GraftWatch
A dashboard for transplant coordinators that reads clinic notes, immunosuppressant levels, and lab trends as one stream, then surfaces which kidney recipients are drifting toward graft failure this month.
Post-transplant coordinators managing 80–300 kidney recipients each
- Plain-text serialization of EHR notes plus structured labs and medication levels into one risk stream
- Per-patient graft-failure risk score refreshed nightly with the three input sentences that moved the needle
- Outreach queue that ranks patients by both risk and overdue clinic visits
- Audit trail every coordinator can show at a QAPI committee meeting
A new arXiv paper shows a single text-serialized LLM outperforms the gradient-boosting system actually deployed at a German transplant center for graft failure prediction — so a transplant-coordinator first product is grounded in fresh published evidence, not a marketing claim.
Johns Hopkins publicly released an EMR-based graft-failure predictor in Nov 2025, multiple 2024-2025 ML papers target the problem, and transplant-coordinator burnout/workload benchmarking is an active research area — confirms real clinical demand.New Tool Predicts Graft Failure After Kidney Transplant ↗Phase-specific kidney graft failure prediction with machine learning ↗
Incumbents (Epic Phoenix, Oracle Health/Cerner Transplant) cover general transplant management rather than AI-driven graft-risk dashboards, and the Hopkins tool appears to be an internal EMR research artifact; however the arXiv authors include DNC Information Management GmbH, signaling a likely commercial productizer in the same niche.Epic Phoenix vs Oracle Health Transplant (Cerner) Comparison ↗Large Language Models as Unified Multimodal Learners for Clinical Prediction ↗
Healthcare SaaS in transplant (Epic, Cerner Transplant modules) commands enterprise pricing and buyers exist, but the addressable market is small (~250 US transplant centers) and healthcare procurement cycles are notoriously long, tempering upside.Transplant Management Selection Guide 2026 ↗
Graft failure in kidney transplantation is a permanent clinical problem and post-transplant surveillance will continue indefinitely; AI tooling for it is a durable, not trend-driven, market.
Build is hard: requires HIPAA-grade PHI handling from clinical notes, probable FDA SaMD classification for a risk-prediction dashboard, and deep Epic/Cerner integration — the ML is proven but productionizing is a regulated multi-year effort.