Why a digital project in healthcare is not run like any other
IT expertise in healthcare is aimed at organisations whose activity never stops: care facilities, social and medico-social care organisations, laboratories, home care providers. A digital project there faces three realities at once. The data processed is among the most protected by law. The information system is made up of many specialised software packages that communicate poorly. And the users are caregivers: every minute spent in front of a screen is taken from time spent with a patient or a resident.
These three realities explain why technically successful projects fail in use. A tool that is accurate but slow to fill in gets bypassed. A data warehouse that is rich but poorly secured cannot be opened to those who need it. An algorithm that performs well but is unexplained is not followed by practitioners.
Health data: what the GDPR, HDS certification and the AI Act require
The GDPR and its Article 9. Regulation (EU) 2016/679 classifies data concerning health among the special categories of data. Processing it is prohibited in principle, except in the cases listed in paragraph 2 of Article 9. Every project therefore starts with two questions: on what basis does the processing rest, and is the impact assessment provided for in Article 35 required?
HDS certification. Article L. 1111-8 of the French Public Health Code governs the hosting of personal health data carried out on behalf of those who produced or collected them, or on behalf of the patient. A host on digital media must hold Health Data Hosting certification. The practical consequence: choosing a host, a cloud provider or an online software service starts with reading its certificate and the activities it covers.
The AI Act. Regulation (EU) 2024/1689 classifies AI systems by level of risk. An AI system that is a component of a medical device, or is itself a medical device, may fall into the high-risk category: qualification must be checked case by case with your regulatory teams. Our entry on the AI Act presents the overall logic of the text.
These frameworks are not dealt with at the end of the project. They determine the architecture (where the data is, who accesses it, what is logged) and are documented as the design progresses.
Four typical workstreams
Choosing and organising hosting. Classifying applications by the sensitivity of their data, checking the certification scope of providers, deciding what stays on premises. Our sovereign cloud offering deals with these trade-offs between trusted cloud, public cloud and in-house hosting.
Making identity and exchanges reliable. A given patient must exist only once, and their information must move from one software package to another without re-entry. This requires identity matching rules, duplicate handling and interfaces based on exchange standards such as HL7 FHIR. It is a matter of data quality before it is a technical matter.
Opening data up for management and research. A health data warehouse serves activity analysis, quality of care and research projects. Its value depends on clear rules: who accesses what, in what form (identifying, pseudonymised or aggregated data), for what purpose.
Putting AI at the service of teams. Help with writing reports, searching protocols, scheduling, interpretation support: use cases are numerous and carry very different levels of risk. AI governance serves to classify them and decide which to pursue first.
The field first: running the project with caregivers
A healthcare project is won or lost on the wards. Before writing specifications, we observe the actual work: the start of a shift, a handover, a round. What caregivers do with the current tool, including what they work around, says more than a requirements document.
Deployment follows the same logic. A volunteer pilot ward, champions chosen from among the caregivers, short training built into schedules rather than added to the workload. Operation in degraded mode is part of the design: what does the night team do when the application stops responding? These dimensions fall under change management, which we build into the project from the framing stage.
Adoption is measured through observable signs: the share of entries made at the point of care rather than at the end of the shift, the continued use of paper, change requests coming from the teams themselves.
IT expertise in healthcare: how we work with you
Our approach follows four stages: frame, prove, industrialise, operate. In healthcare, framing brings together management, caregiver representatives, the security officer and the data protection officer. Proof is obtained in a real ward, on a limited scope. Industrialisation adds what a pilot does not show: performance under load, access traceability, recovery after an incident.
When off-the-shelf software does not cover a need (an interface between two systems, a mobile app for teams), our consultants contribute to building it through custom software development, within your teams, with the same security requirements as the rest of the system. To review your project, contact us.