Health Diagnosis Simulation — Sequential Bayesian Reasoning

Case

A patient presents with chest discomfort. One of four conditions is the true cause — order tests to narrow it down, then diagnose.

Order a Test

Diagnostic Probabilities

Confidence threshold to diagnose 85%

Confirm Diagnosis

Case Stats

Cost accrued

Radiation

0 mSv

Top diagnosis confidence

25.0%

Test Log

No tests ordered yet.

How it works

Each candidate diagnosis D_i starts with a prior probability P(D_i). Every test has a known positive-result rate for each true condition, P(+ | D_i) — its sensitivity toward that specific disease. When a test comes back positive or negative, Bayes' rule updates every candidate at once:

P(D_i | result) = P(result | D_i) · P(D_i)
                  ───────────────────────────
                  Σ_j P(result | D_j) · P(D_j)

The numerator is computed for all four diagnoses, then the whole set is renormalised so the probabilities sum to 1 — the posterior becomes the new prior before the next test. This is the same sequential-updating logic used in real differential-diagnosis reasoning: each test is assumed conditionally independent given the true disease, so its evidence can be folded in one test at a time rather than needing every result at once.

Order a Test — draws a real/positive result from the hidden patient's true condition (the actual per-disease positive rate), then re-runs Bayes' rule above.
Confidence threshold — a visual target; the leading bar turns green once its posterior crosses it, mirroring a clinician's "confident enough to act" cutoff.
Confirm Diagnosis — locks in your answer and reveals whether it matches the hidden true condition.

⚠ Couldn't load the 3D engineThree.js failed to load from the CDN. Check your connection and reload.