Association of Care Environment With HIV Incidence and Death Among Orphaned, Separated, and Street-Connected Children and Adolescents in Western Kenya

Key Points Question What is the association of care environment with incident HIV and death among orphaned, separated, and street-connected children and adolescents in western Kenya? Findings In this cohort study of 2551 orphaned, separated, and street-connected youths in western Kenya, living in an institutional environment was not associated with either death or incident HIV compared with living in a family-based setting. However, living in a street setting was associated with a higher rate of incident HIV and death compared with living in a family-based setting. Meaning This study’s findings suggest that alternatives for orphaned, separated, and street-connected youths in low-income regions are needed when designing public policies for the care of vulnerable children, including deinstitutionalization, to ensure that the rights and well-being of children and adolescents are upheld.


Introduction
The United Nations Children's Fund (UNICEF) defines an orphan as a child younger than 18 years who has lost 1 or both parents to death associated with any cause. 1 There were approximately 140 million orphaned children globally in 2015, including 52 million children in Africa. 1 A total of 15.1 million children worldwide were orphaned by both parents (ie, double orphaned), and 16.6 million children were orphaned because of parental deaths associated with HIV/AIDS, of whom 90% live in sub-Saharan Africa. 1,2 Although most orphaned and separated children and adolescents are cared for by a surviving parent or extended family members, 1 the large number of children requiring care on the subcontinent has produced a proliferation of institutional care settings (eg, orphanages and rescue centers). Recent research has estimated that 650 000 to 1.38 million orphaned and separated children and adolescents live in institutional care environments in sub-Saharan Africa. 3 The appropriateness of institutional care for orphaned and separated children and adolescents has been challenged because data have suggested unfavorable short-and long-term physical and mental health outcomes among children living in these environments. 4,5 Studies have reported deficits and delays in physical growth, psychological health, and cognitive and developmental outcomes among children living in institutional settings. In response, some international organizations have advocated for a global policy of deinstitutionalization. 4,6,7 Most studies supporting this policy have originated in eastern Europe, with relatively few from sub-Saharan Africa and other low-and middle-income countries. 4,5 Sub-Saharan Africa has had the highest dependency ratios (ie, the number of children aged Յ15 years and adults aged Ն65 years per 100 persons of working age) worldwide for decades, and these ratios continue to increase. 8,9 Sub-Saharan Africa also has the largest population of people living in extreme poverty, and the World Bank has estimated that 87% of the world's lowest-income population is expected to live in that region by 2030. 10 Numerous studies from sub-Saharan Africa have found that these constraints on families and traditional caregiving may be associated with worse educational, health, and other outcomes among orphaned children, which have been widely documented in family-based settings. [11][12][13][14][15][16][17][18][19] Thus, despite concerns regarding the safety and care of children in institutions, there is increasing discourse about whether and how institutions might, if well designed, managed, and monitored, be able to provide a safety net or last option for children in low-and middle-income countries when other care options may be unsafe or unavailable, such as in sub-Saharan Africa, where the foster care system is not well developed or monitored. [20][21][22] Contextually relevant, rigorous, and longitudinal data are needed to inform safe and beneficial care and to support policies and interventions for orphaned and separated children and adolescents, including safe and beneficial deinstitutionalization.
Few studies have investigated the association between care environment and physical health among orphaned and separated children and adolescents in sub-Saharan Africa. [23][24][25][26][27] We sought to compare the incidence of HIV and death among orphaned and separated children and adolescents living in family-based and institutional settings as well as youths living in street settings, who constitute a particularly marginalized and overlooked population. 28 The study was conducted in 1 county of western Kenya over almost 10 years. Our primary hypothesis was that living in institutional settings would be associated with higher HIV incidence and death compared with living in familybased settings but that living in either of those settings would be associated with lower HIV incidence and death compared with living in street settings.

Methods
The Moi University College of Health Sciences and the Moi Teaching and Referral Hospital Institutional Research and Ethics Committee, the Indiana University Institutional Review Board, and the University of Toronto Research Ethics Board approved this study. Written informed consent for participation was provided by the head of household, the director of the charitable children's institution, or, in the case of street-connected youths, by the district (now county) children's officer.
Individual written informed assent was provided by each child 7 years or older. Fingerprints were used for children and guardians who were unable to sign or write their names. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies.

Design and Setting
The Orphaned and Separated Children's Assessments Related to Their Health and Well-Being (OSCAR) project was a 2-phase longitudinal cohort study. In-depth details about the OSCAR cohort have been previously reported. 29

Participants
The study enrolled participants 18 years and younger between May 31, 2010, and April 24, 2013, with follow-up until November 30, 2019. The OSCAR cohort comprised participants from communities within 8 administrative locations in Uasin Gishu County, which included 300 randomly selected households (family-based settings) caring for children who were orphaned from all causes, 19 of 21 charitable children's institutions (institutional settings) that were operating in the county at the time of study initiation, and a convenience sample of 100 children who were practicing self-care on the streets (street settings). 29 Orphaned youths were defined as those with a biological mother, father, or both who had died. Separated youths were defined as those with a biological mother or father who was potentially alive but functionally not part of the child's life as reported by the head of household. Street-connected youths were defined as those who spent most of their time (>75%) on the street during the night and/or day for at least the past 3 months.

