Unstructured discharge notes in Electronic Health Records (EHRs) often carry signal complementary to structured medical codes, holding patient-specific evidence that standardized cohort-level codes alone cannot capture. However, this evidence in notes is frequently buried in lengthy, noisy text that is not intentionally written with any specific clinical prediction in mind. Summarization is an obvious mitigation, but generic summaries, tuned for fluency rather than the outcome, routinely omit decisive evidence while retaining plausible but uninformative detail. To this end, we propose RASPER, a Reward-Aligned Summarizer for Prediction in EHR, that optimizes note summarization directly against the downstream clinical task. RASPER employs a tunable LLM-based summarizer to extract task-relevant evidence from discharge notes and trains it via reinforcement learning from prediction feedback, using a reward derived from the downstream predictor's loss. To ground the summarizer, a longitudinal encoder converts structured codes into soft prompts that incorporate each patient's clinical context into note summarization. By rewarding the quality of the resulting multimodal prediction, RASPER encourages the summarizer to retain patient-specific evidence that complements, rather than duplicates, information captured by structured codes. RASPER consistently outperforms strong baselines on both readmission prediction and medication recommendation across MIMIC-III and MIMIC-IV.
Figures & tables
Figure 1: Overview of the proposed RASPER framework. A tunable LLM summarizer transforms discharge notes into standardized visit-level summaries, which are concatenated into a patient-level summary. An LLM-aware predictor combines this patient-level summary with soft prompt embeddings from structured EHR codes for outcome prediction. The summarizer is first initialized with supervised fine-tuning and then optimized with RLPF using prediction-loss-derived reward within the PPO objective.
Figure 2: Prompt used by the tunable LLM summarizer to extract relevant information from discharge notes.
Component
Hyperparameter / Symbol
Value
Soft prompt encoder
P
8
LLM hidden size
D
2304
Max Summary length
Ci
512
Reward normalization
ϵ
1×10−8
Reward clipping
rmax
2.0
PPO clipping
ϵPPO
0.2
Table 2: Hyperparameters used in the proposed framework.
MIMIC-III
MIMIC-IV
Model
Readmission
Medication
Readmission
Medication
AUROC
PRAUC
F1
Jaccard
AUROC
PRAUC
F1
Jaccard
DL-based Methods
Deepr Nguyen et al. (2017)
0.671
0.702
0.255
0.160
0.621
0.651
0.228
0.189
RETAIN Choi et al. (2016)
0.657
0.678
0.253
0.156
0.673
0.698
0.349
0.255
GRASP Zhang et al. (2021)
0.618
0.635
0.153
0.096
0.535
0.578
0.267
0.195
Table 3: Baseline performance comparison on MIMIC-III and MIMIC-IV. Best results are in bold, and underlined results denote the second-best results that are statistically lower at the 95% confidence level.
MIMIC-III
MIMIC-IV
Model Variant
Readmission
Medication
Readmission
Medication
AUROC
PRAUC
F1
Jaccard
AUROC
PRAUC
F1
Jaccard
RASPER w/o EHR encoder
0.661
0.698
0.288
0.187
0.685
0.704
0.353
0.258
RASPER w/o Summarization
0.674
0.712
0.298
0.191
0.691
0.713
0.363
0.262
RASPER w/o RL (GPT only)
0.689
0.725
0.309
0.201
0.708
0.723
0.371
0.272
RASPER w/o RL (SFT only)
0.684
0.721
0.305
0.198
0.701
0.718
0.367
0.269
Table 4: Ablation studies on MIMIC-III and MIMIC-IV. Best results are in bold, and underlined results denote the second-best results that are statistically lower at the 95% confidence level.
Figure 3: Performance comparison across different EHR models on MIMIC-III and MIMIC-IV.
MIMIC-III
MIMIC-IV
Backbone
Readmission
Medication
Readmission
Medication
AUROC
PRAUC
F1
Jaccard
AUROC
PRAUC
F1
Jaccard
Llama-3.2-1B
0.701
0.734
0.321
0.214
0.722
0.737
0.384
0.278
Qwen2.5-0.5B
0.697
0.730
0.315
0.209
0.716
0.731
0.379
0.276
Gemma-2-2B
0.704
0.738
0.326
0.218
0.725
0.742
0.388
0.282
Table 5: Comparison of LLM backbones on MIMIC-III and MIMIC-IV across two tasks.
