Modelling of South African adolescent hypertension: application of frequentist quantile regression on main effect variables and interactions

1 Introduction

Low and middle-income countries (LMICs), especially those in sub-Saharan Africa (SSA), are now bearing a greater burden of hypertension than high-income countries (HICs) (1). With mounting evidence of its rising incidence in developing nations like South Africa, childhood hypertension has become a major global public health concern (2). A recent study based on cross-sectional data from the South African Demographic and Health Survey 2016 comprising of 1,062 adolescents aged between 15 and 19 years reported hypertension prevalence of 26%, where males recorded a prevalence rate of 28.3% compared to females (24.6%) (3). Early-onset hypertension in children is not only an indicator of poor cardiovascular health in youth but also a precursor to adult hypertension and other non-communicable diseases (4, 5). In South Africa, the dual burden of undernutrition and obesity, combined with socioeconomic and racial disparities, further complicates the epidemiology of pediatric hypertension (6).

Continent-wide research shows that hypertension and elevated BP are also substantial in African children and adolescents. An updated systematic review across African studies reported a pooled hypertension prevalence of approximately 7.45% among paediatric populations, with elevated BP prevalence around 11.38%, and combined elevated BP/hypertension up to 21.74% (7).

Recent global systematic reviews suggest that pediatric hypertension is a growing global health issue. A large meta-analysis incorporating 271 studies with data over 3.6 million children and adolescents reported that the global prevalence of paediatric hypertension is around 3.9% for sustained hypertension and occasional (single visit) hypertension is around 11.9% (8). Another global analysis found overall pooled childhood hypertension prevalence of about 4.28%, increasing with age and peaking in early adolescences (9). The same study by Zhou et al. (9), reported that global childhood hypertension nearly doubled over the last two decades (2000–2020), rising from about 3.2 to 6.2%, with obesity identified as a major driving factor.

Beyond estimating prevalence, understanding how risk factors influence different levels of blood pressure is essential, particularly because cardiovascular risk is not uniformly distributed across populations. Traditional modelling techniques such as simple linear, multiple linear or logistic regression typically focus on estimating mean or conditional probabilities, which may obscure crucial information on how covariates affect the entire distribution of blood pressure outcomes. These methods are particularly limited when heteroscedasticity, outliers or non-normal error structures are present, both of which are common in pediatric health data (10). Quantile regression provides a robust alternative by enabling the estimation of conditional quantiles of the response variable, offering insights into how predictors influence various points in the blood pressure distribution, especially the upper quantiles that are critical for hypertension diagnosis.

Despite its advantages, there is limited application of quantile regression in the context of South African child health data, particularly in modelling systolic and diastolic blood pressure across different percentiles. Moreover, interactions between main-effect variables studied in the South African National Income Dynamics Study (NIDS) such as body mass index (BMI), age, gender, race, cigarette consumption, depression, exercise and employment status remain underexplored (11). These interactions may significantly influence hypertension outcomes, but their roles are often omitted due to methodological limitations or interpretational complexity in traditional models. Interactions amongst variables can play a crucial role in predicting diseases such as hypertension (12).

Despite the growing prevalence of childhood hypertension in South Africa, little research has explored how risk factors jointly influence different points of the blood pressure distribution. This gap motivates the application of frequentist quantile regression to uncover both main and interaction effects. Such an approach will offer more nuanced insights into how demographic and health-related predictors interact across the distribution of blood pressure values, potentially informing targeted interventions and policy decisions aimed at early prevention and management.

2 Materials and methods

The data and variables, theoretical models, and data analysis techniques employed in this paper are presented in this section.

2.1 Data and variables

This was a retrospective study using a nationally representative sample drawn from the 2017–2018 Wave 5 Household survey of the South African National Income Dynamics Study. Based on this secondary data, data for South African Male and Female Individuals below 18 years was extracted for analysis.

