The lay abstract featured today (for Statistical Inference for Association Studies in the Presence of Binary Outcome Misclassification by Kimberly A. Hochstedler Webb and  Martin T. Wells) is from Statistics in Medicine with the full article now available to read here.

How to cite

Hochstedler Webb, K.A. and Wells, M.T. (2025), Statistical Inference for Association Studies in the Presence of Binary Outcome Misclassification. Statistics in Medicine, 44: e10316. https://doi.org/10.1002/sim.10316

Lay Abstract

Data used in healthcare and biomedical research is often subject to error. For example, patients may not recall all of their symptoms when responding to a survey or doctors may misdiagnose a patient in their medical records. Myocardial infarction (MI, or heart attack) diagnosis is one such condition that may be reported imperfectly in survey data and medical records. For researchers seeking to understand risk factors for MI, misdiagnosis and misreporting in study datasets may cause unpredictable bias in the results.

In this research paper, the authors developed statistical tools that account for errors in outcome variables in medical studies. These tools take advantage of the fact that researchers may have an idea of the factors that lead to an error in their dataset in the first place.  In the case of MI, one such factor is often patient gender. Because MI symptoms in women tend to be less “visible” than those in men, it is hypothesized that women are misdiagnosed at a greater rate than men.

Using this known mechanism for MI misdiagnosis, the authors demonstrated that the probability that a patient’s MI or lack of an MI was misdiagnosed can be estimated based on the patient’s gender. By accounting for potential errors in the data due to gender-based misdiagnosis, it is possible to analyze the impact of risk factors on true MI diagnosis. The result of this analysis is a more accurate method of studying risk factors for easily misdiagnosed or misclassified medical conditions.

Using a nationwide dataset, the authors applied their methods to study risk factors for MI and correct for potential gender-based misdiagnosis. With this more accurate analysis technique, they found that common risk factors for heart attacks, like smoking and lack of exercise, may be even more important than previously thought. In addition, they estimated rates of MI misdiagnosis across genders – estimating that as many as 40% of MI events in women are missed by doctors.