“Global Insights: Unveiling Network Dynamics Across 29 Nations with JOGH”

Factors related to lifestyle—such as dietary risks, consumption of alcohol and tobacco, inactivity, and a lack of physical activity—considerably affect health outcomes both in the short term and over the longer term. Immediate effects are observed in alterations to physical and psychological well-being [1], sleep [2], and social health [3]. In contrast, sustained poor lifestyle habits significantly heighten the risk of non-communicable diseases, including cardiovascular ailments [4], diabetes, cancer, and chronic respiratory illnesses [5], ultimately influencing life expectancy and mortality rates [6]. The worldwide incidence of non-communicable diseases highlights the essential role of lifestyle in fostering health. Nevertheless, it is complicated and resource-heavy to implement thorough lifestyle interventions covering all aspects of lifestyle at a population level, requiring the identification of the most influential lifestyle choices for targeted actions and efficient resource distribution for comprehensive health promotion, particularly in low- and middle-income nations facing greater challenges and limited resources.

Network analysis is a method that evaluates the impact of one variable within its network group or its associations with another network group while treating all other in-network variables as confounding factors [7]. According to network analysis theory, the central lifestyle and health outcome variables hold the most significance in their networks, providing the greatest predictive ability and the potential for substantial benefits when adjusted. The bridge lifestyle, which is the lifestyle factor most closely linked to the health outcome network, serves a similar role but influences the entire health network [7]. The existence of the most significant lifestyle is corroborated by the interconnected relationships identified in prior studies involving lifestyles, health outcomes, and their interactions. For instance, physical inactivity correlates with other lifestyle habits, such as smoking, alcohol consumption, and unhealthy eating [8,9] Moreover, physical inactivity influences various interim health outcomes, including mental health and sleep quality [10]. Additionally, habits like smoking, drinking, and poor dietary choices also negatively affect other health outcomes, such as mental health and sleep quality [11]. Building on these interconnected patterns, a study employed network analysis to pinpoint central lifestyle factors, health outcomes, and bridge lifestyles from a sample representing multiple countries [12]. However, global methods may fail to recognize important cross-national differences in lifestyle trends and cultural practices, thus limiting the effectiveness of interventions in specific areas. Therefore, pinpointing central lifestyle factors at the national level is critical to customizing interventions for regional contexts.

The premise that central lifestyle factors differ between nations is backed by various influences, including socioeconomic and cultural factors, which affect the coexistence and strength of connections among lifestyle habits. For instance, restricted access to nutritious foods is more common in low-income nations [13], potentially leading to inadequate food availability and a diminished variety of food choices. Conversely, high-income countries often experience elevated rates of sedentary behaviours, such as prolonged periods of sitting and increased screen time [14], attributed to factors like the widespread adoption of digital technologies, the prevalence of desk and office jobs, and a greater dependence on cars for transportation. Additionally, cultural elements, such as the American inclination towards sweet flavours [15,16] and the alcohol restrictions in some regions of Saudi Arabia due to religious and legal mandates [15,16], can increase or decrease the frequency of certain lifestyle practices. The high-density urban setting in Hong Kong may restrict physical activity and escalate stress levels [17]. On the other hand, the mainly rural landscape of Rwanda encourages elevated levels of physical activity [18]. The relationship between these distinctions and country-specific key variables can be elucidated through the marked presence of particular lifestyle factors and health outcomes. These elements may bolster their connections with other concurrent lifestyles or health outcomes, thereby enhancing their central role within a network and rendering each country’s key variables distinctive. Recognizing these varied key lifestyles is vital for crafting effective, tailored local health initiatives, yet remains insufficiently understood.

Hence, this research intends to evaluate the interplay between lifestyle factors and health outcomes across 29 nations and to determine the key variables (central lifestyle, central health outcome, and bridge lifestyle). We propose that each nation possesses unique central lifestyle factors, health outcomes, and bridge lifestyles. The outcomes are anticipated to offer country-specific insights for local resource allocation aimed at effective comprehensive health enhancements.

METHODS

Study design

This was an international cross-sectional study.

Study settings

This study encompassed six regions classified by the World Health Organization, covering 29 nations following the exclusion of Spain due to a limited sample size of 51. Recruitment took place primarily through online platforms to maximize reach and enable voluntary involvement in preferred languages. The published protocol provides thorough details regarding the study design [19].

