gms | German Medical Science

50. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds)
12. Jahrestagung der Deutschen Arbeitsgemeinschaft für Epidemiologie (dae)

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie
Deutsche Arbeitsgemeinschaft für Epidemiologie

12. bis 15.09.2005, Freiburg im Breisgau

Alternatives for logistic regression in cross-sectional studies: A comparison of six models estimating the prevalence ratio

Meeting Abstract

Suche in Medline nach

  • Martin Wolkewitz - Universität Heidelberg, Heidelberg

Deutsche Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie. Deutsche Arbeitsgemeinschaft für Epidemiologie. 50. Jahrestagung der Deutschen Gesellschaft für Medizinische Informatik, Biometrie und Epidemiologie (gmds), 12. Jahrestagung der Deutschen Arbeitsgemeinschaft für Epidemiologie. Freiburg im Breisgau, 12.-15.09.2005. Düsseldorf, Köln: German Medical Science; 2005. Doc05gmds250

Die elektronische Version dieses Artikels ist vollständig und ist verfügbar unter: http://www.egms.de/de/meetings/gmds2005/05gmds076.shtml

Veröffentlicht: 8. September 2005

© 2005 Wolkewitz.
Dieser Artikel ist ein Open Access-Artikel und steht unter den Creative Commons Lizenzbedingungen (http://creativecommons.org/licenses/by-nc-nd/3.0/deed.de). Er darf vervielf&aauml;ltigt, verbreitet und &oauml;ffentlich zug&aauml;nglich gemacht werden, vorausgesetzt dass Autor und Quelle genannt werden.


Gliederung

Text

Background

Logistic regression is commonly used to analyse cross-sectional studies with binary outcome. The odds ratio as an estimate for the prevalence ratio is frequently reported although prevalence ratios can be estimated. However, it may lead to extreme overestimation when the prevalence is greater than 10 %.

The aim of this discussion is to compare different alternatives which calculate the prevalence ratio directly. The comparison was made regarding validity, precision, hypothesis testing and confounding behaviour.

Methods

Several simulations were used to test the performance of regression models including (1) unadjusted Cox / Poisson, (2) Cox with robust sandwich variance , (3) log-binomial, (4) Poisson with scale parameter adjusted by Pearson , (5) Poisson with scale parameter adjusted by deviance and (6) a GEE-approach based on the logistic model. Additionally, the unconditional logistic regression model was performed to notice overestimation.

Results

All six models gave valid point estimates for the prevalence ratio. Cox with robust variance, log-binomial, Poisson with scale parameter adjusted by Pearson and the GEE- logistic model performed well and had similar results concerning precision, power and confounding. The unadjusted Cox / Poisson model produced standard errors that were too large and had less power, especially with increasing prevalence. When testing whether the estimate is equal to the true parameter, the Poisson model adjusted by deviance rejected too often.

Conclusion

When the prevalence ratio is the measure of interest in cross-sectional studies, there are very suitable alternatives for logistic regression. When using the Cox model, adjusting with the robust variance improves precision and power remarkable. Although precautions are needed regarding convergence problems, the log-binomial model is also preferable.


Literatur

1.
Barros AJ, Hirakata VN. Alternatives for logistic regression in cross-sectional studies: an empirical comparison of models that directly estimate the prevalence ratio. BMC Med Res Methodol 2003;3(1):21.
2.
Skov T, Deddens J, Petersen MR, Endahl L. Prevalence proportion ratios: estimation and hypothesis testing. Int J Epidemiol 1998;27(1):91-5.
3.
Zocchetti C, Consonni D, Bertazzi PA. Relationship between prevalence rate ratios and odds ratios in cross-sectional studies. Int J Epidemiol 1997;26(1):220-3.