A Clinical Data Based Framework for Predicting Mortality and Length-of-Stay in Pneumonia Patients

Respiratory diseases are a major cause of death globally, leading to a significant burden on healthcare services. Mortality and length-of-stay prediction are key tasks that help clinicians intervene proactively and develop appropriate treat-ment plans. In our work, we employ a Long Short-Term Memory (LSTM) based model for both mortality and length-of-stay prediction in patients with pneumonia. This model utilizes clinical time-series data of varying lengths, along with static variables from admission information, to effectively capture the sequential information of clinical variables. Additionally, we implement a data splitting strategy to track dynamic changes in model performance at different timestamps, helping to bridge the gap between model development and real-world scenarios. Furthermore, we account for the right-skewed distribution of length-of-stay data during model establishment to enhance its robustness. Extensive experiments on the CAP-AI dataset demonstrate the effectiveness and robustness of our approach for both prediction tasks.