Independent Variables
The primary exposure of interest was care environment (institutional, family-based, or street), which was determined by a participant's living circumstances at enrollment. 21 Sociodemographic characteristics were ascertained through a clinical encounter and included age, sex, orphaned or separated status (maternal, paternal, or both), HIV status (positive, negative, or unknown), and time living in current care environment at baseline (<6 months, 6 months to <2 years, 2-5 years, >5 years, or all of life).

Outcomes
Primary outcomes were incident HIV and death. Counseling and testing for HIV was offered to all participants 18 months and older using rapid fingerstick assays administered by nationally certified HIV counselors. Children younger than 18 months were referred to the local HIV clinic for DNA testing to ascertain HIV status. All-cause death was ascertained at regular intervals by community health workers who visited participating households and documented deaths using standardized death reporting tools. Deaths among children in institutional settings were ascertained through annual assessments, which documented the outcomes of participants who were no longer living at the institutions. Deaths among street-connected youths were ascertained by the project social worker (who maintained extensive networks within the street youth community) and by physical tracing of children.

Data Sources
Data collection was conducted in situ at the participating charitable children's institutions or at the OSCAR project clinic for participants living in family-based and street settings. Participants completed a standardized clinical evaluation annually (or semiannually for street-connected youths), and children 10 years and older completed an additional psychosocial evaluation annually. 31 The clinical encounter was an enhanced well-child care visit that included a complete survey of physical history and a review of symptoms. Household-level data were obtained through annual site assessments administered by the project manager (for charitable children's institutions) or community health workers (for participating households). 32 Site assessments were not conducted for street-connected youths.

Bias
Community health workers were study staff dedicated to following up participants from households in the community to ascertain deaths and other issues on a quarterly basis throughout the lives of the participants. We used our relationships and networks with the charitable children's institutions to ascertain outcomes of children within their environments who were no longer living in the institutions because they were older than 18 years. We conducted dedicated team-based outreach to other cities in Kenya to which street-connected children were known to have migrated, which enabled us to ascertain outcomes directly from participants.
We calculated effect estimates that were adjusted for the potential confounders of age and sex.
We used competing risk regression analysis for HIV incidence because death is a competing risk for this outcome, 33 and we used Cox survival models to assess death and time to incident HIV or death.
The covariate for age was categorized as younger than 12 years vs 12 years and older to account for the onset of puberty and the increased probability of sexual activity after puberty. Robust SEs were calculated to account for clustering by care environment. In a sensitivity analysis, we assessed the effect of censor year to examine bias that may have occurred from differential follow-up times.

Study Size
Before study initiation, we conducted power and sample size calculations to estimate our power to detect 5%, 10%, and 15% differences in the probability of death. Our calculations assumed (1)  05. The calculations indicated that we had 86% to 99% power to detect the prespecified differences. Because the actual enrollment numbers exceeded expectations, the power was expected to be higher using the same assumptions.

Statistical Analysis
For all study participants, demographic characteristics at enrollment were both summarized and stratified by care environment. For participants in family-based and institutional settings, care We conducted a survival analysis to assess the association of care environment with HIV incidence, death associated with any cause, and time to incident HIV or death. For each outcome, we assessed overall survival by care environment using Kaplan-Meier estimates of cumulative incidence and the P value from a log-rank test. Results from models were reported as hazard ratios (HRs) with 2-sided 95% CIs, and P < .05 was considered statistically significant. Participants with HIV-positive status at study enrollment were omitted from the analyses of HIV incidence and time to incident HIV or death. As a secondary analysis, we examined the association of sex and added an interaction term between care environment and sex to the same analyses that were conducted for direct comparisons by care environment. All analyses were performed using R software, version 4.0.1 (R Foundation for Statistical Computing). 34

Participant Sociodemographic Characteristics
The analysis included 2551 participants. Of those, 1230 participants were living in family-based settings, 1230 were living in institutional settings, and 91 were living in street settings (    Abbreviation: NA, not applicable. a Among those with more than 1 study visit. b A total of 3 street-connected youths died but did not have more than 1 study visit. In the unadjusted analysis, living in an institutional setting vs a family-based setting was associated with a lower risk of death (HR, 0.29; 95% CI, 0.09-0.99; P = .047; log-rank P = .03) and Figure. HIV Incidence, Death, and Time to Incident HIV or Death no higher risk of HIV (HR, 1.33; 95% CI, 0.41-4.35; P = .64; log-rank P = .66); this association remained when the analysis was restricted to participants with HIV-positive status at baseline (logrank P = .02) but not when restricted to participants with HIV-negative status at baseline (log-rank P = .10). After adjusting for sex, age, and baseline HIV status ( Because the median years of follow-up were different between participants living in institutional, family-based, and street settings, we performed a sensitivity analysis to explore the associations of different censor times with results. The IQR for follow-up times became more similar across groups as we censored the data at different points from 4 to 9 years after study initiation.