Before data analysis, data cleansing was done. Observations with missing data for any of the study’s variables and individuals older than 18 were eliminated as part of the data cleaning procedure. Missing data cases present some challenges in conducting data analysis because standard modelling techniques just exclude them from the analysis (13). A final data set of 1812 respondents was obtained from 1900 originally observed and used for data analysis.

The study variables included systolic blood pressure and diastolic blood pressure as the outcome variables. The main effect variables were BMI, age, gender_male, race_african, race_coloured, race_asian/indian, race_white, exercises_never, exercises_once or twice a week, exercises_three or more times a week, cigarette consumption_yes, depression_rarely or none of the time, depression_some or little of the time, depression_occasionally or all of the time, perceived health status_poor, perceived health status_good and perceived health status_excellent.

All categorical variables were converted into numerical dummy variables prior to model fitting. For instance, Gender was coded as 0 = Female and 1 = Male. Continuous variables such as BMI were retained in their original metric. Possible study interactions were to be determined by the pairwise hierarchical interaction technique.

2.2 Frequentist quantile regression

In statistical modelling, regression analysis is one of the most popular and effective multivariate methods for determining how a set of factors, X, influences a specific outcome variable, Y. The traditional mean linear regression is concerned with the estimation of the conditional expectation E ( Y ∣ X ).

Equations 1–3 is a process to derive quantile regression, it is ideal to begin from the traditional mean regression. Consider the standard linear model where: x i ∈ ℜ k and β ∈ ℜ k are column vectors of size kand y i ∈ ℜ is a scalar variable:

y i = x i T β + e i (1)

Assuming that E ( ε ∣ x ) = 0, then the conditional mean model is obtained. If Med ( ε ∣ x ) = 0, then the conditional median model is derived. In the conditional mean model, the regression coefficients are obtained by solving:

β ̂ = argmin β ∈ ℜ k ∑ i = 1 n ( y i − x i T β ) 2 (2)

In median regression, the aim is to find an estimate β that minimizes the sum of the absolute deviations:

β ̂ = argmin β ∈ ℜ k ∑ i = 1 n ∣ y i − x i T β ∣ (3)

Quantile regression can now be derived by extending the median case to all other quantiles of interest:

β ̂ = argmin β ∈ ℜ k ∑ i = 1 n ρ τ ( y i − x i T β ) (4)

Where ρ τ ( x ) = x ( τ − I ( x < 0 ) ) and where I ( ⋅ ) denotes the indicator function.

For any τ ∈ ( 0 , 1 ), the loss function ρ τ assigns a weight of τ to positive residuals and a weight of ( 1 − τ ) to negative residuals.

The quantity β τ ̂ is called the τth regression quantile. The case where τ = 0.5 corresponds to minimising the sum of absolute residuals, i.e., median regression. To minimise the objective function in Equation 4, the method of Barrodale and Roberts (14) was implemented.

2.3 Pairwise hierarchical interactions

Bien et al. (15) state that a pairwise interaction model containing all possible pairwise products of the predictors is presented in the following form:

y = β 0 + ∑ l = 1 L i ∑ i = 1 m β i , l , X i , l + ∑ 1 ≤ i < j ≤ m ∑ l = 1 L i ∑ k = 1 L j θ i , j , l , k X i , l X j , k

The interaction between two independent variables X i andX j, occur when the effect X ion ywill vary depending on the level or value ofX j. Pairwise hierarchical interactions in this study shall be performed in a way that satisfies strong hierarchy. Whenever an interaction is estimated to be nonzero and both main effects are included in the model, then it is said to obey strong hierarchy (16). Weak hierarchy is derived whenever either of the interaction’s main effect variables is present (16).

Interaction is in the form of three ways:

• Interaction between two categorical variables.
• Interaction between a categorical variable and a continuous variable.
• Interaction between two continuous variables.

Interaction between two categorical variables.

