Equity-by-design checklist for EHDS secondary use of genomic data and AI

Issue

The European Health Data Space (EHDS) is expected to expand secondary use of electronic health data for public-interest research and policymaking, including population genomics and AI-supported analytics. If equity is not assessed explicitly, secondary use may reproduce or widen urban–rural and minority-group disparities via uneven data coverage and biased model behaviour. We designed an equity-by-design checklist for (i) health data access bodies (HDABs) reviewing permit requests and (ii) applicants seeking secondary use of genomic or other high-dimensional datasets. Using a requirements-to-controls mapping approach, we translated EHDS secondary-use provisions and associated implementation guidance into practical decision points on purpose limitation, representativeness, subgroup performance, transparency/contestability and post-use oversight.

Results

The tool operationalises 12 checks as five permit-stage questions: (1) Is the requested use within authorised secondary-use purposes and proportionate to the stated aim? (2) Has representativeness been evaluated across neighbourhood deprivation, migration background and age/sex strata, with a mitigation plan for gaps? (3) Are AI/ML models tested for subgroup performance and bias, with predefined thresholds and auditable logs? (4) Are outputs sufficiently explainable and contestable for affected groups, including a plan for responsible policy translation? (5) Is there post-use monitoring (model drift, adverse equity impacts) and transparency about permits and key safeguards? A one-page applicant template supports consistent review and reduces administrative burden.

Conclusions

Embedding equity checkpoints into EHDS secondary-use workflows can align genomics/AI innovation with fair public health decision-making.

Key messages

• Equity risks in EHDS secondary use concentrate at representativeness, bias testing, contestability and post-use monitoring.

• A short equity-by-design checklist can help HDABs authorise AI-enabled genomics without widening urban health inequities.