N J Birkett1, J Boulet. 1. Department of Epidemiology and Community Medicine, University of Ottawa, Ontario, Canada.
Abstract
OBJECT: To examine the reliability and validity of a food habits questionnaire developed by Kristal et al in male manual laborers. DESIGN: A cross-sectional baseline survey. SETTING: The community of Ottawa-Carleton, Canada. SUBJECTS: All people working in non-office-based positions for two local governments were invited to a heart-health screening clinic. Male subjects (n = 362) who met risk factor eligibility criteria were interviewed and entered into the study. MAIN OUTCOME MEASURES: The food habits questionnaire developed by Kristal et al was the main focus of analysis. Additional outcomes included a food frequency questionnaire and risk factors for cardiovascular disease. STATISTICAL ANALYSES: Reliability was assessed using Cronbach's alpha and Pearson correlation coefficients; confirmatory factor analysis was also done. Validity assessment included partial correlations. RESULTS: Low internal consistency was found for the five subscales of the questionnaire (alpha = .13 to .53). Confirmatory factor analysis did not reveal the postulated five-factor (subscale) structure. Correlation of the subscale scores with dietary fat intake was low (r = -.09 to -.23), and none of these associations were statistically significant after adjustment for age, body mass index, and education. An alternative scoring system that treated the questionnaire as a unidimensional behavioral checklist produced a higher internal consistency (alpha = .70) and significant correlation with dietary fat intake (r = -.27). CONCLUSIONS: The psychometric properties and scoring of the food habits questionnaire need to be explored in additional populations before the questionnaire is adopted for general use.
OBJECT: To examine the reliability and validity of a food habits questionnaire developed by Kristal et al in male manual laborers. DESIGN: A cross-sectional baseline survey. SETTING: The community of Ottawa-Carleton, Canada. SUBJECTS: All people working in non-office-based positions for two local governments were invited to a heart-health screening clinic. Male subjects (n = 362) who met risk factor eligibility criteria were interviewed and entered into the study. MAIN OUTCOME MEASURES: The food habits questionnaire developed by Kristal et al was the main focus of analysis. Additional outcomes included a food frequency questionnaire and risk factors for cardiovascular disease. STATISTICAL ANALYSES: Reliability was assessed using Cronbach's alpha and Pearson correlation coefficients; confirmatory factor analysis was also done. Validity assessment included partial correlations. RESULTS: Low internal consistency was found for the five subscales of the questionnaire (alpha = .13 to .53). Confirmatory factor analysis did not reveal the postulated five-factor (subscale) structure. Correlation of the subscale scores with dietary fat intake was low (r = -.09 to -.23), and none of these associations were statistically significant after adjustment for age, body mass index, and education. An alternative scoring system that treated the questionnaire as a unidimensional behavioral checklist produced a higher internal consistency (alpha = .70) and significant correlation with dietary fat intake (r = -.27). CONCLUSIONS: The psychometric properties and scoring of the food habits questionnaire need to be explored in additional populations before the questionnaire is adopted for general use.