Clustering of dietary variables and other lifestyle factors (Dutch Nutritional Surveillance System)

Abstract

Study objective: The aim was to investigate whether dietary factors cluster in a favourable or unfavourable way and to characterise the groups identified by lifestyle and sociodemographic variables.

Design and setting: This cross sectional study was based on data of the 1987-1988 Dutch national food consumption survey (DNFCS), obtained from a panel by a stratified probability sample of the non-institutionalised Dutch population.

Participants: 3781 adults (1802 males and 1979 females) of the DNFCS, aged 19 to 85 years, were studied.

Measurements and main results: To estimate dietary intake two day food records were used. Lifestyle factors were collected by structured questionnaire and sociodemographic variables were available from panel information. Cluster analysis was used to classify subjects into groups based on similarities in dietary variables. Subsequently, these groups were characterised by sociodemographic and lifestyle factors as well as by the consumption of food groups. Eight clusters were found. In comparison with the guidelines, the dietary quality in four clusters was poor. The cluster with the poorest dietary intake (high intake of fat, cholesterol, and alcohol; low intake of dietary fibre) showed on average a high consumption of animal products (except milk), fats and oils, snacks, and alcoholic beverages, and a low consumption of fruit, potatoes, vegetables, and sugar rich products. Smoking, body mass index, dietary regimen on own initiative, hours of sleep, gender, age, socioeconomic status, and day of the week were found to discriminate among the clusters.

Conclusions: Cluster analysis resulted in substantial differences in mean nutrient intake and seems useful for dietary risk group identification. Undesirable lifestyle habits were interrelated in some clusters, but an exclusive lifestyle for health risk has not been found.