Exploring clinical associations using '-omics' based enrichment analyses.

Area of Science:

• Computational biology
• Medical informatics
• Clinical data analysis

Background:

• Electronic health records (EHR) generate vast clinical data, similar to the "-omics" data explosion.
• Physician-curated problem summary lists in EHRs offer a rich source of clinical information.
• Hypothesis: Collective physician knowledge in EHRs can reveal significant, literature-absent diagnosis associations.

Purpose of the Study:

• To apply a bioinformatics approach to identify novel associations between clinical diagnoses using EHR data.
• To leverage the "Molecular Concept Map" (MCM) tool for analyzing large-scale clinical problem lists.
• To uncover previously unknown or poorly documented relationships between medical conditions.

Main Methods:

• Utilized the "Molecular Concept Map" (MCM) method, originally for gene expression data, on EHR problem lists.
• Analyzed 1.5 million clinical problem summary entries from 327,000 patients.
• Calculated odds ratios (OR) and p-values for each of the 750,000 identified associations.

Main Results:

• Confirmed known associations (e.g., diabetes with retinopathy, hypertension).
• Identified recently reported associations (e.g., irritable bowel syndrome and vulvodynia).
• Discovered novel associations, such as granuloma annulare with osteoarthritis (OR 4.3) and pyloric stenosis with ventricular septal defect (OR 12.1).

Conclusions:

• Bioinformatics tools developed for "-omics" data are applicable to clinical medicine.
• EHR data analysis can generate hypotheses for further research.
• Findings support clinical care by highlighting potential co-occurring diagnoses for physicians.