Landing Page
Step 1
Step 2
Step 3
Step 4
Step 5
Results
Recommended next steps

CheckFirst

AI-Powered Medical Urgency Triage

What Inspired Us

It usually begins with a quiet question.

A headache that lingers.
A tightness in the chest.
A fever that refuses to leave.

In that moment, a person stands between two risky choices: ignore it and hope it passes, or panic and rush to the emergency room. Both reactions can cause harm.

We realized the real problem is not diagnosis. It is uncertainty.

CheckFirst was built to reduce that uncertainty.

The Problem

Healthcare systems face two dangerous patterns:

Underreaction → Delayed treatment → Severe outcomes
Overreaction → Emergency room overload

People lack a structured, accessible, and explainable way to assess medical urgency before seeking care.

The gap exists between symptom onset and professional consultation.

How We Built It

1. Structured Risk Modeling

We designed a weighted risk scoring function:

$$ RiskScore = (W_s \times Severity) \times AgeFactor \times DurationFactor \times ComorbidityFactor $$

Where:

$W_s$ = base weight of symptom
Severity = user-reported intensity (1–10)
AgeFactor = multiplier for high-risk age groups
DurationFactor = escalation over time
ComorbidityFactor = existing medical conditions

This allows structured reasoning instead of keyword matching.

2. Symptom Interaction Modeling

Symptoms rarely act alone. Certain combinations dramatically increase risk.

We modeled interaction effects as:

$$ CombinedRisk = \sum_{i=1}^{n} W_i + \sum_{i \neq j} Interaction_{ij} $$

This captures high-risk clusters such as:

Chest pain + sweating + shortness of breath
→ Elevated cardiac risk

This makes the system context-aware.

3. Machine Learning Classification

We trained a supervised learning model (Random Forest / Gradient Boosting) using structured triage scenarios.

Input features:

Encoded symptom vectors
Severity score
Duration
Age
Known medical conditions

Output classes:

Monitor at Home
Consult Within 24–72 Hours
Urgent Care Needed
Emergency Attention Required

4. Explainable AI Layer

In healthcare, explanation is essential.

We implemented feature attribution scoring:

$$ Importance(feature_i) = \Delta Prediction\ when\ feature_i\ changes $$

The system displays:

Which symptoms influenced the result
Why escalation was triggered
What changes would increase urgency

Transparency builds trust.

Challenges We Faced

Balancing Safety and Precision

Over-escalation causes panic.
Under-escalation causes harm.

We carefully calibrated thresholds to remain conservative for high-risk patterns while avoiding unnecessary emergency recommendations.

Ethical Boundaries

CheckFirst does not diagnose conditions.
It does not replace medical professionals.
It provides structured urgency guidance.

Ambiguous cases default toward recommending professional consultation.

Bias Mitigation

Medical risk varies across age groups and genders.
We adjusted risk multipliers to prevent systematic underestimation of symptoms in historically underdiagnosed populations.