Let X 1 and X 2 be two categorical variables, then the interaction between these categorical variables is given by:

X 1 : 2 = [ 1 X 1 ] ∗ [ 1 X 2 ]

= [ 1 X 1 X 2 ( X 1 ∗ X 2 ) ]

Whenever there is an interaction between two categorical variables, interactions are taken at each level of the variable.

Interaction between a categorical variable and a continuous variable.

Let X be a categorical variable with L levels and Z be a continuous variable, then the interaction between these two variables can result into one of the following cases (16):

• μ ij = μ(no main effects, no interactions),
• μ ij = μ + θ 1 i (one main effect X 1 or Z 1),
• μ ij = μ + θ 1 i + θ 2 z(two main effects),
• μ ij = μ + θ 1 i + θ 2 z + θ 1 : 2 i z(main effects and interaction).

Interaction between two continuous variables.

Let Z 1 and Z 2 be two continuous variables, then the interaction between these continuous variables is given by:

Z 1 : 2 = [ 1 Z 1 ] ∗ [ 1 Z 2 ]

= [ 1 Z 1 Z 2 ( Z 1 ∗ Z 2 ) ]

2.4 Data analysis

Descriptive statistics were conducted using IBM SPSS version 30. Frequencies on biographic and lifestyle characteristics of respondents. To investigate possible interactions amongst SBP’s and DBP’s risk factors by applying the pairwise hierarchical interactions technique using the R package, namely hierNet (17). hierNet was employed because it allows the simultaneous estimation of main effects and pairwise interactions while enforcing hierarchical constraints, ensuring that interaction terms are included only when corresponding main effects are present.

The optimal regularisation parameter (λ) was selected using cross-validation. The final λ value (lamhat) produced a hierarchical lasso model that retained all main effects and a subset of pairwise interactions while satisfying strong hierarchy constraints. The relatively large number of retained coefficients suggests that the selected λ prioritised predictive performance over model sparsity. Several interaction effects remained in the model, indicating potential synergistic relationships among predictors. Although some coefficients were small in magnitude, they were retained because they improved overall model fit under the selected cross-validation criterion.

Finally, the SBP and DBP frequentist quantile regression models were fitted using the quantreg R package (18). In order to investigate how blood pressure risk variables impact those who are most at risk for hypertension, the upper quantiles (75th and 95th) were computed in this study.

3 Results

The study’s empirical findings are presented in this section. Tables and figures are used to display these findings. Additionally, this part provides an interpretation of the findings.

3.1 Biographical details

Table 1 presents the demographic characteristics of the study respondents, broken down by gender, race, and age. The gender distribution was fairly balanced, with a slightly higher proportion of females (52.4%) than males (47.6%). The vast majority of participants identify as African, making up more than four-fifths (84.2%) of the sample. The coloured group forms the second-largest racial category (12.7%). Asian/Indian (1.2%) and White (1.9%) individuals make up very small proportions of the total.

Table 1. Biographical details.

Most respondents were either 16 or 17 years old, accounting for more than 85% of the sample combined. Only a small proportion (14.4%) were 15 years old. This age distribution indicates that the current study targeted late adolescents.

3.2 Lifestyle characteristics

Table 2 reveals a breakdown of several lifestyle characteristics across different categories, showing the number of individuals (n) and their corresponding percentage within each category. Most participants (61.9%) were sedentary, with fewer than a quarter (21.1%) exercising regularly, which is a potentially concerning trend for long-term health. A substantial majority 1743 (96.2%), report not smoking, while only 3.8% indicated that they do smoke.

Table 2. Lifestyle characteristics.

Most individuals report experiencing depression rarely or none of the time (less than 1 day), accounting for 73.4% of the sample. A smaller group experiences depression some or little of the time (1–2 days) at 20.9%, and a very small percentage (5.7%) experiences it occasionally or all of the time (3–7 days). A vast majority, 81.0%, perceive their health as excellent, and 16.7% consider it good. Only a small fraction, 2.3%, report poor perceived health.

