From data to knowledge: artificial intelligence methods for studying comorbidity in electronic health records

European Physical Journal: Special Topics. 2026. P. 1–24.

Daria Lukianenko, Ragimova A., Mukhorina A., Tyukina T., Zakharov D.

Electronic health records (EHRs) contain vast volumes of clinical information that encode complex relationships between diseases. Traditional approaches to the analysis of interrelated or co-occurring diseases have focused on pairwise associations between diagnoses, missing the higher-order structures that characterise multimorbid patients. The present paper offers a narrative review of existing statistical, machine-learning, and artificial intelligence methods for extracting and analysing complex interrelated or co-occurring diseases from EHRs. The application of diverse approaches based on the analysis of diagnostic codes, of textual data, and of combinations of sources of different modalities within hybrid methods makes it possible to uncover emergent interactions between diseases. The work shows the critical role of natural-language processing of narrative text in extracting clinically relevant information absent from structured codes, which in turn motivates the integrated application of deep-learning methods with graph-based analysis for the identification of disease clusters and patient stratification. The analysis shows that going beyond pairwise comorbidity models enables the discovery of collective disease-progression trajectories, offering new opportunities for personalised medicine.

Keywords: electronic health records, natural-language processing, deep-learning methods

Publication based on the results of: Multidisciplinary study of behavior and decision-making in health population and patients using behavioral, economic, neurocognitive, neuroeconomic, neurocomputational and neural network approaches (2025)