Once the doctor has reviewed the extracted answers, the backend converts them into standardised categories: age bands, BMI ranges, smoking status adjusted for age and cigarettes per day, and blood pressure stages. Anything left blank or marked “not sure” is recorded as missing rather than assumed safe.

These categories are passed to a deterministic rule engine, which checks them against three datasets: the UK Medical Eligibility Criteria (UKMEC) classifications, product-specific contraindication rules for Microgynon 30, and medication interaction rules. Unrecognised medicines are flagged rather than ignored.

When a patient’s information matches a rule, the engine returns the classification, an explanation, follow-up information and a source reference where available. The worst finding sets the outcome: UKMEC category 4 or a product contraindication is red, categories 2 and 3, interactions or missing answers are amber, and no findings is green.

The frontend uses these findings to present a traffic-light summary with the reasons beside it. Because every result traces back to a specific rule and source, the checks are more transparent and repeatable than relying on an AI model to generate a recommendation. The prototype is still being validated and supports clinical review rather than replacing it.