Voice-Assisted Artificial Intelligence Enabled Screening for SARS-CoV-2 Exposure in Cardiovascular Clinics: Primary results of the VOICE-COVID-19-II Randomized Trial.

BACKGROUND: Voice-assisted artificial intelligence (AI)-based systems may streamline clinical care among patients with heart failure (HF) and caregivers; however, randomized clinical trials are needed. We evaluated the potential for Amazon Alexa® (Alexa), a voice-assisted AI-based system, to conduct screening for SARS-CoV2 in a HF clinic.

METHODS AND RESULTS: We enrolled 52 participants (patients and caregivers) from a HF clinic, who were randomly assigned with a subsequent cross-over to receive a SARS-CoV-2 screening questionnaire via Alexa or healthcare personnel. Primary outcome was overall response concordance, as measured by the percentage of agreement and unweighted kappa scores between groups. A post-screening survey evaluated comfort with using the AI-based device. In total, 36 (69%) participants were male, the median age was 51 (34-65) years and 36 (69%) were English-speaking. Twenty-one (40%) participants were HF patients. For the primary outcome, there were no statistical differences between groups: Alexa - Research coordinator group 96.9% agreement and unweighted kappa score of 0.92 (CI 0.84-1.00) versus Research coordinator - Alexa group 98.5% agreement and unweighted kappa score of 0.95 (CI 0.88-1.00) (p-value for all comparisons > 0.05). Overall, 87% of participants rated their screening experience as good or outstanding.

CONCLUSION: Alexa demonstrated comparable performance to a healthcare professional for SARS-CoV2 screening in a group of HF patients and caregivers and may represent an attractive approach to symptoms screening in this population. Future studies evaluating such technologies for other uses among patients with HF and caregivers are warranted. NCT04508972.