Appendix figures & tables1 asset
Supplementary material from the paper’s appendix.
Appendix
Model
Time (s)
Deepr
0.09
RETAIN
0.06
GRASP
0.17
AdaCare
0.05
StageNet
0.11
Adore
0.21
Appendix
Table 6: Inference time comparison for a batch of 12 patients.
Predicting future clinical outcomes from electronic health records (EHR) remains challenging due to the complexity and heterogeneity of patient data. LLMs have shown strong potential for such predictive tasks, yet existing approaches mainly focus on enhancing medical knowledge through distillation or RAG while relying on the model's internal ability to interpret contextual information. In this work, we present ReMedi (Reasoner for Medical Clinical Prediction), a framework for improving clinical outcome prediction from EHR. ReMedi generates rationale-answer pairs using a challenging sample regeneration mechanism for complex clinical questions, which leverages ground-truth answers as hints to enhance reasoning for further fine-tuning and preference tuning. ReMedi integrates ground-truth outcome guidance into the preference data construction loop, regenerating rationale-answer variants. By tuning on these rationale-answer pairs, the model improves its predictive performance. Experiments on multiple EHR prediction tasks demonstrate substantial gains of up to 19.9 percent over state-of-the-art baselines in terms of F1 score, underscoring ReMedi's effectiveness in real-world clinical prediction.
Yushi Cao, Yiming Chen, Hongchao Jiang +2
ASUS Intelligent Cloud Services (AICS), Singapore · NTU Artificial Intelligence Center of Research Excellence (NTU AI-CoRE), Taiwan · National University of Singapore, Singapore
Recent approaches to 30-day hospital readmission prediction rely on pre-trained language models applied to discharge summaries. Although these methods achieve strong performance, they depend on the availability of clinical notes, incur substantial computational costs, and yield representations that lack interpretability. We propose a knowledge-enriched feature representation that augments structured Electronic Health Record (EHR) data with four medical knowledge sources: disease ontology mapping, procedure classification, drug ingredient vocabulary, and organ system laboratory aggregation, without using clinical notes. Each feature dimension corresponds to a named clinical concept, yielding a sparse and interpretable patient representation. The approach is evaluated with six classifiers on a MIMIC-IV v2.2 cohort. Under 20-fold cross-validation, the best configuration achieves an AUROC of 0.743. This performance is comparable to that of previously reported methods on this dataset, including both those using only structured data and those incorporating clinical notes, while requiring considerably less computational cost. Interpretability analysis shows that demographics, organ system labs, drug ingredient features, and first-level ontology disease categories drive prediction, while deeper hierarchy levels contribute negligibly. These findings indicate that knowledge-enriched structured features offer a competitive and efficient alternative to embeddings from clinical notes for 30-day readmission prediction.
Large Language Models (LLMs) have shown strong promise for mining Electronic Health Records (EHRs) by reasoning over longitudinal clinical information to capture context-rich patient trajectories. However, leveraging LLMs for structured EHRs (e.g., standardized diagnosis and medication codes) presents two key challenges. First, translating time-stamped EHR sequences into plain text can obscure both temporal structure and code identities, weakening the ability to capture code co-occurrence and longitudinal regularities. Second, unlike cohort-trained predictive models that learn a shared, task-aligned representation space across patients, LLMs are often applied in a case-isolated inference setting where each patient is processed independently without leveraging population-level patterns. To address these challenges, we introduce RePrompT, a time-aware LLM framework that integrates structured EHR encoders through prompt tuning, without modifying underlying architectures. Specifically, RePrompT recurrently incorporates latent states from prior visits to preserve longitudinal information, and injects population-level information through trainable prompt tokens derived from a cohort-trained, task-aligned EHR encoder. Experiments on MIMIC-III and MIMIC-IV demonstrate that RePrompT consistently outperforms both EHR-based and LLM-based baselines across multiple clinical prediction tasks.