https://orcid.org/0000-0002-7848-7643
Figures
Abstract
Background
Adolescents tend to experience heightened vulnerability to risky and reckless behavior. Adolescents living in rural settings may often experience poverty and a host of risk factors which can increase their vulnerability to various forms of health risk behavior (HRB). Understanding HRB clustering and its underlying factors among adolescents is important for intervention planning and health promotion. This study examines the co-occurrence of injury and violence, substance use, hygiene, physical activity, and diet-related risk behaviors among adolescents in a rural setting on the Kenyan coast. Specifically, the study objectives were to identify clusters of HRB; based on five categories of health risk behavior, and to identify the factors associated with HRB clustering.
Methods
A cross-sectional survey was conducted of a random sample of 1060 adolescents aged 13–19 years living within the area covered by the Kilifi Health and Demographic Surveillance System. Participants completed a questionnaire on health behaviors which was administered via an Audio Computer-Assisted Self–Interview. Latent class analysis on 13 behavioral factors (injury and violence, hygiene, alcohol tobacco and drug use, physical activity, and dietary related behavior) was used to identify clustering and stepwise ordinal logistic regression with nonparametric bootstrapping identified the factors associated with clustering. The variables of age, sex, education level, school attendance, mental health, form of residence and level of parental monitoring were included in the initial stepwise regression model.
Results
We identified 3 behavioral clusters (Cluster 1: Low-risk takers (22.9%); Cluster 2: Moderate risk-takers (67.8%); Cluster 3: High risk-takers (9.3%)). Relative to the cluster 1, membership of higher risk clusters (i.e. moderate or high risk-takers) was strongly associated with older age (p<0.001), being male (p<0.001), depressive symptoms (p = 0.005), school non-attendance (p = 0.001) and a low level of parental monitoring (p<0.001).
Conclusion
There is clustering of health risk behaviors that underlies communicable and non-communicable diseases among adolescents in rural coastal Kenya. This suggests the urgent need for targeted multi-component health behavior interventions that simultaneously address all aspects of adolescent health and well-being, including the mental health needs of adolescents.
Citation: Ssewanyana D, Abubakar A, Newton CRJC, Otiende M, Mochamah G, Nyundo C, et al. (2020) Clustering of health risk behaviors among adolescents in Kilifi, Kenya, a rural Sub-Saharan African setting. PLoS ONE 15(11): e0242186. https://doi.org/10.1371/journal.pone.0242186
Editor: Joel Msafiri Francis, University of the Witwatersrand, SOUTH AFRICA
Received: November 20, 2018; Accepted: October 29, 2020; Published: November 12, 2020
Copyright: © 2020 Ssewanyana et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All data files and variable descriptions files are available from the KWTRP Research Data Repository in the Havard Dataverse database (https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/30IOZ4).
Funding: E.B received funding from Wellcome Trust [084535, 091758] (https://wellcome.ac.uk/); and the Hewlett Foundation through a grant awarded to INDEPTH [IND/0107] (https://hewlett.org/) to support this work. The funding bodies had no role in the study’s design, collection, analysis and interpretation of results, the writing of this manuscript or decision in submission of the paper for publication.
Competing interests: RS received research grants from Shionogi & Co., LTD., Daiichi Sankyo Co., LTD., Chugai Pharmaceutical Co., Ltd., and GlaxoSmithKline K.K. This does not alter our adherence to PLOS ONE policies on sharing data and materials.
Introduction
Adolescents (10–19 years) have a high propensity for risk-taking which can lead to health risk behavior (HRB) clustering or co-occurrence of multiple forms of risky behavior [1, 2]. HRB are specific forms of behavior associated with increased susceptibility to a specific disease or ill health as shown in epidemiological or social data. Examples of HRB include behavior that contributes to unintentional injury or violence, unhealthy dietary habits, inadequate physical activity, sexual behaviors contributing to unintended pregnancy and sexually transmitted diseases, alcohol, tobacco and drug use [3, 4]. Clustering is when the observed proportion of a combination of risk factors exceeds its expected proportion [5]. The mechanisms underlying HRB clustering are poorly understood [2]. Although there has been growing interest in HRB clustering from other parts of the world such as Europe and North America [6–9], there is a dearth of research on this topic among adolescents from sub-Saharan Africa (SSA), which has the largest proportion of young people [10]. Besides, existing research suggests that a multiple-risk-behavior approach is often not reflected in the formulation of policies and interventions to reduce adolescent HRB [7]. Therefore, the need for more holistic and viable HRB assessment, such as clustering of HRB during adolescence has been highlighted as crucial for effective planning of adolescent health promotion interventions [2, 11].
Some of the early literature on adolescent behavior, for example by Jessor, describes a tendency for covariation among risk behaviors which can result in a ‘risk behavior syndrome’ [12, 13]. Jessor’s problem-behavior theory posits that three major systems comprise explanatory variables for problem-behavior and that proneness to such behavior is explained by the extent of balance between the instigations and controls within each system. These systems are: perceived-environment system (consisting of proximal and distal variables of social controls, support and models); personality system (comprising interrelated socio-cognitive variables such as beliefs, expectation, and attitudes); and the behavior system (consisting of both problem behavior such as HRB, and conventional behavior like religiosity and academic involvement) [12]. Specifically within the behavior system, it is theorized that involvement in one problem behavior increases the likelihood of involvement in the other forms of problem behavior. This covariation among risk behavior is attributed to the shared social ecology by youths which may offer socially organized opportunities and normative expectations to learn and engage in multiple forms of risky behavior. Covariation may also occur due to a similarity in psychological meaning and purpose attached to different forms of risk behavior [14, 15]. For example, young people may engage in substance use and risky sexual behavior as a way of affirming their independence and gaining acceptance in their peer networks [12, 13]. Jessor deduces that a consideration of covariation (clustering) of multiple risky behaviors directs attention to influencing the adolescent’s lifestyle as a whole rather than focusing on specific behavior.
