Enhancing mental disorder screening via exercise electrocardiogram data using the portrait method

Mental disorders are globally prevalent yet remain significantly underdiagnosed, presenting a major public health challenge. Electrocardiogram (ECG) data, being widely available, has been implicated in mental disorders detection and shows potential as a screening tool. This study aims to explore the feasibility of using exercise ECG data to automatically screen for mental disorders.

This cross-sectional study analyzed exercise ECG data from the UK Biobank, including individuals with depressive episode (n = 2135), reaction to severe stress and adjustment disorders (RSSAD) (n = 858), and phobic anxiety disorder (n = 143), alongside matched control groups. Diagnoses were derived from linked hospital and primary care records. For each diagnostic category, the dataset was randomly split into training (70 %) and test (30 %) sets. A Portrait method was developed, wherein ECG variables were analyzed for each disorder, and statistically significant variables were used to construct individual portraits based on their dissimilarities from others in the dataset. Disease-specific group-level portraits were then generated by aggregating individual portraits. New portraits were subsequently mapped onto these group portraits to support screening assessment.

The Portrait method demonstrated acceptable performance across all test sets of three mental disorders with an accuracy of 81.98 % for depressive episode (sensitivity 98.68 %, specificity 63.02 %), 87.14 % for RSSAD (sensitivity 70.00 %, specificity 100.00 %), and 82.81 % for phobic anxiety disorder (sensitivity 86.70 %, specificity 79.65 %).

These results underscore the potential of exercise ECG data for identifying three mental disorders and demonstrate the efficacy of the Portrait method for early screening.