Participants and sample size

We enlisted participants aged 18 and above employing convenience and snowball sampling methods. While these strategies facilitated efficient recruitment across diverse geographical areas, they also pose potential selection biases, particularly impacting the representation of socioeconomic and rural/urban diversity within the sample populations. Sample sizes for each network – lifestyle (18 nodes), health outcomes (13 nodes), and bridge (31 nodes) – were established based on the maximum number of edges: 153, 78, and 465, aligning with the guideline of at least three participants per parameter [20]. Consequently, the required sample sizes for each network were 459, 234, and 1134 participants, respectively. Sample sizes across countries varied from 150 to 2238, with the majority meeting these benchmarks. It is worth noting that centrality measures in network analysis can be deemed reliable even if the sample size does not meet initial projections, as long as they pass the stability test via case-dropping subset bootstrap [20]. To ensure reliability, stability tests were performed for all 87 networks, with only those networks that successfully passed these assessments being reported. We acknowledge the limitations inherent in these sampling methods and the sample sizes in certain countries and their potential impact on the

the connections between nodes pass through a particular node, indicating its importance in facilitating communication within the network.

Generalizability of our results

To address these implications, stability assessments through case-dropping subset bootstrap were carried out across all 87 networks, with only those networks that met the stability criteria being reported.

Assessment Methods

Demographic Characteristics

The demographic factors examined in this research encompassed gender, age, country of residence, marital status, highest academic qualification achieved, and employment status.

Evaluation of lifestyle elements and health outcomes

The development of items for the study questionnaire was based on an extensive literature review coupled with collaborative discussions with public health experts, nurses, and dietary specialists. The selected variables pertaining to lifestyle and health outcomes mirror aspects widely recognized in public health studies, emphasizing critical lifestyle factors such as diet, physical activity, and substance consumption, along with health outcomes in physical, psychological, social, and financial wellness. Specifically, the 18 lifestyle elements comprise:

• nutritional (types of foods consumed daily, fruit and vegetable intake, usage of frozen foods or processed products, snacks, soft drinks, juices or other sugary beverages, home-cooked meals, cooking at home, takeout meals, traditional Chinese medicine or natural health products, and oral supplements or vitamins)
• physical activity (frequency, duration, type, and total amount)
• sedentary activities (total sitting time and screen duration)
• dependency behaviors (tobacco use and alcoholic beverage consumption).
• The 13 health-related impacts encompass:
• physical health (weight, appetite, quality of sleep, perceived physical health, and life quality)
• mental health (psychological burden, emotional strain)
• social health (family issues, social support received, social support given, and community involvement)
• financial health (income level and economic pressures).

An initial English version was crafted and its face validity confirmed through thorough reviews that refined the questionnaire for clarity, reduced redundancies, and organized the flow of questions for easier completion. To ensure cultural appropriateness, engagements with experts from various countries were held, followed by translations into several languages using a standardized forward-backward approach. Lifestyle elements and health outcomes were evaluated using individual items, rendering formal psychometric assessment unnecessary [19].

Participants assessed the effect of COVID-19 on 18 lifestyle elements and 13 general health impacts utilizing a 5-point Likert scale, where 1 represented ‘significantly reduced’, 3 indicated ‘no change’, and 5 meant ‘significantly increased’. To maintain consistency in item alignment indicative of healthier lifestyles and improved health outcomes, certain items were reverse-coded by adding a ‘less’ prefix.

Data Collection Methods

Data acquisition was conducted through online survey tools, supplemented by offline electronic PDF formats in regions with restricted internet availability. Both approaches were standardized in content and structure to ensure comparability and uphold data quality across various collection contexts. Comprehensive pre-testing verified the equivalence of the two formats, facilitating consistent data collection in all areas. To incentivize participation, one Hong Kong dollar was donated to the Red Cross for each completed questionnaire. The recorded response rate was 75.2%.

