Retrospective imputation of Glasgow Outcome Scale scores at one month after aneurysmal subarachnoid hemorrhage

Abstract

BACKGROUND: Aneurysmal subarachnoid hemorrhage (aSAH) is a relatively rare, but devastating form of stroke. Fortunately, mortality rates have declined, but there is still a dearth of research on long-term survivorship, in part due to large amounts of missing data in longitudinal datasets. Clinical outcomes scale scores are often missing in the early post-hospital discharge period, due to regional and provider variations in the utilization of these tools in the outpatient setting. Yet, much of the data that is needed to calculate such scores may be available in the medical record. METHODS: Three experienced nurses used a decision tree protocol to impute 1-month Glasgow Outcome Scale (GOS) scores from medical record information in an existing database. Interrater reliability was evaluated, and the imputed scores were tested for associations with other demographic and clinical variables known to impact outcomes. RESULTS: Scores were imputed on 585 patients (72.8% female, mean age of 53 years, 53% African American). The majority of subjects had an imputed GOS score of 4 (n=301; 51.5%). Interrater agreement across the three nurses was estimated to be 0.72 (Fleiss’s Kappa, p <0.0001). Age, stroke severity measures, employment, and prior history of hypertension or diabetes were associated with 1-month GOS scores. CONCLUSIONS: Using this decision tree protocol to derive follow-up outcome scores could be useful for unionization of large data sets, allowing greater trending of recovery over time. This type of data is needed to fill knowledge gaps and allow the development of effective interventions for recovery after aSAH.