BACKGROUND
Medical artificial intelligence (MAI) is developing rapidly. It plays an important role in the screening, diagnosis, and management of various diseases. However, developing and implementing MAI also give rise to many ethical dilemmas. Although MAI ethics frameworks have already been published at home and abroad in response to growing concerns, the adoption is still limited. The effective practice of MAI ethics requires comprehensive knowledge, an active attitude, and practical experience. Therefore, it is necessary to develop a scientifically and specifically designed scale for measuring researchers’ ethics knowledge, attitude, and practice (KAP) in MAI research.
OBJECTIVE
This study aimed to develop and validate an instrument for measuring researchers’ knowledge, attitude, and practice in MAI ethics (KAP-MAIE), and to examine its reliability and validity.
METHODS
The KAP-MAIE scale was developed according to the KAP model and strict methodology for developing a reliable and valid scale. We established a list of items based on literature review and expert consultation. A pilot survey was conducted on 12 MAI researchers to ascertain the applicability of the scale. In total, 269 MAI researchers were invited to participate in the reliability and validity evaluation of the scale. We performed item analysis for content refinement, correlation analysis, exploratory factor analysis (EFA), and confirmatory factor analysis (CFA) for construct validity, and calculated the interrater agreement(IRR), item-level content validity index(I-CVI), and scale-level CVI (S-CVI) for content validity, Cronbach alpha coefficient, and Split-half reliability for internal consistency.
RESULTS
The KAP-MAIE scale includes a different number of items for each section, namely knowledge (10 items), attitude (6 items), and practice (7 items). The EFA resulted in a three-factor solution consistent with three subscales of knowledge, attitude, and practice, accounting for 74.86% of the variance. The second-order CFA demonstrated a good fit to the data (root mean square error of approximation=0.073). Each subscale correlated with overall scale, and subscales also correlated with each other. The correlation coefficient on subscale and overall scale was higher than that between subscales.The inter-rater agreement, I-CVI, and S-CVI were excellent. The scale also had an exceptional internal consistency (Cronbach alpha=0.933, Split-half reliability=0.727).
CONCLUSIONS
The proposed KAP-MAIE scale is a reliable and valid instrument that can measure researchers’ knowledge, attitude, and practice in MAI ethics. This is the first instrument developed for this purpose.