Analysis of Categorical Data on Pregnancy Outcome

Pickering, R. M (1987) Analysis of Categorical Data on Pregnancy Outcome. PhD thesis, University of Glasgow.

Abstract

In this thesis we examine the risk of low birthweight, preterm delivery and changes in the distribution of birthweight standardised for gestational age related to ten maternal, sociodemographic and biological covariates available from routinely- collected maternity discharge records. The information is considered in the form of categorised variables and is analysed using regression models adapted for use with categorical data. A second part of the thesis describes a classification of neonates based on a latent class model for eleven variables measured during the neonatal period. Chapter 1 contains a review of the epidemiology of birthweight and gestational age, focusing on their relationship with the ten covariates and also with perinatal mortality. In Chapter 2, the SMR2 and SMR11 data collection schemes are described, and the variables used in the regression analyses and the neonatal classification are defined. Chapter 3 introduces the binary logistic regression model and several approaches to examining the fit of the model. Ordered logistic regression models for polytomous response data and, in particular, the proportional hazards model for polytomous survival times are described. More general models, where the ordering is relaxed, are considered. A comparison is drawn with various methods related to fitting regression models to categorical data that were not used. The first section of Chapter 4 summarises the main findings relating to the three perinatal outcomes. Later sections describe the results in more detail and examine the impact of changes in model assumptions for each outcome considered Chapter 5 introduces and describes the latent class model for neonates. One to six class models are explored but problems with multiple maxima in the likelihood arose when four or more classes were fitted. The impact of missing data, the stability of the classification between 1978 and 198O, and the effect of repeating the analysis on random halves of the data are examined for the three class model. Chapter 6 discusses the use of SMR2 data in epidemiology, the results of the regression analyses and the cost of fitting regression models to categorical data. The neonatal classification is discussed in the second part of Chapter 6.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Additional Information: Adviser: G D Murray
• Keywords: Biostatistics, Statistics
• Date of Award: 1987
• Depositing User: Enlighten Team
• Unique ID: glathesis:1987-76421
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Nov 2019 14:31
• Last Modified: 19 Nov 2019 14:31
• URI: https://theses.gla.ac.uk/id/eprint/76421