This study presents a comparative evaluation of predictive models for estimating the monthly demand for public hospital cardiovascular emergency services in the Dominican Republic. Using historical data from 2022 to 2024, five models were implemented: SARIMA, Seasonal Linear Regression, Decision Tree, Multilayer Perceptron (MLP), and Long Short-Term Memory (LSTM). One representative hospital was selected from each health region, prioritizing those with the highest cardiovascular emergency volume. The methodology included preprocessing steps such as seasonal differencing, normalization, and sliding window transformations, along with stationarity assessments using ADF and KPSS tests. Model performance was evaluated using RMSE and MAE metrics, complemented by statistical measures such as mean, median, and standard deviation. Results showed that machine learning models, particularly MLP and LSTM, performed better in highly variable environments, while classical models yielded better results in series with well-defined seasonality. The findings offer a robust and replicable framework for hospital planning and resource allocation, supporting data-driven decision-making in public health system.