OBJECTIVES: No specific diagnosis codes exist for either multiple sclerosis (MS) subtypes or MS relapses. This study aimed to (1) identify all original validation studies used algorithms to identify MS in RWD, and (2) assess evidence for algorithm validity.

METHODS: A systematic review was completed according to to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Four key bibliographic databases were searched.

RESULTS: Of 5,056 identified articles, and 21 met inclusion criteria. Most included studies (43%) aimed to identify MS cases; others focused on MS subtypes (28%), relapses (15%), diagnosis and severity (9%), or disability status (5%). Only eight (38%) reported all four algorithm validity metrics: sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). Specificity was most frequently used (76 %), followed by sensitivity (71%), PPV (67 %), and NPV (52 %). Manual validation was the gold standard (76.1%) for sixteen studies, all of which found at least one algorithm with high sensitivity (77–99%), while specificity ranged from 88 to 99.9%, PPV from 38 to 95%, and NPV from 70 to 100%. All the studies aimed at identifying MS cases reported sensitivity (44%–100%) and specificity (87–100%), but only five reported PPV and NPV (38–99% and 17–100%, respectively). All three studies that identified MS relapses used administrative health databases with manual validation as the gold standard, finding sensitivity to range from 70 to 100%, specificity from 87 to 100%, PPV from 64 to 100%, and NPV from 53 to 100%. The best preformed algorithm for MS relapse identification included diagnoses, medications, hospitalization duration, or outpatient visits (sensitivity 70%, specificity 100%, PPV 100%, and NPV 96%).

CONCLUSIONS: Although there are no relapse-specific MS diagnosis codes, there are few validated algorithms to detect MS relapses, and none have been developed for other important RWD sources, such as electronic health records.