Validation of a neurovascular comorbidities index for retrospective database analysis

Abstract

OBJECTIVEComorbidities have a significant effect on patient outcomes. Accounting for this effect is especially important in retrospective reviews of large databases; overpowered studies are at risk for finding significant results because of inaccurate patient risk stratification. The authors previously created a neurovascular comorbidities index (NCI) for patients with an unruptured intracranial aneurysm and found that the model’s ability to predict patient outcomes was statistically significantly improved over that of the routinely used Charlson Comorbidity Index (CCI) and Elixhauser Comorbidity Index (ECI). In this study, the authors aimed to validate use of the NCI over that of the CCI and ECI for risk stratification of patients with other neurovascular diseases.METHODSThe authors queried the National (Nationwide) Inpatient Sample database for the years 2002–2012 to compare the accuracy of the previously validated NCI with that of the CCI and ECI with respect to predicting outcomes for patients who had an arteriovenous malformation, a ruptured intracranial aneurysm, carotid artery stenosis, or dural arteriovenous fistula and who underwent surgical intervention.RESULTSFor patients with an arteriovenous malformation, the NCI outperformed the CCI and ECI in predicting poor outcome, hospital length of stay (LOS), and total cost but was equivalent to the CCI in predicting death. For patients with a ruptured intracranial aneurysm, the NCI outperformed the ECI and CCI in predicting death, poor outcome, LOS, and total cost. For patients with carotid artery stenosis, the NCI outperformed the ECI and CCI in predicting LOS, but it was equivalent to the ECI in predicting death and total cost and inferior to the CCI in predicting poor outcome (p < 0.002 for all). An insufficient number of patients with dural arteriovenous fistula who underwent surgical intervention were available for analysis (n < 10), and they therefore were excluded from study. For 11 of 12 metrics, the NCI was the significantly more efficient model.CONCLUSIONSThe NCI outperforms the CCI and ECI by providing more appropriate and efficient risk stratification of patients regarding death, outcome, LOS, and cost. Given this finding, the NCI should be used for retrospective reviews of patient outcomes instead of the CCI or ECI.