Non-random aggregations of healthy and unhealthy lifestyles and their population characteristics - pattern recognition in a large population-based cohort

Abstract

Background: There is currently no consensus on the number and domains of lifestyle factors to incorporate in research examining lifestyle combinations. A common approach in this field involves creating unweighted lifestyle scores by summing healthy/unhealthy lifestyle scores. However, further exploration and comparison are needed to advance beyond simple summation and investigate more nuanced lifestyle patterns.

Methods: Latent class analysis was performed to identify lifestyle patterns among 112,842 participants aged 18 or older from the Dutch Lifelines cohort (baseline data, collected from 2007 to 2013). Ten lifestyle factors were selected based on the six pillars of Lifestyle Medicine: smoking habits, binge drinking, daily alcohol intake, diet quality, ultra-processed food consumption, long-term stress, physical (in)activity, sleeping, TV watching time and social connections. Lifestyle factors were assessed using validated self-report questionnaires.

Results: We identified five lifestyle patterns: “Healthy but physically inactive” (8.6% of the total population, class 1), “Unhealthy but no substance use” (8.5%, class 2), “Healthy in a balanced way” (37.2%, class 3), “Unhealthy but light drinking and never smoked” (31.6%, class 4) and “Unhealthy” (14.2%, class 5). Socio-demographic characteristics including age distribution, sex, education level, income and employment status differed significantly (nominal p < 0.05) across lifestyle patterns. Multiple comparison analysis showed that healthy lifestyle scores differed within lifestyle pattern pairs. Proportions of unhealthy lifestyle scores are comparable between the “Healthy but physically inactive” and the “Healthy in a balanced way” patterns as well as the “Unhealthy but no substance use” and “Unhealthy but light drinking and never smoked” patterns.

Conclusion: The five identified lifestyle patterns exhibit distinct, non-random clustering of behaviours, each linked to specific socio-demographic characteristics. Understanding these clustering tendencies can help identify target populations and uncover barriers to unhealthy behaviours, aiding the development of tailored health interventions. The overlap in the distribution of unhealthy lifestyle scores between two lifestyle patterns suggests that the latter may provide a more comprehensive perspective on habitual behaviours.

Supplementary Information: The online version contains supplementary material available at 10.1186/s13690-025-01678-1.

Keywords: Combined lifestyle factors; Healthy lifestyle scores; Lifestyle clusters; Lifestyle medicine; Tailored lifestyle intervention.

Fig. 1

A framework showing the study design and specifying the lifestyle factors used for lifestyle pattern recognition. UPF: ultra-processed food; MVPA: moderate-to-vigorous physical activity; LCA: latent class analysis. Age and sex data are available for the entire population. A sample of 112,797 participants is available for multiple comparisons on marital status, 112,565 on education, 111,520 on income and 111,605 on employment status

Fig. 2

Relative contribution of each lifestyle factor to the five lifestyle patterns based on item-response probabilities for the healthy, moderate and unhealthy level. The “Healthy but physically inactive”, “Unhealthy but no substance use”, “Healthy in a balanced way”, “Unhealthy but light drinking and never smoked” and “Unhealthy” correspond to class 1 to class 5, respectively. DQ: diet quality; UPF: ultra-processed food consumption; PA: physical activity; Rec.: recommended; More Rec.: more than recommended; Less Rec.: less than recommended; SC: social connections. Each lifestyle factor has three response levels (A, B, and C, representing healthy, moderate, and unhealthy levels, respectively). Horizontally, the sum of probabilities for the three levels within each lifestyle pattern equals 1. Supplementary Material 3, Table 1, summarises the specific probabilities by lifestyle patterns. A deeper blue colour indicates a higher probability of being in the healthy level, while a deeper orange colour indicates a higher probability of being in the unhealthy level

Fig. 3

Percentages of participants adhering to a cumulative number of lifestyles at a healthy or unhealthy level for the total population and the five lifestyle patterns. A: the entire population. B: the “Healthy but physically inactive” pattern (Class 1). C: the “Unhealthy but no substance use” pattern (Class 2). D: the “Healthy in a balanced way” pattern (Class 3). E: the “Unhealthy but light drinking and never smoked” pattern (Class 4). F: the “Unhealthy” pattern (Class 5). The blue line represents healthy lifestyles, while the orange line indicates unhealthy lifestyles