The largest group falls into the healthy BMI range (18.50–24.99) at 64.3%. Interestingly, 21.2% are categorised as underweight (<18.50). A smaller proportion is overweight (25.00–29.99) at 10.0%, and only 4.5% are classified as obese to morbidly obese (30.00). This distribution suggests a relatively healthy weight profile for most individuals, though a notable portion is underweight.

3.3 Pairwise interactions for DBP

The main effects and interactions retrieved for the DBP model are shown in this section. The pairwise interactions for DBP were conducted using 8 explanatory variables which consists of 2 continuous and 6 categorical variables. Categorical variables were treated in the model as per their respective levels.

It can be deduced from Table 3 that 6 main effects and 0 interactions were obtained as possible predictors of DBP.

Table 3. Main effects and interactions for DBP.

3.4 Frequentist quantile regression estimates for DBP’S risk factors

Table 4 depicts the Frequentist Quantile Regression for DBP’s estimates for risk factors associated with DBP at two upper quantiles (i.e., the Q (75) and the Q (95)) of the DBP distribution. BMI and Race (Coloured) were the only statistically significant predictors of elevated DBP, particularly among individuals in the upper tail of the diastolic blood pressure distribution. Exercises (Never), Exercises (Once or Twice a Week), Depression (Occasionally or All the Time) and Perceived Health Status (Good) did not show statistically significant effects at these higher quantiles as evidenced by confidence intervals spanning zero.

Table 4. Frequentist quantile regression estimates for DBP’s risk factors.

*p-value < 0.05; **p-value < 0.01; ***p-value < 0.001.

3.5 Pairwise interactions for SBP

Similarly to DBP, the pairwise interactions for SBP were conducted using 8 outcome variables consisting of 2 continuous and 6 categorical variables. Categorical variables were utilised in the model as per their respective levels using dummy coding.

The results presented in Table A1 (Appendix A) suggest that 16 main effects and 15 interactions were deduced for the SBP model. These results imply that SBP can be predicted by the 16 main effect variables and the 15 interactions detected.

3.6 Frequentist quantile regression estimates for SBP’s risk factors

Table 5 displays the frequentist quantile regression estimates for SBP’s risk factors at the 75th and 95th percentiles of the SBP distribution. Gender Male, Cigarette Consumption Yes, AgeGender Male and Race AfricanCigarette Consumption Yes had statistically significant influence on SBP at both the higher ends of its distribution. These findings indicate that these factors exert a persistent influence on SBP among individuals with moderately high and extremely high blood pressure levels.

Table 5. Frequentist quantile regression estimates for SBP’s risk factors.

*p-value < 0.05; ** p-value < 0.01; *** p-value < 0.001.

Risk factors that were statistically significant at only the 95th percentile included Perceived Health Status (Good), BMIGender (Male) and Cigarette Consumption Depression (Occasionally or All of the Time). These factors were not significant at the 75th percentile, suggesting their effects become more pronounced at the upper tail of the SBP distribution.

4 Discussion

This study provides valuable insights into the associations between selected main-effect variables, interaction terms, and elevated DBP and SBP at the upper ends of the blood pressure distribution (75th and 95th percentiles) using frequentist quantile regression. The findings demonstrate that several demographic, behavioural, and health-related factors are associated with higher blood pressure among South African adolescents, although the cross-sectional design does not permit causal inferences.

Across all upper quantiles, an increase in BMI was observed to positively associate with elevated DBP only. These results suggest a consistent and robust relationship between higher BMI and elevated DBP levels across the upper spectrum of the DBP distribution. The implication is that overweight or obese individuals are more likely to experience elevated peripheral vascular resistance, increased sympathetic nervous system activity, insulin resistance, and activation of the renin angiotensin aldosterone system (RAAS), all of which contribute to higher DBP, a finding consistent with previous studies that obesity was observed as a significant risk factor for hypertension among children and adolescents (8, 19).

