Transthoracic echocardiography does not improve prediction of outcome over APACHE II in medical-surgical intensive care
Additional Document Info
PURPOSE: To examine the hypothesis that transthoracic echocardiographic findings predict mortality in critically ill patients. METHODS: A retrospective analysis of concurrently collected data for consecutive patients from May 1996 to May 1998 who had transthoracic echocardiography on or within six months of admission to the medical surgical intensive care (MSICU). We examined the role of physiologic, clinical, and echocardiography variables in predicting the mortality of patients admitted to the MSICU. Three logistic regression models were developed: 1) clinical; 2) echocardiographic; and 3) combined clinical with echocardiographic. Univariate and multivariate analyses were performed and the relative strength of clinical and echocardiographic predictors was compared using odds ratio (OR) and receiver-operator-characteristic (ROC). RESULTS: Of 4,070 MSICU patient admissions, 1,093 patients had transthoracic echocardiography; the study group comprised 942 patients with complete clinical and echocardiographic data. The MSICU mortality was 28%. For the combined model, analyses identified left ventricular systolic function (LVSF), [OR 1.26; confidence interval (CI) 1.01-1.57], severe tricuspid regurgitation (TR) (OR 3.72; CI 1.04-13.24), medical diagnosis (OR 1.91; CI 1.15-3.19), and acute physiology and chronic health evaluation (APACHE) II score (OR 1.27; CI 1.23-1.31), as predictors of MSICU mortality. The combined model yielded an area under ROC curve of 0.913. For the clinical model, analyses identified age (OR 1.04; CI 1.02-1.05) and APACHE II (OR 1.32; 1.26-1.35) as predictors of mortality with an area under ROC curve of 0.917. For the echocardiography model, TR (OR 2.40; 1.08-5.38), severe aortic insufficiency (AI) (OR 4.13; CI 1.17-16.29) and pulmonary hypertension (OR 2.05; 1.01-4.09) were identified as predictors of outcome with an ROC curve of 0.536 for this model. CONCLUSION: Statistical models utilizing clinical variables are predictive of mortality in MSICU. Models that include diagnostic transthoracic echocardiography variables do not provide incremental value to predict ICU mortality. These findings may have implications for non-invasive hemodynamic assessment of critically ill patients, and raise the hypothesis that echocardiography-guided interventions may not alter outcome in ICU.