# ClinIT AI Features
ClinIT AI is a clinical decision-support layer that drafts documents, interprets data, and surfaces insights — always requiring doctor review before outputs are committed to the patient record.
---
Core Principle
> 🤖 AI assists. The doctor decides.
> Every AI output is marked PENDING_REVIEW until the treating doctor approves it. Only approved outputs are saved to the patient chart.
---
AI Modules Available
### Prescription & Medication
| Module | How to Access | What it Does |
|---|---|---|
| Smart Prescription Review | Session → Review Rx | Checks for drug interactions, allergy contraindications, dosing concerns |
| Prescribing Defaults | Settings → Prescribing | Personalises default doses/durations per drug class |
### Clinical Documentation
| Module | How to Access | What it Does |
|---|---|---|
| Discharge Summary | Session → AI → Discharge Summary | Generates a structured discharge narrative from session data |
| Referral Letter | Session → AI → Referral Letter | Drafts a professional referral letter with patient history pre-filled |
| Clinical Summary | Patient Chart → AI Summary | Summarises the patient's clinical history as a brief paragraph |
| Patient Profile Brief | Patient Chart → AI Brief | Generates a clinician-facing narrative of the patient's full medical history |
### Specialty AI
| Module | Specialty | What it Does |
|---|---|---|
| Cardiology Risk Narrative | Cardiology | Analyses risk factors and generates a structured CVD risk summary |
| Derm Severity Interpreter | Dermatology | Interprets PASI/SCORAD/IGA trends and treatment response |
| Isotretinoin Monitor | Dermatology | Flags safety concerns during isotretinoin monitoring visits |
| ENT Audiogram Interpreter | ENT | Generates a narrative from pure-tone audiometry results |
| ENT Rhinology Narrative | ENT | Summarises nasal exam and rhinology findings |
| Myopia Predictor | Optical | Predicts myopia progression trajectory |
| Paeds Growth Interpreter | Paediatrics | Analyses growth chart data and flags concerns |
| Physio Outcome Interpreter | Physiotherapy | Interprets outcome score trends and generates a progress narrative |
| Psychiatric Formulation | Psychiatry | Generates a biopsychosocial formulation |
| Mood Trajectory Predictor | Psychiatry | Analyses mood diary entries for patterns |
| OBGYN Pregnancy Coach | OBGYN | Provides visit-specific pregnancy guidance |
| OBGYN Risk Narrative | OBGYN | Summarises obstetric risk factors |
| OBGYN Anomaly Scan Notes | OBGYN | Structures anomaly scan findings into a clinical note |
| Chronic Disease Coach | General | Management suggestions for chronic conditions |
| Dev Screen Interpreter | Paediatrics | Interprets developmental screening results |
### Analytics AI
| Module | Access | What it Does |
|---|---|---|
| No-Show Narrative | Analytics → Appointments | Analyses no-show patterns clinic-wide |
| Recall Priority | Analytics → Patients | Prioritises patients due for recall visits |
| Treatment Sequence | Clinical | Suggests optimal treatment sequencing for complex cases |
| Demand Forecast | Analytics | Forecasts appointment demand by specialty/day |
| Patient LTV | Analytics | Calculates and analyses patient lifetime value |
---
Reviewing AI Outputs
- All AI outputs are stored in the `AiOutput` table with status PENDING_REVIEW.
- A notification appears when an AI output is ready.
- Open the output — read it carefully.
- Click Approve to save it to the patient record, or Discard to reject it.
- Approved outputs are timestamped and linked to your user ID.
---
Privacy & Data
- AI processing uses the Anthropic API (Claude).
- Only the data needed for the specific module is sent (no full database dumps).
- Patient identifiers are minimised — names are replaced with patient IDs in some modules.
- AI interactions are logged in the
AiOutputtable for auditability.
---
Known Limitations
- AI-generated clinical content is not a substitute for clinical judgement.
- Drug interaction checking is illustrative — always verify with a current formulary.
- Specialty AI modules perform better with complete, up-to-date patient data.