Menu
Menu
Menu
Menu
Background: Health risk behaviors during adolescence may cluster into patterns that might be predicted by specific factors, among which HIV may have an important role. Method: In a cross-sectional study conducted between 2017 and 2018, clustering of HRB and its associated factors was investigated in rural Kenya among 588 adolescents (36% perinatally HIV infected; 28% perinatally HIV exposed but uninfected; and 36% HIV unexposed/uninfected). Latent class analysis of 22 behaviors followed by multinomial logistic regression were conducted. Four risk behavior classes were identified. Results: No significant differences were found in behavioral class membership across the three HIV groups (p = 0.366). The risk of membership to the higher risk behavioral classes relative to class 1 (the substance and drug abstinent low risk takers) increased with older adolescent age (p = 0.047), increased among adolescent who experienced mental distress (p < 0.001), and those who felt unsafe in their neighborhood (p < 0.002). Better working memory (p = 0.0037) was found to be protective. Conclusion: The results highlight a need to include screening and interventions for internalizing mental health problems and deficits in executive functioning, as well as steps to involve family members and communities to address psychosocial risk factors in adolescents in Kenya.
The first wave of data collection was conducted between November 2017 and October 2018, providing a baseline assessment for the ongoing longitudinal study, the Adolescent Health Outcomes Study (AHOS). The study was conducted at the Centre for Geographic Medicine Research-Coast at the Kenya Medical Research Institute (CGMRC-KEMRI) and all the participants were residents of Kilifi County at the coast of Kenya. About 1.4 million people resided in Kilifi County by 2016 of whom the majority (61%) were rural dwellers and 22% were aged 10–20 years [53]. Kilifi is termed as a “medium HIV county” with 45 per 1000 affected of whom 6000 (19%) were youth and young adults, aged 15–24 years [50]. About 891 km2 of Kilifi County is covered by Kilifi Health and Demographic Surveillance System (KHDSS) [54], a region with various ongoing CGMRC-KEMRI research activities. Perinatally HIV infected and perinatally HIV exposed but uninfected adolescents and their caregivers were recruited through sequential sampling from all families that attended HIV clinic days at eight HIV treatment and care clinics at health facilities (hospitals and health centres) in Kilifi County. Recruitment was conducted by a trained research assistant in liaison with health workers at the participating HIV treatment facilities. Some of the perinatally HIV exposed but uninfected adolescents and their caregivers were also recruited by visiting families affected with HIV within their community with the assistance of a community health worker based at an HIV clinic. HIV unexposed and uninfected adolescents were randomly sampled among households within the KHDSS using the KHDSS population register [54]. Medical records at the health facilities were used to confirm perinatally HIV-infected adolescents’ HIV status. As part of the eligibility criteria, the adolescents had to be fully aware of their HIV status and that of their biological mother. The adolescent’s awareness of maternal HIV status was an important criterion for abating the ethical dilemma and emotional burden for example, tension and mistrust, that can potentially arise from the abrupt awareness of the source of HIV infection by the adolescents participating in this study and the members of their household. HIV exposure of perinatally HIV exposed but uninfected adolescents was ascertained from maternal medical records (antenatal care cards) confirming HIV infection of the mother during pregnancy. Additionally, recent medical records of the adolescent (if available) were used to ascertain that the perinatally HIV exposed but uninfected adolescent was HIV uninfected. The perinatally HIV exposed but uninfected adolescent also had to be aware of his or her biological mothers’ HIV status for study eligibility. HIV unexposed and uninfected adolescents were not directly tested for HIV but recruitment was restricted to those whose mothers willingly shared their HIV test results at the time of their pregnancy with the participating adolescent. For both perinatally HIV exposed but uninfected adolescents and HIV unexposed and uninfected adolescents, a brief screening checklist was utilized to exclude adolescents who had experienced severe childhood illness or were having recurring health problems so as to minimize the possibility of including HIV-infected adolescents among the control group. Besides, a more detailed assessment of the adolescents’ medical history, symptoms, and concerns was also done by a trained clinician during the assessments for data collection. All eligible adolescent participants had to be accompanied by a legal caretaker during their appointment for data collection at the CGMRC-KEMRI. Monetary reimbursement of 300 Kenyan shillings (about 3 US dollars) and a transport fee reimbursement were given to the accompanying caretaker and a snack was provided to all participants prior to the assessments. Ethical approval to conduct this study was obtained from the Kenya Medical Research Institute Scientific and Ethics Review Unit (KEMRI/SERU/CGMR-C/084/3454). Permission was also obtained from the Kilifi County government, department of health services (HP/KCHS/VOL.VIX/80). The current study comprises 558 (199 HIV unexposed and uninfected, 158 perinatally HIV exposed but uninfected, and 201 perinatally HIV infected) adolescents aged 12–17 years. Initially, 638 potentially eligible adolescent participants (227 HIV unexposed and