DRAS Console — Diabetes Risk Augmenting System

Kinetic SVM 3-class + SVR severity · confidence-scored · human-in-command

DEMO · clinician decides

An endocrinologist reviewing a glucose-tolerance curve and the Diabetes Index on a diagnostic console

Patient OGTT input (2-hour values)

Blood glucose — 2h OGTT (mg/dL)

132

Insulin — 2h OGTT (µU/mL)

150

Age (yrs)

BMI

Enter the patient's 2-hour OGTT glucose & insulin (fasting values would use different thresholds). The model reads insulin against glucose to separate the at-risk mechanism.

Classification & confidence

—

Non-diabetic0%

At-risk (rim)0%

Diabetic0%

Confidence in top class—

Annual hospitalisation risk (reference)—

Kinetic severity: est. damping A (SVR)—

severity index (100 = most severe)0 / 100

A clinician reviewing glucose-tolerance curves, making the final decision

The system flags who needs a glucose-tolerance test. A clinician makes every call.DRAS surfaces its own uncertainty and hands the judgement to a human — augmenting the clinician, never replacing them.

3 · Medical resource planning — outpatient diabetic population

🌍 GLOBAL COVERAGE — resource planning anchored to the real health-system resources of 24 countries worldwide. Choose your market below; annual cost, bed-days, supplies and workforce re-compute in your own economy.

Outpatient diabetic patients in your service

Market (per-capita cost anchor)

Annual care cost (diabetic cohort)—

Expected inpatient bed-days / yr—

Insulin units / yr (supplies)—

Test strips / yr (supplies)—

Cost avoidable via early intervention—

P1 · NAS Nurses for follow-up (@500/nurse)—

P2 · Adherence Adherence checks / yr (quarterly)—

At-risk to pre-empt (est. shadow cohort)—

Ties the individual call to the system view: every diabetic flagged here becomes nurse-follow-up load (Project 1, Nurse Augmenting System) and an adherence-monitoring case (Project 2, Medication Adherence). Catching the at-risk rim before conversion is where the avoidable cost lives.

How the numbers are derived — data vintage & cost method

Data sources & vintage

L1 · OGTT kinetics	Author's doctoral thesis — glucose-response (OGTT) simulations. Damped-oscillator damping A.
L2 · Screening	NIDDK Pima Indians Diabetes dataset (768 subjects), collected c. 1990.
L3 · Population risk	US CDC BRFSS survey, 2015 (diabetes health-indicators, 253,680 records).
L4 · Hospitalisation	UCI Diabetes 130-US Hospitals. Encounters 1999–2008 (101,766 admissions).
L5 · Economics	Country current health-expenditure per capita (WHO / World Bank), compiled mid-2026, converted to SGD.

How costs are derived — illustrative planning estimates, not billed figures:

Annual care cost	= patients × diabetic-cohort multiplier 2.3 × per-capita anchor.
Avoidable cost	= at-risk patients (20%) × per-capita anchor × 50% risk reduction (DPP-style lifestyle intervention, literature).
Bed-days	= admitted (30%) × mean 3-day stay × 1.1 readmission factor (130-hospital 1999–2008 pattern).
Supplies	Insulin units & test-strips scaled from the treated-diabetic fraction.

Cohort is the synthetic DRAS-RP dataset (main.csv). Economic anchors are point estimates; multipliers are directional planning figures, not epidemiological claims.

AEGIS · Human-AI Augmenting Systems — augment, never replace · human-in-command · honest metrics
Classification via SVM (3-class) and continuous severity via SVR (fitting-free damping estimate, CV R² 0.20 — coarse; precise damping needs the full OGTT curve). Calibrated to the synthetic DRAS-RP cohort (main.csv). Screening estimate for reference only — a clinician makes every call. Not for clinical use.