Substance use, behavior resulting in injury, poor personal hygiene, poor diet, and low levels of physical activity make significant contributions to the burden of morbidity and mortality in adolescence [16, 17]. The clustering of these and other forms of HRB may aggravate poor educational attainment, poor social and health outcomes during adolescence, as well as shape an unhealthy or socially problematic lifestyle during adulthood [18]. Harmful use of alcohol, unhealthy diet, tobacco use and physical inactivity are linked to an increased risk of dying from non-communicable diseases (NCDs) like diabetes, cardiovascular diseases, cancers and respiratory conditions [19, 20]. These forms of behavior are found to cluster among adolescents [21–23] and this could potentially exacerbate risk to NCDs. A recent survey of 2540 school adolescents in Seychelles [21] for example, reported a high prevalence of inadequate fruit and vegetable consumption (61%), physical inactivity (83%), high consumption of soft drinks (68%), current alcohol use (48%), and current tobacco use (23%). Moreover, the co-occurrence of 3 or more forms of risky behavior linked to NCDs was 81% within this sample [21]. Another survey among 8 African countries (Kenya, Namibia, Morocco, Swaziland, Tanzania, Uganda, Zambia and Zimbabwe) found high prevalence (25%-63%) of bullying (on at least 1 day within the past 30 days) among the adolescents [24]. This study also found that adolescents’ exposure to bullying was strongly significantly associated with their current cigarette use, current alcohol use, lifetime drug use, and multiple sexual partnerships [24]. These findings on HRB covariation were also corroborated by those from another multi-country survey of 7 African states which reported a high (15.5%) prevalence of early initiation of tobacco smoking among adolescents and its significant association with alcohol and other drug use, unintentional injuries and violencer [25]. In both multi-country surveys, only the associations between the main HRB of interest (bullying [24], early initiation of tobacco smoking [25]) and isolated forms of HRB were investigated as opposed to the approach of clustering of multiple forms of HRB.
Behaviors such as sub-optimal oral hygiene, poor handwashing practices and inadequate body hygiene have been linked to increased risk for communicable diseases like enteric infections [26], acute respiratory infections [27], and trachoma [28]. The clustering of these forms of behavior is also increasingly documented among adolescents [29–31]. For instance, a survey conducted among adolescents from 9 African countries reported sub-optimal tooth brushing among 22% and only 58% of the sample reported regular handwashing after using the toilet [31]. Some studies among adolescents have also reported an overlap between behaviors linked to communicable diseases and NCDs [32, 33]. This behavioral overlap (i.e. of both forms of behavior linked to communicable diseases and NCDs) may partly explain the double burden of NCDs and communicable diseases currently faced in low resource settings like SSA [34]. In general, the existing body of research indicates an urgent need for more holistic approaches to HRB research among adolescents as this could fill knowledge gaps on how to appropriately design and implement HRB intervention programs.
With up to a third of all deaths in Kenya attributable to NCDs, there is evidence of a growing burden of NCDs in this region, just like in many other low- and middle-income countries (LMICs) [35]. Studies conducted among the Kenyan adolescent sub-population highlight a high burden (61% among girls and 36% among boys) of sub-optimal engagement in physical activity [36]; early onset (10 years) and general increase in alcohol, tobacco and other drug use [37]; sub-optimal hygiene practices [31, 38]; and a high burden of bullying [24], physical and sexual violence (4% - 48%) [39]. However, it is common that individual HRB is mostly reported in isolation within the Kenyan context thus a missed opportunity to identify the most vulnerable adolescents and the most crucial cross-cutting factors to target when simultaneously intervening multiple forms of HRB. Noteworthy, a holistic approach to HRB reporting such as HRB clustering, aligns with the strategic objective of promoting lifestyle and implementing interventions to reduce priority modifiable risk factors (physical inactivity, unhealthy diet, harmful use of alcohol and tobacco use, and violence and injuries) as stipulated in Kenya’s national strategy for the prevention and control of non-communicable diseases [40, 41]. A better understanding of patterns of HRB co-occurrence will inform interventions and adolescent health promotion policies.
Variations in the sources of vulnerability to HRB clustering may exist across Kenyan counties. Adolescents who reside in poorer settings such as Kilifi County may experience disproportionately poorer behavioral health. Indeed, previous studies have reported abject poverty (about 58% of Kilifi residents live below the poverty line [42]), poor educational outcomes (high drop-out and poor transition between grades), detrimental cultural practices, and poor accessibility to good quality adolescent-friendly services as some of the sources of vulnerability among youths in Kilifi county [43–45]. These numerous sources of vulnerability are likely to predispose adolescents in this rural setting to multiple forms of HRB and injustices. Thus, there is an urgent necessity for more comprehensive research with a multiple-risk-behavior approach to guide a more cohesive and efficient approach to adolescent risk in this context. The objectives of this study, which utilizes data collected in a Kilifi county which is a rural setting at the Kenyan coast, were: (i) to identify clusters of health risk behavior among adolescents living in Kilifi based on 5 behavioral categories (injury or violence, substance use, hygiene, physical activity, and dietary behavior); and (ii) to identify the factors associated with health risk behavior clustering.
Materials and methods
Study design
This was a cross-sectional survey conducted between August and December 2014, among young people aged 13–24 years who were residents of the area covered by the Kilifi Health and Demographic Surveillance System (KHDSS). Using the KHDSS population register, stratified random sampling based on age and sex was utilized to draw a sample of 2072 study participants who were then contacted by the study team through household visits. Of the 2072 potential study participants, consent and/or assent were obtained from 1524 (73.6%) aged 13–24 years. Those whose consent and/or assent were not obtained was due to their decline to participate in the study (3.0% of the total sample), having trans-migrated or out-migrated from the KHDSS (11.7%), absence from home during the household visit (8.1%), and other reasons such as procrastination by the participant (3.6%). All the survey questions were administered via an Audio Computer-Assisted Self–Interview (ACASI) programmed in any of the 3 languages of English, Swahili and Giryama. The data were collected during household visits prior to which a participant would be instructed in detail on how to use the ACASI and left to complete the interview in private on a touch-screen laptop with a pair of headphones.
Study setting
The residents of the KHDSS mainly belong to the Mijikenda ethnic group, and their native language is one of the Mijikenda languages (commonly Giriama) and Swahili. It is estimated that nearly 58% of the residents of Kilifi County live below the poverty line and a third have not attained any formal education [42, 46]. By 2016, about 61% of the 1.4 million residents of Kilifi County were rural dwellers. About 22% of the residents were adolescents. Local home business and subsistence farming are the major sources of livelihood, however, the transport industry (such as motorcycle taxi and tuk-tuk) has become a significant source of livelihood for many unemployed youths [43].
Participants
The current study focuses on the 1060 adolescents aged 13–19 years (70% of the total 1524 13-24-year-olds who participated in the survey). The focus on the 13–19 year age-group was to ensure that this specific study succinctly focuses on the adolescence developmental period. Written parental or guardian consent as well as adolescents’ assent were obtained from participants aged 13–17 years. Written consent was directly sought from adolescents aged 18–19 years and those who were less than 18 years but were married (4%). Ethical approval was obtained from the Kenya Medical Research Institute National Scientific and Ethical Committee (Number 2823) and the Ethics Committee of the London School of Hygiene and Tropical Medicine.
Measures
Five behavioral categories of injury or violence-related behavior, substance use, hygiene behavior, physical activity, and dietary behavior were utilized in this study. All items/questions used to assess these behavioral factors (see S1 File) were borrowed from the Global School–Based Student Health Survey (GSHS) 2013 core questionnaire modules [47].
