The Clinical Problem
AI models can suggest dangerous drug combinations — confidently.
Adverse drug events (ADEs) cause over 125,000 deaths annually in the US. A significant proportion are preventable — triggered by drug-drug interactions, contraindicated combinations, or dosing errors that a rule-based check would have caught.
General-purpose AI language models score well on average clinical reasoning — but fail catastrophically on specific adversarial cases: beta-blockers in cocaine toxicity, NSAIDs in advanced CKD, PDE5 inhibitors co-prescribed with nitrates. These are the cases where the model is most confident and most wrong.
Adversarial Safety Evaluation
All test cases sourced from published clinical guidelines (AHA, KDIGO, FDA labeling). Every combination below was intercepted.
24-Gate Safety Engine
Six safety dimensions. All deterministic.
Drug-Drug Interactions
Pairwise DDI databases + CYP450 metabolic pathway conflict detection
Organ Function Awareness
Renal eGFR dosing (KDIGO) · Hepatic Child-Pugh classification · QTc risk
Drug-Disease Contraindications
Patient conditions vs. drug labels — e.g. beta-blocker contraindicated in asthma
Population Safety
Pediatric weight-based dosing · Beers Criteria · Pregnancy FDA categories
Allergy Cross-reactivity
Structural similarity detection beyond simple drug name matching
Cumulative Burden Scoring
Anticholinergic Cognitive Burden (ACB) across full medication list + ATC duplication
How It Works
From transcript to validated suggestion in <2 seconds.
Clinical Reasoning
AI analyzes transcript, current medications, diagnoses, allergies, and organ function to build a full clinical picture.
Formulary Search
AI models identify appropriate candidates from a comprehensive clinical formulary based on diagnosis and patient context.
Cross-Validation
Candidates are cross-referenced against the clinical picture to filter pharmacologically inappropriate choices.
24-Gate Safety Pipeline
Each candidate passes all 24 deterministic safety gates. CRITICAL gates hard-block. CAUTION gates flag with clinical detail.
Evidence Ranking
Surviving candidates ranked by clinical evidence tier — first-line, second-line, adjunct — with transparent validation labels.
Clinician Reviews
Doctor reviews the validated, ranked list. All suggestions are drafts. Every override is acknowledged and logged.
FAQ
Clinical AI safety — answered.
How does ELLEXMED's AI Suggest medication engine work?
When a doctor clicks 'AI Suggest', the system performs clinical reasoning on the patient's transcript, current medications, conditions, allergies, and organ function data. AI models identify candidate medications from a comprehensive formulary. Each candidate is then passed through 24 deterministic safety gates — rule-based checks that cannot hallucinate — covering DDI, drug-disease contraindications, renal/hepatic dosing, QTc prolongation, allergy cross-reactivity, and more. Surviving candidates are ranked by clinical evidence tier and returned to the doctor as a reviewed, labelled list.
What is a deterministic safety gate and why does it matter?
A deterministic safety gate is a rule-based check that produces a guaranteed outcome (PASS, CAUTION, or CRITICAL) based on verified clinical data — it does not use probability or estimation. This matters because AI language models can confidently suggest dangerous drug combinations in edge cases. ELLEXMED's gates intercept those suggestions before they reach the clinician, regardless of what the underlying AI model outputs.
What does ELLEXMED's 90% MedQA score mean?
MedQA is a standardized clinical reasoning benchmark derived from US Medical Licensing Examination (USMLE) questions. ELLEXMED achieved 90% accuracy (18/20 correct) on a clinical question sample set. More importantly, the safety pipeline architecture raises the reasoning floor — eliminating entire categories of dangerous answers rather than reducing individual error rates.
What is AISI and why does it matter for medication AI?
AISI (AI Safety Institute) is a UK government body that evaluates AI systems for safety under adversarial pressure — including scenarios designed to induce dangerous outputs. For clinical AI, this means testing whether the system correctly refuses or flags high-risk prescriptions that a baseline model would accept. ELLEXMED's safety pipeline is evaluated against this framework.
What drug-drug interactions does ELLEXMED check?
ELLEXMED checks pairwise pharmacokinetic interactions from curated pharmacological databases, plus CYP450 metabolic pathway conflicts (which cause most clinically significant DDIs). A neural vector-similarity layer provides additional coverage for interactions not present in pairwise databases, using 1,084 drug-pair vectors calibrated at ALERT threshold ≥0.70 and CAUTION threshold ≥0.67.
What happens when a doctor overrides an AI safety warning?
Every CRITICAL or CAUTION override is acknowledged by the clinician and timestamped in ELLEXMED's immutable audit log. The override records the gate that was triggered, the drug involved, and the clinician's identity. This creates a verifiable clinical decision trail for compliance and review purposes.
Is AI Suggest a replacement for clinical judgment?
No. Every AI Suggest output is explicitly a draft for clinician review — the system informs, the doctor decides. The AI ranks and validates candidates; the clinician has full authority to accept, modify, or reject every suggestion. ELLEXMED is designed as a clinical copilot, not an autonomous prescribing system.
What organ safety checks does AI Suggest perform?
AI Suggest checks renal function (eGFR-based dose adjustment per KDIGO guidelines), hepatic safety (Child-Pugh classification for hepatic metabolism), and QTc prolongation risk from FDA drug label data. Medications requiring dose reduction for renal or hepatic impairment are automatically flagged or adjusted.
Start with 14 days free.
AI Suggest is included in all ELLEXMED plans. No credit card required.