A Bit About This Project
For reputational and financial reasons, hospitals tend to avoid readmissions. Readmission is a healthcare metric that tracks patients who are re-admitted to a hospital within 30 days. This is generally avoided by hospitals and is in the best interest of patients due to many complications, such as reduced coverage during readmissions from insurances like Medicare, which can cause financial issues for hospitals and patients themselves (2012). Hence, we believe that by analyzing the most prominent groups of readmitted patients, we can lower the readmission rates by identifying those that might require additional outpatient treatment, longer stays, or other preventative measures.
​
Project objective/proposal:
Our project objective/proposal is analyzing demographic, diabetic diagnoses, and medication histories to understand their impact on readmission types.
Environment: Google Colab
Language: Python
Machine Learning Models: Logistic Regression, Decision Tree, Random Forest
Work Process
Step 1
Step 2
Step3
Step4
Data cleaning
- Replacing '?' to 'None'
- Dropping expired records
- Remapping
EDA
One-hot-encoding
- Grouping three diagnoses by the sum of unique values
- Breaking out race
​
Fitting models
- Logistic regression
- Decision tree (80% accuracy, 18% recall)
- Random Forest