This article is more than 5 years old. Information may no longer be current.
Administrative data algorithms accurate in detection of RA
Among the general population, administrative data algorithms had a high degree of accuracy for detecting patients with rheumatoid arthritis, according to study results.
Researchers used a random sample of 7,500 adult patients from the Electronic Medical Record Administrative data Linked Database in Ontario, Canada, to perform a retrospective chart abstraction. Sensitivity, specificity and predictive value for more than 100 administrative data algorithms for rheumatoid arthritis (RA) case ascertainment were computed and compared using physician-reported diagnoses as the reference standard.
The researchers identified 69 patients with RA, for a lifetime RA prevalence of 0.9%. Although all algorithms had specificity of 97% or greater, sensitivity ranged from 75% to 90% among physician billing algorithms, according to the researchers.
Most algorithms had adequate positive predictive value, despite the low prevalence of RA. Specifically, the researchers found a sensitivity and positive predictive value of 78% and specificity and negative predictive value of 100% for the algorithm of “[one hospitalization RA diagnosis code] or [three physician RA diagnosis codes with one or more by a specialist over 2 years].”
Disclosure: Financial support was provided by the Canadian Institutes of Health Research (operating grant 119348) and ICES.