Patient preferences for the use of AI in chest X-ray result processing and communication

INTRODUCTION: Artificial Intelligence (AI) tools are increasingly used in imaging such as X-rays, but there is limited evidence about how patients view the use of AI in the diagnostic process. METHODS: Using the Technology Acceptance Model (TAM) to select attributes and the Health Belief Model (HBM) to interpret between-group differences this study evaluates patients' preferences regarding the use of AI in processing and communication of chest X-ray (CXR) results in the context of the NHS in the UK. A choice-based conjoint experimental (CBC) design was employed to examine various attributes influencing patient preferences across three samples 1) respondents with lung cancer diagnosis, 2) respondents who have had chest X-rays for any reason and 3) respondents with no recent CXRs. Furthermore, word-emotion association analysis was used to assess and compare the emotions levels in the open-text responses across the samples. RESULTS: A total of 440 respondents completed the study. Overall, respondents expressed support for using AI in X-ray processing, believing it would reduce waiting times. Waiting time and the decision on the next step of the care pathway accounted for the largest share of preference with a combined relative importance; the share of choice variation attributable to these attributes, of ∼67% in Samples 1 and 2 and 60% in Sample 3. Lung cancer patients (Sample 1) significantly favoured combined Human and AI processing (mean part-worth +13.3, p < 0.001), whereas Samples 2 and 3 favoured Human-only to Human and AI (mean part-worth -5.7, p = 0.008 and -12.0, p < 0.001 respectively for the attribute level Human and AI). CONCLUSION: Overall, patients welcome the supplementary use of AI in their care, with lung cancer patients having higher acceptance. IMPLICATIONS FOR PRACTICE: Understanding patient preferences can inform healthcare providers about optimising the diagnostic pathway, improving patient experiences, and potentially alleviating patient anxiety.