Identification of Patients in Need of Advanced Care for Depression Using Data Extracted From a Statewide Health...

Area of Science:

• Artificial Intelligence in Healthcare
• Clinical Decision Support Systems
• Mental Health Informatics

Background:

• Depression is a prevalent global mental illness with significant health and economic impacts.
• Distinguishing between mild and severe depression is crucial for appropriate care, yet challenging for primary care providers.
• Severe depression necessitates specialized care from mental health professionals.

Purpose of the Study:

• To develop predictive decision models for identifying patients requiring advanced depression care.
• To utilize comprehensive patient data, including diagnostics, behavior, demographics, and visit history.
• To apply these models within a large public health system (Eskenazi Health).

Main Methods:

• Random forest decision models were trained using merged structured and unstructured patient data.
• Data included diagnostic, behavioral, demographic, and past visit history.
• Models predicted advanced care needs for the overall patient population and specific high-risk subgroups.

Main Results:

• Models achieved high predictive accuracy, with Area Under the Curve (AUC) scores ranging from 78.87% to 94.43% for different patient groups.
• Sensitivity ranged from 68.79% to 83.91%, and specificity from 76.03% to 92.18%.
• The models demonstrated effectiveness in identifying patients needing advanced depression care.

Conclusions:

• Automated screening for advanced depression care needs is feasible using diverse patient datasets.
• These models can be integrated into clinical workflows to facilitate preventative care.
• Improved access to specialized mental health services can be achieved through these AI-driven tools.