Challenges
- Help prevent the hospitalization/re-hospitalization of patients post dialysis by leveraging predictive analytics
- Build a model to predict the likelihood of hospitalization for high-risk patients within 30, 60 & 90 days of treatment at the hospital
- Facilitate optimal utilization of slots and related resources allocated to patients with higher propensity of hospitalization
Solutions
- Analyzed data captured during a patient visit, including demographic details, vitals diagnosis, lab tests, and medical conditions currently associated with the patient
- Developed classification machine learning models to predict patient hospitalization rate using advanced ML and DL algorithm
- Analyzed important features and derived new ones to augment the model performance
- Interpreted the reason behind hospitalization of each patients predicted positive by the model using model interpretability techniques
Tools & Technologies
Python, Anaconda, Spyder, Sokit Learn
Key benefits
- Predicted the likelihood of patients’ readmission with 72% ROC-AUC score
- Increased utilization of dialysis slots by 70%
- Helped client improve their rating by keeping a check on hospitalization of CKD patients
- Made efficient utilization of resources by providing better care to high-risk patients
- Substantial medical cost savings with the help of preventive care