Injury or violence-related behavior
Participants were asked about the number of times they were seriously injured in the past 12 months (response options ranged from ‘0 times’ to ‘12 or more times’) and in the analysis the data was categorized into 3 levels of “Never”, “Once” and “More than once” for easier interpretation of the results.
Bullying or victimization was assessed by asking participants how many days they were bullied during the past 30 days (response options ranged from ‘0 days’ to ‘all the 30 days’). In our analysis, this item was re-categorized to “0 days”, “1–5 days” and “6 or more days” for simplicity.
Substance use.
Cigarette smoking was assessed by asking participants how old they were when they first tried smoking a cigarette (responses were categorized to “Never smoked”, “13 years or less” and “14 years and above” so as to capture early onset of smoking behavior).
Participants were also asked on how many days they had smoked during the past 30 days (categorized to “No” if the response was ‘0 days’ and “Yes” if the participant had smoked during any of the 30 days so as to capture current/recent smoking behavior).
Alcohol consumption was assessed by asking participants how old they were when they first had a drink of alcohol other than a few sips (categorized to “Never drank alcohol” (or only a few sips), “13 years or less” and “14 years and above” in order to capture early onset of alcohol consumption). The participants were also asked if they had at least one drink containing alcohol during the past 30 days (categorized to “No” if the response was ‘0 days’ and “Yes” if the participant had drunk at least one alcoholic drink during any of the 30 days so as to capture current/recent alcohol consumption).
Lifetime use of marijuana was probed by asking how many times the respondent had used marijuana in their lifetime (categorized to “No” if the response was ‘0 times’ and “Yes” if the participant had ever used marijuana).
In the survey questions, locally relevant names and examples of substances were used.
Hygiene behavior.
Oral hygiene was assessed by asking participants how many times per day they cleaned or brushed their teeth in the past 30 days (response options ranged from ‘did not clean or brush teeth’ to ‘4 or more times per day’). The responses were re-categorized to “Poor” if the response was ‘did not clean or brush teeth’ or ‘did so less than daily’ and to “Adequate” if the response ranged from ‘once a day’ to ‘4 or more times per day’.
The participants were asked how often they washed their hands after using the toilet or latrine in the past 30 days (response options ranged from ‘never’ to ‘always’). We re-categorized responses to: “Poor” if the response was ‘never’ or ‘rarely’; to “Sometimes” if the response was ‘sometimes’; and to “Adequate” if the response was ‘most of the time’ or ‘always’.
Physical activity.
Participants were asked how many days they were physically active for a total of at least 60 minutes per day during the past 7 days. Their responses were categorized into “0 days”, “1–4 days” and “5–7 days” for easier interpretation of the results. They were also asked about the time they spent on a typical or usual day sitting and watching television, chatting with friends or doing other sitting activities (response options ranged from ‘less than 1 hour per day’ to ‘more than 8 hours per day’) and we reclassified their response options to “4 hour or less” and “5 or more hours” [48].
Dietary behavior.
Daily fruit consumption was assessed by asking respondents how many times per day they usually ate fruits during the past 30 days (response options ranged from ‘did not eat fruits’ to ‘5 or more times per day’). We re-categorized the response options to “0 or less than 1 time”, “1–2 times”, and “3 times or more”.
Participants were also asked how many days they ate food from a fast-food restaurant during the past 7 days. Their responses were re-categorized to “0 days”, “1–4 days” and “5–7 days”. While asking both questions, locally relevant examples of fruits and fast foods were used.
Other variables (social demographics and co-variates).
We hypothesized that the demographic characteristics of a respondent, the degree to which their parents/guardians monitored their behavior [49], and symptoms of depression in the past two weeks might be associated with risk group membership [50].
Respondents’ age, sex, area of residence, and school attendance were ascertained during the survey. Participants’ age was generated by computing the difference between the date of survey completion by the participant and their date of birth verified from written records such as birth certificates or healthcare documents. School attendance was assessed by asking the participants if they were currently attending school. The response options were “No” and “Yes”. Upon completion of the consenting process at the participant’s household, their area of residence was selected by the research assistant from the computer’s uploaded list of sub-locations within the KHDSS.
Parental monitoring during past 30 days was assessed by asking the respondents how often their parents or guardians knew what they did during their free time in the past 30 days (item borrowed from GSHS). The response options were “Never”, “Rarely”, “Sometimes”, “Most of the time”, and “Always”.
Participants’ depressive symptoms over the past two weeks were assessed using the major depression inventory (MDI); which has been found reliable (Cronbach’s alpha = 0.83) and valid (Loevinger’s H coefficient for MDI’s 10 items was 0.38 confirming a unidimensional structure) to use among adolescents in this setting [51]. Both total raw scores from MDI and diagnostic classifications (i.e. 0–20 (No or doubtful depression, 21–25 (mild depression), 26–30 (moderate depression) and 31–50 (severe depression) [52] are utilized in our analysis.
Statistical analysis
All analyses were conducted in the STATA15 software package (StataCorp LLC). Latent class analysis (LCA) [53] was done to identify clustering based on the 13 behavioral factors among the five behavioral categories, as described earlier. LCA is a more person-centred finite mixture modeling approach and its more recent extensions allow for manipulation of ordinal, nominal, continuous and count data [54]. LCA classifies individuals from a heterogeneous population into smaller more homogeneous groups called latent classes or groups. Membership to these non-directly observed (latent) groups is inferred from the participants’ varying patterns of responses in the data. Participants were placed in respective latent classes based on their posterior probabilities of belonging to a particular latent class. For each cluster, the proportion (%) of specific behavioral factors was computed and differences across clusters were tested using Chi-square tests [55]. Six models varying from one to six latent classes were generated so as to select the model with the best goodness of fit indices. The Akaike information criterion (AIC) [56] and Bayesian information criterion (BIC) [57] were utilized to identify the best models. Lower AIC and BIC values indicate better model fitting [58]. Entropy was also assessed to indicate the level of separation between classes. Higher values of normalized entropy represent a better fit whereby values greater than 0.80 indicate that the latent classes are highly discriminating [59, 60]. Meaningfulness and interpretability of the classes was also taken into account while determining the suitable number of classes. Descriptive analysis of means and percentages were used to summarize the socio-demographic characteristics of the adolescents in each cluster.
Stepwise ordinal logistic regression with nonparametric bootstrapping [61] at 95% confidence interval was utilized to identify factors associated with behavioral cluster (latent class) membership. The initial model was fitted with the variables of age (in years), sex, education level, current school attendance (yes/no), depressive symptoms over the past 2 weeks, residence (rural/urban/peri-urban) and level of parental monitoring in the past 30 days. The specified re-sampling clusters in this stepwise ordinal logistic regression model were the sub-locations used during stratified random sampling.
