The Determinants of AI Adoption in Healthcare: Evidence from Voting and Stacking Classifiers

Abstract

Artificial intelligence (AI) has emerged as a disruptive force in the healthcare industry, driving new breakthroughs that promise to enhance treatment outcomes while simultaneously lowering costs. Artificial intelligence in healthcare has demonstrated promise to help doctors and patients at each step of the healthcare system, from an accurate diagnosis to urgent monitoring of patients and self-management of long-term illness. Despite physician and administrative interest, the use of these technologies in healthcare institutions remains limited. We hypothesized that risks such as black box issue, error rate, and legal risks halt the adoption.  Similarly, technical combability in healthcare centers stemming from cloud adoption, the presence of IT skills in healthcare, and digitalized healthcare records significantly explain the AI adoption in healthcare. To test our hypotheses, we applied Ensemble Voting Classifier and Stacking Classifier algorithms. The ensemble voting classifier outperforms the stacking classifier in terms of accuracy. Our findings indicate that majority of healthcare institutions with limited technological compatibility and high perceived risks have no plans to use artificial intelligence at this time. The majority of healthcare institutions with moderate risk perceptions and moderate technical combability are indecisive about integrating artificial intelligence. Healthcare facilities with good technological combability and low (AI) perceived risks are either uncertain or eager to use artificial intelligence approaches. Both classifiers yielded almost identical results, demonstrating the validity of our empirical findings.