The “Race (Coloured)” variable was associated with DBP, but only at the highest quantile Q (0.95). This indicates that being of “Coloured” race is linked with a substantially higher DBP, specifically for individuals who are already at the extreme upper end of the DBP distribution. Similar racial disparities in hypertension have been documented in South Africa, where differences in socioeconomic conditions, access to healthcare, dietary patterns, environmental exposures, and behavioural risk factors contribute to unequal cardiovascular outcomes (3). Furthermore, previous analyses of South African hypertension data reported a higher prevalence of elevated blood pressure among Coloured populations than among several other racial groups, supporting the present findings (20). This finding underscores the importance of considering racial disparities in health outcomes, particularly among those with severe BP. A study conducted in the United States of America (USA) found that ethnic and racial disparities in the prevalence and treatment of hypertension do exist (21).

Gender (Male) showed a substantial positive association with SBP at both quantiles, suggesting that men are more likely to have elevated SBP at the upper end of the distribution. The increased effect at the 95th percentile indicates an even more pronounced difference in SBP for male adolescents at very high SBP levels. These findings are consistent with extensive evidence showing that men generally exhibit higher blood pressure levels and greater hypertension prevalence during early and middle adulthood (22). However, this finding of the current study does not support the previous research conducted in South Africa, which found that being a female adolescent had a significant effect on hypertension (3).

Cigarette consumption (Yes) was positively related to higher SBP at both quantiles, reinforcing smoking as a critical risk factor for hypertension. Smoking contributes to endothelial dysfunction, oxidative stress, arterial stiffness, inflammation, and increased sympathetic nervous system activity, all of which elevate blood pressure and cardiovascular risk (23). This present finding seems to be consistent with other research, which found that tobacco exposure is a modifiable risk factor associated with the presence of elevated blood pressure in US children and adolescents (24).

The interaction between Age and Gender (Age*Gender Male) was also significantly positive at both quantiles, indicating that the effect of ageing on SBP is more pronounced in male adolescents. In other words, as males age, their SBP tends to increase more steeply than females, especially among those with already elevated SBP. This outcome observed in this study mirrors those of the previous studies that male adolescents are more likely than female adolescents to have hypertension, and this disparity widens with age, especially during and after puberty (25). Males often have greater systolic and diastolic blood pressures, with a more noticeable increase during puberty, possibly as a result of the influence of androgens like testosterone (26). Clinically, this finding suggests that male adolescents may benefit from earlier and more frequent blood pressure monitoring during adolescence, particularly during pubertal transition. Early identification of accelerated increases in SBP among males may facilitate timely lifestyle interventions and reduce the likelihood of progression to adult hypertension and future cardiovascular disease.

The interaction between being African and smoking (Race African*Cigarette Consumption Yes) is statistically significant and notably large at both quantiles. This finding is particularly notable as it suggests that among individuals who smoke, those identified as African experience a substantially greater increase in SBP, especially at the highest SBP levels. This implies a synergistic association between race and smoking on SBP risk within this specific group suggesting that racial and behavioural factors jointly influence hypertension risk and may reflect broader social and environmental determinants of health (20). This finding corroborates the previous work of Liu et al. (27), who suggested that smoking and being Black together may raise the risk of hypertension more than the impact of each factor alone. From a clinical perspective, this interaction identifies a particularly vulnerable subgroup that may require intensified smoking cessation programmes, targeted cardiovascular risk screening, and culturally sensitive prevention strategies. Such targeted interventions may help prevent the development of severe hypertension and its associated complications, including stroke, heart failure, and chronic kidney disease.

Several variables became significant only at the 95th quantile, suggesting that their effects are concentrated among individuals with extremely elevated SBP. The effect of the predictor interaction of BMI and male gender was statistically significant only at the 95th quantile, which implies that obesity has a deleterious impact on SBP among males with severe hypertension. The significance of this interaction only at the extreme upper tail highlights the value of quantile regression in uncovering risk factors that may remain hidden in conventional analyses (28).

