Development of a screening algorithm for borderline personality disorder using electronic health records

Area of Science:

• Digital Medicine
• Psychiatric Diagnostics
• Machine Learning in Healthcare

Background:

• Borderline Personality Disorder (BPD) is prevalent but underdiagnosed, leading to delayed treatment.
• Automated screening of Electronic Health Records (EHRs) can identify undiagnosed BPD patients.
• Current diagnostic challenges necessitate innovative identification strategies.

Purpose of the Study:

• To develop and validate an automated algorithm for screening Borderline Personality Disorder (BPD) using structured EHR data.
• To improve the identification of patients with BPD who might otherwise be missed.
• To provide a tool to support clinical decision-making in BPD diagnosis.

Main Methods:

• Developed a two-step screening algorithm integrating rule-based selection and machine learning (ML).
• Utilized a large-scale, de-identified US-based EHR database.
• Validated the algorithm against clinical expert ratings on random EHR samples.

Main Results:

• The BPD screening algorithm demonstrated high consistency with clinical expert ratings.
• Achieved an Area Under the Receiver Operating Characteristic (AUROC) of 0.837.
• Reported high accuracy (0.820), positive predictive value (0.717), and specificity (0.922).

Conclusions:

• The developed EHR-based algorithm is a promising tool for identifying potential Borderline Personality Disorder (BPD) cases.
• This digital health resource can facilitate clinical decision-making and improve BPD diagnosis rates.
• Further development and integration of such tools can advance digital medicine in psychiatry.