Background: Acute Decompensated Heart Failure is a major burden worldwide; with in-hospital mortality ranging between 3%-7%1, there has been interest for risk stratification models for prognostication purposes. Using data from the Acute Decompensated Heart Failure National Registry (ADHERE), Greg Fonorow and associates were able to identify the best predictors of in-hospital mortality using classification and regression (CART) analysis and developed a risk stratification model2. These parameters are serum creatinine, blood urea nitrogen, and systolic blood pressure. However, because the registry from which it was based was primarily composed of Caucasians and African-Americans; it is unclear whether it can be applied to Filipinos. The objective of the study was to determine whether the CART analysis model can be applied in the local setting.
Methods: The author conducted a prospective cohort involving 334 patients admitted at the Philippine Heart Center for acute decompensated heart failure. In accordance with the CART analysis model, patients were categorized into 5 risk groups based on serum blood urea nitrogen, serum creatinine, and systolic blood pressure determined on admission. These 5 groups, in order of increasing mortality risk are: low, intermediate 3, intermediate 2, intermediate 1, and high risk groups. The patients were followed-up from the time of admission at the emergency room until hospital discharge, hospital transfer or in-hospital death.
Results: 40 patients expired out of the 334 patients enrolled in the study. All of the independent variables in this study showed significant difference (p-value <0.005) in terms of mortality; with higher serum BUN and creatinine and lower systolic blood pressure portending higher mortality rates. In terms of risk groups, there was increasing mortality rate from low to high risk groups. The odds ratio for mortality risk was 3.84 for higher risk groups with a p-value of <0.005. Hosmer-Lemeshow was tested for validation of the risk stratification; and was found to be at 0.70. The area under the curve (ROC) range at 95% confidence interval was between 0.76252 and 0.89105, with mean of 0.8268; indicating good accuracy.
Conclusion: The Classification and Regression Tree (CART) analysis model can be used in the local setting for risk stratification of patients admitted for acute decompensated heart failure.