Predicting child and adolescent mental health emergency department revisits: a machine-learning approach compared to a clinician-derived baseline
Bibliographic record
Abstract
BACKGROUND: Predicting child and youth mental health (CYMH) emergency department (ED) revisits (RVs) is critical for improving patient outcomes and optimizing use of resources. Previous CYMH ED RV studies have used statistical methods with research cohorts and produced varying results. Our aims were to develop a predictive algorithm incorporating machine learning (ML) with electronic health records (EHR) and validate it against a clinician-driven algorithm in a proof of concept project. METHODS: Data were retrospectively collected from a tertiary care pediatric hospital’s EHR from November 2017–November 2023, yielding 12,700 ED encounters from 8,696 patients, 8–18 years of age. The feature set comprised patient demographics, visit-level variables, laboratory results, procedure codes, and medication records. A mapping of 230 International Classification of Diseases (ICD)-10 codes into 28 Diagnostic and Statistical Manual (DSM)-5 categories was performed and a logistic regression (LR) ML model developed. Both tasks used clinical expert input. Seven clinical experts then independently assigned weights to 191 variables using a custom-designed application to create a structured clinician-weighted baseline for comparison to the ML algorithm. Both models were evaluated using AUROC and F1 score as primary metrics with precision and recall as secondary. LR coefficients and odds ratios were the primary interpretability outputs, while SHapley Additive exPlanations (SHAP) were used for supplementary visualization across four age strata. RESULTS: The LR machine learning model achieved an AUROC of 0.78, outperforming the structured clinician-weighted baseline (AUROC range: 0.54–0.64) Detailed analysis revealed that predictors such as past ED RV count, psychotherapeutic medication history, substance use history, and prior outpatient MH visits were consistently influential. CONCLUSIONS: This proof of concept project demonstrates that ML can provide complementary, clinically interpretable predictions of CYMH ED RV. Alignment between model-derived predictors and clinician-weighted features supports interpretability and lays a foundation for further development. Future steps include enhancing sensitivity, expanding feature sets, and conducting prospective silent-mode validation to refine performance. CLINICAL TRIAL REGISTRATION: Not applicable.
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 | 0.015 | 0.038 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| 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".