Depression is common, globally on the rise, and correlates with voice disorders. The overall aim of the study is to identify voice acoustic features in healthy controls (HC) compared to people with Major Depressive Disorders (MDD) in a German sample. The study aims to identify (1) speech eliciting tasks and (2) acoustic features to differentiate the HC and MDD groups, and (3) to determine associations between acoustic features and sociodemographic (age, sex) as well as clinical data (BDI-II, NEO-FFI).

Twenty participants with a diagnosis of mild or moderate MDD and twenty HC complete a hierarchy of speech tasks (vowel production, counting, passage reading, picture description, video commentary) with varying valence (neutral, negative, positive, except for vowel/counting). Stimuli from validated databases were pre-tested in a German and Cantonese sample to determine a cross-cultural set of stimuli. Outcome measures include M/SD of fundamental frequency (fo), smoothed cepstral peak prominence (CPPs), and low-to-high harmonic ratio (LHR); semitones (ST SD); and the coefficient of variation (CoV) of F1 and F2.

To date, 17 participants in the HC group and 12 in the MDD group (10 = moderate), with a mean age of 25.2 years (SD =3.8), were included in the analyses. Preliminary data show significant group differences in BDI-II and NEO-FFI Neuroticism and Extraversion scores (p <0.001). There was a significant main effect of task for fo M (p < 0.001), CPPs SD (p < 0.007), and LHR SD (p = 0.036) with higher values for passage reading. The interaction effect of group by task by valence was significant for fo SD and ST SD besides various other significant two-way interactions. Linear regression of passage reading corrected for multiple comparisons showed significant sex effects for fo M and CPPs SD for all valence conditions and for fo SD for the negative and neutral condition.

Data collection continues. Initial data indicate that tasks differ in vocal engagement and that variability in voice features will be of primary interest. The findings are expected to contribute to the development of language-agnostic tools that leverage voice and speech biomarkers for early detection and monitoring of MDD in otolaryngologic and psychiatric settings.