Genetics

This post covers Damotte, Lizee and colleagues' study in Multiple Sclerosis Journal, which identified 4,142 MS patients in the UCSF electronic record system, extracted clinical measures from free-text notes, and compared them against a prospective research cohort for patients present in both. It covers how the patient-finding algorithm trades completeness for precision, the per-variable accuracy of extraction with year of onset as the weakest, the high agreement for EDSS and walking speed, the finding that 28.6% of research-defined secondary progressive patients remain recorded as relapsing-remitting in the clinic, the reproduction of known MS natural history, and the information bias toward moderately disabled patients that the authors measure themselves.