Abstract TMP61: Preliminary Results of a Population Based Outcomes Pilot Study the Greater Cincinnati/Northern Kentucky Stroke Study

    loading  Checking for direct PDF access through Ovid


Background: We previously followed a cohort of survivors longitudinally with direct interview for outcomes. We then proved that we could obtain similar outcomes by tracking people in a health information exchange followed by first phone call contact at 3 months, thus minimizing contact. This study sought to determine the feasibility of estimating population-based post-stroke outcomes using information available in the electronic medical record (EMR) without any patient contact.Methods: Our study is a retrospective population-based epidemiology project that ascertains hospitalized strokes via ICD-discharge codes (ICD-9 430-436, ICD-10 I60-I67, G45-G46). Our study region encompasses 5 greater Cincinnati counties; study period 1/1/15-12/31/15. For this pilot outcomes study, we identified all ischemic strokes that presented to a system of 4 hospitals and performed phone calls at 3 and 6 months to determine current place of residence and functional outcomes including the modified Rankin Score (mRS) and the Euroqol (EQ-5D). Simultaneously, our lead Study Coordinator reviewed all available EMR records and used this information to estimate outcome status blinded to phone call results. We compared the “gold-standard” of interview-determined place of residence, mRS and EQ-5D to those estimated from EMR using the either the Kappa statistic or interclass correlation (ICC), as appropriate.Results: See Table for details. Using standard definitions of Kappa/ICC, estimation of 3- and 6-mo place of residence was “almost perfect” and estimation of mRS grade and EQ-5D was “substantial”. Censoring observations where no EMR information was available did not reduce Kappa/ICC.Conclusion: This work shows promise in using EMR information to accurately estimate post-stroke outcomes without patient contact, to allow a population-based estimation of outcome. Future work will involve machine-learning to improve our accuracy in outcomes estimation from EMR information.

    loading  Loading Related Articles