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 Videos

Bias arising from missing data in predictive models.

Marc H Gorelick1

  • 1Department of Pediatrics, Section of Emergency Medicine, Medical College of Wisconsin, Milwaukee, WI, USA. mgorelic@mcw.edu

Journal of Clinical Epidemiology
|September 19, 2006
PubMed
Summary
This summary is machine-generated.

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 Age, Weight, and Dose and Clinical Response to Probiotics in Children with Acute Gastroenteritis.

The Journal of nutrition·2020
Same author

Factors Associated With Nonadherence in an Emergency Department-based Multicenter Randomized Clinical Trial of a Probiotic in Children With Acute Gastroenteritis.

Journal of pediatric gastroenterology and nutrition·2020
Same author

Perspectives on Urinary Tract Infection and Race.

JAMA pediatrics·2020
Same author

Lactobacillus rhamnosus GG versus Placebo for Acute Gastroenteritis in Children.

The New England journal of medicine·2018
Same author

The Pediatric Emergency Care Applied Research Network Registry: A Multicenter Electronic Health Record Registry of Pediatric Emergency Care.

Applied clinical informatics·2018
Same author

Randomised controlled trial of <i>Lactobacillus rhamnosus</i> (LGG) versus placebo in children presenting to the emergency department with acute gastroenteritis: the PECARN probiotic study protocol.

BMJ open·2017
Same journal

Harms Reporting Was Frequently Incomplete or Discordant in Biomedical Randomized Trials Published in 2023: A Meta-epidemiological Study.

Journal of clinical epidemiology·2026
Same journal

Using an Open Science Checklist in Grant Proposal Reviews to Predict Reproducibility of Funded Publications.

Journal of clinical epidemiology·2026
Same journal

A comparison of five statistical methods used to analyse longitudinal EORTC QLQ-C30 quality of life scores in randomised controlled trials: a simulation study.

Journal of clinical epidemiology·2026
Same journal

Sample Size Determination for Decision-centered Pragmatic Trials.

Journal of clinical epidemiology·2026
Same journal

Many multicenter randomized controlled trials do not account for center effect: a methodological review.

Journal of clinical epidemiology·2026
Same journal

Patient Acceptability of the Modified Zelen Approach to Randomized Trials - A Survey of the CAPS THA Cohort.

Journal of clinical epidemiology·2026
See all related articles

Handling missing data in predictive models significantly impacts results. Common methods like complete case analysis, assuming missingness as normal, or imputation can introduce bias in odds ratios and model performance.

Area of Science:

  • Statistics
  • Biostatistics
  • Health Informatics

Background:

  • Missing data is a common challenge in developing predictive models.
  • Inconsistent predictor variable measurement can compromise model validity.

Purpose of the Study:

  • To evaluate the impact of three common missing data handling techniques on predictive model outcomes.
  • To assess the bias introduced by different methods in estimating odds ratios and model discrimination.

Main Methods:

  • A Monte Carlo simulation study utilized simulated data.
  • Logistic regression predicted hospital admission using white blood cell count (WBC), fever, and procedures (PROC).
  • Simulations involved deleting WBC data (15-85%) under various missingness patterns, analyzed via complete case (CC), missing at normal (MAN), and imputation.

Related Experiment Videos

Main Results:

  • The MAN and imputation approaches showed substantial over/underestimation of WBC odds ratios (OR) and bias toward the null for fever OR.
  • CC analyses revealed bias toward the null for WBC OR, away from the null for PROC OR, and variable bias for fever OR.
  • All tested analytic approaches demonstrated significant bias in overall model discrimination estimates.

Conclusions:

  • Common methods for handling substantial missing data can lead to biased odds ratio estimates and reduced predictive model performance.
  • The validity of predictive models is affected by how predictor variables with inconsistent measurements are handled.