AI AutomationHealthTech / Predictive Analytics

CareIQ — Clinical Risk Stratification & Readmission Predictor

Domain / Scope
Clinical Triage / Predictive Analytics
Year
2026
Duration
18 weeks
Primary metric
41% readmission reduction
41%
Reduction in 30-day readmission rate (18.4% → 10.9%)
0.91
AUC-ROC score — top-decile predictive accuracy for readmission models
2.1M
Patient records utilized in model training and cross-validation
< 180ms
P99 inference latency at time of clinical discharge
3
Hospital systems fully integrated with live EHR discharge workflows
76%
Of high-risk patients reached within 24 hours of discharge

The Problem

System Bottlenecks & Technical Friction

Unplanned 30-day hospital readmissions create severe clinical and financial burdens. Traditional discharge planning relies primarily on static protocol checklists rather than individualized patient risk stratification. High-risk patients frequently receive standard discharge instructions without timely post-discharge follow-up. Clinical teams need real-time, explainable risk scoring embedded directly into EHR discharge workflows so care coordinators can prioritize interventions.

Our Engineering Approach

Architecture Design & Implementation

CareIQ is a predictive ML platform generating real-time 30-day readmission risk scores at the point of discharge. An XGBoost classifier is trained on 2.1M de-identified records incorporating 47 clinical and social determinant variables. Integrated SHAP (SHapley Additive exPlanations) values provide clinical explainability by showing clinicians the top contributing factors for each score. AWS SageMaker hosts the low-latency inference endpoint, while automated workflows trigger structured check-ins for high-risk cohorts.

Technical Architecture

System breakdown & stack.

8 core subsystem modules
01

XGBoost classifier trained on 2.1M patient records with 47 clinical + SDOH features

02

SHAP explainability layer providing per-prediction factor attribution for clinical transparency

03

AWS SageMaker real-time inference endpoint with p99 latency < 180ms

04

FastAPI score ingestion API integrated into EHR discharge workflows

05

Twilio Programmable Voice for automated multi-touch post-discharge outreach calls

06

PostgreSQL patient tracking database with 90-day rolling readmission outcome data

07

React nursing dashboard with risk stratification queue, call logging, and escalation workflows

08

Docker-based model retraining pipeline running monthly on new discharge outcome data

XGBoostSHAPAWS SageMakerTwilioReactPostgreSQLFastAPIDocker

Engagement Timeline

Engineering delivery schedule.

Total: 18 weeks
Weeks 1–3

Patient data extraction, de-identification, feature engineering across 47 clinical + SDOH variables

Weeks 4–7

XGBoost model training, hyperparameter tuning, SHAP explainability integration, validation on holdout set

Weeks 8–11

SageMaker deployment, EHR discharge API integration at 3 hospital sites

Weeks 12–14

Twilio outreach automation, call script design, nursing dashboard build

Weeks 15–18

Parallel validation with control group, outcome tracking, go-live and 30-day result measurement

More Case Studies

Related AI projects

Want results like these for your business?

Book a 30-minute scoping call with a senior engineer. We'll scope your AI project, define the architecture, and give you a fixed-price proposal within 5 business days.