Co-occurrence of and factors associated with health risk behaviors among adolescents: a multi-center study in sub-Saharan Africa, China, and India

Introduction

Health risk behaviors (HRBs) are prevalent during adolescence (10–19 years old), typically occurring between the ages of 10 and 19 years old.1, 2, 3 These behaviors encompass a wide range of activities, including smoking, poor nutrition, alcohol consumption, and physical inactivity, collectively known as “SNAP” risk factors.⁴ These factors can contribute to the development of non-communicable diseases later in their life and significantly increase the morbidity and mortality in low-income countries.^4,5 The World Health Organization's Global Burden of Disease study estimated that HRBs account for approximately two-thirds of all premature deaths in adults that are associated with childhood risk behaviors.⁶ Besides, risky sexual behaviors also pose a substantial health concern for adolescents on a global scale. For example, it was estimated that 21%–64% unmarried adolescents did not use contraception during their sexual activity in sub-Saharan Africa.⁷ Such behaviors can not only lead to early pregnancy and childbirth complications which are the primary causes of mortality among girls aged 15–19 years, but also result in sexually transmitted diseases which are infected by one million people every day.^8,9 Specifically, there were 93,000 African adolescents aged 15–19 infected with HIV in 2022, accounting for over 70% of HIV new infections globally.¹⁰

Co-occurrence of HRBs refers to the phenomenon where adolescents engage in multiple types of HRBs simultaneously.¹¹ This can include various combinations of substance use, unhealthy dietary habits, physical inactivity and risky sexual behaviors. The clustering of HRBs poses significant health risks during adolescence and later during adulthood as the cumulative effects of multiple unhealthy behaviors can increase vulnerability to chronic diseases including physical and psychosocial health issues.¹² In reality, HRBs seldom occur in isolation but tend to co-occur, often in a synergistic manner.¹¹ Some classical theories have illustrated the mechanism of HRBs co-occurrence. Gateway theories posit that involvement in one form of risk behavior can increase the exposure and desire for additional risky behaviors, and diminish the perceived danger of other risky behaviors.¹³ The problem behavior theory posits a set of shared risk factors for engaging in risky behaviors, with protective factors diminishing the likelihood of problem behavior while risk factors augment it.¹³ Substantial evidence suggests when HRBs co-occur among adolescents, their combined effects can be more detrimental than when they occur independently.¹¹ For instance, the heightened prevalence of four common risk factors has been associated with a significant 3.35-fold increase in non-communicable disease mortality.¹² These effects are particularly pronounced in low- and middle-income countries (LMICs).¹⁴

Consequently, preventing the co-occurrence of HRBs assumes paramount significance. Addressing individual risk behaviors in isolation is unlikely to yield enduring results,¹⁵ and addressing multiple HRBs concurrently is proved to be more cost-effective, particularly in light of constrained healthcare resources.¹⁶ Currently, most health policies regarding adolescent HRBs are managed in a nonoverlapping risk-specific way.¹⁷ Shifting to preventive approaches that focus on shared risk factors is essential for effectively addressing multiple risk behaviors.^17,18 For instance, school-based education has been advocated as an effective approach to mitigate multiple HRBs in adolescents by teaching life skills that target these risk factors and by fostering personal assets like positive mental health and health-related attitudes.^17,18 Additionally, family-targeted interventions addressing family connectedness, parental attitudes and involvement in HRBs have been trialed in various countries to mitigate multiple HRBs in adolescents.^17,18 However, the majority of such interventions aimed with multiple HRBs simultaneously are primarily available in only high-income countries.17, 18, 19

