Voice of the Patient

See article by Bayrampour et al

When it comes to measurement tools, psychiatry is a medical discipline at a disadvantage. We have no machines to help diagnose depression, there are no laboratory tests for anxiety, and there is no scanner to locate the root of psychosis in the brain. There are instruments, to be sure, a wealth of patient-completed questionnaires and structured clinical interviews, though most have limited use in clinical practice and all remain subjective and prone to various forms of bias. Fifty-four years after Dr Fred Plum’s famously dour quote,¹ “Schizophrenia is the graveyard of neuropathologists,” progress toward the objective measurement of brain health remains scarce—but it is not entirely absent.

There is hope in the form of a new generation of digital biomarkers. While these tools have been critiqued as being largely unsubstantiated hype,² a new study by Bayrampour et al³ offers something more. As described in their article, “Using Speech to Develop Multivariable Prediction Models for Major Depressive Disorder and Generalized Anxiety Disorder During Pregnancy,” a novel set of digital tools may help accurately diagnose depression and anxiety in pregnancy. These tools rely on a modality familiar to every clinician: the clinical interview.

Specifically, their study utilized recordings of the Structured Clinical Interview for DSM-5. Originally collected as part of a study of anxiety in pregnancy, these recordings were digitized and parsed into discrete features and then analyzed through machine learning and other techniques. While the approach taken to testing different speech features was technically complex and involved a number of specialized software tools, many of the variables described are conceptually straightforward, eg, pauses in speech, vocal pitch, and irregularities in vocal tone.

Although issues such as diarization (accurate identification of multiple speakers in a recording) and interpretation of concepts such as harmonics-to-noise ratio are of technical interest, the real value of this work lies in making the case for speech as a biomarker. With a small set of acoustic features predicting either major depressive disorder or generalized anxiety disorder, a set of clear phenotypes emerges from the data. In some cases, their accuracy exceeds established risk factors, including planned versus unplanned pregnancy, education, pregnancy complications, and household income.

While the study was focused on pregnancy, one of the conclusions the authors draw is that the most predictive features are those that work across populations, such as the length of pauses in patient speech. This observation is strikingly consistent with prior published findings, particularly those of Cohen et al (2024), demonstrating the role of speech latency extracted from interviews collected during a clinical trial of bipolar depression.⁴ While Cohen et al focused on speech latency as a tool for enrichment and patient selection in clinical trials, their work aligns well with Bayrampour and colleagues’ approach, leveraging vocal features to aid in screening among vulnerable populations.

Speech production is a complex behavior requiring cognitive, social, emotional, and neuromotor resources.⁵ It is also central to most psychiatric assessments. During a structured or semistructured clinical interview, patients may be asked about a range of topics, from their symptoms and daily life to role functioning and sexual health. Formulating and verbalizing responses strains multiple cognitive resources, and so the interview functions as a challenge test with a high degree of ecological validity. When paired with digital technologies, speech in a semistructured interview becomes a novel, quantified, objective measure of performance, one nested in the simplest and most traditional form of clinical evaluation.

Being mindful of the critiques of digital biomarkers and their tendency to overpromise, it can at least be said the speech markers presented by Bayrampour and colleagues³ are a promising new source of data in mental health assessment and diagnosis. Yet, our next great advance may not come from genetics, three-dimensional imaging, or molecular findings. In the end, it may be the voice of the patient that finally refutes Dr Plum and guides psychiatry out of his proverbial graveyard and into a more objective, better quantified future.

Article Information

Published Online: September 9, 2026. https://doi.org/10.4088/JCP.26com16630
© 2026 Physicians Postgraduate Press, Inc.
J Clin Psychiatry 2026;87(4):26com16630
Submitted: July 27, 2026; accepted July 28, 2026.
To Cite: Opler M. Voice of the patient. J Clin Psychiatry 2026;87(4):26com16630.
Author Affiliations: Clario, a part of Thermo Fisher Scientific, Philadelphia, Pennsylvania; The PANSS Institute, New York, NY.
Corresponding Author: Mark Opler, PhD, MPH, Clario, 1818 Market St, Ste 2600, Philadelphia, PA 19103, United States ([email protected]).
Financial Disclosure: Dr Opler is a full-time employee of Clario Inc. and a shareholder of Thermo Fisher Scientific.
Funding/Support: None.