PetriDish AutoimmuneClinical decision support · not diagnosticDeterministic rule engine · v0.3

Reduce missed high-risk autoimmune presentations.
Shorten the path to rheumatology.

Without over-ordering ANA panels. Without replacing your clinician. Deterministic. Auditable. Every claim resolves to an evidence object with a rule version and a required review flag.

Try the triage belowIntended use statement
NEVER OUTPUT
  • "Patient has lupus" or any final diagnosis
  • Definitive exclusion of TB or any infection
  • Autoimmune-disease probability without pre-test calibration
  • Prescribing decisions without clinician review
ALWAYS OUTPUT
  • Urgency tier · Red / Amber / Yellow / Abstain
  • Differential categories with supporting features
  • Missing critical data · what to collect next
  • Recommended confirmatory / reflex tests
  • Evidence object per triggered rule
  • requires_clinician_review: true
Live · structured intake

4-tier disposition · reproducible · rule-versioned

Structured intake · Module A input
Symptoms (toggle)
Labs & prior workup
Awaiting triage input…
Fill the intake and press Run triage. The disposition and full evidence chain appear here.
The clinical journey

Symptoms → Investigations → Referral → Treatment safety → Monitoring

A
Structured triage
Symptoms + history + baseline labs → Red/Amber/Yellow/Abstain
B
Infection safety gate
TB · HBV · HCV · HIV · pregnancy · vaccination · never says 'excluded'
C
ANA reflex
ICAP pattern → disease-specific confirmatory bundle · never converts ANA+ to disease probability
D
Disease pathway
SLE · Sjögren · SSc · APS · Myositis · RA · Vasculitis · phenotype-specific
P
Thiopurine safety
TPMT + NUDT15 · CPIC-graded · NUDT15 is the dominant Indian-risk gene
M
Longitudinal monitoring
Complement · dsDNA · CBC · urine protein trends · missingness-aware
Who this is for
  • Rheumatology clinics · shorten path to referral, reduce missed high-risk cases
  • Diagnostic labs · sit inside your ANA report as the reflex + clinical-action layer
  • Primary-care networks · structured triage before scarce specialist time is consumed
  • Immunology researchers · trace-audited evidence chain, versioned rules, reproducible
Validation roadmap
  • Retrospective silent study · 100–200 historical cases
  • Prospective shadow pilot · rheumatology clinic + diagnostic lab
  • Multi-centre validation · 600–1,000 cases · 3+ sites · adjudicated by 2 rheumatologists
  • Primary endpoint: sensitivity for urgent / appropriate referral
  • Safety endpoints: false-negative rate on organ-threatening presentations, TB miss rate, abstention rate

One workflow. One thesis. Traceable end-to-end.

Ready to pilot at one rheumatology clinic or diagnostic lab. Retrospective concordance audit · zero risk · your data stays in your building.

Request clinical pilotMethodsGitHub