Clinically, this finding suggests that obese male adolescents represent a high-risk group for severe elevations in systolic blood pressure. Consequently, weight-management interventions, nutritional counselling, and regular cardiovascular assessments should be prioritised for overweight and obese males to prevent the progression of hypertension and the early onset of cardiovascular target-organ damage.

The perceived health status (good) had an inverse relationship with the SBP at the 95th percentile, indicating that people with more positive perceptions of their health status had lower systolic blood pressures within the more hypertensive group. Despite the fact that self-assessed health is subjective, it is now seen as an important factor in predicting cardiovascular events due to its ability to assess the broad aspects of health beyond clinical parameters.

The interaction between cigarette consumption and depression (occasionally or all of the time) was also significant at the 95th quantile. This finding suggests that smoking and depressive symptoms jointly exacerbate systolic blood pressure among individuals with severe hypertension. The coexistence of depression and smoking has been linked to increased stress responses, poorer health behaviours, and greater cardiovascular risk. Individuals experiencing both risk factors may therefore represent a particularly vulnerable subgroup requiring integrated behavioural and mental health interventions. The clinical relevance of this finding lies in the need for multidisciplinary approaches to hypertension management. Screening for depressive symptoms in adolescent smokers with elevated blood pressure may enable healthcare providers to address both psychological and behavioural risk factors simultaneously. Integrated smoking cessation and mental health support programmes may therefore be more effective in reducing cardiovascular risk than interventions targeting either condition alone.

These results reinforce the importance of adopting targeted hypertension prevention strategies that focus on high-risk groups rather than relying solely on population-average effects. Interventions aimed at reducing obesity, promoting smoking cessation, improving cardiovascular screening among men, and addressing racial inequalities in healthcare access may yield substantial benefits among individuals located at the upper tail of the blood pressure distribution. Furthermore, the significance of several interaction effects demonstrates that hypertension risk factors rarely operate independently, highlighting the need for multifactorial approaches to hypertension prevention and management.

5 Implications of study findings and recommendations

5.1 Weight management remains crucial

The positive association between higher BMI and both DBP and SBP at the upper tail underscores overweight/obesity as an important risk factor of extreme blood pressure in youth. Interventions promoting healthy weight through nutrition, physical activity, and school/community programs are likely to reduce the prevalence of severe hypertension risk.

5.2 Targeted racial considerations for severe BP

The selective association of Race (Coloured) with DBP at the 95th percentile points to the need for targeted monitoring and culturally tailored interventions for youths at greatest risk of extreme DBP. Public health efforts should ensure equitable access to screening, preventive care, and early treatment options for marginalized groups.

5.3 Gender-specific risk trajectories

The stronger association of male gender with elevated SBP at the upper quantiles, and the Age*Gender interaction indicating faster SBP rise with age in boys, suggest adolescence-specific strategies. Programs focused on male youths; including cardiovascular health education, safe engagement in physical activity, and smoking prevention may help curb steep SBP increases during puberty.

5.4 Smoking cessation and prevention as a high-impact lever

Positive links between cigarette use and higher SBP at multiple quantiles reinforce tobacco control as a critical public health priority. Early prevention, cessation support, and policy measures to reduce youth smoking could substantially lower extreme SBP risk.

5.5 Complex interplay of race and smoking

The large, significant Race African*Cigarette Consumption Yes interaction indicates compounded risk for SBP among African youths who smoke. This highlights the necessity of integrated interventions that address both racial health disparities and tobacco use, with culturally appropriate messaging and resources.

Finally, the findings also have important implications for policy development and surveillance strategies. Quantile-specific findings reveal that risk factors may differentially affect those at the extreme ends of BP distributions. Public health surveillance should include percentile-based monitoring to identify high-risk subgroups and to evaluate targeted interventions’ effectiveness on preventing extreme hypertension.

