Predicting Ultra-High Risk Outcomes Using Linguistic and Acoustic Measures From High-Risk Social Challenge Recordings: mHealth Longitudinal Cohort Exploratory Study
Bibliographic record
Abstract
Background: Early detection of individuals at ultra-high risk (UHR) for psychosis is critical for timely intervention and improving clinical outcomes. However, current UHR assessments, which rely heavily on psychometric tools, often suffer from low specificity. Speech-based machine learning prediction models can potentially be used to improve prognostic accuracy. However, existing studies often used long, open-ended speech tasks, which limit scalability. The High-Risk Social Challenge (HiSoC) is a short 45-second speech task designed to measure social functioning in individuals with UHR. If the HiSoC task is able to capture predictive signals, it may serve as an effective and scalable speech task for future prediction models. Objective: The study aims to explore whether linguistic and acoustic features extracted from the HiSoC task are associated with UHR outcomes and if they are predictive of different UHR outcomes. Methods: Audio recordings of HiSoC task responses were collected from 41 participants with UHR enrolled in the Longitudinal Youth at Risk Study. A total of 12 individuals converted to psychosis, 15 remitted from UHR status, and 14 maintained UHR status. The responses from the converted group were obtained within 12 months of psychosis onset, while the responses from the remitted and maintained groups were collected at baseline. Linguistic features analyzed included words per minute, articulation rate, dysfluency, and sequential coherence. Acoustic features comprised the mean and SD of fundamental frequency, the mean and SD of intensity, and HF500. Feature differential analysis was conducted via multivariate linear regression. Linear support vector machines were trained as outcome prediction models. Nested cross-validation was used to estimate the generalizability error. The models were principally evaluated on balanced accuracy (BA). Results: The converted group exhibited lower words per minute (adjusted P=.02) and higher dysfluency (adjusted P=.004) compared to the remitted group. No significant differences were found in articulation rate, sequential coherence, or acoustic measures across the outcome groups. Two models outperformed random guess, namely the models using linguistic variables (BA 0.741, 95% CI 0.521-0.882) and linguistic and acoustic variables (BA 0.851, 95% CI 0.508-0.944). Conclusions: Linguistic features extracted from a short speech task exhibit a measurable difference between the outcome groups. Our findings support the feasibility of using signals extracted from the HiSoC task recordings to predict remission in participants with UHR.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.002 | 0.003 |
| Meta-epidemiology (narrow) | 0.000 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".