VOICE-COVID-II : Alexa successful in SARS-CoV-2 symptoms screening

The pandemic put enormous pressure on healthcare systems. In the VOICE-COVID-II trial, the language-based AI model Alexa successfully recognised SARS-CoV-2 symptoms in patients with heart failure (HF) and their caregivers.

Prof. Abhinav Sharma (McGill University Health CentreMontreal, Canada) presented the VOICE-COVID-II trial (NCT04508972), which compared the accuracy of voice-based data captured by Alexa with that of human healthcare professionals regarding SARS-CoV-2 symptom screening in patients with HF and their caregivers [1]. A previous study has already shown the promise of this system in healthcare-based applications [2].

The single-centre, open-label, crossover, randomised-controlled trial, enrolled 52 patients with HF or caregivers and randomised them 1:1 to either a healthcare professional or the Alexa system. This was followed by a cross-over, so patients previously assigned to the healthcare professional were assigned to Alexa and vice versa. Both the AI-based technology and the healthcare professional performed a 5-point questionnaire including questions regarding preferred language,  purpose of the visit, and symptoms of SARS-CoV-2. The primary outcome of the trial was the overall concordance level for the 5-point screening questionnaire between the AI system and the healthcare professional.

The baseline characteristics revealed a predominantly male group with a median age of 51 years; 40% of the participants were patients with HF. Baseline characteristics were evenly distributed between groups.

In total, 520 questions were delivered and there was an overall agreement of 97.7% between AI and healthcare professional. This agreement was regardless of whether patients were first contacted by the professional and then switched to Alexa or vice versa. In addition, an unweighted kappa score of 0.93 (95% CI 0.88–0.99) suggested highly correlated results.

An additional post-screening survey showed that most patients and caregivers (83%) felt comfortable with the protocol and found it is easy to use. However, 25% of participants had privacy concerns regarding AI technology.

As Prof. Sharma pointed out, it is essential to demonstrate the ability of language-based AI models to complete simple tasks before moving on to more complex ones. “Our study is one of the first randomised trials of voice-based AI technologies among patients with HF and their caregivers who are often not considered in digital studies,” Prof. Sharma concluded.

1. 
   
   1. Sharma A, et al. Amazon Alexa to screen for SARSCoV2 symptoms. Session Late breaking clinical trials: acute heart failure and patients monitoring, Heart Failure 2023, 20–23 May, Prague, Czechia.
   2. Sharma A, et al. Eur H J Dig Health 2021;2:521–7.

Copyright ©2023 Medicom Medical Publishers



Posted on 8 August 202327 March 2024