Overall, there was 9.9% of missing data for only 9 out of the 13 behavioral variables included in our analysis. Little’s MCAR test [62] indicated that these data were missing completely at random (P>0.05). Multiple imputation was used for missing data on these individual behavioral variables due to non-response [63].
Results
The final sample comprised 1060 adolescents with a mean age of 15.7 years (SD = 1.9), though the majority (66.1%) were in the older adolescent age-group of 15–19 years. The proportion of males among older (15-19years) and younger (13-14years) adolescents was similar (i.e. 52.6% versus 53.5%). Most (79.2%) of the participants had a primary school level of education. Most adolescents (77.4%) resided in rural settings of Kilifi County.
Three behavioral clusters were optimal among the 6 different models on the basis of normalized entropy values (i.e. highest entropy value) and the BIC and AIC criteria (i.e. lowest BIC and AIC) (see Table 1).
Behavioral characteristics of the clusters. Results on the behavioral characteristics of each cluster are presented in Table 2.
Cluster 1 (low risk-takers) was characterized by adolescents who reported abstinence from substance use (alcohol, tobacco, marijuana); much better oral hygiene and handwashing behavior as compared to other cluster members; and reported the highest daily fruit intake (85% reported eating fruits 3 times or more per day). Their reported involvement in injury and violence was also much lower than that reported by adolescents in cluster 3 and more adolescents in cluster 1 reported engaging in at least an hour of physical exercise on 5–7 days in the past week compared to those of cluster 2 (p = 0.03). However, the adolescents of cluster 1 (low-risk takers) consumed fast foods more frequently than adolescents from other clusters and spent more sedentary hours on a typical day than adolescents in cluster 2.
Cluster 2 (moderate risk-takers) comprised adolescents that reported very low substance use, had the least reports of serious injury (18%), least hours of sedentary activities on a typical day, and reported much lower consumption of fast foods than those of cluster 1 (p<0.001). Their reported oral hygiene and handwashing behavior was only better than that of adolescents in cluster 3. Adolescents in cluster 2, however, had the least daily fruit consumption (only 6% reported eating fruits 3 times or more) and the lowest proportion (25%) of adolescents who engaged in active physical activity for 5–7 days per week.
Cluster 3 (high risk-takers) was characterized by having the highest reports of substance use behavior (both early initiation and current use), poorest oral hygiene and handwashing practices, and its members reported the highest involvement in injury and violence (44%). Adolescents in cluster 3 also reported the second-lowest daily fruit consumption and spending more sedentary time than those in cluster 2.
Socio-demographic characteristics of behavioral clusters
The socio-demographic characteristics of adolescents in each cluster are summarized in Table 3.
The composition of these clusters ranged from 9% (high risk-takers) to 68% (moderate risk-takers) of the 1060 adolescents who participated in the study. There were significant age differences across the clusters with the mean age being highest (16.6 years) among adolescents in cluster 3 (high risk substance users) and lowest (15.3 years) for cluster 1 (low risk takers). The majority of the high risk-takers (cluster 3) were male adolescents (83%; (p<0.001)). In the other clusters, the composition by sex was almost balanced. The reported burden of mild to severe depressive symptoms among high risk-takers (8%) was 4 times higher than that in cluster 2 and twice that in cluster 1 (p = 0.004). In comparison to clusters 1 and 2, there was a greater proportion of adolescents (30.6%) in cluster 3 (High risk-takers) who were currently not attending school. Adolescents in the high risk-takers’ cluster also had the highest reports (56.1%) of sub-optimal parental monitoring during the past 30 days (i.e. rarely or never). Overall, 77.4% of the adolescents in each cluster were rural residents and there were no significant differences across the clusters.
Factors associated with behavioral cluster membership
Table 4 summarizes the results of factors associated with cluster membership.
Age was associated with cluster membership with the odds of belonging to a higher risk-taking cluster increasing with the age of the adolescent. Sex was also strongly associated with cluster membership (p<0.001). The odds of belonging to one of the higher risk clusters (i.e. moderate or high risk-takers) relative to the cluster 1 increased with higher scores on the MDI (a measure of depressive symptoms) (p<0.001). Parental monitoring was also associated with cluster membership in that the odds of belonging to one of the higher risk clusters (i.e. moderate or high risk-takers) relative to the cluster 1 were lower for adolescents whose parent/guardian always or most of the time knew what the adolescents did in their free time compared to those whose parent/guardian never knew what they did during their free time. The odds of belonging to a higher risk-taking cluster were higher among out of school adolescents (p = 0.001).
Discussion
This study demonstrates that there is clustering of injury, substance use, poor hygiene, and suboptimal physical activity and dietary behavior among adolescents living in Kilifi, Kenya. Three distinct latent HRB clusters were identified. Three studies on HRB clustering conducted among adolescents from the Netherlands [64], China [65] and the USA [66] generated four behavioral clusters. The difference in the number of clusters generated may be due to the variation in the forms of behavioral factors upon which clustering was conducted and due to disparity in contextual factors. The variation in magnitude of risk-taking across the clusters in our study is consistent with results from the Dutch study which showed that one of their 4 clusters comprised higher risk behavior coupled to negative health outcomes [64]. In our study, the 2 healthier clusters (clusters 1 and 2) had some occurrence of injury and physical inactivity but the distinctive characteristic of the moderate risk cluster (cluster 2) was their poor hygiene. We propose that the differences in hygiene may indicate that adolescents in the moderate risk cluster were more disadvantaged, but unfortunately data to elucidate this were not collected.
Another key finding was that risky behaviors associated with communicable and non-communicable diseases clustered together. The clustering of poor hygiene, smoking, and inadequate fruit/vegetable consumption was also seen among Chinese adolescents [67] and clustering of substance use and sexual risk behavior among adolescents in the USA [66]. The implications of these findings are that behavioral interventions should simultaneously address communicable and NCDs-related risk behaviors of adolescents. Family-based multiple component interventions targeting adolescents and their parents have been shown to address communicable and NCDs-related HRB simultaneously [11]. Such interventions for instance comprise components which reinforce adolescents’ refusal and problem solving skills, increase parental awareness of the risks faced by their adolescents and help to improve adolescent-parent communication [11]. However, more research is needed to more fully understand the specific mechanisms that underlie the clustering of communicable and NCD-related risk behaviors as this could support the design of the most effective interventions.
We found a high burden of injury and violence, which overlapped with high levels of substance use, predominantly among adolescents in Cluster 3 (high risk-takers). A strong association between exposure to bullying and substance use (alcohol, tobacco, and drugs) among adolescents has been reported from a multi-site study among 8 African countries, Kenya inclusive [24]. Substance use, especially alcohol use, has been described as an important cross-cutting predisposing factor to unintentional injury and violence during adolescence [68]. With the trends towards increasing and earlier age of initiation of substance use among Kenyan youths [37], there is need for more comprehensive substance use prevention and harm reduction interventions that concurrently address injury and violence.
