Machine Learning-Enhanced Data Quality Validation for Electronic Health Record Integration

The harmonization and interoperability of electronic health records (EHRs) are critical for enabling comprehensive, patient-centred care. Data quality issues, including incompleteness, inconsistency, duplication, and semantic inaccuracies, significantly compromise the clinical utility of EHR systems. Nevertheless, the clinical utility of EHR systems. Conventional rule-based validation methods have limited flexibility and scalability, especially with the increasing prevalence of complex and massive healthcare data sources. To consolidate the EHRs from independent health information systems, this paper presents a machine learning (ML) empowered data quality validation framework for automating quality control, enriching data authenticity, and mitigating integration flaws. The study presents a layered architecture that utilizes machine learning (ML) algorithms to detect anomalies in the data, infer missing values, and provide automated quality scores. The model leverages supervised and unsupervised algorithms trained on historical EHR integration failures and common data integrity violations. Algorithms such as random forests, support vector machines, and autoencoders are applied to identify structural and semantic errors in semistructured data (e.g., patient demographics, medication lists, and laboratory results). Additionally, NLP models are utilized to review unstructured notes and discharge summaries, ensuring that the context remains consistent.