Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jun 11, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Evaluating Large Language Models for Automated Data Cleaning and Feature Engineering in Clinical Datasets.

Shimul A Babli1, Pulkit Gairola2,3, Chukwuemeka E Ogbu4

  • 1Internal Medicine, Islami Bank Medical College and Hospital, Rajshahi, BGD.

Cureus
|June 10, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Association Between Low Vitamin D Levels and Depression: A Systematic Review and Meta-Analysis.

Cureus·2026
Same author

The Role of Inflammatory Markers as Predictors of Response to Antidepressants in Major Depressive Disorder: A Meta-Analysis.

Cureus·2026
Same author

Association Between Smokeless Tobacco Use and Cardiovascular Disease Outcomes: A Systematic Review and Meta-Analysis With Subgroup Evaluation of Nawar-Type Products.

Cureus·2026
Same author

Physiological Characterisation of Severe Heart Failure Using Long-Duration Ambulatory ECG: A Retrospective Exploratory Analysis.

Cureus·2026
Same author

GLP-1 Receptor Agonists in the Management of Post-bariatric Weight Regain and Dysglycemia: A Systematic Review and Meta-Analysis.

Cureus·2026
Same author

Biological aging across the metabolic dysfunction-associated steatotic liver disease spectrum: A systematic review.

iLIVER·2026

Large language models (LLMs) effectively automate clinical data preprocessing, outperforming traditional methods for data cleaning and feature engineering in electronic health records (EHRs). Domain expert oversight is crucial for ensuring data integrity and clinical plausibility.

Area of Science:

  • Clinical Informatics
  • Biomedical Data Science
  • Artificial Intelligence in Healthcare

Background:

  • Electronic health records (EHRs) present significant data quality challenges (missing values, outliers, inconsistencies) for clinical research and machine learning.
  • Traditional data cleaning pipelines require extensive manual effort and domain expertise.
  • Large language models (LLMs) show promise for automating complex tasks but their application in structured clinical data preprocessing is underexplored.

Purpose of the Study:

  • To evaluate the efficacy of LLMs in automating structured data cleaning and feature engineering for clinical datasets.
  • To compare LLM performance against traditional rule-based pipelines and assess downstream predictive task performance.

Main Methods:

  • Three LLMs (GPT-4, Claude 3.5 Sonnet, Gemini 1.5 Pro) were assessed on five data cleaning and two feature engineering tasks.
Keywords:
clinical data cleaningelectronic health records (ehrs)feature engineeringlarge language models (llms)machine learning in healthcare

More Related Videos

Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
05:47

Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems

Published on: June 13, 2025

Related Experiment Videos

Last Updated: Jun 11, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems
05:47

Evidence-based Knowledge Synthesis and Hypothesis Validation: Navigating Biomedical Knowledge Bases via Explainable AI and Agentic Systems

Published on: June 13, 2025

  • Publicly available clinical datasets (MIMIC-IV, eICU, NHANES) were utilized.
  • LLM-generated scripts were benchmarked against a rule-based pipeline; downstream performance was evaluated using XGBoost and logistic regression for in-hospital mortality prediction (AUROC, F1 score).

    • Main Results:

    • Claude 3.5 Sonnet achieved the highest data cleaning F1 score (0.90), surpassing GPT-4 (0.89), Gemini 1.5 Pro (0.85), and the rule-based baseline (0.77).
    • Data preprocessed by Claude 3.5 Sonnet yielded the best XGBoost AUROC (0.851) for mortality prediction on MIMIC-IV, outperforming other LLMs, rule-based cleaning, and no cleaning.
    • LLM-engineered features improved AUROC by 0.015-0.025 over manually constructed features.

    Conclusions:

    • LLMs offer substantial automation for clinical data cleaning and feature engineering, achieving performance on par with or better than manual methods.
    • Despite high performance, human expert oversight is essential to validate clinical relevance and prevent undetected data errors.