Patients support AI for diagnostic imaging but balk at using it in triage decisions

Patients support deploying artificial intelligence in diagnostic imaging but balk at using it for triage decisions, according to new survey results. 

The technology continues to gain momentum in radiology and other specialties, though data is limited around how consumers feel about its proliferation. European researchers recently aimed to close this information gap, sharing their findings Jan. 5 in the journal of Insights into Imaging. 

Based on a survey of 200 patients, 92% said they expect to benefit from AI in healthcare, though 73% admitted their knowledge about it is limited. About 87% of those polled are in favor of using AI for diagnostics, and 73% said so for therapy-related decisions. However, only 28% approved of the technology’s implementation in patient triage, which the authors labeled as a novel research finding. 

“A possible explanation for the rejection of the use of AI in the field of triage may be that the subject of ‘triage’ is a topic associated with fear and uncertainty,” Dr. Omid Nikoubashman, a professor and radiologist with University Hospital Aachen, Germany, and colleagues noted. “It is conceivable that uncertainty towards the mechanisms of triage, combined with uncertainty towards AI, amplifies disapproval in this field,” they added later. 

Researchers administered their voluntary, paper-based questionnaire at their academic medical center (University Hospital RWTH Aachen), targeting patients who had a diagnostic imaging appointment sometime between 2022 and 2023. Those responding were an average age of 49 and split evenly between men and women. The survey was written in plain language, took less than 10 minutes to complete and included a mix of multiple- and single-choice questions, also covering sociodemographic factors such as education, migration background and employment status. Respondents also had a range of disorders, including orthopedic and oncologic conditions.

About 84% of respondents said they wanted to be informed about the use of artificial intelligence in at least one of the mentioned care domains (diagnostics, therapy, triage). Those with higher education, greater self-assessed knowledge about AI and personal experience with the technology showed greater approval of it. 

Nearly 73% of patients who supported the general use of AI in healthcare in this study rejected its use for triage. The authors cited previous research, noting that resistance to medical AI may be caused by difficulty in understanding both algorithms and human decision-making in healthcare. “Interestingly,” they added, those with a higher degree of knowledge about artificial intelligence tended to support its use for triage. Nikoubashman and colleagues also noted that most participants approved of the idea of AI as a support system for physicians, rather than a substitute.

The study did not unearth any noteworthy findings related to how different demographic groups view AI. Previous studies determined that seniors, women, and those with lower levels of education and technical affinity have less trust in healthcare artificial intelligence. However, the study’s sample size may have been too small to suss out such details. 

“Nonetheless, it still makes sense to pay special attention to the mentioned patient group[s], because [they are] generally at risk of being marginalized in healthcare,” the authors noted. “All in all, our results, with approval and disapproval of AI depending on the field of application, emphasize that rather broad efforts must be undertaken to address the concerns of all patients, not only to improve their understanding and thus acceptance of AI in healthcare, but also to protect potentially marginalized groups from possible disadvantages.”