Kintsugi is developing smarter mental healthcare infrastructure using voice biomarkers to streamline access to care.
The surge in demand for healthcare access has transformed infrastructure to handle volume, but not intelligently. Virtual health has not solved for the $200B cost of undiagnosed depression, yet. - Depression and anxiety are hard to track with paper tests. - Chronic conditions are 80% comorbid with depression. Depression directly increases treatment costs 2-3X. - 30-40% of undiagnosed patients are being missed and health payors & systems are paying for these costs. The current standard of care for mental health screening is lacking: the process of screening for mental health with PHQ-9 and GAD-7 is administratively burdensome given its lengthy process and does not fit well into existing clinical workflows, especially for virtual care. Further, the screening decision is made by the practitioner based on his/her subjective perception of the patients’ needs leading to screening biases and inconsistencies. This results in low overall rates of screening (less than 5% of primary care visits) with significant variance across populations. African American patients are half as likely to be screened as white counterparts. Similarly, elderly patients are also half as likely to be screened as middle-aged patients. Screening with an objective tool such as KiVA™ that can be seamlessly embedded into clinician workflows has the potential to increase screening rates dramatically as well as improve consistency of screening across genders, ages and ethnic backgrounds. The non-intrusive nature of KiVA™ means that screening can be conducted in the background with no disruption to the clinical workflow. This in turn leads to a frictionless, enhanced patient experience.
The Kintsugi Voice Biomarker API (KiVA™) is a cloud-based software API that can aid in the detection of clinical depression and anxiety from a short raw audio voice sample. Samples can be as short as 20 seconds, and the scoring scales against the PHQ-9 and GAD-7 clinical modules. KiVA™ is a versatile clinical decision support tool that seamlessly integrates into call centers, telehealth platforms, care management platforms, and remote patient monitoring apps. By utilizing KiVA™ screenings, practitioners can improve various steps in their care cycle: - KiVA™ screening provides early detection and gives healthcare providers the ability to manage the severity and ultimately cost of mental health disorders (e.g. by reducing hospitalizations and emergency visits). - KiVA™ screening scores allow for risk stratification by severity and a plan for 95% of individuals who are not in the most severe 5%. The whole spectrum from mild to moderate to severe can be treated with the appropriate tools and clinical services. - Medication titration and adherence, faster route to prescription efficacy. Separation of concerns: depression and anxiety are often comorbid with chronic conditions. - By using the tool over time, healthcare providers can gain insight on factors affecting changes to depression and anxiety scores throughout the patient treatment cycle. Kintsugi provides a unique platform to analyze voice biomarker signals for risk population segmentation and data enrichment to pre-existing virtual visit workflows. The Kintsugi voice journaling app, loved by users in over 250 international cities, has KiVA™ embedded to help consumers track their well-being over time.
As a National Science Foundation awarded company, our patent-pending technology is meaningfully advanced. There are several competitors in the market, but currently our product is the only one that works in a production environment with voice samples as short as 20 seconds. - Allows for real-time scoring: KiVA™ can run in real-time using only 20 seconds of audio and returning the results to the practitioner with low latency. - Language agnostic: Since the ML models that power the technology were trained on the world’s largest annotated voice dataset with international users from 250 international cities, the technology is language agnostic. - KiVA™ is deployed as an API endpoint in a large Payor's highest volume care management platform (20MM calls a year). - Secure endpoint that only requires an audio file to score results. - API-first platform for flexible workflow integrations - FDA De Novo pending
Before integrating KiVA™ into their clinical call centers, the executive team at a large Provider Network wanted to look closer at the problem of undetected depression across their patient population. We analyzed over 16K audio files of clinical call center data. Along with the audio data, we received demographic information (age, race, veteran status, etc.), past PHQ-9 screening scores, and patient diagnosis data. We analyzed relative screening rates across the selected demographics, and ran the audio files through our KiVA™ API solution in order to determine how many patients in the selected group should be flagged for depression. Our results were a strong indicator of the value Kintsugi can provide to healthcare organizations. Across all age and demographic groups, KiVA™ identified more than a 30% increase in incremental depressed patients. These incremental patients were in addition to patients who had already been diagnosed with depression. As a result of this analysis, we are currently working through our Statement of Work, and will be going live over the next 2-3 months in their clinical call centers, focusing on identifying Medicare patients with depression comorbid with chronic health conditions.
Kintsugi’s team has deep clinical, technical, and operational expertise. The NSF has awarded us multiple SBIR grants for novel AI technologies to detect depression and anxiety from 20 seconds of speech. FDA Breakthrough, DeNovo, and patents pending, Kintsugi enables healthcare organizations to identify, triage, and care for patients at scale.
