Adding Stroke Severity Measure Improves Mortality Risk Models
(HealthDay News) – Adding stoke severity to a hospital 30-day mortality risk model improves model discrimination and hospital performance rankings.
Gregg C Fonarow, MD, from the University of California in Los Angeles, and colleagues analyzed data from 782 hospitals participating in Get With The Guidelines-Stroke, representing 127,950 fee-for-service Medicare beneficiaries with ischemic stroke. Participants had a documented score for the National Institutes of Health Stroke Scale (NIHSS). Performance was compared for claim-based hospital mortality risk models, with and without inclusion of NIHSS scores.
The researchers observed significantly better discrimination in the hospital mortality model with NIHSS scores compared to the model without (C statistic, 0.864 vs. 0.772). The model with NIHSS scores ranked 26.3% of the top 20% or the bottom 20% of performers differently than the claims model without NIHSS scores. Reclassification by the model with NIHSS scores changed 57.7% of hospitals initially classified as "worse than expected" to "as expected." After the addition of NIHSS scores, both net reclassification improvement and integrated discrimination improvement indexes demonstrated significant enhancement (93.1 and 15.0%, respectively). The addition of NIHSS scores also improved explained variance and model calibration.
"These findings suggest that it may be critical to collect and include stroke severity for optimal hospital risk adjustment of 30-day mortality for Medicare beneficiaries with acute ischemic stroke," the authors conclude.
The study was funded by the Get With The Guidelines-Stroke program, with support provided by the pharmaceutical industry; several authors disclosed financial ties to the pharmaceutical and biotechnology industries.