Another key finding is that age and sex are important socio-demographic characteristics associated with the co-occurrence of HRB. Similar to our findings, older age and being male were associated with membership of riskier behavioral clusters among Chinese adolescents [67] and adolescents from Bahamas [69], though these factors were not associated with cluster membership among Dutch adolescents [64]. Our findings on the association of age and sex with co-occurrence of HRB highlight the importance of ensuring that interventions designed to address multiple HRB during adolescence are both developmentally appropriate and gender sensitive.
Severity of depressive symptoms among adolescents from Kilifi was associated with higher risk clusters. This relationship may be bi-directional, however, investigating the directionality was beyond the scope of this cross-sectional study. This finding is consistent with results from other adolescent studies that have shown a link between depressive symptoms and engagement in multiple forms of HRB such as substance use, risky sex and sedentary behavior during adolescence [70–72]. A plausible mechanism linking depressive symptoms to HRB clustering is experience of early childhood adverse life events. Indeed, Kenyan studies have found that adolescents who experienced adverse events such as extreme household poverty, living with caregivers who abuse substances, living with sick caregivers and experience of family conflict were more likely to engage in various forms of delinquent and health risk behavior [45, 73, 74], as well as experience psychological problems like anxiety and depression [74, 75]. These findings highlight the need for inclusion of the prevention and management of common mental health disorders in interventions addressing HRB. Moreover, it provides the impetus for longitudinal work to investigate causal pathways and identify the most strategic timing and aspects of intervention.
We found that parental monitoring of adolescents was associated with a lower risk of being a member of a higher risk cluster, though this relationship may also be in either direction. This finding underscores an important and urgent need for rigorous research to examine parenting behavior and its underlying factors so as to identify specific barriers and opportunities for promoting supportive parenting behavior among adolescents’ caregivers in Kilifi. Other studies report that parent/caregiver knowledge of their children’s whereabouts, activities, and associates (parental monitoring) builds resilience, moderates peer-influenced risk behavior and its effects persist into late adolescence [76]. This finding is suggestive that parental monitoring interventions may address HRB among adolescents in Kilifi. Some potential components of such interventions include fostering parent-child communication, improving parental engagement in adolescents’ activities like school activities, and teaching adolescents interpersonal skills [77, 78]. However, these intervention components require testing and tailoring to the context of Kilifi.
Finally, we found that school attendance played a protective role against membership to the higher risk clusters. Indeed, the school environment is a key setting for promoting child and adolescent health and safety as has been reflected in Kenya’s national school health policy and strategic implementation plan [79, 80]. Nutrition health, sanitation and hygiene, disease prevention, child rights, gender issues, and life skills are among some key thematic areas prioritized for strategic implementation in Kenyan school environment [80]. Thus, adolescents who attend school are more likely to benefit from these various health promotion services and structures which potentially offers significant protective benefits against multiple forms of HRB. Conversely, our finding also emphasizes an important need to improve efforts targeting adolescents who do not attend school, as these may face considerable risk for multiple forms of HRB. There is need for multi-sectoral coordination and poverty-intervention programs that seek to change individuals, families, school systems and public policies so as to address poor educational outcomes [79, 81]. Also from a broader perspective, an urgent need to address underlying determinants of adolescent HRB, such as household poverty, complacency in law enforcement, and potentially harmful social norms and customs like early marriages and patriarchy, has been emphasized among other studies conducted in Kilifi [45, 82]. Flexible adolescent-friendly service delivery models that reach beyond the school and health facility environment are required in this setting.
Strengths and limations
The main strength of this study is the breadth of risky behaviors that were considered and the use of a robust analytical approach (latent class analysis) to understand HRB profiles based on a variety of behavioral outcomes simultaneously. To the best of our knowledge, there has not been any study applying this approach in understanding adolescent HRB within Kenya and very few elsewhere in SSA. Although permission was specifically requested to explore sexual behavior of the study participants, ethical approval was granted to collect data on sexual behavior only among the older youths aged 18 years and above. Thus, sexual risk behavior was not included as it was only assessed in older youths who are not part of our adolescent sample. Nonetheless, our study included both communicable and NCD-related behaviors which reflects the double burden of morbidity within SSA [83]. The study had an acceptable (70%) participant response.
However, HRB was self-reported and this may have introduced some level of social-desirability bias, which may have been different according to age, sex or other risk factors. Nonetheless, the use of ACASI likely reduced reporting bias [84]. There were a few participants (less than 3%) that needed some minimal help, especially with maneuvering from one item to the next in some of the sections of the ACASI. Its plausible that during assistance by the research assistant, some level of self-desirability bias may have been introduced. Our categorization of participants’ residence into rural, peri-urban and urban setting based on the KHDSS sub-locations may reflect the general socio-economic status of the sub-locations. However, our study lacked household-level measures of socio-economic status such as household income, parental level of education or employment, which may have an important role in HRB clustering. Its also probable that re-categorization of HRBs may have introduced some level of assessment bias especially through assortment of precise details within the data. Also, this being a cross-sectional study, it was not possible to establish the direction of the associations that were found. Besides, a more in-depth understanding of the mechanisms underlying clustering of HRB in this sub-population would require a more explorative research approach to complement the quantitative findings from this study.
We used the AIC and BIC criteria to identify the optimal latent classes. However, these criteria have some limitations, especially the AIC [85]. Nonetheless, the values of the BIC (which is considered superior to AIC) in our analysis indicated that our model choice (3 latent classes) was most optimal. Also, latent class analysis is a person-centered approach and as such, our results may not be fully generalizable to other samples of adolescents.
Conclusions
There is constellation of HRB related to communicable and non-communicable diseases among adolescents in rural coastal Kenya and thus multi-component health behavior interventions which simultaneously address all HRB may contribute to greater public health benefit. These interventions are also likely to be more effective when explanatory factors and mechanisms underlying HRB clustering are identified so that adolescents’ specific needs are addressed and beneficial intervention components are tailored to the local context. To tailor interventions to this adolescent population rapid situation analyses on adolescents’ and parenting needs within the study setting are needed in order to identify and incorporate relevant intervention components. Some promising interventions components include building parenting skills and communication within the family, strengthening adolescents’ problem-solving skills, strengthening resilience in the face of adversity, strengthening family connectedness and providing pyscho-education which can target contextually relevant sources of mental illness. There is also need for improved targeting of adolescents who drop out of school as well as those at risk of school dropout [86, 87]. This is because majority of the adolescent health research and interventions have habitually targeted school-attending adolescents [11]. Besides, the urgent need for increased prioritization of research and investment in programmes focusing on Kenya’s most-at-risk adolescents is well documented [88]. It is plausible that the adolescents who were vulnerable to multiple forms of HRB within our study setting comprised most-at-risk adolescents, whose specific needs necessitate further investigation so as to tailor responsive interventions and services. Adolescent health services or interventions targeting HRB need to stretch beyond focusing on HRB by also including mental health management. Future studies need to examine mechanisms that underlie the clustering of communicable and NCD-related HRB during adolescence.
