Predictor
Portfolio project
1.0 Readmission Intelligence

Turn discharge data into clear readmission signals

A focused clinical AI demo that estimates 30-day readmission risk, compares model outputs, and explains what moved the score.

XGBoost risk score
48%
Moderate risk
Age
65
Medications
5
Prior stays
1
Chronic conditions
Prior visits
Length of stay
Follow-up within 72 hours
1.1 Signals

Inputs that explain the patient

Age
Length of stay
Medications
5+ medications
Chronic conditions
High-risk condition
Prior visits
Prior inpatient stays
1.2 Workflow

From raw profile to interpretable action

Patient Data
01
Feature Engineering
02
XGBoost Model
03
Risk Score
04
SHAP Explanation
05
Clinical Action
06

Estimate risk quickly

Convert discharge parameters into a probability score between 0% and 100%.

Compare model behavior

Show XGBoost beside logistic regression so the prediction has a baseline.

Explain the drivers

Surface the features that push the prediction higher or lower.

1.3 Models

Built to compare and explain

XGBoost

Primary

Captures nonlinear interactions between utilization, medications, and conditions.

Logistic Regression

Baseline

Provides a simpler statistical baseline for comparison.

SHAP-style impact

Explainability

Breaks the final score into directional feature contributions.

Educational use only

This demo should support, not replace, clinical judgment. Actual readmission risk depends on many factors not captured by the model, including social determinants of health, care coordination, adherence, and clinical assessment.

Ready to explore the model

Open the predictor

Launch demo