To enhance the interventions of multiple HRBs among adolescents in LMICs, there is a pressing need to consolidate scientific evidence regarding their potential co-occurrence of HRBs. In Asia, a cross-sectional study involving 3578 adolescents across 11 districts in Malaysia revealed that one-third of adolescents exhibited the co-occurrence of two or more HRBs.²⁰ Similarly, in Nepal, a cross-sectional survey of 1108 adolescents revealed a prevalence of 40.7% having co-occurrence of HRBs.⁵ In China, a cross-sectional study involving 4625 adolescents indicated that 67.7% of males and 58.1% of females were engaged in multiple HRBs.²¹ The situation is even more pronounced in sub-Saharan Africa, home to the largest proportion of adolescents and youths, where an estimated 89.2% of adolescents were reported to engage in at least 2 HRBs, based on data pooled from nine countries.²² In Liberia, a cross-sectional study found 65.2% of middle school students were at moderate or high risk of having multiple HRBs,²³ while in Kenya, this figure stood at 77.1%.²⁴ In Ghana, a cross-sectional survey with 1763 senior high school students aged 11–19 years found that 94.8% had more than one HRBs.²⁵ Across the world, a global school-based research with 304,779 adolescents aged 11–17 years old from 89 countries found that over 30% had 3 or more HRBs.²⁶ These findings highlight the importance of designing effective and efficient interventions that take advantage of the age of opportunities and the triple dividend of investing in adolescent health.^27,28

It is also imperative to identify the specific risk factors contributing to this complex phenomenon. Several multi-center studies have underscored variations in the influence of sex and age on the co-occurrence of HRBs.^22,26,29 Notably, both male and female adolescents had evident physical inactivity and poor diet, but males were more likely to have substance use.²⁶ And older adolescents had higher odds of reporting co-occurrence of HRBs.²⁶ Besides, a series of national-level evidence from African countries like Liberia, Ghana, and Kenya examined risk factors among distinct clusters of adolescents based on their HRBs and lifestyles.23, 24, 25 It was found that school attendance and level of parental monitoring were significantly associated with the risk of adolescent HRBs clustering.²⁴ A lower socio-economic status was also associated with higher risk of co-occurrence of HRBs among adolescents.²³

However, there is still a critical knowledge gap on HRBs among adolescents in LMICs, where 90% of all adolescents reside in.³ Most of evidence was inferred from school-based sampling frameworks, normally focusing on middle-school adolescents, resulting in a gap of understanding for out-of-school, migrant, younger, and vulnerable adolescents. Therefore, the objectives of this study, as a multi-center analysis integrating adolescents from both schools and communities, are to describe the prevalence of co-occurrence of HRBs among adolescents in six sub-Saharan African countries and two Asian countries, encompassing smoking, alcohol consumption, poor dietary habits, physical inactivity, and risky sexual behavior, as well as identify the shared risk factors.

Methods

Results

Table 1 presents the overall and by country demographic characteristics and HRBs of participants. Of the 9697 participants included in the analysis, 52.28% were females and the median age of the participants was 14.00 years (interquartile range [IQR] 12.00–16.00). The distribution of age and sex among the participants in each country is shown in the Supplementary Table S1. The majority of the individuals were enrolled in an educational institution (88.42%), were not engaged in paid work in the past 12 months (80.14%), and lived with their parents (72.66%). Based on the MacArthur Scale of Subjective Social Status, around two thirds of the participants belonged to the middle social status (65.88%). Of the 9697 participants, 8779 (90.53%) participants had at least one HRB, and 5267 (54.32%) had two or more HRBs concurrently.

