Clinical decision making improved by flexible guideline reasoning models
Latent Decision Flow Rewards for Clinical Guideline Reasoning
Machine LearningArtificial Intelligence
Summary
Clinical guidelines help doctors decide what treatments to give by following decision steps, but for each patient, several different decision paths can lead to the same right answer. The authors show that previous training methods for AI to follow these guidelines struggled with multiple paths and sometimes gave less accurate answers. They created a new method called GUIDER that trains AI models to generate and choose among many reasonable decision flows, improving both the correctness of answers and adherence to guidelines. This method was tested on several clinical datasets and outperformed previous approaches.
What this means in practice
- •For clinical ai developers: Build clinical AI tools that produce guideline-consistent decisions while allowing multiple valid reasoning paths to improve accuracy.
- •For healthcare data teams: Enhance patient case decision support systems by integrating flexible reasoning models that optimize guideline adherence and final decision correctness.
Authors
Gaurav Bhatt, Marinka Zitnik
Abstract
Clinical practice guidelines encode evidence-based recommendations as conditional decision flows, which makes them a natural source of process supervision for training reasoning models. Two properties of clinical reasoning make this supervision difficult to use. A decision flow consistent with a guideline does not by itself determine the correct answer for a given patient, which we call the path--answer gap, and several valid decision flows can support the same clinical decision, which we call path multiplicity. Current training approaches address parts of this problem. Answer rewards supervise the final decision, and process rewards computed against a reference flow supervise one valid decision flow per case. Supervised fine-tuning on annotated flows teaches guideline-consistent reasoning but constrains the model to one trajectory per case and can reduce answer accuracy. We introduce GUIDER, a method that jointly optimizes a decision flow controller and a reasoning policy. The controller learns to generate plausible decision flows from guideline-derived supervision, and the policy learns to generate a decision flow and a final answer for each patient case. GUIDER computes the process reward for the policy over several controller-generated flows, which accounts for path multiplicity, and gates this reward on answer correctness, which addresses the path--answer gap. Joint training allows the process supervision to evolve with the policy. On MedGUIDE, GUIDER with Qwen3-14B achieves 82.5 weighted accuracy and 73.5 guideline adherence, 3.3 and 3.1 points above the strongest clinical reinforcement learning baseline. With Qwen3-14B and Gemma4-12B backbones, GUIDER also improves over CRPO, MRPO, and FaithMed on the out-of-domain benchmarks AMEGA-LLM, RealMedQA, and CancerGUIDE.