AI Copilot

AI model architecture

Clinit uses a model fallback chain to guarantee AI availability even when primary model endpoints are congested:

Primary: GPT-4o-mini — fastest, most cost-efficient, used for most clinical AI

Fallback 1: GPT-3.5-turbo — triggered if primary times out (>8s)

Fallback 2: Llama 3.3-70b — open-source fallback via OpenRouter

Fallback 3: Gemma 3-27b — secondary open-source

Fallback 4: Gemma 3-12b — final fallback before error response

The client adds clinic context (specialty, country), and strips any PHI before sending to the model.

Clinical AI routes

Twenty AI-powered routes available in Clinit:

CLINICAL DOCUMENTATION

ai/clinical-summary — Structured session summary from vitals + notes + medications

ai/session-summary — Concise session summary for patient-facing portal view

ai/soap-notes — SOAP note generation from bullet-point dictation

ai/discharge-summary — Full discharge summary from session data

ai/referral-letter — Professional referral letter; specialty-specific sections auto-populated

ai/patient-brief — 3-bullet brief for doctor to read before the patient enters

PRESCRIPTIONS & SAFETY

ai/smart-prescription-review — Drug interactions, contraindications, dose errors, pregnancy/renal flags

ai/drug-interactions — Full medication list cross-check on save

CARDIOLOGY

ai/cardiology-risk-narrative — ESC 2021 plain-language narratives for SCORE2, GRACE, CHA₂DS₂-VASc, HAS-BLED, HEART

DERMATOLOGY

ai/derm-severity-interpreter — PASI/SCORAD/DLQI interpretation + EADV/BAD step-up recommendation

ai/isotretinoin-monitor — Monthly safety monitoring alerts against iPLEDGE thresholds

PAEDIATRICS

ai/peds-growth-interpreter — Z-score interpretation; failure-to-thrive and obesity flags

ai/dev-screen — Developmental milestone screening with delay flags

PHYSIOTHERAPY

ai/physio-outcome-interpreter — DASH/KOOS/VAS interpretation + treatment progression suggestion

OPTICAL

ai/myopia-predictor — Progression trajectory from historical SE and axial length measurements

GENERAL MEDICINE

ai/chronic-disease-coach — Action plans for DM, HTN, Asthma, COPD, CKD, HF per NICE/WHO guidelines

OPERATIONS

ai/demand-forecast — Appointment volume prediction for next 4 weeks

ai/noshow-narrative — No-show risk prediction + personalised outreach message draft

ai/recall-priority — Hygiene recall prioritisation by BPE score, last visit, and risk factors

ai/dashboard-insight — Daily clinic insight card for the home dashboard

ai/treatment-sequence — Dental treatment plan phase sequencing by clinical priority

AI prescription review — 5 safety checks

When a prescription is saved, Clinit runs five parallel AI checks before writing to the database:

Check 1 — Drug-drug interactions

Compares every new medication against the patient active medication list + all medications in the current session. Uses a curated interaction database updated quarterly.

Check 2 — Dose range validation

Validates the entered dose against published dose ranges for the patient age, weight (if recorded), and renal function (eGFR from latest lab if available). Flags doses below the therapeutic minimum or above the safe maximum.

Check 3 — Contraindications

Checks the drug against the patient active problem list. Example: warns if an NSAID is prescribed for a patient with active peptic ulcer disease.

Check 4 — Pregnancy and lactation flags

For female patients of childbearing age (15–49), flags FDA Pregnancy Category C, D, or X drugs and drugs contraindicated in breastfeeding.

Check 5 — Renal and hepatic adjustment

If eGFR < 60 or AST/ALT > 3× ULN appears in the patient lab history, checks whether dose reduction or monitoring is required for the prescribed drug.

Result display:

✅ Green — all checks passed

⚠️ Amber — caution; consider alternative

🔴 Red — contraindicated; strong recommendation to change

The prescribing doctor can acknowledge any warning and proceed, with the acknowledgement logged in the audit trail.