Back to Search
Start Over
Eliminating systematic bias from case-crossover designs.
- Source :
- Statistical Methods in Medical Research; Oct/Nov2019, Vol. 28 Issue 10/11, p3100-3111, 12p
- Publication Year :
- 2019
-
Abstract
- Case-crossover designs have been widely applied to epidemiological and medical investigations of associations between short-term exposures and risk of acute adverse health events. Much effort has been made in literature on understanding source of confounding and reducing systematic bias by reference-select strategies. In this paper, we explored the nature of bias in the ambi-directional and time-stratified case-crossover designs via simulation using actual air pollution data from urban Edmonton, Alberta, Canada. We further proposed a calibration approach for eliminating systematic bias in estimates (coefficient estimate, 95% confident interval, and p-value). Bias check for coefficient estimation, size check and power check for significance test were done via simulation experiments to show advantages of the calibrated case-crossover studies over the ones without calibration. An application was done to investigate associations between air pollutants and acute myocardial infarction hospitalizations in urban Edmonton. In conclusion, systematic bias in a case-crossover design is often unavoidable, leading to an obvious bias in the estimated effect and an unreliable p value in the significance test. The proposed calibration technique provides an efficient approach to eliminating systematic bias in a case-crossover study. [ABSTRACT FROM AUTHOR]
Details
- Language :
- English
- ISSN :
- 09622802
- Volume :
- 28
- Issue :
- 10/11
- Database :
- Complementary Index
- Journal :
- Statistical Methods in Medical Research
- Publication Type :
- Academic Journal
- Accession number :
- 138594950
- Full Text :
- https://doi.org/10.1177/0962280218797145