JAMA Network Open | Global Health
After refitting the same models, the original results and conclusions were found to be relatively robust to censoring time (eTable 3 in the Supplement).

Discussion
The findings of this cohort study suggest that among orphaned, separated, and street-connected children and adolescents in this region of Kenya, there was little or no association between care environment and HIV incidence or death, with the exception of street-connected youths, for whom the data suggested a substantial need for intervention. Our findings regarding the associations of institutional care compared with family-based care were inconsistent with those of a previous metaanalysis, which found that children in institutional care had worse outcomes than those in familybased care. 5 There may be several explanations for this discrepancy. First, most of the studies included in the meta-analysis were conducted in Eastern Europe, and their results may therefore have been specific to geographic region. Second, the list of studies included in the meta-analysis was not published. Therefore, we could not be certain that data collected by the authors of the present study and other researchers working in low-and middle-income countries, which suggest that orphaned and separated children living in institutional environments have similar or even improved outcomes compared with children living in family-based settings, were included in the meta-analysis. 23,25,32,[35][36][37][38][39][40][41] Third, the comparator population used in the meta-analysis to conclude that institutionalization had adverse consequences comprised children who were being raised in a family or were included based on other standard norms derived from typically developing peers; thus, youths in the comparator population were not necessarily orphaned. 5 Our data suggest that, when compared directly with orphaned and separated youths in family-based settings, the differences between care environments were minimal, and living in an institutional setting may be associated with some protective benefits. 23,24,31,32,42 Our data are consistent with the findings of another study, which reported that the quality of care within care environments was more important to a child's well-being than differences between care environment types. 25 Our finding that living in a street setting vs a family-based setting was associated with HIV incidence and death was consistent with other, albeit limited, data from studies conducted in sub-Saharan Africa. 43-45 These results highlight the substantial inadequacies of family-based care for many vulnerable young people, given that extreme poverty, family conflict, and child abuse and neglect are the primary reasons children migrate to the street. 28 Although cash transfer programs for households caring for orphaned children have been implemented in Kenya and elsewhere, with generally positive impacts, these programs remain insufficient given the substantial need. 26,27,[46][47][48][49][50][51][52][53] The World Bank reported that only 18% of individuals in the lowest income quintile residing in low-income countries are covered by social safety nets. 54 Given the already large number of children and adolescents living in the streets (who migrated there mostly because of poverty, family conflict, and violence or abuse 28 ) and the lack of support for households caring for orphaned and separated children, 32,55 caution from governments and other stakeholders is warranted when considering rapid and widespread deinstitutionalization.
As a last option, institutional environments, when regulated and monitored, have the potential to provide evidence-based care for larger numbers of children and adolescents, although their cost-effectiveness needs to be further investigated. 5,56 A previous study described the ways in which the orphan dilemma, particularly in sub-Saharan Africa, requires a diversity of care environments to meet the needs of children and uphold their rights. 32

Strengths and Limitations
This study has strengths. First, it directly compares the HIV and survival outcomes of orphaned and separated youths in institutional environments with those in a random sample of households providing family-based care in the same geographic region, paying careful attention to sex equity.
Second, it includes street-connected children and adolescents, who constitute a small but important population of especially high-risk and overlooked youths. Our data highlight that this population is in greater need of safety, care, and support relative to either of the other 2 groups. Third, this cohort study was powered to investigate the associations between care environment and death and followed up participants for almost 10 years. Fourth, HIV and death outcomes were ascertained by project staff and were not dependent on self-reporting.
This study also has limitations. First, the study is observational and cannot ascertain causality.
Second, the study does not examine long-term outcomes; the median follow-up of participants living in institutional settings was shorter than that of participants living in family-based settings, likely because Kenya mandates the exiting of institutions at age 18 years (with rare exceptions for youths finishing secondary school). Third, the relatively low number of deaths and incident HIV cases prohibit us from reaching firm conclusions or adjusting for multiple covariates. Nevertheless, these data are among the only prospective data available about these important outcomes among highly vulnerable populations, and they provide contextually relevant information that suggests further study is warranted. over the many years of the study. We also acknowledge all of the residents of these locations, particularly the many households caring for vulnerable members of their community, notably orphaned and separated children and adolescents, children with physical and mental disabilities, and older adults. We particularly thank the county and district children's officers for their dedication to protecting the children of Uasin Gishu County, Kenya, as well as the Uasin Gishu Children's Services Forum. This study would not have been possible without the willing participation of the children and adolescents and their guardians, and we acknowledge and thank them. The enthusiasm and commitment of the staff and volunteers of the Orphaned and Separated Children's Assessments Related to Their Health and Well-Being (OSCAR) project did not go unnoticed, and we thank them for their diligent efforts and engagement in this project.