Using LCA, we identified three distinct clusters with the lowest AIC, as shown in Supplementary Figure S1. The distribution of HRBs across the three clusters is depicted in Table 2. The observed P-values indicate statistically significant differences in HRBs among three clusters (P < 0.05). Furthermore, Supplementary Table S2 specifically displays the P-values derived from the pairwise comparisons performed using multiple chi-square tests among different clusters. Cluster 1 was regarded as a “Healthy” group. It was comprised of 2689 (27.73%) participants characterized by the absence of HRBs, where all members were physically active, 93.53% were at low to moderate risk of poor diet quality, none of the members had ever smoked, 82.93% had not consume alcohol ever, and 95.80% didn't engage in risky sexual behavior. Cluster 2 was regarded as at high risk of “Physical inactivity and poor diet quality” group. It consisted of 6609 (68.16%) participants with physical inactivity and poor diet quality. Among them, 100% were physically inactive, 55.14% were at moderate risk and 7.91% were at severe risk of poor diet quality. Other HRBs were distributed scarcely, with 1.21% participants ever smoked cigarettes, 10.68% consumed alcohol, and 4.46% engaged in risky sexual behavior. Notably, among all 6813 physically inactive participants, 6609 (97.01%) were distributed in cluster 2, and among all 725 participants with high risk of poor dietary habits, 523 (72.14%) were also in cluster 2. Cluster 3 was regarded as at high risk of “Smoking, alcohol consumption and risky sexual behavior” group. It comprised of 399 (4.11%) participants with smoking, alcohol consumption, and risky sexual behavior. Among them, 54.64% had smoked cigarettes, 86.97% had drunk alcohol, and 65.16% had engaged in risky sexual behavior. We also used questions about substance use during last 30 days as a sensitivity analysis. The cluster result was shown in Supplementary Table S3. Similarly, it can be inferred that physical inactivity was also grouped with poor diet in Cluster 2, and risky sexual behavior was also grouped with smoking and alcohol consumption in Cluster 3.

Table 3 shows the results of regression models. Compared to Cluster 1, membership in Cluster 2 was negatively associated with older age (aOR 0.90, 95% CI 0.82–0.99) and having a high socio-economic status (aOR 0.80, 95% CI 0.68–0.93), but positively associated with being female (aOR 1.15, 95% CI 1.05–1.26), not being engaged in paid work (aOR 1.26, 95% CI 1.11–1.42), and living alone (aOR 1.46, 95% CI 1.21–1.76). After adjusting countries as the covariate, the role of older age (aOR 0.96, 95% CI 0.86–1.07), living alone (aOR 0.99, 95% CI 0.80–1.22) and high socio-economic status (aOR 1.01, 95% CI 0.85–1.19) were not significant, while other factors have similar results. Membership in Cluster 3 was negatively associated with being female (aOR 0.43, 95% CI 0.34–0.55), not being engaged in paid work (aOR 0.67, 95% CI 0.52–0.86), but positively associated with older age (aOR 3.99, 95% CI 3.06–5.21), not being enrolled in an educational institution (aOR 1.98, 95% CI 1.51–2.59), and living with guardians other than parents (aOR 1.93, 95% CI 1.49–2.50). After adjusting countries as the covariate, the significance of the factors kept the same.

Discussion

We examined the co-occurrence of health risk behaviors (HRBs) and its determinants among adolescents in six sub-Saharan African countries and two Asian countries. We found a high burden of HRBs among adolescents in these countries. Three distinct clusters of HRBs were identified: “healthy,” “physical inactivity and poor dietary quality,” and “smoking, alcohol consumption and risky sexual behavior”. Moreover, our findings demonstrated that several factors including sex, age, educational enrolment, engagement in paid work, and family cohabitants were associated with the presence of HRBs.

Our findings indicated a remarkably high prevalence of HRBs, with more than 90% of adolescents engaged in at least one HRB, with nearly 55% engaged in two or more HRBs. This level of co-occurrence of HRBs aligns with previous reports from high-income countries such as the United States of America, United Kingdom, and Netherlands, reporting the prevalence of two or more HRBs co-occurrence between 50% and 70%.¹¹ Our findings corroborate a study conducted across 89 countries, which similarly reported a high prevalence of co-occurrence of HRBs, affecting 82.4% of adolescents.²⁶ This pervasive co-occurrence of HRBs can be attributed, in part, to a dearth of health knowledge among adolescents, contributing to the high prevalence of poor dietary quality and physical inactivity in sub-Saharan Africa. Additionally, risk behaviors such as cigarette use and excessive alcohol consumption may be perceived as less harmful by society and policymakers, resulting in uncertainty about acceptable tolerance levels for these risk behaviors within the community.⁴⁰

