top of page

Vocal Biomarkers: How Ellipsis Health reads between the lines in Mental Health Diagnosis

Can the subtle inflections in our voice unravel the complexities of our mental health?

Understanding our mental state remains one of the most challenging aspects facing the medical community. But at Ellipsis Health, we are pioneers in spotting valuable hidden cues — vocal biomarkers — that shine a light on our internal state.

Until the dramatic impact of machine learning in healthcare, psychoanalysis relied on traditional self-reporting methods and lengthy assessments with a mental health provider. While these still play an important role, self-reporting relies heavily on a person’s ability to articulate their feelings, which can be influenced by social stigmas and personal awareness and provider visits are subject to availability and financial limitations. Vocal biomarkers can democratize assessments, by covering unsaid, unrecognized elements of our communication, making these clues to clinicians more expansive, accurate, and reliable.

The impacts and applications of this breakthrough technique are vast and Ellipsis Health is leading the way in unlocking that potential. We have already been recognized by CB Insights’ Digital Health 150 as a leader in screening, monitoring, and diagnostics, with a healthy list of peer-reviewed studies and approved patents.

On a more practical level, we are already an integral part of health plans and systems such as Cigna and Ceras Health, giving healthcare providers a fuller picture of a patient’s mental state before one-to-one consultations.

Join us as we detail the process behind using vocal patterns that demonstrate clues to a person’s internal state and what this breakthrough means for mental health screening in the future.

Understanding vocal biomarkers in mental health

Before jumping into the minutiae of how our technology works, we need to answer the obvious question: What are vocal biomarkers in the context of mental health assessment and diagnosis?

The vocal biomarkers refer to the aspects of speech behind the actual words spoken (semantics) and the way those words are spoken (acoustics, i.e. variations in tone, shifts in pitch, the rhythm of sentences, and the tempo of our dialogue).

The ultimate breakthrough we have achieved through this technology is objectivity. While clinicians can measure a patient’s blood pressure, heart rate, and oxygen levels to get a picture of their physical health, mental health assessment has commonly relied on self-reported PHQ and GAD-7 surveys which are more subjective. Vocal biomarkers represent the first measurable vital sign for mental health, allowing clinicians to assess issues accurately, providing decision support to aid in their diagnosis and treatment.

Ellipsis Health’s approach to vocal biomarkers

Spotting the potential in the use of vocal biomarkers for mental health assessment, Ellipsis Health brought together a team of experts from a range of disciplines, including artificial intelligence, machine learning, behavioral health, and neuroscience.

This combination of skill sets has allowed us to make progress that was simply not possible in the era before the machine learning capabilities available today. Only highly trained clinicians could detect and analyze elements of speech behind the words themselves, which is a slow, subjective process that relies on professionals who remain in short supply.

Now, we can analyze short patient recordings with vast deep-learning models to spot trends common to people suffering from depression and anxiety. This accelerates the identification of disorders, leading to early detection and intervention on a much wider scale than previously possible.

Navigating the challenges in implementing vocal biomarker technology

As with any breakthrough technology, vocal biomarkers in the mental health sphere present unique challenges to overcome.

Ensuring reliability across diverse populations

One of the primary challenges in the realm of vocal biomarkers — and in fact in machine learning algorithms in general — is ensuring the reliability and accuracy of our tools across different languages and cultural backgrounds. Currently, Ellipsis Health’s technology is proficient in English across varied demographics, having run expansive studies of up to 11,000 individual speakers in the USA. This runs against the grain of comparable studies, which are typically based on small corpora.

Expanding our capabilities to include more languages and dialects is a crucial goal, but it requires a careful, meticulous approach. We are committed to growing our data sets with diverse voices to ensure that our technology is inclusive and effective for a broad range of users.

Upholding privacy and consent in a digital age

We live in an era where data security is paramount — especially when dealing with mental healthcare. As effective as it is in screening for mental health issues, collecting voice recordings is a sensitive process, so Ellipsis Health has an unwavering commitment to privacy and consent. Our technology strips every recording of personal health information and identifiable details to give patients the confidence to speak freely.

We adhere to stringent data privacy laws and regulations, namely HIPAA, GDPR, and SOC2, ensuring that our users’ voice data is handled with the utmost care and confidentiality. With an eye to international expansion and rigor, we are also in the process of attaining HITRUST certification, a global risk management framework.

The future of vocal biomarkers in mental healthcare

At Ellipsis Health, we know that the journey of vocal biomarkers in mental healthcare is just getting started. Having already proved the efficacy of the technology, we are now focused on continually enriching our datasets, making algorithms more robust and inclusive.

But that’s only part of the battle.

According to the CDC, over 50% of people who die by suicide are not diagnosed with a mental health condition, so we have a desperate need for efficiencies in screening. Vocal biomarkers are already proving to be a decisive tool in accelerating the diagnosis, treatment, and monitoring of mental health issues, relieving pressure on healthcare systems where physician burnout jumped from 38% to 63% from 2020 to 2021.

These streamlined solutions benefit payers too. Screening from their home gives members much-needed treatment before issues escalate. From a financial perspective, a recent Evernorth study from the Journal of the American Medical Association (JAMA) shows that behavioral outpatient care can save payers newly diagnosed with conditions such as anxiety, depression, and substance abuse, among other issues. By closely monitoring progress and minimizing the risk of costly crisis events, payers can save more than $2,500 over the first 15 months post-diagnosis, extending to a reduction of $3,300 over 47 months.

Therefore, we’re continuing to build on collaborations with health insurance companies such as Cigna International Health, healthcare providers, and case management companies like Ceras Health to reach the widest number of patients and members possible.

As we continue to explore new possibilities of vocal biomarkers in mental health, we want to keep you in the know. Follow Ellipsis Health on LinkedIn for all the latest breakthroughs, partnerships, and applications in this exciting field.


bottom of page