Supporting information
S1 Questionnaire. Young people health survey questionnare (English studio version).
https://doi.org/10.1371/journal.pone.0242186.s002
(DOCX)
S2 Questionnaire. Young people health survey questionnare (Giriyama studio version).
https://doi.org/10.1371/journal.pone.0242186.s003
(DOCX)
S3 Questionnaire. Young people health survey questionnare (Swahili studio version).
https://doi.org/10.1371/journal.pone.0242186.s004
(DOC)
References
- 1. Steinberg L (2004) Risk taking in adolescence: what changes, and why? Ann N Y Acad Sci 1021: 51–58. pmid:15251873
- 2. Spring B, Moller AC, Coons MJ (2012) Multiple health behaviours: overview and implications. J. Public Health 34: i3–i10. pmid:22363028
- 3.
DiClemente RJ, Hansen WB, Ponton LE (2013) Handbook of adolescent health risk behavior. Berlin: Springer.
- 4.
Currie C, Zanotti C, Morgan A (2012) Health Behaviour in School-aged Children (HBSC) study: international report from the 2009/2010 survey. Copenhagen: WHO Regional Office for Europe.
- 5. Schuit AJ, van Loon AJM, Tijhuis M, Ocké MC (2002) Clustering of lifestyle risk factors in a general adult population. J Prev Med 35: 219–224.
- 6. Leech RM, McNaughton SA, Timperio A (2014) The clustering of diet, physical activity and sedentary behavior in children and adolescents: a review. Int J Behav Nutr Phys Act 11: 4. pmid:24450617
- 7. Hale DR, Viner RM (2012) Policy responses to multiple risk behaviours in adolescents. J. Public Health 34: i11–i19. pmid:22363026
- 8. Meader N, King K, Moe-Byrne T, Wright K, Graham H, et al. (2016) A systematic review on the clustering and co-occurrence of multiple risk behaviours. BMC Public Health 16: 657. pmid:27473458
- 9. Parker KE, Salmon J, Costigan SA, Villanueva K, Brown HL, Timperio A (2019) Activity-related behavior typologies in youth: a systematic review. Int J Behav Nutr Phys Act 16: 44. pmid:31097036
- 10.
The World Bank. Chart: The World's Youngest Populations Are in Africa. 2016. Available from: https://blogs.worldbank.org/opendata/chart-worlds-youngest-populations-are-africa.
- 11. Hale DR, Fitzgerald-Yau N, Viner RM (2014) A systematic review of effective interventions for reducing multiple health risk behaviors in adolescence. Am J Public Health 104: e19–e41. pmid:24625172
- 12. Jessor R (1991) Risk behavior in adolescence: a psychosocial framework for understanding and action. J. Adolesc Health 12: 597. pmid:1799569
- 13. Jessor R (1987) Risky Driving and Adolescent Problem Behavior: An Extension of Problem-Behavior Theory. Alcohol Drugs and Driving 3: 1–11.
- 14. Siegel AW, Cousins JH, Rubovits DS, Parsons JT, Lavery B, Crowley CL (1994) Adolescents' perceptions of the benefits and risks of their own risk taking. J Emot Behav Disord 2: 89–98.
- 15. Xiao Y, Romanelli M, Lindsey MA (2019) A latent class analysis of health lifestyles and suicidal behaviors among US adolescents. J Affect Disord 255: 116–26. pmid:31150941
- 16. Gore FM, Bloem PJ, Patton GC, Ferguson J, Joseph V, et al. (2011) Global burden of disease in young people aged 10–24 years: a systematic analysis. Lancet 377: 2093–2102. pmid:21652063
- 17. Kyu HH, Pinho C, Wagner JA, Brown JC, Bertozzi-Villa A, et al. (2016) Global and national burden of diseases and injuries among children and adolescents between 1990 and 2013: findings from the Global Burden of Disease 2013 Study. JAMA Pediatr 170: 267–287. pmid:26810619
- 18. Kipping R, Campbell RM, MacArthur G, Gunnell D, Hickman M (2012) Multiple risk behaviour in adolescence. J. Public Health 34: i1–i2. pmid:22363025
- 19.
World Health Organization (2018) Noncommunicable diseases. Geneva. Available from: https://www.who.int/news-room/fact-sheets/detail/noncommunicable-diseases.
- 20. Lee I-M, Shiroma EJ, Lobelo F, Puska P, Blair SN, Katzmarzyk PT, et al. (2012) Effect of physical inactivity on major non-communicable diseases worldwide: an analysis of burden of disease and life expectancy. Lancet 380: 219–29. pmid:22818936
- 21. Pengpid S, Peltzer K (2019) Prevalence and correlates of behavioral non-communicable diseases risk factors among adolescents in the Seychelles: results of a national school survey in 2015. Int J Environ Res Public Health 16: 2651. pmid:31349537
- 22. Tassitano RM, Dumith SC, Chica DAG, Tenório MCM (2014) Aggregation of the four main risk factors to non-communicable diseases among adolescents. Rev Bras Epidemiol 17: 465–78. pmid:24918417
- 23. Idowu A, Fatusi AO, Olajide FO (2016) Clustering of behavioural risk factors for non-communicable diseases (NCDs) among rural-based adolescents in south-west Nigeria. Int J Adolesc Med Health 30: 1. pmid:27299197
- 24. Brown DW, Riley L, Butchart A, Kann L (2008) Bullying among youth from eight African countries and associations with adverse health behaviors. Ped Health 2: 289–300.