Validation and Rigor

To bolster internal validity, validation inquiries were included within the questionnaire. Participants were prompted with the question, ‘Where does the sun emerge each day?’ Additionally, to ensure cultural relevance in Nigeria, the inquiry ‘What is your STATE capital?’ was implemented. Each country’s questionnaire underwent pilot testing with at least 10 participants prior to distribution. Findings from these pilot tests revealed that participant responses fell within expected ranges and no confusion relating to the questions was noted; thus, no adjustments to the questionnaire were warranted. Furthermore, to minimize social desirability bias, the survey was conducted anonymously. Additionally, data quality checks were performed prior to analysis to eliminate responses indicative of misreporting or misunderstanding. These strategies collectively aim to mitigate potential biases and augment the accuracy and reliability of our dataset.

Statistical Evaluation

Data collected for our investigation were transferred to a Microsoft Excel database to enable stringent quality oversight. This involved discarding incomplete or duplicate responses to uphold data integrity. We also resolved data inconsistencies by filtering out responses that did not correspond with validation queries, such as incorrect information regarding sunset timings or capital cities, thereby enhancing the precision of our dataset and the credibility of our analysis. Subsequent evaluations were conducted using R Statistical Software (v4.1.1; R Core Team, Vienna, Austria). The network analyses included four aspects: topological overlap evaluation, network estimation, stability assessment, and calculations of centrality and bridge centrality measures.

Evaluation of Topological Overlap

The ‘networktools’ R package’s ‘goldbricker’ function was utilized to identify distinct variables to avert artificial correlations among similar variables within a network. A significance threshold of 0.25 for inclusion and a significance level of 0.01 were established [21].

Network Estimation

For each nation, three networks were examined: one consisting of all residual lifestyle factors, another comprising all remaining health outcomes, and a bridge network that interconnected the two. Pairwise correlations were calculated using partial correlation analysis to adjust for confounding impacts from other nodes. The least absolute shrinkage and selection operator was employed to reduce edges and eliminate minor correlations, while the extended Bayesian information criterion assisted in selecting the tuning parameter, resulting in a more interpretable and sparse network [22]. Network estimation and visualization were accomplished using the ‘bootnet’ and ‘qgraph’ R packages, respectively [22]. In the visual displays, nodes represented network items and edges illustrated their interconnections, with edge thickness denoting the strength of association – blue indicating positive and red indicating negative correlations.

Network Stability Assessment

The stability of the network was examined using the ‘bootnet’ package [22]. Edge weight stability was evaluated with 95% confidence intervals (CIs) obtained from nonparametric bootstrapping; narrower CIs signified a higher degree of network reliability [22]. Centrality stability was quantified by the correlation stability coefficient (CS-C), determined using case-dropping subset bootstrap methods. A CS-C value exceeding 0.25 was viewed as acceptable, with values greater than 0.5 being preferred [22].

Central Node, Centrality, Bridge Node, and Bridge Centrality

A central node holds a substantial influence in a network through its connections with other nodes, while bridge nodes serve as pivotal points connecting distinct clusters within another network [23,24]. In network analysis, the betweenness centrality quantifies how frequently the connections between nodes traverse through a specific node, highlighting its significance in facilitating inter-node communications within the network.

a node appears on the most direct paths among various pairs, highlighting its supplementary function. Closeness centrality determines a node’s typical distance from all others, illustrating network reachability. Strength centrality aggregates the edge weights associated with a node, reflecting its connectivity. In contrast, expected influence centrality accumulates edge weights without absolute scaling, representing prevalent influences in networks comprising both positive and negative edges. Considering our goal to detect nodes that impose considerable cumulative impacts, especially within networks featuring negative interactions, we emphasize expected influence centrality to accurately identify central nodes and effectively recognize bridge nodes [25]. The ‘qgraph’ library in R computed the expected influence index, which considers both positive and negative edges to highlight central nodes with the highest scores. Bridge nodes were evaluated employing the ‘networktools’ library in R, which employs the bridge expected influence (one-step) index that combines a node’s edges to nodes in alternate networks [23]. Variations in centrality between pairs of nodes were analyzed through Wilcoxon tests using 1000 bootstrapped indices, produced via the ‘bootnet’ library in R. Adjustments for multiple comparisons were implemented using Holm-Bonferroni corrections.

