NIKard prepares for EHDS with a Health Research Platform on AWS

Cardiovascular disease causes 36% of all deaths in Poland. Across the WHO European Region it kills 10,000 people a day. Part of the answer to those numbers is already sitting in hospital systems — decades of diagnoses, treatments, outcomes and images, recorded patient by patient. Almost none of it is ready for research.

The National Institute of Cardiology (NIKard) in Warsaw has more of that material than almost anyone in the region. Founded in 1979, it is Poland's top reference center for cardiology and treats the country's most complex cardiac cases. Four decades of that work left it with 17.5 million diagnostic studies and 530 terabytes of medical imaging, growing by 350 gigabytes a day.

Forty years of health data, just out of reach

Every hospital's data is scattered: clinical records live in the hospital information system (HIS), medical images in the picture archiving and communication system (PACS), and finance in the enterprise resource planning system (ERP). Each of these systems has its own data model and its own access rules. None of them was built to answer a research question. So every study starts with months of getting to the data. Preparing the patient cohort alone consumes up to 80% of the time set aside for the study. Institutions, in turn, have no single process for research requests and no one place that shows who asked for what and what data they received. Nor do they have the procedures or the resources to give a research team a properly secure, isolated, controlled and audited place to work, so every project reinvented all three.

On top of all this, there is a deadline. Using clinical data for research and development, rather than only for treating the patient it came from, is called secondary use. From March 2029, under the European Health Data Space (EHDS) regulation, it becomes an obligation for every holder of health data.Each institution will have to describe its datasets, answer data requests with anonymized statistics, make the data available in a secure environment, and keep a record of what it shared and with whom.

When NIKard became one of Poland's Regional Digital Medicine Centers, funded by the Medical Research Agency (Agencja Badań Medycznych, ABM), it decided to solve all of this at once and years ahead of the deadline.

Solution A research platform the institute owns

Following a public tender, NIKard turned to Chaos Gears, an AWS Premier Tier Services Partner with a track record in regulated, data-heavy environments, to build that platform with them. Our engineers and the institute's Centre for Digital Medicine worked as a single team from the first analysis through to the production launch. The result is implementation of the Health Research Platform: a medical research environment in the AWS cloud that integrates medical data from different sources and lets many research projects run in parallel, designed from the start against European legal and regulatory requirements. The platform is in production and handles the whole study lifecycle - from the first question and the feasibility check, cohort selection and approval, and finally secure research environments with standardized data, imaging, AI models and scalable compute.

One source of truth about the patient

The platform connects to the hospital's existing systems (HIS, PACS, ERP) without changing them. Data flows one way only, from the hospital to the platform. Once inside the platform, data is translated into the two standards medical research actually works in: Health Level Seven Fast Healthcare Interoperability Resources (FHIR) and the Observational Medical Outcomes Partnership Common Data Model (OMOP CDM). It is then gathered in one place and made searchable by patient criteria and medical terminology, including International Classification of Diseases (ICD) codes. Every morning the platform loads what happened in the hospital the day before: new visits, test results and hospital stays are added and matched to the existing patient's records, so researchers work on data that is one day old at most. The result is a single source of truth about patients, covering forty years of hospital stays and diagnostic studies at the institute, ready for modern medical research.

From research question to patient cohort

The researcher could start on their own. They check whether a study is feasible at all: how many patients match the criteria, what the age and sex breakdown looks like. The answer comes in minutes and shows counts, not records. Then they submit a request. The institute's administrator approves the study and its cohort criteria. Once approved, the platform assembles the cohort and delivers it in FHIR and OMOP CDM, together with the imaging the study was granted. The whole process now takes hours instead of weeks.

Medical imaging: 530 terabytes, fetched on demand

Cardiology runs on images, and NIKard's PACS archive holds more than 530 terabytes of them, growing by about 65,000 studies a year. To save on cost, the platform does not copy the archive into the cloud at once. It syncs the metadata and fetches the images a study needs on demand. Every DICOM file is anonymized before it reaches a research project. Researchers review the anonymized images in a browser-based viewer, and can annotate them for training AI models, with independent annotators and an adjudicator who settles disagreements.

Private laboratory in the cloud for every research study

For each approved research project and the cohort released to it, a Virtual Research Environment is created: an isolated workspace in the cloud, set up on request, holding only the cohort data covered by the data entrustment agreement. The research team logs into the platform access portal or virtual desktops and has everything it needs to analyze the data, train models and work with the latest AI models on the market. Computing power comes from the cloud, so the institute's own servers are no longer the limit. Each environment has its own monitored budget with costs visible separately, and compute stops automatically at an agreed threshold. When the research ends, the environment is archived and shut down.

Secure by default, auditable by design

Patient data is anonymized and stays in the institute's own AWS accounts in the European Union, encrypted at rest and in transit. Access follows least privilege and uses the identities the institute already manages. Every operation on the data is logged: who, when, what. A General Data Protection Regulation (GDPR) audit, an internal audit or an ethics committee gets its answers straight from that log.

Outcome When data stops being the barrier

It took five months from the pre-implementation analysis to the production launch. A feasibility check that used to mean weeks of waiting now takes minutes. Anonymized patient cohorts in FHIR and OMOP CDM reach the research environments within hours, together with DICOM imaging. Computing power is provisioned dynamically and billed separately for each research project.

EHDS makes secondary use of health data an obligation from March 2029. NIKard is preparing for it three years early: its clinical data is already anonymized, standardized in FHIR and OMOP CDM, and in use for research today. The institute is among the first research centers in Poland in this position.

Researchers are already working with the platform, and our cooperation did not stop at launch. We continue to develop it together with the institute's Centre for Digital Medicine, as new studies come through it and as the ground keeps moving — EHDS guidance, data standards and AI tooling are all changing faster than any single release can anticipate.

Our solution: Health Research Platform

We built the Health Research Platform for hospitals, research institutes and laboratories sitting on decades of their own clinical data — to put it to work now, and to be ready for EHDS. See the solution.