INTRODUCTION: The incidence of dementia generally and Alzheimer’s disease specifically is increasing (1), largely because of the increased life expectancy of the global population, making this a major source of healthcare expenditure. Despite the increasing prevalence of dementia across the world, the disease is often diagnosed late in its progression. This is partly due to the heterogeneity of the disease both in symptom onset and progression.
No single cognitive test exists that can accurately diagnose dementia, the subtype of Alzheimer’s disease, or the preclinical stage of mild cognitive impairment (MCI) across a culturally diverse population. Instead, diagnoses are reached through a combination of clinician-administered tests, including assessments of medical and family history, cognitive function, other functional behaviors, peripheral biomarkers (e.g., nutritional deficiency), and, increasingly, brain imaging. Reliance on these diagnostic tools has led to significant health disparities in diagnosing, treating, and studying dementia and Alzheimer’s disease across the U.S. and around the world.
Even within high-resourced environments, by the time traditional Alzheimer’s symptoms of declining memory are noteworthy, the neurodegenerative trajectory is believed to be on a near-irreversible course. While there is no definitely curative drug treatment for the disease at present, delaying onset by just 5 years could potentially cut societal prevalence in the U.S. by 50% (2, 3). Early detection, and thus early intervention, could improve quality of life, helping alleviate symptoms and slow the progression of disease (4, 5).
Dementia is an insidious disease that takes up to decades to develop, and its nature provides the opportunity for prediction through subtle clinical changes that may appear years before a person meets the criteria for diagnosis. Increasingly, the use of digital biomarkers is being explored for screening and diagnosis, while ‘digital therapeutics’ are also emerging (5, 6). Digital biomarkers are physiological and behavioral measures collected from participants through digital tools that can be used to explain, influence, or predict health-related outcomes (7). The deep penetration of smartphones with voice recorders, coupled with the fact that production of speech involves multiple cognitive domains, suggests that voice-based digital biomarkers could open possibilities for a scalable, economical (automated transcription costs ∼$1 per 15-minute sample) (8), and widely accessible (due to real-time administration and scoring of neuropsychological tests) test to detect changes in individuals who have not yet met the threshold for clinical symptoms (9).
The goal of this study was to examine whether metrics extracted from digital audio recordings could serve as potential digital voice biomarkers, through the development of a predictive algorithm for earlier detection of cognitive dysfunction, and thereby improve outcomes.
Read the full publication here.