In our study, the majority of participants exhibited a pattern of physically inactivity and consumed unhealthy foods, which aligns with a substantial body of literature exploring the interconnectedness between diet and physical activity. This trend may be partially attributed to the socio-economic status of both households and neighborhoods.⁴¹ Previous studies have consistently demonstrated that higher socio-economic status can positively influence both diet quality and physical activity levels simultaneously.⁴² Besides, our findings regarding the impact of school enrollment are in harmony with existing research that underscores the effectiveness of interventions targeting both dietary habits and physical activity. Consequently, there has been a persistent call for the implementation of school-based programs aimed at addressing these two interconnected HRBs.⁴³

While the number of participants who smoked cigarettes, consumed alcohol, and engaged in risky sexual behavior was relatively small, these three HRBs exhibited a tendency to co-occur. This clustering of HRBs during adolescence is a cause for heightened concern, as it aligns with gateway theories, suggesting that the increasing association between risk behaviors during adolescence may pave the way for the aggregation of more risky behaviors as individuals transition into adulthood.⁴⁴ Previous studies have consistently reported a robust association between former smoking and alcohol use.⁴⁵ Some researchers have posited that the association may be attributed to the physiological effects of nicotine stimulation, while others have suggested the possibility of reverse causation, wherein alcohol disinhibits smoking behavior.⁴⁶ Furthermore, smoking and alcohol use have been jointly regarded as a means of socialization. However, individuals who use these substances and exhibit better psychosocial functioning than non-users may paradoxically engage in riskier sexual behaviors, often characterized by reduced condom use and an elevated risk of sexually transmitted diseases.⁴⁷

Our study also found that socio-demographic factors were significantly associated with the co-occurrence of HRBs. Consistent with prior research, female adolescents were physically less active than their male counterparts.⁴⁸ This discrepancy may attributed to less social support girls received to engage in physical activities.⁴⁹ Besides, compared to boys, girls also perceived less enjoyment when taking physical education.⁵⁰ Moreover, an analysis focused on obesogenic clusters also identified a cluster characterized by poor nutrition status and physical inactivity that was specific to girls in America.⁵¹ Similar to south Africa, this is primarily attributed to dietary habits like increased exposure to sugar sweetened beverages and salty food, changes in snacking habits, and also physical activity habits like sedentary lifestyles.^52,53 Our study also identified significant correlates between socioeconomic factors and HRBs. Adolescents who lived without parental care or in low socio-economic settings were at a high risk of poor diet quality and physical inactivity. These findings are consistent with the earlier reports from Europe where adolescents from low socioeconomic settings were more likely to engage in multiple HRBs.54, 55, 56 Importantly, this finding is particularly salient for LMICs, where poverty-related issues and poor nutrition remain a critical public health concern.⁵⁷ Consistent with previous research, our results indicated sex and age differences in HRBs. Male and older adolescents were more likely to smoke cigarettes and alcohol, as well as engage in risky sexual behaviors.⁵⁸ Similar HRB patterns have been observed in high-income countries such as the United Kingdom and Australia.^54,59 Moreover, attending at an educational institution emerged as a protective factor against HRBs. According to the Social Control Theory, strong bonds with educational institutions can foster traditional social values and discourage substance use.⁶⁰ The Social development model also supports the importance of schools in preventing HRBs.⁶⁰ What's more, our findings also indicated that adolescents engaged in paid work and those living with non-parental guardians were more prone to engaging in these HRBs. Prior studies have similarly revealed that factors such as low parental monitoring, peer influence, and socio-economic disadvantage contribute to an increased risk of substance use and risky sexual behaviors.⁴⁶

