AI
A Neurosymbolic Multimodal Framework for 30-Day Diabetic Readmission Prediction: Integrating Structured EHR Data, Dense Clinical Embeddings, and Knowledge-Grounded Symbolic Extraction
Publication: cs.AI updates on arXiv.org
Author: Una Suman Kumar Patro, Sibangi Subhadarsani, Narendra Badam
Source published: 20/11/2026, 05:00:00 UTC
Briefing created: 20/11/2026, 05:54:57 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 20/11/2026, 05:03:25 UTC.
Briefings
A neurosymbolic multimodal framework improves diabetic readmission prediction by integrating structured EHR data with dense clinical embeddings and knowledge-grounded symbolic extraction. Evaluated on the Diabetes 130-US Hospitals dataset (101,766 encounters), the model achieved the highest sensitivity (0.6667) and lowest Brier score (0.0977) among tested approaches, with Shapley additive explanations identifying prior utilization as the dominant risk factor.
Key points
- The study proposes a neurosymbolic multimodal framework fusing structured EHR data, dense clinical embeddings, and knowledge-grounded symbolic extraction to address unstructured Hinglish phrasing in clinical notes.
- On the Diabetes 130-US Hospitals dataset (101,766 encounters), the model achieved the highest sensitivity (0.6667) and lowest Brier score (0.0977) among tested approaches, with Shapley additive explanations identifying prior utilization as the dominant risk factor.
Context
Diabetes mellitus patients face substantial clinical and economic burdens from unplanned 30-day readmissions. Existing models typically rely solely on structured EHR data, ignoring free-text clinical notes that often contain code-mixed Hindi-English phrasing and negation, which general-purpose language models parse unreliably.
Limitations
The evaluation is limited to the public Diabetes 130-US Hospitals dataset (101,766 encounters) with a 80/20 split. Results are based on a single preprint (arXiv:2611.25236v1) and represent a single snapshot of performance on a specific cohort, with no validation on external data or clinical impact assessment.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 20/11/2026, 05:54:57 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeifpe7w7muumwpsolcemli37diyd3dhthcx4k2eyvwx6s3wmjlvbmu