Related Experiment Videos
Can structural equation models inform blood pressure research?
1Division of Biostatistics and Epidemiology, Department of Community Medicine, University of Connecticut School of Medicine, Framington, CT 06030-6205, USA.
Objective:
To show how structural equation models might be used to better understand the ways in which risk factors influence blood pressure.
Methods:
Nine measurements on 2009 women and 1518 men for whom there was complete data both at time 1 and at time 8 of the Framingham Heart Study were used to test a hypothetical model of how risk factors such as age, obesity, smoking, vital capacity, and heart rate influence each other and blood pressure. The hypothetical model was translated into structural equations and tested against the data.
Results:
The hypothetical model fits the data for women at time 1 very well with a chi2=15.41 which, with 14 degrees of freedom, has P=0.32 and indicates there is no difference between the covariance structure generated by the hypothetical model and the covariance structure generated by the data. The same model was tested for women at time 8 and for men at times 1 and 8 also and fit almost as well. Age and percentage of ideal weight of subjects exert the strongest influence on systolic blood pressure, whereas the effect of age on diastolic blood pressure seems less consistent. Smoking has no direct effect on blood pressure, but it does have a small effect on heart rate and a negative effect on obesity, suggesting, perhaps, that, while it has no direct effect, it does play an indirect role.
Conclusions:
Structural equation models can be used by researchers trying to understand how risk factors can influence blood pressure in complex ways. The methodology is especially appropriate for testing competing conceptual models.