GLP-1 RA Eligibility Study in a Multiethnic Asian Cohort - SingHealth DUE Parametric Synthetic Dataset

This dataset is a synthetic clinical dataset (n = 481) generated to support machine learning model development for GLP-1 receptor agonist (GLP-1 RA) eligibility classification in a multiethnic Asian population. It was generated by parametric sampling with marginal distributions anchored to publicly reported aggregate prescribing statistics from Leow et al. (2024; doi:10.22541/au.173535301.14921524), derived from a real-world GLP-1 RA prescribing population across SingHealth institutions, Singapore. Eligibility labels were assigned deterministically using a prespecified rule-based algorithm derived from the Health Promotion Board–Ministry of Health (HPB/MOH) Singapore Clinical Practice Guidelines on Obesity and SingHealth Drug Usage Evaluation criteria. The dataset contains 11 variables: three continuous clinical features (BMI_val, HbA1c_val, SBP_val), six binary clinical features (T2DM, HTN, Dyslipidaemia, CVD, Prediabetes, On_OAD), and two outcome columns (label, label_enc). The encoded label column uses 1 for GLP-1 RA Eligible and 0 for GLP-1 RA Ineligible / Generic Alternative. The label reflects the study-specific deterministic rule hierarchy used in the associated manuscript. In this hierarchy, patients with BMI 25.0–26.9 kg/m² and cardiometabolic comorbidity were assigned to the Generic Alternative pathway because this BMI range falls below the Asian-adapted obesity pharmacotherapy threshold used for GLP-1 RA eligibility under the comorbidity-weighted indication. Eligibility labels were generated from clinical features using deterministic rules and were not randomly sampled from class prevalence. The dataset was generated using random seed 42 and contains no real patient-level records. It was generated from aggregate-level prescribing statistics and the rule-based procedure described in the Supplementary Methods of the associated manuscript. Associated manuscript: Madathil AK, Khoo CM, Chan ECY. Equitable GLP-1 RA Eligibility Gatekeeping Using an Interpretable Machine Learning Framework in a Multiethnic Asian Cohort. [Under submission], 2026. Code repository: https://github.com/arjunkmadathil/GLP-1_GatekeepingStudy Source statistics: Leow JL et al. Real-world utilisation, efficacy, and safety of injectable GLP-1 receptor agonists: a drug use evaluation in Singapore's largest healthcare cluster. Authorea Preprints. 2024. doi:10.22541/au.173535301.14921524.