STEP 01 Why It Matters
STEP 02 Designed around Stakeholder Needs
WHY THIS PROJECT MATTERS

"Repeated readmissions can indicate gaps in discharge planning, care coordination, medication review, patient education, or post-discharge follow-up."

Thirty-day readmission is also a hospital quality indicator. Under CMS HRRP, excess readmissions can reduce Medicare payments by up to 3%, while AHRQ/HCUP reported an average readmission cost of approximately $15,200.

$15,200 Estimated average readmission cost
Up to 3% CMS Medicare reimbursement reduction risk
11.16% 30-day readmission rate in project dataset

That is why this project asks: can hospitals identify high-risk diabetic patients before discharge — and act earlier?

Source: CMS HRRP and AHRQ/HCUP, cited in project report.

DESIGNED AROUND STAKEHOLDER NEEDS

Designed around Stakeholder Needs

The system serves two distinct user personas, each with different needs and objectives.

HOSPITAL MANAGER
Required Information Nodes:
PLATFORM SOLUTION

Executive BI Dashboard

Aggregates SQL queries into live high-level trends, financial exposures, and clinical penalty metrics.

Open Live BI Dashboard
DISCHARGE PLANNER
Required Information Nodes:
PLATFORM SOLUTION

AI-Powered Discharge Portal

Predicts 30-day readmission risk, maps drivers, and coordinates care checklists via clinical copilot.

Open Patient AI Portal
BEYOND THE MODEL

one more thing...

Why and How this AI Copilot is different and useful?

EVIDENCE PIPELINE

From question to grounded answer

The Copilot does not answer from vibes. A RAG layer retrieves patient context, then model evidence and SHAP signals are assembled before generating a discharge-focused response.

01

User prompt

The planner asks: "Review whether Olivia Patel is ready for discharge"

02

Patient record enters the pipeline

Encounter details, labs, diagnosis history, medication status, and discharge context are retrieved.

03

SHAP values support reasoning

Model drivers become evidence: prior utilisation, A1c context, age group, and diagnosis burden.

04

Evidence-grounded output

The answer explains risk, highlights uncertainty, and suggests what to review before discharge.

Clinical AI Copilot

Hello! How can I assist you with this patient today?

I can review structured clinical records, translate risk predictions, and evaluate discharge readiness. Ask a free-form question or use the quick actions below:

Checklist Phases Evaluated
  • 1 Historical Risk Profile
  • 2 Current Encounter Complexity
  • 3 Labs & Glycemic Review
  • 4 Medication & Diabetes Regimen
  • 5 Discharge Readiness & Care Transition
Suggested Inquiries
Review whether Olivia Patel is ready for discharge
Review whether Olivia Patel is ready for discharge
Checking 5-phase checklist...
Historical Risk Profile
Current Encounter Complexity
Labs & Glycemic Review
Medication & Diabetes Regimen
Discharge Readiness & Care Transition
Review whether Olivia Patel is ready for discharge
Overall Assessment Medium Risk

This patient is stratified as Medium Risk (11.55%) with prior utilization level low and discharge planned to home.

1 Historical Risk Profile
Review

The patient's advanced age and recent history of inpatient utilization are visible factors in the risk assessment that should be reviewed during discharge planning.

Evidence Observed
  • Age group: senior adult
  • Recent inpatient utilization: present
  • Prior utilization level: low
Model evidence SHAP audit trail Show
Model Explanation (SHAP)
  • Age and recent inpatient utilization contribute to the medium-risk profile.
2 Current Encounter Complexity
Clear

The clinical burden remains at a medium level with a three-day length of stay; current documentation appears adequate for the complexity level.

Evidence Observed
  • Length of stay: 3 days
  • Clinical burden level: Medium
  • Current documentation: adequate
3 Labs & Glycemic Review
Review

While maximum glucose levels remain normal, the lack of HbA1c testing is worth checking as a planning consideration for long-term diabetes management.

4 Medication & Diabetes Regimen
Review

Medication reconciliation is complete, but the recommendation for diabetes education should be reviewed before discharge to ensure adherence with the current repaglinide regimen.

Evidence Observed
  • Active medication count: 14
  • Diabetes meds used: Yes
  • Insulin status: No
  • Medication change during stay: No
  • Key diabetes medications: repaglinide: Steady
  • Medication reconciliation status: Completed
  • Diabetes education status: Recommended
  • Reviewed by pharmacist: Oliver Reed, PharmD
Model evidence SHAP audit trail Hide
Model Explanation (SHAP)

The model explanation highlights insulin_No as a risk-decreasing driver. The model explanation highlights Diabetes medication used as a risk-increasing driver.

5 Discharge Readiness & Care Transition
Clear

Discharge to home is currently planned with standard follow-up, and no major social support concerns were identified in the planning notes.