Data classification is crucial in improving medical decision-making efficiency within Medical Cyber-Physical Systems (MCPS). In this study, a hybrid data classification method is proposed for classification of heart disease data. The proposed method uses two-level data classification that integrate ensemble learning and fuzzy logic. At first level, XGBoost algorithm is used to categorize the data into two classes viz. critical and non-critical. At the second level, key clinical characteristics are extracted for each class and the severity index data are calculated. The further classified using fuzzy logic classifiers depending on the severity indices. The proposed model is tested for three different datasets to compare performance and has an accuracy of 99.50%, which is relatively higher than the algorithms compared. This hybrid methodology can improve medical decision making efficiency by providing efficient classification in real-time.