Welcome to our awesome tool that's like a superpower for hospitals, helping predict if a patient might need to return within 30 days of leaving. It's all about keeping patients healthier and out of the hospital!
We built this to spot patients who might be at risk of coming back to the hospital soon after discharge. By catching these risks early, doctors can provide extra care to keep patients on the road to recovery.
We used a super-smart machine-learning tool called XGBoost, which digs through patient data to find hidden patterns. To make it the best it could be, we used Optuna to test 150 different setups, tweaking everything until we found the perfect formula—like tuning a guitar for the perfect sound!
Our model learned from a big dataset filled with details like where patients went after discharge, how long they stayed, their gender, lab test results, and how serious their condition was. We used 80% of the data to train the model and 20% to test it, making sure it handled the rare readmission cases with extra care.
We played with settings like how deeply the model thinks, how fast it learns, and how much data it looks at each time. After 150 experiments, we locked in the best setup for top-notch predictions.
Our final model is a champ! Here's how it performed:
It nails predictions 93.2% of the time—super reliable!
When it flags a patient for readmission, it's right 51.97% of the time.
It catches 62.2% of the actual readmission cases.
The overall score, balancing precision and recall, is a solid 0.5663.
For non-readmitted patients (class 0), it's nearly perfect with 97% precision and 96% recall. For readmitted patients (class 1), it's got 52% precision and 62% recall, handling a tough, smaller group of cases.
This tool is a big deal for hospitals! It helps pinpoint patients who need a little extra TLC, which could mean fewer return visits and better health outcomes. It's not just tech—it's a way to make healthcare better and brighter!