Voice-based screening for SARS-CoV-2 exposure in cardiovascular clinics

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Clinical Trial Data Capture

Background:

• Voice-based AI assistants offer potential for clinical data capture, but feasibility in healthcare settings is unproven.
• Assessing SARS-CoV-2 exposure risk factors and symptoms is crucial in cardiovascular clinics.

Purpose of the Study:

• To explore the feasibility and accuracy of Amazon Alexa for screening SARS-CoV-2 exposure risk factors and symptoms.
• To evaluate the interrater reliability between Alexa and healthcare providers in symptom screening.

Main Methods:

• Prospective enrollment of 215 participants in quaternary care cardiovascular clinics.
• Screening by healthcare provider followed by Amazon Alexa, with comparison using Cohen's Kappa statistic.
• Participant satisfaction survey with Alexa's performance.

Main Results:

• High interrater reliability between Alexa and healthcare providers (Cohen's Kappa: 0.989 simple, 0.992 weighted).
• Alexa misidentified only one screening response out of 645 questions.
• Participants rated Alexa highly (mean score 4.4/5).

Conclusions:

• AI-driven voice assistants are feasible for multilingual data collection in SARS-CoV-2 exposure screening.
• Integration of such AI tools in cardiovascular healthcare and clinical trials is warranted.
• Voice assistants demonstrate potential for efficient and accurate clinical data acquisition.