Background: In most of Taiwan’s medical institutions, congestion is a serious problem for emergency departments. Due to a lack of hospital beds, patients spend more time in emergency retention zones, which make it difficult to detect cardiac arrest. Objective: We seek to develop a Drug Early Warning Scoring Model (DEWSM), including drug injections and vital signs as these research important features. We use it to predict cardiac arrest in emergency departments via drug classification and medical experts’ suggestion. Methods: We propose this new model for detecting cardiac arrest via drug classification and by using a sliding window, and apply learning-based algorithms to time-series data for a DEWSM. To evaluate the proposed model, we use the area under the receiver operating characteristic curve (AUROC). Results: We identify the two important drug predictors: bits (intravenous therapy), and replenishers and regulators of water and electrolytes (fluid and electrolyte supplement). We verify feature selection, in which accounting for drugs to improve the accuracy and demonstrate that thus accounting for the drugs significantly affects prediction. Also, we show that CPR events can be predicted four hours before the event. Conclusion: Our study used a sliding window to account for dynamic time-series data consisting of the patient’s vital signs and drug injections. The experimental results of adding the drug injections were better than only vital signs. In addition, we using LSTM method as the main processing time series data, it was the bases for comparison of this research.