Early detection of Sick Building Syndrome (SBS) in healthcare environments is vital to safeguard occupant health, especially in regions like Iraq where structured indoor air quality (IAQ) monitoring systems are largely unavailable. This study introduces a novel deep learning framework for real-time SBS prediction, developed using the first publicly available IAQ dataset collected from Baqubah Teaching Hospital, Diyala, Iraq. The dataset spans eight months and includes ten environmental parameters such as CO₂, TVOC, PM₂.₅, PM₁₀, CO, O₃, temperature, humidity, air quality index, and light intensity. The proposed framework employs a hybrid 1D-CNN-BiLSTM model designed to capture both local spatial correlations and bidirectional temporal dependencies in IAQ signals. To enhance its suitability for real-world deployment, particularly on resource-limited devices, the model undergoes a series of multi-level optimization techniques that significantly improve efficiency while maintaining high predictive accuracy. Experimental evaluations against multiple deep learning baselines demonstrate that the optimized model achieves a strong balance between accuracy and computational performance. These results highlight the framework’s practicality for continuous, real-time monitoring of SBS risk factors. Beyond its technical contribution, this research provides the first IAQ dataset from Iraq and establishes a foundation for intelligent environmental health management in healthcare buildings across developing regions.