Ascent: An Agentic System over the Model Context Protocol for Real-World Clinical Data Analysis

Angelo Ziletti; Leonardo D'Ambrosi; Melanie Tuchardt; Tim Kondziella

cs.AISep 21, 2026

Ascent: An Agentic System over the Model Context Protocol for Real-World Clinical Data Analysis

Authors: Angelo Ziletti, Leonardo D'Ambrosi, Melanie Tuchardt, Tim Kondziella

Organizations: Bayer AG, Germany

Abstract

Answering epidemiological questions from real-world clinical data requires medical coding, schema-aware SQL, and validation of implicit choices about populations, denominators, and time. We present Ascent, an agentic system that exposes medical coding, question answering, and cohort analysis through a shared Model Context Protocol tool surface for standardized and native schemas. We introduce EpiTrap, a dataset testing whether systems avoid recognized pharmacoepidemiological errors, and compare a fixed pipeline with agents across models and orchestrators. With capable models, agents improve accuracy over the fixed pipeline by an average of 27 and 20 percentage points on native and standardized schemas, respectively. These gains require more tool calls and longer runtimes. Experience from real projects highlights the system's value for feasibility assessment, diagnostic iteration, and expert-guided analysis.

Figures & tables

Panel of Figure 1 — Figure 1: Ascent architecture. MCP clients and model-agnostic orchestrators access shared analytical services and governed data stores through a common tool surface.

Figure 2: Example Ascent session through a chat interface. After loading an appropriate skill, the system surfaces consequential choices about the prevalence definition and target database.

Question (system sees only this)	The recognized error the reasoning must avoid
Count patients with severe hyperglycemia (blood glucose above 250 mg/dL) at any measurement.	Unit harmonization. Glucose is recorded in mixed units, so applying one threshold to raw values misclassifies measurements. Correct: harmonize units, remove implausible values, then apply the threshold.
Estimate the one-year incidence of gout among adults with at least one year of continuous enrollment.	Prevalence–incidence conflation. A pre-existing gout diagnosis is not an incident case. Correct: exclude prevalent cases via a baseline washout.
Measure the cervical-cancer screening rate among adult women.	Denominator ineligibility. Women with documented absence of the cervix (e.g. following total hysterectomy) are ineligible for routine screening. Correct: exclude them from the denominator using prior diagnoses and procedures.

Table 1: Representative EpiTrap questions , spanning data-quality, estimand-definition, and denominator-eligibility errors. The system sees only the left column; it passes if its reasoning avoids the error in the right, whatever path it takes.

Baseline: no agent loop, no framework
	Native, $n{=}156$			OMOP, $n{=}153$
Configuration	Acc.	Time	Calls	Acc.	Time	Calls
mech. pipeline	50%	231 s	3.0	49%	180 s	3.0
Claude Agent SDK
Sonnet 5, plain	75%	565 s	9.4	75%	275 s	8.8
$\hookrightarrow$ no scaffold	76%	197 s	16.8	84%	177 s	18.6

Table 2: EpiTrap results across two data models, native Optum EHR and OMOP-CDM. Acc. is rubric-based accuracy; Time and Calls are per-question means. “No scaffold” rows withhold the specialized text-to-SQL tools, leaving the agent to write SQL itself.

Data Analysis Agents

Data analysis agents use language models and tools to explore datasets, run analyses, and produce answers or reports grounded in evidence.

See all 106 papers

Large Language Models for Structured Clinical Data Analysis: Dual-Agent Grounding and Validation
Sep 28, 2026Erfan D. Dehkalani, Seetha Shankaran, Abbot R. Laptook +3Multi-Agent LLM Systems LLM Grounding
Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital, and Harvard Medical School, Boston, MA, USA · Department of Pediatrics, Wayne State University School of Medicine, Detroit, MI, USA · Department of Pediatrics, Women & Infants Hospital of Rhode Island, and Warren Alpert Medical School of Brown University, Providence, RI, USA +1
ClinLens: Towards Long-Horizon Coding Agents for Longitudinal Multimodal Clinical Data Science
Jul 28, 2026Yuan Zhu, Ethan B. Liu, Frank Nie +1Long-Horizon Agent Evaluation AI Agent Benchmarks
Shandong University, China
PhysicianBench: Evaluating LLM Agents in Real-World EHR Environments
May 4, 2026Ruoqi Liu, Imran Q. Mohiuddin, Austin J. Schoeffler +10AI Agent Benchmarks Electronic Health Records
Stanford University

Sep 28, 2026J/K move · Enter open · S save

Large Language Models for Structured Clinical Data Analysis: Dual-Agent Grounding and Validation

Erfan D. Dehkalani, Seetha Shankaran, Abbot R. Laptook, C. Michael Cotten+2

Fetal-Neonatal Neuroimaging and Developmental Science Center, Boston Children’s Hospital, and Harvard Medical School, Boston, MA, USA · Department of Pediatrics, Wayne State University School of Medicine, Detroit, MI, USA · Department of Pediatrics, Women & Infants Hospital of Rhode Island, and Warren Alpert Medical School of Brown University, Providence, RI, USA+1

Multi-Agent LLM Systems LLM Grounding LLM Agent Verification Text-to-SQL

Open paper AskPDF arXiv

Objective: To develop and characterize CLEAR-Med, a dual-agent framework for natural-language analysis of structured clinical data that separates SQL-based invocation from independent validation. Methods: CLEAR-Med uses one agent to translate a question into executable Structured Query Language (SQL), retain the executed query and database result, and produce a draft. Deterministic checks and a separately invoked cross-provider Validation Agent then accept the draft, request one bounded repair, or abstain. We formalized the system as a bounded selective pipeline and evaluated CLEAR-Med's configuration and scalability, and the Invocation Agent's accuracy and consistency on a 25-query development benchmark, using a harmonized 21-site neonatal hypoxic-ischemic encephalopathy table containing 532 de-identified infant records and approximately 1,300 variables. Results: CLEAR-Med completed all six nominal scalability configurations, including 500x1300. Across 25 development-benchmark queries repeated five times, the Invocation Agent answered 83 of 125 responses correctly (66.4%; query-cluster bootstrap 95% CI, 48.0-83.2%), compared with 15 of 125 (12.0%; 95% CI, 3.2-22.4%) for the ungrounded ChatGPT baseline, a paired improvement of 54.4 percentage points (95% CI, 36.8-72.0%). Conclusion: CLEAR-Med provides a general architecture for traceable analysis of structured clinical data: numerical claims remain linked to executed SQL, and unresolved cases can fail closed. The reported experiments characterize CLEAR-Med's configuration and scalability and the Invocation Agent's accuracy, while the formal analysis establishes the encoded-property guarantee of the complete control flow; a prospective full-pipeline evaluation of the validation and abstention stages is the next stage of this work.

Abstract

Data Analysis Agents

Explore similar work

Large Language Models for Structured Clinical Data Analysis: Dual-Agent Grounding and Validation

ClinLens: Towards Long-Horizon Coding Agents for Longitudinal Multimodal Clinical Data Science

PhysicianBench: Evaluating LLM Agents in Real-World EHR Environments