Smartphone app identifies, predicts depressive symptoms with vocal biomarkers
Click Here to Manage Email Alerts
Key takeaways:
- Each participant used the app an average of 3.2 times per week.
- Participants reported improved mental health symptoms and better functioning at school, work and social activities at the end of the study.
A smartphone app demonstrated the ability to evaluate vocal biomarkers in a 30-second voice sample to identify and predict elevated mental health symptoms, according to a study published in Frontiers in Psychiatry.
“The ability to collect mental health data from patients between clinic visits could transform how we monitor symptoms and optimize treatment plans,” said Lindsey Venesky, PhD, NCSP, a licensed psychologist and clinical director at the Cognitive Behavior Institute in Pittsburgh, said in a related press release from Sonde Health. “Voice-based health tracking technology can provide accurate insights into a client’s mental health status over time and can do so seamlessly and unobtrusively, with little added effort for clients.”
Venesky and colleagues conducted a cohort study in which they asked participants to use a smartphone app developed by Sonde Health to record 30-second voice samples for 4 weeks. They also completed the M3 Checklist at the beginning and end of the 4-week period to report mental health symptoms, including those related to depression, anxiety, PTSD and bipolar disorder, and their impact on school, work and social interactions.
The app analyzed the voice recordings and provided researchers with a mental fitness vocal biomarker (MFVB) score based on vocal patterns like jitter and shimmer, pitch variability, energy variability, vowel space, phonation duration, speech rate and pause duration. Scores ranged from 0 to 100, where 80 to 100 indicated “excellent” mental fitness, 70 to 79 was labeled as “good,” and 0 to 69 was labeled as “pay attention.”
The researchers used two approaches when analyzing the association between MFVB scores and M3 results. The closest-MFVB approach compared a single voice recording that was conducted nearest in time to the M3 survey, whereas the time-weighted approach analyzed changes in MFVB scores over time as participants continued using the app. Then they calculated relative risk estimates for worse M3 symptom severity in each MFVB score category and compared them with relative risk ratios.
The study included 104 participants (73% female; aged 16 to 80 years; 93% white) who used the app at least once. Eighty-one people completed the offboarding survey. Anxiety was the most common diagnosis (38%), followed by trauma and stress-related disorders (31%) and depression disorders (20%).
Participants completed 1,336 voice recordings, for an average of 3.2 per participant per week, as well as 185 M3 assessment, 177 of which were included in the analysis.
The relative risks for elevated mental health symptoms included 0.82 (95% CI, 0.61-1.1) for those with an excellent MFVB score, 1.02 (95% CI, 0.85-1.24) for the good group and 1.25 (95% CI, 0.99-1.58) for the pay attention group with the closest-MFVB method, with similar results observed for the time-weighted method.
The relative risk ratio comparing the pay attention and excellent groups was 1.53 (95% CI, 1.09-2.14) for closest-MFVB and 2 (95% CI, 1.21-3.3) for time-weighted MFVB, indicating that participants in the pay attention category are more likely to experience elevated mental health symptoms, according to the researchers.
The MFBV analysis was most effective for identifying depression symptoms, researchers wrote. The relative risk ratio among people with depression was 1.78 (95% CI, 1.08-2.93) for closest-MFBV and 2.6 (95% CI, 1.45-4.66) for time-weighted MFVB.
Participants also experienced a 13% reduction in elevated mental health symptoms from the beginning of the study to the end (65% vs. 53%). At the end of the study, 25% to 30% reported better functioning at work, school and in social situations. Among participants who had significant impairment in these areas at baseline, 35% to 40% reported improvement.
“This study further validates our voice-based health tracking platform as an objective indicator of mental well-being,” said Erik Larsen, PhD, MS, senior vice president of clinical development and customer success at Sonde Health, said in the press release. “The results show vocal biomarkers can provide meaningful insights into mental health in a preventive, scalable way. We believe the MFVB can foster stronger awareness about individuals’ mental wellbeing, thereby encouraging them to cultivate healthy habits and proactively mitigate their mental health risks.”
Reference:
- Study Confirms Sonde Health’s voice-based mental fitness solution accurately identifies individuals with elevated mental health symptoms. https://www.sondehealth.com/news-blog/sonde-healths-voice-based-mental-fitness-solution-accurately-identifies-individuals-with-elevated-mental-health-symptoms. Published March 11, 2023. Accessed March 14, 2023.