RESULTS

Demographics and item description

Following the examination of 19,094 responses, 16,461 were deemed eligible for analysis. Exclusions involved vacant or incomplete responses (1940), duplicates (116), inconsistent submissions (450), entries from non-participating nations (126), and records lacking age or gender data (1). Table 1 provides the sociodemographic characteristics, averages, standard deviations, and abbreviations related to lifestyle aspects and health outcomes, while Table S1 in the Online Supplementary Document presents a country-wise breakdown.

Table 1.  Sociodemographic of respondents from 29 countries (n = 16 461)

WordPress Data Table

H – health result, L – lifestyle, SD – standard deviation

Unique items persisted for network evaluation

Redundancy evaluation through the Goldbricker technique removed L15 (Table 1), L16, and L17 due to overlap with L18, which depicts exercise habits more accurately. The concluding network encompassed 15 lifestyle elements and 13 health results without further redundancies.

Lifestyle networks across various nations

Network resilience

The bootstrapped 95% CI evaluation validated the precision of edge weights in the lifestyle network, with tight intervals indicating trustworthy estimates (Figure S1 in the Online Supplementary Document). Furthermore, CS-C values for anticipated effects, generated from a case-dropping subset bootstrap, varied between 0.21–0.75 (Table 2; Figure S2 in the Online Supplementary Document). Excluding the United States, these figures exceeded the 0.25 mark, confirming the interpretability of 28 nation-specific lifestyle networks.

Table 2.  Overview of network metrics across 29 countries (n = 16 461)

“`WordPress Data Table

CS-C – coefficient of correlation stability, H – health outcomes, L – lifestyle, NA – not applicable, PC-C – coefficient of partial correlation

Network configuration and central lifestyle

Taking Mainland China as a case study, Figure 1, Panel A illustrates the lifestyle network where 49.5% of its 150 edges have nonzero values, indicating robust connectivity. The most significant links were between (the edge linking Lifestyle 7 and Lifestyle 6 (L7–L6), partial correlation coefficient = 0.85), L11–L12 (partial correlation coefficient = 0.75), and L13–L14 (partial correlation coefficient = 0.66) (Table 2). The central element, home cooking (L7), was validated by centrality index charts and bootstrapped difference assessments (P<0.05) (Figure 1, Panel B). For the other 27 nations, Figure S3 in the Online Supplementary Document displays the network configurations, with nonzero edge percentages varying from 15.2 in Guatemala to 68.6% in Hong Kong (Table 2). Additionally, Figure S4 in the Online Supplementary Document includes centrality index diagrams and outcomes of centrality bootstrapped difference evaluations. Table S2 in the Online Supplementary Document presents the partial correlation coefficients for all edges across all nations.

Figure 1. Network configuration and centrality difference evaluations of lifestyles (Panels A–B), health outcomes (Panels C–D), and combined (Panels E–F) in mainland China. The abbreviations for nodes in Panels A, C, and E are available in Table 1. In Panels B, D, and F, a grey cell signifies that there is no significant difference between the respective two variables. A dark cell indicates that there exists a significant difference between the corresponding two variables at the 5% significance level. A white cell showcases the value of expected influence or anticipated bridge influence.

Health outcome networks among countries

Network stability

Narrow confidence intervals across nations validated the accuracy of estimated edge weights (Figure S5 in the Online Supplementary Document). Moreover, CS-C values (Table 2; Figure S6 in the Online Supplementary Document) varied from 0.21 to 0.75. Australia was the sole country under the 0.25 threshold, resulting in 28 interpretable health outcome networks.

Network configuration and central health outcome

For Mainland China, (Figure 1, Panel C) reveals a network configuration with 51.3% (40 out of 78) of edges being nonzero. The most substantial connections were found between H7–H6 (Table 1) and H5–H4 (both 0.67), along with H2–H1 (0.42) (Table 2). Quality of life (H5) was recognized as the most impactful health outcome, corroborated by centrality index and bootstrapped difference assessments (P<0.05) (Figure 1, Panel D). For the other 27 nations, Figure S3 in the Online Supplementary Document delineates the network configurations, with nonzero edge percentages ranging from 15.4 in the UK to 68.0% in Indonesia (Table 2). Figure S4 in the Online Supplementary Document encompasses centrality index diagrams and findings of centrality bootstrapped difference evaluations. Table S3 in the Online Supplementary Document provides the partial correlation matrix for all edges across all networks.