uninfected, 185 perinatally HIV exposed but uninfected, and 226 perinatally HIV infected) had shown interest in participating, but 560 ultimately took part. Of these 560, data on HRB outcomes from 2 participants was completely missing, therefore they were excluded. Non-response (n = 78) was mainly attributable to silent refusal (i.e., lack of follow-up in making or attending visits (n = 39, 50%), or to direct refusal of further contact with the research team (n = 9, 11.5%), to the failure to meet inclusion criteria (n = 7, 8.9%), and to participant relocation (n = 2, 2.6%)). Some eligible participants were not scheduled for an assessment after attaining the required sample size quotas (n = 21, 27%). Overall, the non-respondents did not differ by HIV group composition (34.6% HIV unexposed and uninfected, 34.6% perinatally HIV exposed but uninfected and 30.8% perinatally HIV infected) and sex distribution (p = 0.75). However, among the non-respondents, the HIV unexposed and uninfected group was the oldest (mean = 13.9 years, SD = 1.7, p = 0.002). Informed consent was obtained from all the individual participants included in the study. Written parental or guardian consent as well as adolescents’ assent were obtained. HRB, the primary outcome of this study, was assessed using an audio-computer assisted self-interview (ACASI) of the Kilifi Health Risk Behavior Questionnaire (KRIBE-Q) in Swahili language. The KRIBE-Q was previously adapted and validated for use among the adolescent sub-population in Kilifi and is a reliable (Gwet AC1 = 0.82) measure for adolescents’ HRB [55]. Contextually relevant examples and explanations of HRB were utilized in the interviews for clarity. In summary, the reported behaviors comprised the following: Six injury and violence-related behaviors were reported: (i) was engaged in physical fights within the past 12 months; (ii) was seriously injured within the past 12 months; (iii) experienced dating violence (physical and/or sexual) within the past 12 months; (iv) was forced to have sexual intercourse; (v) was bullied within the past 12 months; and (vi) experienced suicidal behavior (ideation and/or attempt) within the past 12 months. Three behaviors reported on sexual risk behavior were (i) early sexual debut (dichotomized initiation of sex before 14 years or at 14 years and above); (ii) engagement in transactional sex (victim and/or perpetrator) in the past 12 months; and (iii) condom nonuse during most recent sex. Tobacco and alcohol use were categorized into one group as both are examples of licit substances that are largely accessible within the Kenyan context [56]. Marijuana and Khat were grouped together in a second category as central nervous stimulants [57], which was fairly accessible by youths within the study setting [58]. A third categorization was for other drug use. The five substance use behaviors reported were (i) lifetime use of tobacco or alcohol products, (ii) recent use (past 30 days) of tobacco or alcohol products, (iii) lifetime use of marijuana and Khat products, (iv) recent use (past 30 days) of marijuana and Khat products, and (v) lifetime use of other drugs. Two indications of poor oral hygiene and poor general body hygiene were reported. Respondents were asked how often they cleaned or brushed their teeth in a regular week and how often they washed their entire bodies with water and soap during a regular week. Gambling behaviors were captured by an item asking if the adolescent ever spent much more than they planned on gambling activities within the past 12 months. Three behaviors on physical activity and sedentary behavior were reported: (i) number of days of vigorous physical activity (at least 10 min at a time) during a regular week, (ii) number of days of moderate physical activity (at least 10 min at a time) during a regular week, and (iii) number of hours spent on sedentary activities during a regular day. Contextually relevant examples and explanations of sedentary behavior and vigorous and moderate forms of physical activity were utilized in the interviews for purposes of clarification. Dietary behaviors were captured by assessing the frequency of (i) fruit and vegetable consumption and (ii) fatty or fast food intake during a regular week. Three core EF domains, namely working memory, inhibitory control, and cognitive flexibility [59], were assessed. All EF assessments were administered by a research assistant (trained in psychological and cognitive assessment) in quiet and properly lit rooms, arranged to minimize any form of distractions. Standardized procedures for administration of each EF test were followed. Five trails of the comprehensive trail making test (CTMT) were administered in numerical order following the standardized procedure [60]. Raw scores for each trial (number of seconds taken by an examinee to complete the trial) were recorded by the assessor. T-scores per trial were obtained from the CTMT examiner’s manual [60] and performance summarized by the average T-scores for all completed trials. The Stroop color and word test (SCWT) [61] was administered to assess inhibitory control. Study reports have shown a link between deficits in the ability to inhibit cognitive interference (inhibitory control) and impulsivity [62, 63]. From the three sections of the test (each timed for 45 s), the raw word score (words correctly identified), raw color score (correctly identified colors), and the raw color-word score (correctly identified color of ink for a contrasting name of color) were recorded by the assessor per examinee. The interference score was computed by subtracting the raw color score from the raw color-word score. Participants were administered the backward digit span and the letter-number sequencing (LNS) subsets. These tasks