnosis cost, time, method, waiting time and accuracy. Attributes with the most substantial effect on choices were the accuracy and payments, especially the preferences for diagnosis ‘accuracy’ attribute, was constant across classes. Except for ‘class 1’, in which people highlight the diagnosis methods. All attributes had a significant effect on choices in the expected direction.
CONCLUSIONS: Considerable segments of respondents had fixed preferences for either diagnosis option. Applying latent class analysis was essential in quantifying preferences for attributes of diagnosis choice. People’s preference to the “Accuracy” was palpable. AI will have a potential market, however, accuracy and diagnosis expense are needed to be taken into consideration.
METHODS: A DCE approach was used. Attributes from different dimensions which have been hypothesized were diagnostic method; waiting periods; diagnosis time; the precision of diagnosis; Follow-up support service, and diagnostic expenses. With data from the DCE component, a restricted latent class model was estimated with fixed size of each class to determine discrete ‘classes’ of diagnosis preferences. For statistical analysis, we construct generalized logit and mixed logit models with the 428 datasets.
RESULTS: 55.74 per cent of the respondents are opted for AI diagnosis regardless of the description of the clinicians. Logit models presented the evident patients' preference of 'AI+clinician' diagnosis method. And clearly, the higher accuracy is, the more patients would prefer. In addition, the most acceptable latent class model is consisting of three latent classes of respondents, defined by different preferences for the diagnosis cost, time, method, waiting time and accuracy. Attributes with the most substantial effect on choices were the accuracy and payments, especially the preferences for diagnosis ‘accuracy’ attribute, was constant across classes. Except for ‘class 1’, in which people highlight the diagnosis methods. All attributes had a significant effect on choices in the expected direction.
CONCLUSIONS: Considerable segments of respondents had fixed preferences for either diagnosis option. Applying latent class analysis was essential in quantifying preferences for attributes of diagnosis choice. People’s preference to the “Accuracy” was palpable. AI will have a potential market, however, accuracy and diagnosis expense are needed to be taken into consideration.
Conference/Value in Health Info
2020-11, ISPOR Europe 2020, Milan, Italy
Value in Health, Volume 23, Issue S2 (December 2020)
Code
SP3
Topic
Health Technology Assessment, Medical Technologies, Methodological & Statistical Research, Patient-Centered Research
Topic Subcategory
Artificial Intelligence, Machine Learning, Predictive Analytics, Decision & Deliberative Processes, Digital Health
Disease
No Specific Disease
Explore Related HEOR by Topic