Introducing Co-Medic Research: Identify patient populations in primary care

From patient data to population research

Primary care holds a large amount of patient data, lab results, diagnoses, prescriptions, vital signs, specialist letters. Access to this data at population level has always been a bottleneck. Identifying a cohort like "patients with elevated NT-proBNP and signs of heart failure who are not receiving RAAS-inhibiting therapy" typically requires combining custom database queries with extensive manual chart review.

Co-Medic Research makes population and cohort analysis directly actionable, providing straightforward insight into the full practice population based on coded EHR data supplemented with concepts from unstructured clinical text. The result is a structured, queryable population table ready for analysis.

How it works

Every data point in the practice, queryable

Each column in a population table maps to a specific type of clinical data. The system currently supports more than 20 column types, including:

Enrichments: when structured data is not enough

Not every clinical finding is correctly coded. A diagnosis mentioned only in a discharge letter, a lab value buried in a specialist letter, a medication change described in free text... this is the reality of primary care data.

Data coverage across a practice population

When there are gaps in structured data, Co-Medic can run targeted enrichments to fill them from unstructured sources such as specialist letters and clinical notes.

Every enriched value links back to the source document. Nothing is a black box, you can always verify where a data point came from.

A practical example: diabetes screening

A practice wants to get a complete picture of its diabetes population. They build a population query with five criteria, Type 2 diabetes diagnosis, HbA1c, eGFR, LDL cholesterol, and metformin use, and evaluate every patient in the practice at scale.

Each column retrieves data from the patient's clinical record, lab results via LOINC code, medication via ATC classification, and diagnoses via ICPC-2 and SNOMED-CT. Where data is missing, Co-Medic can run targeted enrichments to extract values from unprocessed letters and reports.

From question to cohort

You do not need to configure every criterion by hand. Co-Medic Research includes an agent that helps you build population queries through natural language. Describe your research question: "Find patients with CKD stage 3+ who are not on SGLT2 inhibitors without a recorded contraindication", and the agent translates that into criteria, enrichments, and thresholds.