Identifying individuals with multiple sclerosis in an electronic medical record

Area of Science:

• Health Informatics
• Neurology
• Epidemiology

Background:

• Electronic medical records (EMRs) offer a valuable resource for assessing and enhancing healthcare quality.
• The increasing adoption of EMRs presents opportunities for studying individuals with Multiple Sclerosis (MS).

Purpose of the Study:

• To develop and validate an algorithm for identifying patients with MS within EMR data.
• To assess the accuracy of EMR data for MS patient identification.

Main Methods:

• Utilized a large dataset of 73,003 adult patients from 83 primary care physicians in Ontario.
• A reference standard of 247 MS patients was established via chart abstraction.
• Assessed algorithm accuracy using cumulative patient profile (CPP) data, prescriptions, and billing codes.

Main Results:

• An algorithm using CPP data achieved high accuracy: 91.5% sensitivity, 100% specificity, 98.7% positive predictive value (PPV), and 100% negative predictive value (NPV).
• Incorporating MS-specific medication prescriptions and specific billing codes marginally improved sensitivity to 92.3% with a PPV of 97.9%.

Conclusions:

• EMR data can reliably identify patients diagnosed with MS.
• This validated algorithm facilitates the identification of MS patient cohorts in primary care, supporting quality of care research and clinical decision-making.