Structured EHR Data for Pediatric Anxiety Risk Prediction

Eric W. Lee published a dataset on figshare in May 2026, containing structured electronic health record data used to evaluate machine learning models for predicting pediatric anxiety disorders. The data was used to train models like XGBoost, LSTM, and RETAIN, achieving AUROC scores up to 0.857. The dataset is 9.5 KB in size and is available under a CC-BY-4.0 license.

Use Cases

• Benchmarking machine learning models for clinical risk prediction based on structured EHR features.
• Training deep learning models that leverage longitudinal patient histories for anxiety onset classification.
• Comparing the predictive value of area-based measures of health versus standard EHR data for mental health outcomes.
• Developing age-stratified modeling approaches for pediatric developmental stages.

Strengths

• Dataset is associated with a published study reporting specific model performance metrics, such as an AUROC of 0.857 for a Dipole model.
• Data is structured for direct use in training machine learning models, as described in the associated research methodology.
• License is clearly stated as CC-BY-4.0, facilitating open use.

Limitations

• Row count is unknown, which may limit suitability assessment for large-scale modeling.
• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is very small at 9.5 KB, indicating a limited scope of data.

Provenance

Source

figshare

Collection Method

Retrospective, case-control study using structured EHR data from 30-day windows prior to diagnosis.

Freshness

Last updated 2026-05-12 17:43:10; freshness should be verified.