Real-time Remote Asthma Monitoring Through Smartphone Voice Analysis

Overview

This study examines the impact of home-based monitoring of respiratory function in asthma patients via a smartphone-based vocal biomarker platform. Previous work from cross-sectional studies has indicated that brief voice samples, analyzed by machine learning models, can predict the presence of respiratory conditions (asthma, COPD, ILD, COVID-19 and persistent cough) with an accuracy of approximately 70%. The present study seeks to extend these findings to establish whether the same vocal biomarker models can accurately track changes in respiratory function in asthma patients, and whether this capability, when incorporated into a smartphone app similar to those used for home-based asthma management, can improve relative level of asthma control.

The study aims to investigate whether home-based monitoring of respiratory function in asthma patients using a smartphone-based vocal biomarker platform can improve the level of asthma control. The study will involve 70 patients aged 18 years and above with a primary diagnosis of asthma or allergy with asthma as a comorbidity, with an Asthma Control Test (ACT) score of less than or equal to 19 at baseline. The primary objective of the study is to assess whether vocal biomarker scores can provide asthma patients with real-time objective information on their respiratory function using voice samples recorded on their personal smartphone device. The secondary objectives of the study include determining whether smartphone apps incorporating vocal biomarker capabilities can improve asthma control, assessing patient engagement with the apps, and examining the impact of the apps on healthcare utilization in asthma patients. Exploratory objectives include determining how the participant's digital health literacy phenotype can assist in creating engagement with asthma management apps and developing care team dashboards for providers to assist in periodic evaluation of patient status using a vocal biomarker platform. The study will explore voice recordings for potential new machine learning model development to assess whether different vocal biomarker combinations can provide superior respiratory function monitoring ability than the established Respiratory Symptom Risk scores that were developed from cross-sectional asthma data. The primary endpoints of the study are performance-related measures of Respiratory Symptom Risk score, and the secondary endpoints include the proportion of subjects with ACT improvement and the proportion of participants providing voice samples. The study will be conducted for three months in clinics providing care to asthma patients, typically asthma/allergy clinics. The Sonde Health App will be used to provide vocal biomarker respiratory symptom risk scores to patients on their personal smartphone device and be used for data collection. The findings of the study could help establish whether the same vocal biomarker models can accurately track changes in respiratory function in asthma patients and whether incorporating this capability into smartphone apps can improve the level of asthma control.