have been previously modified for use with children, adolescents, and older populations within the study setting [64]. Scores were the total number of correct sequences (total correct raw scores) for both LNS and backward digit span tasks. The adolescents’ age, sex, current educational level, and orphanhood status were captured and ascertained in the presence of their caretaker as well as from records such as birth certificates. The adolescents’ caretakers were also asked about their household socio-economic status using an assets index that has been extensively utilized in the Kilifi context [65]. Adolescents’ weight and height were measured according to recommended procedures [66]. Body mass index for age (BMI for age) and height for age were then computed using the WHO standards [67]. Items assessing parent-to-adolescent interaction, peer-to-peer relationship, and school attachment were from measures previously utilized in adolescent sub-populations [68–70]. The most suitable items for each of the three components were selected based on factor analytic approach. Additional items were taken from the KRIBE-Q [55] and assessed “household food insecurity” in the past 30 days (asking if one went hungry because there was not enough food at home), “feeling unsafe in their neighborhood,” “experience of mental distress” (feeling sad/hopeless almost every day for 2 weeks or more in a row) within the past 12 month, and “use of tobacco products by their caretaker/s.” A blood sample was taken from the perinatally HIV-infected adolescents and analyzed for CD4/CD8 cell count and HIV viral load concentrations. All analyses were conducted in the STATA15 software package (StataCorp LLC). First, latent class analysis (LCA) [71] was performed to identify behavioral classes based on the 22 behaviors described above. Five models varying from one to five latent classes were generated and the Akaike information criterion (AIC) and Bayesian information criterion (BIC) were utilized to select the model with the best goodness of fit indices and lowest BIC values [72]. Assignment of participants to respective latent classes was based on their posterior probabilities of class membership. Entropy was also measured to indicate the level of separation between classes. Values of normalized entropy greater than 0.80 indicate that the latent classes are highly discriminating [73]. The HRB composition, socio-demographic factors (sex, age, socio-economic status, education level, household food insecurity), biological factors (BMI, height for age, medical history, HIV status, CD4/CD8 cell count, HIV viral load concentrations), and psychosocial factors (mental distress, orphanhood, caregiver-adolescent interaction, caregiver tobacco use, school attachment, peer-peer relationship, feelings about neighborhood safety) of each class from the optimal model were summarized using descriptive statistics of proportions (%) and means. Bivariate analyses (Chi-square test or Fisher exact test for categorical variables and analysis of variance [ANOVA] with Bonferroni correction for post-hoc analysis for continuous variables) were used to test for significance of differences in HRB composition, socio-demographic, and biological and psychosocial factors across the behavioral classes. ANOVA with Bonferroni correction for post-hoc analysis was performed to identify differences in EF outcomes across classes. A directed acyclic graph (DAG) of the hypothesized exposure-outcome relationship was generated in DAGitty V2.3 open access software [74] (see Fig. 1). The exposure-outcome relationship relies on prior knowledge from other empirical studies and assumed contributory effects which were explained in the introduction section of this article. Using the DAG, we identified variables for adjustment (also known as minimum sufficient adjustment sets) in estimating the effect of perinatal HIV status on HRB clustering. DAGs are increasingly utilized in modern epidemiology and are found to be crucial in advancing investigation of causal relationships which involve multiple interrelated variables [75, 76]. DAGs are also useful for avoiding the introduction of collider bias (i.e., conditional associations introduced by selected covariates) and identifying confounding [77]. Multinomial logistic regression was conducted to investigate the association between perinatal HIV infection and HRB clustering while controlling for the variables identified from the DAG model as minimum sufficient adjustment sets. Multiple imputation was used for data missing at random on individual behavioral variables due to non-response [78]. A directed acyclic graph (DAG) conceptualizing the effect of perinatal HIV infection on health risk behavior clustering among adolescents. HIV: perinatal HIV infection, HRB clustering: Health risk behavior clustering, EF: Measure of executive functioning domains of working memory, inhibitory control and cognitive flexibility, Low_SES: Low household socio-economic status, Sex: sex of the adolescent, Age: age of the adolescent, HIV_biomarkers: HIV treatment outcomes (CD4/CD8 cell count and HIV viral load concentrations), Poor_anthropometry: poor adolescent anthropometric measures of body mass index and height for age, Orphanhood: being an orphan, Education: adolescent’s current educational level, Food_insecurity: household food insecurity in the past 30 days, Mental_distress: Experience of mental distress within the past 12 months, Caregiver_substance use: Use of substances by the caretaker, Insecurity: Feeling unsafe in their neighborhood, Parent_adolescent interaction: Parent-to-adolescent interaction, Peer-peer: Peer-to-peer relationship, School_attachment: School attachment, Exposure, : Outcome, : Other Variables, : Causal path, : Biasing path
N/A