Bridge networks across nations

Network stability

Figure S7 in the Online Supplementary Document illustrates the accuracy of estimated edge weights for each nation with narrow confidence intervals, denoting high precision. Additionally, CS-C values for bridge networks spanned 0.00–0.75 (Table 2; Figure S8 in the Online Supplementary Document). Ten countries, including Australia, Egypt, Italy, Lebanon, Macau, Republic of Sudan, Rwanda, Saudi Arabia, Singapore, and South Africa, fell below the 0.25 threshold, yielding 19 interpretable bridge networks.

Network configuration and bridge lifestyle

Figure 1, Panel E depicts Mainland China’s network where 144 of 378 edges are nonzero. The total amount of exercise (L18) emerged as the most significant bridge lifestyle, considerably surpassing other nodes as indicated by centrality bootstrapped difference evaluations (P<0.05) (Figure 1, Panel F). The corresponding bridge edge is L18–H1 (weight loss). For the other 18 countries, Figure S3 in the Online Supplementary Document exhibits network configurations, where nonzero edge percentages range from 6.1 in Guatemala to 46.6% in Indonesia (Table 2). Figure S4 in the Online Supplementary Document contains centrality index diagrams and results of centrality bootstrapped difference evaluations. Table S4 in the Online Supplementary Document provides the partial correlation matrix for all networks.

Summary of the central and bridge variables in networks across 29 nations

Nine categories of central lifestyle were identified among 28 countries:

• home cooking (L7) in Guatemala, Lebanon, Macau, Mainland China, Nigeria, Philippines, Republic of Sudan, Rwanda, Saudi Arabia, South Africa, and Egypt
• food types in daily meals (L1) in Burundi
• less tobacco smoking (L11) in Indonesia and Malaysia
• reduced alcohol consumption (L12) in Rwanda and Vietnam
• shorter sitting duration (L13) in Canada, Rwanda, and Singapore
• lower snack consumption (L4) in Brazil, Chile, Mexico, South Korea, and the UK
• fewer sugary drinks (L5) in Hong Kong, India, Libya, Mexico, and Thailand
• having meals at home (L6) in Egypt and Libya
• utilizing alternative medicine or natural health products (L9) in Australia.

Six categories of central health outcomes were identified among 28 nations:

• social support received (H9) in Indonesia, Malaysia, and Philippines
• physical health (H3) in Burundi
• sleep quality (H4) in Egypt, Guatemala, Saudi Arabia, and Thailand
• quality of life (H5) in Brazil, Chile, India, Italy, Mainland China, Mexico, and USA
• reduced mental…

“`html

Figure 2 illustrates the spread of primary lifestyles, health outcomes, and bridge lifestyles throughout all participating nations.

Figure 2.  Primary lifestyles, primary health outcomes, and bridge lifestyles categorized by nation.

DISCOURSE

This investigation utilised network analysis to explore intricate connections between lifestyle components and health outcomes across 29 nations, identifying crucial elements that could act as potential predictive indicators or intervention targets. We distinguished four primary lifestyle categories: dietary behaviours, utilization of natural health products, substance consumption, and inactive lifestyles; alongside four critical health outcomes: physical wellness, emotional turmoil, social support, and quality of life. Furthermore, three bridge lifestyle factors – daily meal varieties, home cooking, and overall physical activity levels – were noted as key areas for targeted resource investment to improve health outcomes network-wide, given the established cause-and-effect relationships between lifestyle and health. Adjusting central lifestyle components can enhance the wellness of all within-network lifestyles and promote long-term health benefits. The primary health outcomes indicate prospects for interventions beyond lifestyle alterations, such as improved utilization of healthcare services. Bridge lifestyles present specific strategies for boosting overall health through targeted modifications. Shared primary or bridge variables across different regions underscore the potential for collaborative strategies that harness common strengths while addressing mutual challenges.

