Article Text

Download PDFPDF

Cutting edge methodology
P1-2 Structural equation modelling in a fetal mortality case-control study in Brazil
  1. G Alencar,
  2. M Almeida
  1. Department of Epidemiology, Public Health School, University of São Paulo, Sao Paulo, Brazil


Introduction The outcome in the categorical form is very common and broadly used in epidemiology, specially in case-control studies, where the outcome can be disease/death/health conditions of a Yes/No form. It is also common to measure binary or categorical covariates.

Methods Latent variables as constructs/factors are used with many binary indicators in a structural equation model framework. The estimation of structural equation model are based in the assumption of a covariance matrix structure with normal multivariate distribution to the measured variables (indicators). Maximum likelihood estimator or generalised least squares are usually employed. Mean and variance-adjusted weighted least squares and maximum likelihood estimators are also applied with binary variables, usual in case-control studies, and ORs estimates are presented.

Results The comparison of different estimation methods is evaluated and showed similar results in fetal mortality study based in Sao Paulo City, Brazil. ORs were available for ML only.

Conclusion The effect of socioeconomic situation and negative acceptance of pregnancy on fetal deaths is through prenatal care, after excluding the non-significant direct path from nSES to fetal deaths. There was also an effect of nAoP on fetal deaths. The model included also a direct effect of intra-uterine growth retardation to fetal death. Fit indexes for the final model were considered reasonable (CFI=0.95, TLI=0.95, RMSEA=0.05, WRMR=0.99).

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Funding FAPESP, CNPq, CAPES.