Methodological Issues In Oral Health Related Quality Of Life Research Using The Oral Health Impact Profile (OHIP-14)

Abstract

This study aims to identify and address the methodological issues that may occur in the analysis of data for assessing Oral Health Related Quality of Life using the OHIP-14 instrument. The four main methodological issues addressed in this work are the handling of missing data, presence and management of floor and ceiling effects, number of dimensions in OHIP-14 and the responsiveness of OHIP items to change. A total of 360 participants who came for dental treatment at King’s College London Dental Hospital, Denmark Hill, London participated in this study. Baseline data were collected from participants at the time of treatment. Data were also collected at two follow-ups, two and four months after baseline. At baseline, data were collected from all the 360 participants whereas in the first and second follow-ups, 89 and 75 patients respectively provided data. Different techniques for managing missing data, namely completed case, Item mean, subject mean, interpolation, regression, trend, EM algorithm and multiple imputation were tested. The floor and ceiling effects were handled using the Tobit model. Structural Equation Modelling was used to test the existence of one, three, six and seven factor models and these models were compared. The missing data in OHIP items followed a missing completely at random (MCAR) pattern. The mean values obtained from different missing data handling techniques were similar. No significant difference in mean OHIP scores was observed between dropout and non dropout cases and the dropouts followed a Missing At Random (MAR) pattern. Education, Profession and treatment needs significantly predicted (p<0.05) the change in OHIP scores. There was a greater floor effect than the ceiling effect. Use of the Tobit model, to adjust for floor and ceiling effects showed improved estimates for the effect of predictors. The comparison of Ordinary Least Squares (OLS) and Tobit model revealed that the Tobit model fitted the data well. OHIP-14 has good psychometric properties with the Cronbach’s alpha value of 0.93 for measuring the OHRQoL. None of the four models identified from the literature (one, three, six and seven factor models) fitted the data well. OHIP-14 was responsive to change and the individuals were classified as “Improved”, “No Change” and “Worsened” groups. The results were tested with national data from the Adult Dental Health Survey 2009, UK which showed similar results. In conclusion, the missing data in OHIP items can be handled either by multiple imputation or EM algorithm and OHIP-14 items suffer from floor and ceiling effects which can be handled with the Tobit model. As none of the four models reported in the literature fitted the data well, further research is required to explore the dimensions of OHIP-14. OHIP-14 is responsive to change and can be used to measure the treatment effect over a period of time.

Date of Award	2018
Original language	English
Awarding Institution	King's College London
Supervisor	Francis Hughes (Supervisor) & Wei Gao (Supervisor)