In 28 lifestyle networks, four primary lifestyle categories surfaced: dietary behaviours, use of natural health products, substance consumption, and sedentary behaviours, indicative of complex cultural, economic, and public health interactions. In particular, Australia’s emphasis on natural health products as a primary lifestyle, likely propelled by robust cultural acceptance and preference for alternative health methods [26], fosters preventive lifestyles like nutrition and exercise. Conversely, the high alcohol intake observed in Indonesia and Malaysia is swayed by prevalent binge drinking in Malaysia and increasing risks tied to illicit home-produced alcohol in Indonesia [27,28]. Notably, tobacco use held a significant presence in Rwanda and Vietnam, probably due to cultural practices and elevated smoking rates [29,30]. Decreased sitting times, particularly relevant in Rwanda and Singapore, aligns with urban growth and pandemic-induced spikes in sedentary conduct [31,32]. Moreover, dietary behaviours, such as specific food selections, diminished intake of snacks and sugary beverages, home cooking, and dining at home, are crucial in Rwanda and 23 additional countries. The significance of diet within the lifestyle network may be elucidated by its fundamental role in everyday routines and its profound ties to cultural and economic circumstances [33,34], along with its coexistence with other lifestyle components through behavioural and psychological channels, such as self-monitoring and discipline. These pivotal behaviours not only forecast comprehensive lifestyle patterns but also act as catalysts for broader changes, suggesting that efforts focused on these domains could substantially affect entire lifestyle networks. On a global scale, effective approaches to tackle these dietary issues might include enhancing nutritional education and awareness, improving food accessibility and affordability, and subsidizing nutritious foods or imposing levies on unhealthy options to capitalize on diet’s centrality for holistic enhancements in overall healthy lifestyles.

Within 28 health outcome networks, we pinpointed four main categories of primary health outcomes: physical wellness (including sleep quality), emotional distress, social support, and overall quality of life. Social support received played a central role in Indonesia, Malaysia, and the Philippines, reflecting their shared cultural values [35,36]. Physical wellness featured prominently in Burundi, indicating direct concerns regarding bodily health, while sleep quality was central in Egypt, Guatemala, Saudi Arabia, and Thailand, possibly linked to hot climates affecting sleep [37]. Emotional turmoil or psychological burden became significant health issues in countries like Canada, Hong Kong, India, Lebanon, Libya, Macau, Nigeria, Republic of Sudan, Rwanda, Singapore, South Africa, South Korea, UK, and Vietnam, driven by multifaceted factors. Particularly, political strife and economic uncertainty considerably impacted Lebanon, Libya, Sudan, and Nigeria [38], while rapid urbanization and economic pressures affected Singapore, Hong Kong, South Korea, Macau, and others [39]. Quality of life, a multifaceted outcome incorporating physical health, mental health, and social connections [40], emerged as significant in Brazil, Chile, India, Italy, Mainland China, Mexico, and the USA. This underscores the diverse demands for health outcome interventions in each nation, aiming at physical, psychological, and social elements beyond lifestyle modifications to effectively enhance comprehensive health.

In 19 countries with a comprehensible bridge network, our research identified three notable bridge lifestyles considerably influencing overall health outcomes: daily food varieties in Libya, home cooking in Nigeria, and exercise across other nations. In the framework of network analysis, bridge lifestyle variables act as vital connectors between the lifestyle network and health outcome network. These variables showcase significant influence within the lifestyle network while simultaneously forming strong links to health outcome networks. Focusing on these bridge components in interventions can yield widespread, synergistic advantages by enhancing both the specific variables and correlated lifestyle behaviours, resulting in improved overall health outcomes. Libya’s dependency on certain meal types, affected by its desert climate, substantial reliance on food imports, and political unrest, underscores the essential role of dietary choices in health outcomes [41,42]. In Nigeria, home-cooked meals may influence health through nutritional oversight of diets and economic relief owing to cost-efficiency [43,44]. The remaining 17 countries or regions identified exercise as their bridge lifestyle. The well-documented benefits of physical activity on mental and physical health may clarify this consistent outcome.

“`across multiple countries [45], indicating that tactical improvements in exercise availability could result in significant health benefits. Worldwide, successful approaches to enhance exercise availability encompass upgrading public facilities, establishing workplace health programmes, endorsing community projects, embedding physical activity into city planning, executing educational campaigns, ensuring inclusivity, collaborating with health service providers, utilizing digital platforms, and promoting research and assessment. Given the well-documented cause-and-effect link between lifestyle elements and health results, focused resource distribution and heightened actions on bridging lifestyles are vital for maximizing these changes across varied contexts.

