ABSTRACT
Background: People with COVID-19 may experience impairing, and sometimes persisting, symptoms that would require enhanced surveillance. Our objective was to train an artificial intelligence-based model to predict the presence of COVID-19 symptoms and derive a digital vocal biomarker for easily and quantitatively monitoring symptom resolution. Methods: We used data from 272 participants in the prospective Predi-COVID cohort study recruited between May 2020 and May 2021, with both voice recordings and simultaneous COVID-19 related symptom assessment during the first two weeks of follow-up. A total of 6473 voice features were derived from recordings of participants reading a standardized pre-specified text. Models were trained separately for Android devices (3gp audio format) and iOS devices (m4a audio format). A binary outcome (symptomatic versus asymptomatic) was considered, based on a list of 14 frequent symptoms: dry cough, fatigue, sore throat, loss of taste and smell, diarrhea, fever, respiratory problems, increase in respiratory problems, difficulty eating or drinking, skin rash, conjunctivitis or eye pain, muscle pain/unusual aches, chest pain, overall pain level. Performances of the predictive models were evaluated according to the AUC, balanced accuracy, precision, recall, F1 score, and Matthews correlation coefficient. Findings: A total of 1775 audio recordings were analyzed (6.5 recordings per participant on average), including 1049 corresponding to symptomatic cases and 726 to asymptomatic ones. The best performances were obtained from Support Vector Machine models for both audio formats. We observed an elevated predictive capacity for both Android (AUC=0.91, balanced acc=0.85) and iOS (AUC=0.85, balanced accuracy=0.77) as well as good calibrations (Brier Scores = 0.11 and 0.16 respectively for Android and iOS). The vocal biomarker derived from the predictive models accurately discriminated asymptomatic and symptomatic individuals with COVID-19 (t-test P-values<0.001 for both Android and iOS devices). Interpretation: In this prospective cohort study, we have demonstrated that using a simple, reproducible task of reading a standardized pre-specified text of 25 seconds enabled us to derive a vocal biomarker for monitoring the resolution of COVID-19 related symptoms with elevated accuracy and calibration. An external validation, especially in other languages, would now be required before integrating such a vocal biomarker in telemonitoring solutions or in clinical practice.Trial Registration: This study was registered in ClinicalTrials.gov (NCT04380987)Funding: The Predi-COVID study is supported by the Luxembourg National Research Fund (FNR)(Predi-COVID, grant number 14716273), the André Losch Foundation, and the LuxembourgInstitute of Health.Declaration of Interest: Ethical Approval: This study was approved by the National Research Ethics Committee ofLuxembourg (study number 202003/07) in April 2020.