Clinical pathway data engineering framework (CPDEF): a reproducible methodology for constructing machine...

Healthcare analytics, clinical pathway analysis, predictive modeling, and artificial intelligence applications require large-scale and well-structured datasets derived from routine hospital operations. However, hospital information systems often store diagnoses, procedures, reimbursement classifications, and hospitalization outcomes in heterogeneous formats that are difficult to integrate into analytical workflows. This article presents the Clinical Pathway Data Engineering Framework (CPDEF), a reproducible healthcare data engineering methodology for constructing machine learning-ready case-mix datasets from hospital information systems. CPDEF integrates ICD-10 diagnoses, ICD-9-CM procedures, INA-CBG classifications, and length-of-stay information into a standardized analytical repository. The framework was applied to 165,391 inpatient hospitalization episodes collected between November 2016 and April 2026 and produced a repository containing 32 standardized variables suitable for healthcare analytics and artificial intelligence applications. Validation demonstrated deterministic workflow execution, structural consistency, coding-system integrity, and compatibility with standard machine learning workflows, supporting the reproducibility of the proposed methodology. CPDEF provides a transparent and scalable workflow for transforming heterogeneous hospital data into machine learning-ready analytical resources.