Maintaining optimal performance and patient care depends heavily on the psychological health of healthcare workers (HCWs). Current AI-based sentiment analysis methods for estimating HCW stress levels which frequently yield skewed results are impacted by anomalies and are imprecise. This chapter offers a novel hybrid deep learning (DL) model called SAE-CABG, which combines attention mechanism-based bidirectional gated recurrent units (Bi-GRU) and convolutional neural networks (CNN) with stacked auto-encoders (SAE). After feature extraction with SAE, sentiment analysis with CNN and Bi-GRU with attention mechanisms is used to weight and categorize different types of stress experienced by healthcare workers. In order to improve generalization and avoid overfitting, dropout layers are included. The proposed SAE-CABG method outperforms existing prediction techniques by improving precision by 2%, recall by 2%, and F-measure by 3% when evaluated on the Human Stress Prediction dataset. It also regularly reduces training time by up to 47.4%, lowers the misclassification rate by up to 91.7%, and achieves an increase in prediction accuracy of up to 2.32%.