Severity Levels
RxLabelGuard classifies every drug interaction into one of five severity levels. Severity is determined by analyzing the language and context of the FDA label's drug interaction, warnings, contraindications, and boxed warning sections.
Must not be used together. Absolute clinical prohibition.
- Label signals
- Label uses phrases like "is contraindicated," "must not be used," or "do not co-administer."
- Suggested alert tier
- Hard stop — block or require override
Potentially life-threatening interaction. Requires clinical intervention or monitoring.
- Label signals
- Boxed warnings mentioning specific drugs, language about life-threatening outcomes or hospitalization risk.
- Suggested alert tier
- Prominent warning — require acknowledgment
May require monitoring, dosage adjustment, or alternative therapy.
- Label signals
- Language about dose adjustment, increased monitoring, or clinically significant but manageable consequences.
- Suggested alert tier
- Advisory alert — display with recommendation
Minimal clinical significance. Monitor but typically safe to co-administer.
- Label signals
- Pharmacokinetic observations with minimal clinical impact. Advisory language without specific intervention requirements.
- Suggested alert tier
- Informational — log, display on request
Insufficient data in FDA labeling to classify severity.
- Label signals
- Interaction mentioned but without enough clinical context to assign a severity level.
- Suggested alert tier
- Flag for clinical review
How Severity Is Determined
Severity classification uses a combination of deterministic keyword analysis and AI-based classification. The deterministic layer catches clear-cut cases (explicit contraindication language, boxed warning mentions). The AI layer handles nuanced prose where severity must be inferred from clinical context.
Both layers retain the source text snippet as evidence, so downstream consumers can verify the classification against the original FDA label.
Mapping to Clinical Alert Tiers
Most EHR and CDS systems use a tiered alert model (hard stop, warning, advisory, informational). The severity levels map naturally to these tiers, but the exact mapping depends on your organization's clinical governance policies. The "Suggested alert tier" above is a starting point — your clinical team should define the final mapping.
API Response Example
{
"severity": "major",
"mechanism": "Increased anticoagulant effect and bleeding risk",
"recommendation": "Monitor INR closely if coadministered",
"evidenceSnippet": "Aspirin increases bleeding risk when used with warfarin.",
"labelSection": "drug_interactions",
"splSetId": "abc-123"
}