Important Dates
- Competition window (AgentDS): October 18–26, 2025
AgentDS Healthcare Data Challenge @ IEEE ICHI 2026
The IEEE International Conference on Healthcare Informatics (ICHI) will co-host the AgentDS Healthcare Data Challenge, a shared task evaluating human–AI collaboration on realistic, privacy-preserving healthcare problems (e.g., 30-day readmission, ED cost forecasting, discharge readiness). The Healthcare track is part of the broader AgentDS competition spanning six domains (healthcare, commerce, insurance, retail banking, manufacturing, food production); teams participating in Healthcare are automatically enrolled in AgentDS competition.
Fast Facts
- What: Healthcare data challenge with three sub-challenges — 30-day readmission, ED cost forecasting, and discharge readiness.
- How evaluated: Transparent metrics with a public leaderboard during the competition window at agentds.org/leaderboard.
- Who can join: Students, researchers, and industry practitioners worldwide.
Tasks (Healthcare)
- 30-Day Readmission Prediction: Identify patients at risk for early readmission to guide transitional care.
- ED Cost Forecasting: Estimate episode-level costs to support capacity planning and budgeting.
- Discharge Readiness Assessment: Recommend readiness to discharge to reduce length of stay while maintaining safety. (Exact task definitions, metrics, and starter materials are provided on the AgentDS site agentds.org/domains/healthcare.)
Data
- Synthetic & privacy-free: Datasets are designed to mirror realistic hospital patterns while avoiding sensitive or identifiable patient information.
- Reproducible access: Uniform data packages and starter code are provided for all registered teams.
- Fair use: External data use is allowed only as specified in the rules; sharing challenge data outside your team is prohibited.
Participation & Registration
- Register (free): agentds.org/register
- Challenge details: agentds.org/domains/healthcare
- Support: Starter code, free compute resources (JupyterHub) are provided for registered participants.
Submission, Publication & Recognition
- Top-performing and methodologically strong teams may be invited to submit an extended abstract or short paper to an ICHI 2026 challenge/workshop venue.
- Standard peer review applies; invitation does not guarantee acceptance.
- If accepted, papers are included in ICHI proceedings and made available via IEEE Xplore in line with ICHI’s publication policy.
Rules of Conduct (Preview)
- Eligibility & teams: Open to global participants; team composition must be declared and fixed per the posted rules.
- Data use: Use only the provided datasets and any permitted external resources per rules; do not share raw data beyond your team.
- Reproducibility: Document preprocessing and training clearly; code sharing is encouraged (e.g., via a public repo after results freeze) to support reproducibility.
- Leaderboard etiquette: No test set leakage or manual tweaking on hidden labels; follow all evaluation protocols exactly.
- Integrity: Submissions must be your team’s original work; violations may result in disqualification and notification to organizers. (Full legal terms and enforcement details will be posted on the AgentDS site.)
Awards
- Certificates and recognition for top Healthcare teams; additional distinctions may be announced jointly by AgentDS and ICHI organizers.
- Prize: $500 for 1st, $200 for 2nd, $100 for 3rd, and $50 for 4th-10th.
Organizers & Contact
- AgentDS Organizing Committee (University of Minnesota & partners) jie@agentds.org in coordination with the IEEE ICHI 2026 data challenge co-chairs: Jie Ding and Yu Huang.
- Sponsorship/industry contact: saaristo@umn.edu
