Large language models (LLMs) show strong reasoning abilities, but full retraining for the medical domain is often infeasible because of lacking data or compute resources. We present DeepICD-R1, a framework for efficient medical reasoning fine-tuning that unites hierarchical rewards with distilled supervision. We reformulate ICD-10-CM prediction as a reinforcement learning problem and design a hierarchical outcome-based reward that reflects the ICD code structure across chapter, category, and full-code levels. In parallel, we publish a large-scale distilled dataset of over 90k reasoning traces derived from MIMIC-IV admission notes, integrating clinical validation and official coding guidelines. Fine-tuning smaller instruction-tuned LLMs with this data and GRPO reinforcement yields consistent gains in diagnostic accuracy and reasoning coherence. Extensive ablations confirm that hierarchical supervision and verifiable outcome rewards enable competitive, domain-specialized reasoning models without additional pretraining, providing a reproducible foundation for clinical NLP research. Keywords: Clinical NLP, Large Reasoning Model, GRPO, Supervised Fine-Tuning

LLM agents acting in structured environments fail in operational rather than conversational ways, and reliability depends on procedural knowledge of the environment. Prior self-improvement methods accumulate natural-language guidance without checking that each new item preserves previously correct behavior, so a note that fixes one trajectory can silently regress another. We introduce GRASP (Gated Regression-Aware Skill Proposer), which treats agent improvement as a sequence of edits to a bounded skill library, admitting each candidate only if it produces a net improvement on a balanced held-out probe under a hard regression budget. We evaluate GRASP across five base models (gpt-oss-120b, DeepSeek V4 Flash, Gemini 3.1 Flash Lite, GPT-4.1, GPT-5.4) on two FHIR-based clinical benchmarks. On MedAgentBench, GRASP lifts gpt-oss-120b from 40.6% to 88.8%, exceeds the strongest of five self-improvement baselines by 21.0 points, and improves every other base model by 17.2 to 40.3 points. Ablations attribute the gain to comparative proposal generation, the acceptance gate, and the hard regression budget rather than to skill writing itself, which without validation is no better than using no skills. The mechanism generalizes beyond the clinical domain, improving agents on three of four non-clinical environments and remaining flat only where the action space is open-ended. Frozen libraries transfer across models, where skills from a stronger model improve weaker executors beyond what they learn for themselves while the reverse does not, an asymmetry that no ungated baseline reproduces.

Clinical Decision Support Systems assist medical professionals in providing optimal care for patients.A prominent data source used for creating tasks for such systems is the Medical Information Mart for Intensive Care (MIMIC).MIMIC contains electronic health records (EHR) gathered in a tertiary hospital in the United States.The majority of past work is based on the third version of MIMIC, although the fourth is the most recent version.This new version, not only introduces more data into MIMIC, but also increases the variety of patients.While MIMIC-III is limited to intensive care units, MIMIC-IV also offers EHRs from the emergency department.In this work, we investigate how to adapt previous work to update clinical outcome prediction for MIMIC-IV.We revisit several established tasks, including prediction of diagnoses, procedures, length-of-stay, and also introduce a novel task: patient routing prediction.Furthermore, we quantitatively and qualitatively evaluate all tasks on several bio-medical transformer encoder models.Finally, we provide narratives for future research directions in the clinical outcome prediction domain. We make our source code publicly available to reproduce our experiments, data, and tasks.