This research highlights target areas that necessitate increased resources and focused interventions but does not clarify which types of strategies are the most effective. It presupposes that nations will implement context-validated strategies. While all-encompassing interventions for every lifestyle element are optimal, they may only be practical in financially robust environments. Network theory posits that prioritizing core lifestyles, pivotal health outcomes, and bridging lifestyles yields more significant health advancements than merely addressing other elements independently, which is especially pertinent for settings with limited resources. For instance, affluent Macau, moderately affluent mainland China, and impoverished Sudan recognize home cooking as a fundamental lifestyle element. In resource-rich Macau, directing considerable resources toward home cooking can enhance other lifestyles, whereas additional funding can specifically target other factors. In resource-constrained Sudan, assigning a larger share of resources to home cooking can optimize effectiveness. Likewise, focusing on key health outcomes and bridging lifestyles can comprehensively enhance health results. Policymakers should create cost-efficient programmes centered on these critical variables and implement context-specific strategies.

Limitations

Several constraints were acknowledged. Initially, our online recruitment could introduce selection bias, potentially underrepresenting individuals with low socio-economic status and restricted digital proficiency, which may distort the relationships between lifestyle and health outcomes. Future studies should employ stratified sampling and offline data gathering to enhance representativeness and corroborate findings. Secondly, our dependence on self-reported data is susceptible to recall bias, social desirability bias, and inaccurate self-evaluation, likely skewing the observable relationships among variables in the network. Although this was mitigated by validation queries and stringent data quality controls, these limitations warrant caution when interpreting the network configurations and their implications. Future studies should integrate more precise evaluations, such as wearable technology or 24-hour recall, to improve validity. Thirdly, methodological challenges inherent in network analysis, including assumptions of sparsity and the potential for overfitting, may influence the reliability and generalizability of the identified network structures. The estimation techniques of the network rely on statistical models that may not fully encapsulate the intricate relationships between variables, and minor alterations in data can result in distinct network arrangements. Thus, our network analysis findings should be understood with these methodological limitations in mind. Fourth, partial correlations within networks were determined by individual perceptions of lifestyle and health outcome alterations driven by the external factor of the COVID-19 pandemic. Although we expect these alterations will reverse in equal magnitude but opposite direction, maintaining the relationships among variables post-pandemic, these findings require future validation during non-pandemic times. Fifth, the cross-sectional nature of our study during the COVID-19 pandemic confines our findings to correlational insights and prohibits the analysis of temporal network dynamics. By evaluating outcomes in relation to pre-pandemic conditions instead of their current status, the study’s applicability to the post-pandemic era may be limited. However, if we posit a rebound effect of equal magnitude and opposite direction, the insights could retain their significance in a post-pandemic landscape. To substantiate and extend these insights, we recommend conducting longitudinal studies and randomized controlled trials following the pandemic. Finally, small sample sizes in several countries restricted network interpretability and robust comparisons, especially regarding bridging lifestyles. Expanding sample sizes in future research could remedy this challenge.

CONCLUSIONS

This research outlined central lifestyle, health outcome, and bridging lifestyle variables across 29 nations. Adjustments in these identified variables possess the capacity to greatly enhance comprehensive lifestyle and holistic health outcomes, as well as holistic health consequences through lifestyle, respectively. Their significant impact highlights the necessity for targeted interventions and enhanced resource allocation to attain effective, thorough health improvements. The variability in these variables across various nations—shaped by differing cultural, socio-demographic, political, and economic factors—underscores the importance of tailored strategies that capitalize on the unique effects of central or bridging lifestyles. Furthermore, the similarities found among several countries point to potential opportunities for collaborative initiatives to advance global health.

Additional material

Online Supplementary Document

Acknowledgements

We would like to express our gratitude to Miss Bobo Chan for her exceptional administrative support of the project.

Ethics statement: This research was sanctioned by the Institutional Review Board of the University of Hong Kong—the Hospital Authority Hong Kong West Cluster (reference no: UW 20-272). Informed consent was secured from all participants, ensuring they were informed of the study’s aims, data gathering methods, and their rights. To safeguard data confidentiality and manage sensitive health information discreetly, all participants completed the survey anonymously, and data storage and analysis were executed in strict adherence to the Declaration of Helsinki, maintaining the highest standards of data security and confidentiality.

Data availability: Data are accessible upon reasonable request from the corresponding author.