How DrugPair builds a checker result
Updated 16 July 2026. DrugPair uses a layered educational workflow; it is not a complete clinical decision-support system.
1. Normalize the items
Medicine names are matched against public terminology such as RxNorm where available. Foods, drinks, fruits, and supplements are kept as distinct item types.
2. Retrieve compact evidence
The application may query openFDA labeling, NIH Dietary Supplement Label Database data, USDA FoodData Central, Open Food Facts, and other public datasets. Availability and coverage differ by item.
3. Apply curated safety rules
A small local rule set can surface known high-priority combinations. These rules are not exhaustive and must not be interpreted as proof that an unlisted combination is safe.
4. Generate a plain-language explanation
An AI provider may summarize the reduced evidence under strict instructions. The AI is an explanation layer, not the evidence source. When required analysis is unavailable, DrugPair returns an unavailable state rather than a reassuring demo result.
5. Show limitations and next questions
Results can miss interactions and cannot account for every dose, condition, allergy, pregnancy, laboratory result, or personal risk. Users should confirm important decisions with a licensed pharmacist or prescriber.