REGRESSION-MODELS FOR ORDINAL RESPONSES - A REVIEW OF METHODS AND APPLICATIONS

Background. Epidemiologists are often interested in estimating the ris k of several related diseases as well as adverse outcomes, which have a natural ordering of severity or certainty. While most investigators choose to model several dichotomous outcomes (such as very low birthwe ight versus normal and moderately low birthweight versus normal), this approach does not fully utilize the available information. Several st atistical models for ordinal responses have been proposed, but have be en underutilized. in this paper, we describe statistical methods for m odelling ordinal response data, and illustrate the fit of these models to a large database from a perinatal health programme. Methods. Model s considered here include (1) the cumulative legit model, (2) continua tion-ratio model, (3) constrained and unconstrained partial proportion al odds models, (4) adjacent-category legit model, (5) polytomous logi stic model, and (6) stereotype logistic model. We illustrate and compa re the fit of these models on a perinatal database, to study the impac t of midline episiotomy procedure on perineal lacerations during labou r and delivery. Finally, we provide a discussion on graphical methods for the assessment of model assumptions and model constraints, and con clude with a discussion on the choice of an ordinal model. The primary focus in this paper is the formulation of ordinal models, interpretat ion of model parameters, and their implications for epidemiological re search. Conclusions. This paper presents a synthesized review of gener alized linear regression models for analysing ordered responses. We re commend that the analyst performs (i) goodness-of-fit tests and an ana lysis of residuals, (ii) sensitivity analysis by fitting and comparing different models, and (iii) by graphically examining the model assump tions.