A comparison of ordinal regression models in an analysis of factors associated with periodontal disease.

Abstract

Aim:The study aimed to determine the factors associated with periodontal disease (different levels of severity) by using different regression models for ordinal data.Design:A cross-sectional design was employed using clinical examination and ‘questionnaire with interview’ method.Materials and Methods:The study was conducted during June 2008 to October 2008 in Dharwad, Karnataka, India. It involved a systematic random sample of 1760 individuals aged 18-40 years. The periodontal disease examination was conducted by using Community Periodontal Index for Treatment Needs (CPITN).Statistical Analysis Used:Regression models for ordinal data with different built-in link functions were used in determination of factors associated with periodontal disease.Results:The study findings indicated that, the ordinal regression models with four built-in link functions (logit, probit, Clog-log and nlog-log) displayed similar results with negligible differences in significant factors associated with periodontal disease. The factors such as religion, caste, sources of drinking water, Timings for sweet consumption, Timings for cleaning or brushing the teeth and materials used for brushing teeth were significantly associated with periodontal disease in all ordinal models.Conclusions:The ordinal regression model with Clog-log is a better fit in determination of significant factors associated with periodontal disease as compared to models with logit, probit and nlog-log built-in link functions. The factors such as caste and time for sweet consumption are negatively associated with periodontal disease. But religion, sources of drinking water, Timings for cleaning or brushing the teeth and materials used for brushing teeth are significantly and positively associated with periodontal disease.