A multimodal approach to depression diagnosis: insights from machine learning algorithm development in primary care.

Area of Science:

• Medical Informatics
• Psychiatry
• Machine Learning

Background:

• General practitioners (GPs) are crucial for early depression identification.
• Existing screening tools like Patient Health Questionnaire-9 may yield false positives.
• There is a need for more accurate diagnostic support in primary care settings.

Purpose of the Study:

• To develop and validate a machine learning model, Clinical 15, for improved depression diagnosis in primary care.
• To assess the model's accuracy and utility in identifying depression and its subtypes.

Main Methods:

• A two-step machine learning model (Clinical 15) was developed and trained on 581 participants.
• Data integrated self-reported questionnaires from patients presenting to GPs.
• Nested cross-validation framework and Gaussian mixture model clustering were employed.

Main Results:

• Clinical 15 achieved a balanced accuracy of 88.2%.
• The model utilizes a traffic light system (green, yellow, red) for case categorization.
• Four depression subtypes were identified, including an Immuno-Metabolic cluster.

Conclusions:

• Clinical 15 offers a sensitive and specific tool to assist GPs in depression diagnosis.
• The model identified patients in the Immuno-Metabolic cluster, aiding in subtype characterization.
• Further validation via randomized controlled trials and assessment of clinical integration are planned.