Related Experiment Video
Updated: Jun 9, 2026
05:37
An R-Based Landscape Validation of a Competing Risk Model
Published on: September 16, 2022
Comparison of methods to handle missing values in a continuous index test in a diagnostic accuracy study - a
Katharina Stahlmann1, Bastiaan Kellerhuis2,3, Johannes B Reitsma2
1Institute of Medical Biometry and Epidemiology, University Medical Center Hamburg Eppendorf, Hamburg, Germany. k.stahlmann@uke.de.
BMC Medical Research Methodology
|May 27, 2025
Summary
Complete case analysis (CCA) and standard multiple imputation (MI) methods are recommended for missing data in diagnostic accuracy studies with missing completely at random (MCAR) data. Other methods show promise with larger sample sizes or higher correlations.
Area of Science:
- Medical Statistics
- Diagnostic Test Evaluation
- Biostatistics
Background:
- Missing values in index tests can bias diagnostic accuracy studies.
- Complete case analysis (CCA) and single imputation are common but potentially flawed methods.
- Simulation studies are crucial for evaluating methods handling missing data.
Purpose of the Study:
- To compare the performance of various methods for estimating the area under the curve (AUC) of a continuous index test with missing values.
- To assess the impact of different missing data mechanisms (MCAR, MAR, MNAR) on AUC estimation.
- To provide recommendations for handling missing index test data in diagnostic accuracy studies.
Main Methods:
- Simulated data for diagnostic accuracy studies with varying sample sizes, prevalences, and covariate correlations.
- Induced missing values in the continuous index test under different proportions and missingness mechanisms.
- Compared seven methods (including multiple imputation (MI), empirical likelihood, inverse probability weighting) against CCA for AUC estimation.
Main Results:
- CCA performed well with small sample sizes under missing completely at random (MCAR).
- All methods performed well with large sample sizes.
- Augmented inverse probability weighting and standard MI methods were effective under specific conditions (higher prevalence/larger sample size).
- Most methods showed bias under missing not at random (MNAR), especially with low correlation or small sample size/prevalence.
Conclusions:
- Standard MI and CCA are recommended for MCAR data with small and large sample sizes, respectively.
- CCA limitations should be discussed for small sample sizes with missing at random (MAR) or MNAR data.
- Augmented inverse probability weighting and MI are viable alternatives with increased sample size and/or correlation.
- All evaluated methods were biased under MNAR with low correlation.
Related Concept Videos
01:24Kaplan-Meier Approach
The Kaplan-Meier estimator is a non-parametric method used to estimate the survival function from time-to-event data. In medical research, it is frequently employed to measure the proportion of patients surviving for a certain period after treatment. This estimator is fundamental in analyzing time-to-event data, making it indispensable in clinical trials, epidemiological studies, and reliability engineering. By estimating survival probabilities, researchers can evaluate treatment effectiveness,...
01:20Comparing the Survival Analysis of Two or More Groups
Survival analysis is a cornerstone of medical research, used to evaluate the time until an event of interest occurs, such as death, disease recurrence, or recovery. Unlike standard statistical methods, survival analysis is particularly adept at handling censored data—instances where the event has not occurred for some participants by the end of the study or remains unobserved. To address these unique challenges, specialized techniques like the Kaplan-Meier estimator, log-rank test, and Cox...
01:09Censoring Survival Data
Survival analysis is a statistical method used to analyze time-to-event data, often employed in fields such as medicine, engineering, and social sciences. One of the key challenges in survival analysis is dealing with incomplete data, a phenomenon known as "censoring." Censoring occurs when the event of interest (such as death, relapse, or system failure) has not occurred for some individuals by the end of the study period or is otherwise unobservable, and it might have many different reasons...
01:13Accuracy and Errors in Hypothesis Testing
Hypothesis testing is a fundamental statistical tool that begins with the assumption that the null hypothesis H0 is true. During this process, two types of errors can occur: Type I and Type II. A Type I error refers to the incorrect rejection of a true null hypothesis, while a Type II error involves the failure to reject a false null hypothesis.
In hypothesis testing, the probability of making a Type I error, denoted as α, is commonly set at 0.05. This significance level indicates a 5% chance...
In hypothesis testing, the probability of making a Type I error, denoted as α, is commonly set at 0.05. This significance level indicates a 5% chance...
01:19Expected Frequencies in Goodness-of-Fit Tests
A goodness-of-fit test is conducted to determine whether the observed frequency values are statistically similar to the frequencies expected for the dataset. Suppose the expected frequencies for a dataset are equal such as when predicting the frequency of any number appearing when casting a die. In that case, the expected frequency is the ratio of the total number of observations (n) to the number of categories (k).
01:19The Mantel-Cox Log-Rank Test
The Mantel-Cox log-rank test is a widely used statistical method for comparing the survival distributions of two groups. It tests whether a statistically significant difference exists in survival times between the groups without assuming a specific distribution for the survival data, making it a non-parametric test. This flexibility makes the log-rank test particularly valuable in medical research and other fields where the timing of an event, such as death or disease recurrence, is of interest.