Adjusting the Oral Health Related Quality of Life Measure (Using Ohip-14) for Floor and Ceiling Effects

ABSTRACT Introduction The influence of floor (lowest) and ceiling (highest) effects on the outcome measure is of serious concern particularly when the outcome is measured using Likert scales. Conventional regression methods yield biased results and hence tobit regression is to be used to adjust for these effects. This paper is an attempt to use tobit regression in finding the predictors of oral health related quality of life after adjusting for floor and ceiling effects. Methods A sample of 360 participants were asked to self asses their oral health related quality of life (OHRQoL) using Oral Health impact profile with 14 items which forms the data for this study. Apart from descriptive statistics, Ordinary Least squares regression and tobit regression were used to find the significant predictors of OHRQoL and the results of both methods were compared. Results The sample comprised of 41.1% men and 58.9% women. Majority of the participants (68.3%) were whites. The average item difficulty was 0.4 and the average item easiness was 0.03. The floor and ceiling values for the composite scores were 14 and 56 respectively. Age and gender were not statistically significant both in Ordinary Least Squares (OLS) regression and Tobit regression. Full time employment, student and retired have significantly lower scores in OLS but only retired had significantly lower scores in the tobit model. Conclusion Tobit model, after adjusting for floor and ceiling effect, gives higher values for the predictors and the OLS model underestimates the effects of predictors on OHIP scores.