• Haifeng Zhu and Michael D Hill.
• Department of Clinical Neurosciences, University of Calgary, Foothills Hospital, Rm 1242A, 1403 29th Street NW, Calgary, Alberta, Canada T2N 2T9.
• Neurology. 2008 Jul 22; 71 (4): 283-7.

BackgroundAdjustment for comorbidity is an important component of any clinical outcome study using administrative data. The Elixhauser Index is a relatively newer comorbidity index for use with administrative data and has not been used to assess prognosis in patients with stroke. Similarly, an International Classification of Diseases (ICD)-10 coding algorithm has been rarely reported for Elixhauser Index.ObjectiveTo evaluate whether the Elixhauser Index provides a useful comorbidity adjustment for predicting in-hospital case-fatality in stroke outcome studies and to compare the degree of consistency using ICD-9-CM and ICD-10 coding algorithms.MethodsPatients who had stroke from 1998 to 2000 (cohort A in the ICD-9-CM data) and 2003 to 2005 (cohort B in the ICD-10 data) in a large Canadian city were identified from the Hospital Discharge database. The performance of two coding algorithms for predicting the in-hospital case-fatality was assessed using multivariable logistic regression models. The C-statistic was used to compare the performance of each coding algorithm in predicting in-hospital case-fatality.ResultsAmong 2,465 patients with stroke in the ICD-9-CM data (cohort A) and 2,987 patients with stroke in the ICD-10 data (cohort B), there was no difference in model performance using ICD-9-CM (C-statistic was 0.717) as compared with ICD-10 coding algorithms (C-statistic was 0.721; p = 0.83). Elixhauser comorbidity adjustment provided a better prediction of in-hospital case-fatality compared to reduced models including only age and gender (p < 0.0001) for both coding models.ConclusionThe Elixhauser Index provides similar comorbidity adjusted risk estimates using both ICD-9-CM and ICD-10, and may be useful for predicting risk-adjusted in-hospital case-fatality in stroke outcome studies.

35 keywords...

hide…