Course 3
Predictive Modeling in Clinical Data Science
This course focuses on predictive modeling techniques applied to clinical data, which may include electronic health records (EHR), medical histories, lab results, and sensor data to predict disease outcomes and patient risk factors.
Week 1
Introduction to Clinical Data Science
- Types of clinical data: EHR, medical histories, and sensor data.
- Challenges in working with clinical datasets: data privacy, missing values, and imbalanced data.
Week 2
Data Wrangling and Cleaning in Clinical Datasets
- Data cleaning techniques: handling missing data, duplicates, and outliers.
- Feature engineering and creating meaningful features from clinical data.
Week 3
Machine Learning Models for Predictive Analysis
- Overview of regression models, decision trees, random forests, and SVMs for predicting clinical outcomes.
Week 4
Survival Analysis for Clinical Data
- Cox proportional hazards model and Kaplan-Meier estimator for time-to-event data.
- Predicting patient survival or time to relapse in cancer patients.
Week 5
Handling Imbalanced Data and Model Validation
- Techniques to handle imbalanced data in clinical datasets.
- Model validation using cross-validation and ROC analysis.
Week 6
Final Project: Predicting Disease Risk or Outcome
- Build a predictive model to assess patient risk or predict outcomes (e.g., diabetes, heart disease, cancer).