- 25. Peltzer K (2011) Early smoking initiation and associated factors among in-school male and female adolescents in seven African countries. Afr Health Sci 11: 320–328 pmid:22275919
- 26. Curtis V, Cairncross S (2003) Effect of washing hands with soap on diarrhoea risk in the community: a systematic review. Lancet Infect Dis. 3: 275–81. pmid:12726975
- 27. Rabie T, Curtis V (2006) Handwashing and risk of respiratory infections: a quantitative systematic review. Trop Med Int Health 11: 258–67. pmid:16553905
- 28. Delea MG, Solomon H, Solomon AW, Freeman MC (2018) Interventions to maximize facial cleanliness and achieve environmental improvement for trachoma elimination: A review of the grey literature. Plos Neglect Trop D 12: e0006178. pmid:29370169
- 29. Peltzer K, Pengpid S (2014) Oral and hand hygiene behaviour and risk factors among in-school adolescents in four Southeast Asian countries. Int J Environ Res Public Health 11: 2780–92. pmid:24608901
- 30. McKittrick T, Jacobsen K (2015) Oral hygiene and handwashing practices among middle school students in 15 Latin American and Caribbean countries. West Indian Med 64: 266. pmid:26426181
- 31. Pengpid S, Peltzer K (2011) Hygiene behaviour and associated factors among in-school adolescents in nine African countries. Int J Behav Med 18: 150–159. pmid:20593258
- 32. Austregésilo SC, de Goes PSA, de Sena Júnior MR, Pazos CTC (2019) Clustering of oral and general health risk behaviors among adolescents. Prev Med Rep 15:100936. pmid:31372328
- 33. Olczak‐Kowalczyk D, Tomczyk J, Gozdowski D, Kaczmarek U (2019) Excessive computer use as an oral health risk behaviour in 18‐year‐old youths from Poland: A cross‐sectional study. Clin Exp Dent Res 5: 284–93. pmid:31249710
- 34. Marquez P, Farrington J (2012) AFRICA'S NEXT BURDEN: Non-infectious disease. BMJ 345: 24–7.
- 35.
World Health Organization (2016) Global Health Estimates 2015: Deaths by Cause, Age, Sex, by Country and by Region, 2000–2015. Geneva.
- 36.
Kenya National Bureau of Statistics (KNBS) and ICF International (2014) Kenya Demographic and Health Survey, Key Indicators.: Rockville, MD.
- 37.
National Authority for Campaign against Alcohol and Drug Abuse (2012) Rapid situation assessment of the status of drug and substance abuse in Kenya, 2012. Available from: https://scadkenya.files.wordpress.com/2017/01/rapid-assessment-drug-and-substance-abuse-situation-kenya-20122.pdf; 2012
- 38. Okemwa K, Gatongi P, Rotich J (2010) The oral health knowledge and hygiene practices among primary school children aged 5–17 years in a rural area of Uasin Gishu District, Kenya. J. Dent. Oral Hyg 7: 187
- 39.
UNICEF (2012) Violence against Children in Kenya: Findings from a 2010 National Survey. Summary Report on the Prevalence of Sexual, Physical and Emotional Violence, Context of Sexual Violence, and Health and Behavioral Consequences of Violence Experienced in Childhood. Nairobi. Available from: https://www.unicef.org/esaro/VAC_in_Kenya.pdf
- 40.
Kaneda T, Osewe N, Mbau-Simba L (2017) Promoting Healthy Behaviors Among Youth to Tackle Kenya’s Growing Noncommunicable Diseases Epidemic. Washington, DC: Population Reference Bureau.
- 41.
Ministry of Health, Kenya (2015) Kenya national strategy for the prevention and control of non-communicable diseases 2015–2020. Nairobi. Available from: https://www.who.int/nmh/ncd-task-force/kenya-strategy-ncds-2015-2020.pdf
- 42.
Ngugi, E, Kipruto S, Samoei, P (2013) Exploring Kenya's inequality: pulling apart or pooling together? Nairobi: Kenya National Bureau of Statistics and Society for International Development
- 43.
National AIDS Control Council (NACC) (2016). Kilifi County HIV and AIDS Strategic Plan (2016–2020): A Healthy and Productive Population. Nairobi. Available from: http://nacc.or.ke/mdocs-posts/kilifi-county-hiv-aids-strategic-plan/
- 44.
National Council for Population and Development (NCPD) (2017). 2015 Kenya National Adolescents and Youth Survey. Nairobi. Available from: http://www.ncpd.go.ke/2015-national-adolescent-and-youth-survey-preliminary-report/.
- 45. Ssewanyana D, Mwangala PN, Marsh V, Jao I, van Baar A, et al. (2018) Socio-ecological determinants of alcohol, tobacco, and drug use behavior of adolescents in Kilifi County at the Kenyan coast. J. Health Psychol 1: 14.
- 46. Scott JAG, Bauni E, Moisi JC, Ojal J, Gatakaa H, et al. (2012) Profile: the Kilifi health and demographic surveillance system (KHDSS). Int J Epidemiol 41: 650–657. pmid:22544844
- 47.
World Health Organization (2013) Global School-Based Student Health Survey (GSHS). 2013 Core questionnaire module. Available from: http://www.who.int/chp/gshs/GSHS_Core_Modules_2013_English.pdf?ua=1
- 48. Ekelund U, Steene-Johannessen J, Brown WJ, Fagerland MW, Owen N, Powell KE, et al. (2016) Does physical activity attenuate, or even eliminate, the detrimental association of sitting time with mortality? A harmonised meta-analysis of data from more than 1 million men and women. Lancet 388: 1302–10. pmid:27475271
- 49. DiClemente RJ, Wingood GM, Crosby R, Sionean C, Cobb BK, et al. (2001) Parental monitoring: Association with adolescents' risk behaviors. Pediatrics 107: 1363–1368. pmid:11389258
- 50. Bannink R, Broeren S, Heydelberg J, van’t Klooster E, Raat H (2015) Depressive symptoms and clustering of risk behaviours among adolescents and young adults attending vocational education: a cross-sectional study. BMC Public Health 15: 396. pmid:25896828
- 51. Otiende M, Abubakar A, Mochamah G, Walumbe D, Nyundo C, et al. (2017) Psychometric evaluation of the Major Depression Inventory among young people living in Coastal Kenya. Wellcome Open Research 2: 113 pmid:29862324
- 52. Bech P, Timmerby N, Martiny K, Lunde M, Soendergaard S (2015) Psychometric evaluation of the Major Depression Inventory (MDI) as depression severity scale using the LEAD (Longitudinal Expert Assessment of All Data) as index of validity. BMC Psychiatry 15: 190. pmid:26242577
- 53.
Hagenaars JA, McCutcheon AL (2002) Applied latent class analysis: Cambridge University Press.
- 54. Haughton D, Legrand P, Woolford S (2009) Review of three latent class cluster analysis packages: Latent Gold, poLCA, and MCLUST. Am. Stat 63: 81–91.
- 55. McHugh ML (2013) The chi-square test of independence. Biochem Med 23: 143–149. pmid:23894860
- 56. Akaike H (1974) A new look at the statistical model identification. IEEE Trans. Autom 19: 716–723.
- 57. Schwarz G (1978) Estimating the dimension of a model. Ann. Stat 6: 461–464.
- 58. Henson JM, Reise SP, Kim KH (2007) Detecting mixtures from structural model differences using latent variable mixture modeling: A comparison of relative model fit statistics. Struct Equ Modeling 14: 202–26.
