OBJECTIVES:  To develop and validate operational EHR- and claims-based algorithms for RRMS patient identification in a US Integrated Delivery Network (IDN) healthcare system. METHODS:  IDN data (2010-2014) were queried for the study inclusion criteria: MS diagnosis, age ≥ 18 years, ≥ 1 year baseline history, and no other demyelinating diseases. The EHR-based algorithm used natural language processing (NLP). The claims-based algorithms were developed using (1) combinations of: MS diagnosis, specific symptoms during a neurology visit, disease modifying therapies (DMT), brain/spinal MRI; and (2) rule out progressive MS (P-MS) through: (option A) medications for P-MS; (option B) MS severity/progression from adapted Kurtzke Functional Systems Scores; and (option C) P-MS defined by Gilden, 2011. Random samples of NLP-based medical chart reviews were the “gold standard” for algorithm validation and positive predictive value (PPV) calculations. RESULTS:  Of 3,111 MS patients identified, 2,960 (95%) were by claims-based, 990 (32%) by EHR-based, and 839 (27%) by both algorithms. RRMS was established in 2,213 (71%) patients overall. Of 2,960 claims-based, the three algorithm options identified 2,271 (77%) RRMS patients. Of 990 EHR-based patients, RRMS was identified in 837 (85%). An average 19.3 documents per patient were included for NLP-based chart review. The combined claims- and EHR-based algorithms had a PPV (95% CI) of 93% (82%-98%). The claims-based algorithms to identify RRMS had PPV (95% CI) of 88% (78%-94%), 89% (76%-95%), and 89% (79%-95%) for options A, B and C, respectively. CONCLUSIONS: Both the combined claims- and EHR-based and the claims-based algorithms had excellent PPV for identifying RRMS among patients with documented MS subtypes. Traditional medical chart reviews will support the NLP-based chart reviews, particularly for patients without clinical notes of MS subtypes. The claims- and EHR-based algorithms to identify RRMS and NLP-based chart reviews are promising methods for future research.