Accounting for Selection Bias in Studies of Acute Cardiac Events.

Pubmed ID: 29801736

Journal: The Canadian journal of cardiology

Publication Date: June 1, 2018

Link: https://ac.els-cdn.com/S0828282X18300163/1-s2.0-S0828282X18300163-main.pdf?_tid=128cb0e2-a23a-4ad6-a971-28310067c246&acdnat=1541163204_775d96ae927618884adef30f1f093343&link_time=2024-12-12_10:02:25.141118

MeSH Terms: Humans, Male, Female, Aged, Risk Factors, United States, Cohort Studies, Middle Aged, Smoking, Risk Assessment, Hospitalization, Follow-Up Studies, Mortality, Obesity, Diabetes Mellitus, Dyslipidemias, Biomedical Research, Selection Bias, Emergency Medical Services, Acute Coronary Syndrome

Authors: Harper S, Banack HR, Kaufman JS

Cite As: Banack HR, Harper S, Kaufman JS. Accounting for Selection Bias in Studies of Acute Cardiac Events. Can J Cardiol 2018 Jun;34(6):709-716. Epub 2018 Jan 31.

Studies:

Abstract

BACKGROUND: In cardiovascular research, pre-hospital mortality represents an important potential source of selection bias. Inverse probability of censoring weights are a method to account for this source of bias. The objective of this article is to examine and correct for the influence of selection bias due to pre-hospital mortality on the relationship between cardiovascular risk factors and all-cause mortality after an acute cardiac event. METHODS: The relationship between the number of cardiovascular disease (CVD) risk factors (0-5; smoking status, diabetes, hypertension, dyslipidemia, and obesity) and all-cause mortality was examined using data from the Atherosclerosis Risk in Communities (ARIC) study. To illustrate the magnitude of selection bias, estimates from an unweighted generalized linear model with a log link and binomial distribution were compared with estimates from an inverse probability of censoring weighted model. RESULTS: In unweighted multivariable analyses the estimated risk ratio for mortality ranged from 1.09 (95% confidence interval [CI], 0.98-1.21) for 1 CVD risk factor to 1.95 (95% CI, 1.41-2.68) for 5 CVD risk factors. In the inverse probability of censoring weights weighted analyses, the risk ratios ranged from 1.14 (95% CI, 0.94-1.39) to 4.23 (95% CI, 2.69-6.66). CONCLUSION: Estimates from the inverse probability of censoring weighted model were substantially greater than unweighted, adjusted estimates across all risk factor categories. This shows the magnitude of selection bias due to pre-hospital mortality and effect on estimates of the effect of CVD risk factors on mortality. Moreover, the results highlight the utility of using this method to address a common form of bias in cardiovascular research.