- 59. Berlin KS, Williams NA, Parra GR (2014) An introduction to latent variable mixture modeling (part 1): Overview and cross-sectional latent class and latent profile analyses. J Pediatr Psychol 39: 174–87. pmid:24277769
- 60. Tein J-Y, Coxe S, Cham H (2013) Statistical power to detect the correct number of classes in latent profile analysis. Struct Equ Modeling 20: 640–57. pmid:24489457
- 61. Austin PC, Tu JV (2004) Bootstrap methods for developing predictive models. Am Stat 58:131–7.
- 62. Little RJ (1988) A test of missing completely at random for multivariate data with missing values. J Am Stat Assoc 83: 1198–1202.
- 63. Graham JW, Olchowski AE, Gilreath TD (2007) How many imputations are really needed? Some practical clarifications of multiple imputation theory. Prev Sci 8: 206–213.
- 64. Busch V, Van Stel HF, Schrijvers AJ, de Leeuw JR (2013) Clustering of health-related behaviors, health outcomes and demographics in Dutch adolescents: a cross-sectional study. BMC Public Health 13: 1118. pmid:24305509
- 65. Ji Y, Xu H, Zhang Y, Liu Q (2018) Heterogeneity of adolescent health risk behaviors in rural western China: A latent class analysis. Plos One 13: e0199286. pmid:29944679
- 66. Connell CM, Gilreath TD, Hansen NB (2009) A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents. J Stud Alcohol Drugs 70: 943–951. pmid:19895772
- 67. Zhang Y, Ji Y (2015) Clustering of Risk Behaviors and their Social Determinants among Primary School Learners in Beijing, China: A Cross-sectional Study. Chin Med J 128: 1567.
- 68. Sleet DA, Ballesteros MF, Borse NN (2010) A review of unintentional injuries in adolescents. Annu Rev Public Health 31: 195–212. pmid:20235851
- 69. Cole M, Stanton B, Deveaux L, Harris C, Cottrell L, Clemens R, et al. (2007) Latent class analysis of risk behaviors among Bahamian young adolescents: Relationship between values prioritization and latent class. Soc Behav Pers 35: 1061–76.
- 70. Katon W, Richardson L, Russo J, McCarty CA, Rockhill C, et al. (2010) Depressive symptoms in adolescence: the association with multiple health risk behaviors. Gen. Hosp. Psychiatry 32: 233–239.
- 71. Asarnow JR, Zeledon LR, D’Amico E, LaBorde A, Anderson M, et al. (2014) Depression and health risk behaviors: towards optimizing primary care service strategies for addressing risk. J Prim Health Care 4: 152. pmid:25309826
- 72. Bannink R, Broeren S, Heydelberg J, van’t Klooster E, Raat H (2015). Depressive symptoms and clustering of risk behaviours among adolescents and young adults attending vocational education: a cross-sectional study. BMC Public Health 15: 1–9. pmid:25563658
- 73. Kabiru CW, Elung’ata P, Mojola SA, Beguy D (2014) Adverse life events and delinquent behavior among Kenyan adolescents: a cross-sectional study on the protective role of parental monitoring, religiosity, and self-esteem. Child Adolesc Psychiatry Ment Health 8: 24. pmid:25210535
- 74. Wambua GN, Obondo A, Bifulco A, Kumar M (2018) The role of attachment relationship in adolescents’ problem behavior development: a cross-sectional study of Kenyan adolescents in Nairobi city. Child Adolesc Psychiatry Ment Health 12: 27. pmid:29881453
- 75. Khasakhala LI, Ndetei DM, Mathai M, Harder V (2013) Major depressive disorder in a Kenyan youth sample: relationship with parenting behavior and parental psychiatric disorders. Ann Gen Psychiatry 12: 15. pmid:23663452
- 76. Lopez‐Tamayo R, LaVome Robinson W, Lambert SF, Jason LA, Ialongo NS (2016) Parental Monitoring, Association with Externalized Behavior, and Academic Outcomes in Urban African‐American Youth: A Moderated Mediation Analysis. Am J Community Psychol 57: 366–379. pmid:27237941
- 77. Stanton B, Li X, Galbraith J (2001) Parental Underestimates of Adolescent Risk Behavior: A Randomized, Controlled Trial of a Parental Monitoring Intervention. Year B. Psychiatry Appl. Ment. Health 2001: 12–13.
- 78. Hawkins JD, Kosterman R, Catalano RF, Hill KG, Abbott RD (2005) Promoting positive adult functioning through social development intervention in childhood: Long-term effects from the Seattle Social Development Project. Arch Pediatr Adolesc Med 159: 25–31. pmid:15630054
- 79. Wasonga J, Ojeny B, Oluoch G, Okech B (2014) Kenya comprehensive school health policy: lessons from a pilot program. J Public Health Africa 5: 313. pmid:28299114
- 80.
Ministry of Public Health and Sanitation, Kenya and Ministry of Education, Kenya (2009) National School Health Policy. Nairobi. Available from: https://www.prb.org/wp-content/uploads/2018/05/National-School-Health-Policy-2009.-without-cover..-Kenya.pdf.
- 81. Abuya B, Oketch M, Musyoka P (2013) Why do pupils dropout when education is ‘free’? Explaining school dropout among the urban poor in Nairobi. Compare 43: 740–762.
- 82. Forde S, Kendall-Taylor A (2007) Moving the Goalposts: The Impact of Soccer on the Lives and Prospects of Girls and Young Women Rural Kenya. Ufahamu 33: 2–3.
- 83. Boutayeb A (2006) The double burden of communicable and non-communicable diseases in developing countries. Trans R Soc Trop Med Hyg 100: 191–199. pmid:16274715
- 84. Brown J, Swartzendruber A, DiClemente R (2013) Application of audio computer-assisted self-interviews to collect self-reported health data: an overview. Caries Research 47: 40–45. pmid:24107606
- 85. Nylund KL, Asparouhov T, Muthén BO (2007) Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Struct Equ Modeling 14: 535–569.
- 86. King N, Dewey C, Borish D (2015) Determinants of primary school non-enrollment and absenteeism: results from a retrospective, convergent mixed methods, cohort study in rural Western Kenya. PLoS One 10: e0138362. pmid:26371885
- 87.
Sabates R, Westbrook J, Akyeampong K, Hunt F (2010) School drop out: Patterns, causes, changes and policies. Sussex. Available from: https://unesdoc.unesco.org/ark:/48223/pf0000190771.
- 88.
National AIDS and STIs Control Programme (2016) Most at risk adolescents: Invisible or ignored? Situational analysis of policy and health care for most at risk adolescents in Kenya. Nairobi. Available from: https://lvcthealth.org/wp-content/uploads/2019/04/MOST-AT-RISK-ADOLESCENTS-INVISIBLE-OR-IGNORED-SITUATIONAL-ANALYSIS-OF-POLICY-HEALTH.pdf