Patient-Similarity Cohort Reasoning in Clinical Text-to-SQL
This addresses the challenge of automating EHR analytics for healthcare professionals, but it is incremental as it builds on existing text-to-SQL methods with a new benchmark.
The paper tackled the problem of clinical text-to-SQL by introducing CLINSQL, a benchmark requiring reasoning over EHR data, and found that state-of-the-art models like GPT-5-mini achieved only 74.7% execution accuracy, indicating performance far from clinical reliability.
Real-world clinical text-to-SQL requires reasoning over heterogeneous EHR tables, temporal windows, and patient-similarity cohorts to produce executable queries. We introduce CLINSQL, a benchmark of 633 expert-annotated tasks on MIMIC-IV v3.1 that demands multi-table joins, clinically meaningful filters, and executable SQL. Solving CLINSQL entails navigating schema metadata and clinical coding systems, handling long contexts, and composing multi-step queries beyond traditional text-to-SQL. We evaluate 22 proprietary and open-source models under Chain-of-Thought self-refinement and use rubric-based SQL analysis with execution checks that prioritize critical clinical requirements. Despite recent advances, performance remains far from clinical reliability: on the test set, GPT-5-mini attains 74.7% execution score, DeepSeek-R1 leads open-source at 69.2% and Gemini-2.5-Pro drops from 85.5% on Easy to 67.2% on Hard. Progress on CLINSQL marks tangible advances toward clinically reliable text-to-SQL for real-world EHR analytics.