The prevalence of clustering of health behaviors and associated physical and mental health outcomes

By Hugo Francisco de SouzaAug 23 2023Reviewed by Sophia Coveney

In a recent study published in the journal BMC Public Health, researchers conducted a systematic review of the literature investigating the associations between lifestyle, health behaviors, and health outcomes.

They used clustering analyses to evaluate how combinations of sedentary behavior (SB), physical activity (PA), and diet affect mental and physical health. Their results highlight that unhealthy and mixed behavioral clusters were at higher risk of adiposity and cardiovascular disease (CVD) and depicted lower cardiorespiratory and mental health.

An (outdoor) game a day keeps the doctor away

Non-communicable conditions, including overweight, obesity, and their comorbidities, have been increasing globally at an alarming rate. In the United Kingdom (UK) alone, the prevalence of obese adults has more than tripled in the last 20 years, with childhood prevalence growing from 21.0% in 2019-20 to 25.5% the following year.

Habits targeting PA, SB, and diet, when inculcated at an early age, are maintained through childhood and into the adult years. Given that each of those modifiable health behaviors has independently been associated with long-term health conditions, research into the synergistic effects of these behaviors is needed.

Additionally, evidence suggests that health behaviors have synergistic effects more significant than the sum of each behavior in isolation, especially in depression and anxiety risk, psychological distress, and weight changes.

Previous research into these relationships has focused on each behavior in isolation or has clustered behaviors under broad and often vague classifications such as total PA, total SB, or just diet. These studies generally treat additional health behaviors such as alcohol consumption and smoking as subsumed under a health behavior cluster rather than treating them as independent variables, thus reducing study reliability and sensitivity and in extreme cases, resulting in confounding outcomes.

About the study

The present study aimed to collate information from longitudinal, cross-sectional, and cohort studies on associations between health behaviors and health outcomes. Their study was designed to evaluate outcome prevalence based on derived clusters of age, sex, and socioeconomic variables, thereby establishing links between them and the synergistic effects on health behavior choices.

This review was designed using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines. The search strategy was developed following the Population Exposure Context Outcome (PECO) concepts, wherein ‘Population’ referred to the age cluster (children, adolescents, and young adults), ‘Exposure’ referred to the health behavior cluster (PA, SB, and diet), and ‘Outcomes’ referred to observed mental and physical health outcomes.

Reviewers employed this search strategy on three online databases, PubMed, Scopus, and Web of Science, from databased initiation until July 2022. Inclusion criteria were age (between 5-24 years), study methodology (must include clustering or patterning effects of at least a single domain of PA, SB, and diet), and publication language (papers published before 24^th July 2022 in English).

Studies that did not evaluate clustering, and those that included interventions, including randomized controlled trials, were excluded from the dataset.

Identified studies were title, abstract, and full-text screened using the Covidence review management software. Two independent reviewers selected appropriate records and extracted general information, participant characteristics, and study characteristics from publications thus chosen.

Data on clustering models used and health outcomes recorded were additionally extracted for use in this review.

Bias assessment was conducted using recommendations from the Cochrane Handbook for observational studies. Every included study was assessed for bias risk in the following domains:“(1) selection bias, (2) performance bias, (3) detection bias, (4) attrition bias, (5) selective reporting bias, and (6) other factors that may increase the risk of bias.”

Since significant heterogeneity existed between studies, meta-analytic approaches could not be employed in this study. Hence, results were reported narratively.

Study findings

The PRISMA-derived search strategy identified 21,282 records, 4,167 of which were duplicates. Title and abstract screening excluded 16,814 records, and full-text screening finally identified 53 papers to be included in this review. The studies were carried out in 18 countries, with the USA (n = 9), Brazil (n = 9), and Australia (n = 5) forming the bulk of the sampled population.

Studies were predominantly cross-sectional in their study design (n = 49), with four longitudinal studies included. The total sample cohort comprised 778,415 individuals between the ages of 5 to 25. Forty-nine studies provided clustering data on PA, SB, and diet, of which 20 investigated the associations between these health behaviors and health outcomes.

“Risk of bias was conducted for all included studies. For both study types, between 5 and 30% had a high-risk judgment across all domains, while low-risk judgment varied between ~50–90%. Some of the domains had an unclear judgment due to lack of information (~5–55%)”

Clustering analysis methods varied between studies, but most followed the Ward and k-means method. A total of 173 unique health behavior clusters were identified from all studies, which were broadly classified as ‘healthy’ (n = 29), ‘unhealthy’ (n = 46), and ‘mixed’ (n = 98) clusters.

“A healthy cluster was typically [characterized] by good diet quality, high PA, and low SB, while an unhealthy cluster was [characterized] by poor diet quality, low PA and high SB. The majority fell into the mixed cluster, which included one or more healthy [behaviors] coexisting with one or more unhealthy [behaviors] (e.g. high PA, high FV, and high SB). In the healthy lifestyle clusters, only two clusters (high MVPA high FV low SSB low screen time and high MVPA low energy dense food/drink low SB low TV) were reported in two studies, while the unhealthy lifestyle cluster low MVPA low FV high screen time was reported in three studies”

Health behavior prevalence was computed, and it found that for healthy clusters, 24 clusters showed 0-30% prevalence, 18 showed 30-60% prevalence, and only one showed a prevalence between 60-100%. Unhealthy clusters showed prevalence at 0-30% for 55 clusters, 30-60% for 21 clusters, and 60-100% for seven clusters. Mixed clusters were not found to follow a particular structure, but were found to include combinations of healthy and unhealthy clusters coexisting.

Conclusions

In the present review, a sample cohort of 778,415 individuals from 53 studies revealed 172 unique behavior clusters. For this review, the clusters were divided into ‘healthy,’ ‘unhealthy,’ and ‘mixed.’

Overall, the majority of participants examined fell into the mixed clusters, which is in line with previous reviews’ findings, and supports the need for multicomponent interventions addressing several unhealthy [behaviors] simultaneously. It is also noteworthy that high PA and high SB most frequently clustered together, refuting the displacement hypothesis that assumes that time spent on one activity cannot be spent on another (i.e., SB displaces PA)

Alosaimi et al. (2023)

This review identified that unhealthy behaviors were associated with increased adiposity in young people, in line with previous findings. Contrasting previous work, however, this review also elucidates the negative impact of mixed behaviors of adiposity, which previous work refuted. Excess screen time was identified as the most probable variable associated with increased weight gain, irrespective of the cluster. It is likely that excess screen time counteracts the beneficial effects of PA and a healthy diet.

Socioeconomic status was negatively associated with adiposity and positively associated with mental and physical health. Sex also played a role, with girls being more likely to belong to an unhealthy or mixed cluster characterized by a good diet but poor PA and SB.

The current review also found that clusters with higher screen time had greater risk of individual and clustered cardiovascular risk scores, which were predominantly seen in older boys, worse fitness levels, and greater psychosocial risks, mostly in girls.

Alosaimi et al. (2023)