Missing Data Analysis in Longitudinal Studies: Findings from a Quality of Life Study in Malaysian Tuberculosis Patients

This was a longitudinal health-related quality of life (HRQoL) study conducted among smear positive pulmonary tuberculosis (PTB) patients at the chest clinic of Penang General Hospital between March 2010 and February 2011. Primary objective of this study was to propose an appropriate statistical method in analysing longitudinal HRQoL data. Eligible patients (i.e., a case of new smear positive PTB, literate and 18 years and above) were asked to self-complete the SF-36v2 questionnaire at the start of the treatment, after the intensive phase and at the end of the treatment. The mean physical component summary (PCS) and mental component summary (MCS) scores, ranging from 47 to 53, were considered equivalent to the general population norms. Repeated measures ANOVA (with single imputations) and linear mixed model were used to analyse the data. The study’s findings indicated compromised health among the study participants even at the end of treatment. In repeated measures ANOVA analysis, level of education, diabetes, being alcoholic and cough with sputum were the significant predictors of PCS, whereas none of the covariates explained a significant variance in the MCS scores. In linear mixed model, ethnicity, marital status, being a smoker, productive cough and ≥3 TB-related symptoms were the significant predictors of PCS. Similarly, covariates such as ethnicity, hypertension, being a smoker, monthly income ≥1,000 Malaysian Ringgit and ≥3 TB-related symptoms significantly explained variance in the MCS scores. According to different findings obtained from both methods and the limited assumption in applying repeated measures ANOVA, linear mixed model was preferred to analyse this data.