Felix Menne, Elisa Mallick, Johannes Tröger, Alexandra König, Janna Schulze, Diana Immel, Simon Barton, & René Hurlemann
*Presented at ISCTM 2025
Abstract
Methodological Issue Being Addressed: Evaluating treatment effects in psychiatry often relies
on subjective assessments, like questionnaires, which can introduce bias and limit diagnostic
accuracy—particularly in clinical trials and personalized treatment contexts. Objective measures
are needed to enable a more accurate, transdiagnostic approach to mental health. Automated
speech analysis shows promise as a non-invasive, objective tool for tracking affective and
behavioral changes. However, developing scalable and sensitive speech measures that reliably capture subtle fluctuations in symptom severity remains a challenge but are crucial for enhancing clinical trials
and supporting their use in relapse prevention, monitoring, and treatment assessment, ultimately
paving the way for broader adoption in psychiatric practice.
Introduction: This study aimed to evaluate the potential of speech measures to monitor short-term
changes in depressive symptoms in individuals with Major Depressive Disorder (MDD), compared to
healthy controls. Specifically, we sought to determine whether certain speech features reflect
symptom fluctuations over time and distinguish individuals with increasing versus decreasing
symptom severity. This approach could indicate whether speech biomarkers could be employed for
tracking symptom progression and treatment response, with potential applications in early relapse
detection in clinical practice.
Methods: Participants included 22 individuals with MDD and 22 healthy controls, recruited at the
Karl-Jaspers Clinic of Psychiatry, University Hospital Oldenburg, Germany. Depressive symptoms
were assessed using the Beck Depression Inventory (BDI) at two points, two weeks apart. Based on
BDI changes, participants were grouped as “decliners” (higher BDI scores at the second time point,
n=13) and “non-decliners” (lower or unchanged BDI scores, n=31). Automated speech analysis was
conducted on recordings from both time points, examining specific speech features. Participants
completed three storytelling tasks (positive, neutral, and negative) and a “PaTaKa” task, which
involved rapid repetition of the syllables “pa-ta-ka” for 10 seconds. The primary analysis focused on
detecting significant differences in speech characteristics between decliners and non-decliners,
assessing speech biomarkers’ sensitivity to short-term symptom changes. Group differences were
evaluated using the non-parametric Kruskal-Wallis test, with Benjamini-Hochberg correction applied
to p values.
Results: The change in BDI between T0 and T1 was significant within the groups of decliners
(p<0.001) and non-decliners (p=0.01). After adjusting for multiple comparisons, no significant
group differences were found in storytelling-based speech features between the two groups.
However, significant differences were observed between decliners and non-decliners in temporal
features derived from the PaTaKa task (see Table 2).
Conclusion: These findings suggest that temporal features from the PaTaKa task are more
sensitive to short-term changes in depressive symptoms than storytelling-based features. This
heightened sensitivity may result from the PaTaKa task’s focus on motor coordination and speech
timing, which directly reflect psychomotor slowing—a core symptom of depression. By isolating
rhythm and articulation speed, the PaTaKa task effectively captures subtle psychomotor changes
that may go undetected in more complex tasks. Also, psychomotor symptoms might be more
sensitive to treatment response than other symptoms such as anhedonia. These results underscore
the importance of task selection in developing speech biomarkers for monitoring symptom
progression in MDD.