6 Conclusion

This study demonstrates the value of frequentist quantile regression in identifying risk factors associated with elevated diastolic and systolic blood pressure among adolescents, particularly at the upper ends of the blood pressure distribution. The findings revealed that higher BMI, male gender, cigarette consumption, race (coloured) and perceived health status significantly contribute to elevated blood pressure at either the upper ends of the DBP or SBP distributions in adolescents. Notably, interaction effects such as age with gender, race African with cigarette consumption, BMI with gender and Cigarette Consumption with Depression (Occasionally or All of the Time) highlighted the complex and multifactorial nature of hypertension risk. Several predictors were significant only at the 95th percentile, indicating that their influence is concentrated among individuals with severely elevated blood pressure and would likely remain undetected using conventional mean-based regression approaches.

The results underscore the importance of targeted interventions aimed at reducing obesity, discouraging smoking, promoting mental health, and improving cardiovascular screening among high-risk adolescent groups. The observed racial and gender disparities further suggest the need for context-specific public health strategies that address both behavioural and social determinants of health. Overall, this study provides evidence that hypertension risk factors exert heterogeneous effects across the blood pressure distribution and demonstrates that quantile regression is a powerful analytical approach for identifying vulnerable subpopulations.

Early identification and targeted management of these risk factors may contribute significantly to reducing the burden of adolescent hypertension and preventing adverse cardiovascular outcomes later in life.

7 Alignment of study findings with South Africa’s National Policies

The study findings align closely with the priorities and recommendations outlined in South Africa’s national policies as follows:

7.1 Early detection management

National Adolescent & Youth Health Policy_2017 encourages routine screening and youth-friendly services (29). Ideal Hospital Realisation and Maintenance Framework Manual 2018 supports standardized clinical assessments, including BP measurement (30). Integrating BP screening into adolescent care strengthens Primary Health Care (PHC) and reduces long-term disease burden.

7.2 Raise public awareness

National Development Plan (NDP) 2030 calls for prevention of non-communicable diseases (NCDs) through education and behavioural change (31). Awareness campaigns directly support national priorities to shift from curative to preventive healthcare.

7.3 Targeted interventions for obesity and weight management

Strategy for the Prevention and Management of Obesity in South Africa 2023–2028 (32). An introduction to the revised food-based dietary guidelines for South Africa (33). Tackling obesity in adolescence reduces lifetime risk of hypertension and diabetes.

7.4 Gender-specific health strategies

National Adolescent & Youth Health Policy 2017 recognizes gender and social vulnerabilities (29). Gender-responsive programming improves equity and effectiveness of interventions.

Anti-Smoking Campaigns for Youth. National Adolescent & Youth Health Policy 2017 emphasizes substance use prevention (29). Preventing tobacco use among adolescents protects long-term cardiovascular health.

7.5 Data-driven resource allocation

National Health Insurance (NHI) Policy: Supports evidence-based planning and equitable resource distribution (34). Data-informed decision-making ensures efficient use of limited resources.

8 Study limitations

• The risk factors included in the study are the ones measured in the 2017–2018 South African National Income Dynamics Study (NIDS) Wave 5 Household Survey. Consequently, other potentially important determinants of adolescent hypertension, such as dietary sodium intake, family history of hypertension, genetic predisposition, and other clinical or environmental factors, were not available for inclusion in the models.
• NIDS Wave 5 Household Survey conducted in 2017 to 2018 is the latest data set available in South Africa. As a result, the findings may not fully reflect more recent changes in adolescent health behaviours, obesity patterns, smoking prevalence, or healthcare access that have occurred since the data were collected.
• Although the analyses were adjusted for a range of demographic, behavioural, and health-related variables, the possibility of residual confounding cannot be excluded. Unmeasured or imperfectly measured factors may have influenced the observed associations and interaction effects. Consequently, the findings should be interpreted as evidence of statistical associations rather than definitive causal effects.

Generative AI statement

The author(s) declared that Generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Mingjun Zhong, University of Aberdeen, United Kingdom

Reviewed by: Stefanne Madalena Marques, Universidade Federal de Catalão, Brazil

Abebaw Agegne Engda, Debark University, Ethiopia