AIDAVA Documents

From Research Wiki

All available documents from the AIDAVA project, with summaries. Downloaded from aidava.eu/results.

Fact Sheet (PDF)[edit]

Download PDF

Key facts from official fact sheet:

  • Acronym: AIDAVA
  • Full title: AI powered Data Curation & Publishing Virtual Assistant
  • Programme: Horizon EUROPE / HORIZON-HLTH-2021-TOOL-06-03
  • Contract number: 101057062
  • Duration: 48 months (01/09/2022 - 31/08/2026)
  • Project funding: EUR 7,720,615
  • Coordinator: Universiteit Maastricht, Remzi Celebi — remzi.celebi@maastrichtuniversity.nl

Partner list from fact sheet:

  • Austria: Medizinische Universitaet Graz
  • Belgium: B!loba, European Cancer Patient Coalition, European Heart Network AISBL, Katholieke Universiteit Leuven, The European Institute for Innovation through Health Data
  • Bulgaria: Sirma AI EAD (note: listed as Sirma AI, which owns Ontotext)
  • Estonia: Sihtasutus Phoja-Eesti Regionaalhaigla
  • Germany: Averbis, EURICE
  • Netherlands: Universiteit Maastricht
  • Romania: Egnosis SRL
  • Spain: Data for Patients

Note: Bulgaria partner listed as "Sirma AI EAD" in the fact sheet — Sirma AI is the parent company of Ontotext.

Project Overview (PDF)[edit]

Download PDF

Architecture:

  • Backend (Automation): Orchestration of different technologies to maximize automation in cleaning personal health data
  • Frontend: User involvement — dialogue with user, tailored to individual skills, with explanations
  • Output: Curated Data IDENTIFIED (Personal Health Record) + Curated Data ANONYMIZED (Population Patient Registries)

This dual output — identified personal records for the patient AND anonymized population registries for researchers — is central to the "curate once, reuse many times" vision.

Periodic Flash — November 2022[edit]

Download PDF

Project kick-off, October 2022 in Maastricht.

Two use cases defined:

  1. EU Breast cancer registry maintenance across 3 university hospitals in the consortium
  2. Cardiac patients personal longitudinal record to identify risk of cardiac failure

Two paradigm shifts:

  1. Patients are in control of their data through a "data intermediary" (defined in EU Data Governance Act)
  2. Patients are supported by an intelligent virtual assistant

Patients from EHN and ECPC onboarded November 25, 2022 to start developing requirements.

Periodic Flash — March 2023[edit]

Download PDF

Patient requirements workshop completed, February 2023.

Patient incentives identified:

  • Main incentive: ensure treating physician has smooth access to complete medical record
  • Happy to share data "in a controlled way"

Patient requirements for data sharing:

  1. Want to know who will access data and optionally obscure some information
  2. Want easy consent mechanism for all parties
  3. Want to know who uses data and for what purpose
  4. Reward should be envisioned for commercial use of shared data
  5. Concern about data donation — do they lose control?

Timeline:

  • First release of system: summer 2024
  • Second version: mid 2026

Periodic Flash — September 2023[edit]

Download PDF

First year completed successfully.

8 Patient Consultants from ECPC and EHN participated in:

  • Online meetings, face-to-face workshop
  • Definition of use cases
  • Identification of personas and key features (health/digital literacy levels)
  • Specification of business requirements
  • Assessment of evaluation schedule and documentation

AIDAVA will continue working with Patient Consultants on evaluation and testing of preliminary prototypes before site patient testing in summer 2024.

Periodic Flash — November 2024[edit]

Download PDF

G1 prototype tested May 2024. Evaluation from July 2024 in three countries.

Test sites:

  • North Estonian Medical Center (Estonia)
  • Graz University Medical Center (Austria)
  • Maastro Clinics, Maastricht University Medical Center (Netherlands)

Challenges:

  • Ethics approvals time-consuming
  • Finding patients interested in testing was difficult
  • Doctors on summer vacation delayed recruitment

Patient feedback on G1:

  1. Usability on computer and phone was good
  2. System questions not formulated appropriately
  3. Some functionalities didn't work properly, others needed adding
  4. Benefit was not evident for many patients — needs better explanation (e.g. value of IPS and clean data)

Next steps:

  • Enriching prototype with powerful AI solutions
  • Working with patient organisations to clarify benefits
  • G2 prototype testing planned for early 2026

Periodic Flash — March 2025 (Latest)[edit]

Download PDF

G1 evaluation report finalized January 2025. Big thank you to patients in Estonia, Austria, Netherlands.

Evaluation results:

  • 83 patients recruited; 70 completed
  • Overall performance of AIDAVA moderately acceptable by patients
  • Explanations are suboptimal
  • Approximately 45% of documents curated automatically
  • Average time to curate a document is too high (20 minutes)
  • Patient showed interest in AIDAVA-like systems mainly in countries without a central health data portal

What they're working on:

  1. In-depth analysis of users' feedback and improvements including features that improve benefits for patients
  2. Validation of business requirements, identification of ones met, prioritization of ones to implement
  3. Development and testing of prioritized requirements
  4. Amendment to study protocol and Ethics Committee Approval
  5. External communication on results

Timeline:

  • G2 delivery: end 2025
  • G2 testing: early 2026
  • Project end: August 2026

Key Insights from Documents[edit]

What works:

  • Usability of the system is good (computer and phone)
  • Patients are interested in maintaining their health records
  • 45% automation rate is a solid baseline for a first prototype

What needs improvement:

  • Explanation system (why should I do this?)
  • Curation time (20 min/document is too slow)
  • Clearer benefits for patients (especially in countries with existing central portals)
  • Question formulation in the system

For competitive analysis:

  • The project is struggling with patient engagement and perceived value
  • 45% automation means 55% still requires manual work
  • 20 min/document is a significant bottleneck
  • Countries with existing health data portals (like Estonia) show less patient interest
  • G2 will add more AI but won't be tested until early 2026

See Also[edit]