This study has multiple methodological, theoretical, and practical strengths. From the methodological aspect, we used LCA to identify the clusters of HRBs among adolescents. In contrast to traditional methods that involve merely counting the number of different behaviors, LCA enabled us to reduce the dimensionality of variables, facilitating a more explainable and practical analysis. Additionally, LCA is adept at categorizing individuals into exclusive subgroups or latent classes based on multiple categorical variables, offering more nuanced and dynamic insights than other alternative methods.⁶¹ From a theoretical aspect, this study contributes significantly to addressing the dearth of evidence regarding the co-occurrence of HRBs in LMICs. Specifically, it differs from many previous ones that primarily collected adolescent data in schools. This study used two population-based settings to collect data including the Health and Demographic Surveillance System and schools. In addition, this study uses data from multiple sites to understand the clustering of risk behaviors. From a practical aspect, our findings offer valuable insights for informing prevention in LMICs. The identification of distinct clusters of HRBs and their factors provides a crucial framework for designing tailored preventive policies and programs. By addressing the interplay between social determinants and health behaviors, we can effectively prevent the burden of multiple lifestyle risk factors and their long-term health and well-being consequences. Specifically, policies should support adolescent girls with more physical activity and higher quality of diet, while motivate boys to stay away from substance use and risky sexual behavior. Our research underscores the importance of implementing behavior change communication initiatives in both schools and communities to prevent multiple HRBs. To address the research gap in this field, more efforts should be made to invest in and develop scientific evidence in Sub-Saharan Africa to better understand the co-occurrence and interplay of various HRBs among adolescents. It is imperative to establish comparative studies across varying income levels to gauge the prevalence of HRB co-occurrence, clusters, and associations. Additionally, expanding community-based evidence inclusive of all adolescent age groups is crucial for a comprehensive understanding.

Our study has some limitations. First, our survey involved school-based and community-based settings. Although the survey strategies and procedures were strictly controlled to be the same in both settings, there is an omission of out-of-school children in school-based settings compared to community-based settings. To examine whether different sampling settings affect the clustering of HRBs, we conducted two separate LCAs for school- and community-based participants. As shown in Supplementary Tables S4 and Table S5, the clustering patterns among participants from these two different settings did not vary. Secondly, all HRBs and socio-demographical information were self-reported by the participants, introducing the potential for recall and social desirability biases. In particular, participants may have provided inaccurate estimates while responding to sensitive questions related to age at first sexual intercourse, contraceptive methods usage, smoking cigarette, and alcohol consumption. Thirdly, the age distribution participants in China were younger than others because it was based on an existing open prospective cohort study focused on younger adolescents. Consequently, direct comparisons of adolescent HRBs among countries, especially for age-related features like substance use and risky sexual behavior, were not feasible. Therefore, a sensitivity analysis was performed by excluding the Chinese participants from the analytical sample, however, the three clusters of HRB were consistent, regardless of their distribution (Supplementary Table S6). Specifically, the percentage if participants in Cluster 1 decreased and the percentage of participants increased in Cluster 3. Last but not least, there is room for improvement in the measurement HRBs among teenagers. Concerning substance use, if participants had ever smoked cigarettes or had alcoholic drink, they may have already quit these behaviors. But if measuring their behavior with a shorter time window like last 30 days, we may potentially exclude individuals who had such a habit but not took it during the survey. We conducted sensitivity analyses using both measurements to estimate this bias.

This multi-country study provides valuable insights into the high prevalence of co-occurrence of HRBs among adolescents in LMICs. Our findings reveal distinct clustering patterns for five different HRBs and highlight the socio-demographic covariates associated with each cluster. Our study sheds light on the potential shared risk factors of HRBs and underscores the need for tailored interventions that target multiple HRBs in early life. We hope that our findings can inform evidence-informed policies and interventions to promote healthy behaviors and reduce the risk of non-communicable diseases among adolescents in LMICs.

Contributors

WWF led the study design and instrument development and all authors contributed to it. XL completed the statistical analysis and wrote the first draft. HZ examined and verified the data. YD, SS, and KT revised the manuscript. All authors had full access to all the data in the study and had final responsibility for the decision to submit for publication.

Data sharing statement

The data that support the findings of this study are available upon reasonable request from the ARISE network. The data is not publicly available due to research ethics board restrictions.

Declaration of interests

We declare no conflicts of interest.

Appendix A. Supplementary data