1. Equity of Services: Do CBOs reach the most vulnerable children? Presenter: Alexa R. Yakubovich alexa.yakubovich@spi.ox.ac.uk Lorraine Sherr, Lucie D. Cluver, Sarah Skeen, Imca S. Hensels, Ana Macedo, Mark Tomlinson Children & HIV: Equity Now! Symposium 16 July 2016 1

2. A perfect storm… Sub-Saharan Africa 2013: 3.2 million children living with HIV 2015: Estimated 25 million AIDS- orphaned children Millions of children in HIV-affected households 2

3. A perfect storm… Overlapping adversities:  Stigma, poverty, family movement and separation Limited resources 3

4. CBOs: Stemming the tide? Local actors responsive to local needs International funding Availability and provisions for children and families 4

5. CBOs: Stemming the tide? Many CBOs implemented without evaluation Tend to be small Limited research capacity Random allocation often seen as unethical 5

6. CBOs: What and for whom? Are CBOs with psychosocial services reaching the most vulnerable HIV-affected children? Are these CBOs associated with positive psychosocial outcomes? If so, how might CBOs positively affect children? No known large- scale studies of existing services No known studies in LMIC 6 No known large- scale studies of existing services

7. Match making Two prospective observational studies designed in close collaboration: 1.Specifically recruited CBO attenders 2.Random, cross-province sample with no regular access to CBO services Matched data for children aged 9-13 Today’s focus: Time 1 7

8. Match making Child Community Care Study Young Carers Study 24 CBOs in SA, randomly selected from all funded by 11 partner organisations Consecutively sampled 446 children, approx. 35/CBO 1402 children selected through door-to-door household sampling 2 urban, 2 rural health districts with >30% HIV prevalence in Mpumalanga and WC 8

9. Match making Child Community Care Study Young Carers Study Ethics: Universities of Oxford, Cape Town, & KZN; provincial Health & Education Departments Face-to-face interviews with trained fieldworkers Refusal rate <1%Refusal rate <2.5% Ethics: University College London, Stellenbosch University, CBO funding agencies 9

10. Combined measures Reach: Age Gender Informal housing Household size Employment School enrolment Community violence Provision of care to young and sick people HIV status of caregiver Outcomes: Depressive symptoms Suicidal ideation Post-traumatic symptoms Child abuse (emotional and physical) Domestic conflict and violence Parental praise Process: CBO Family strengthening Positive child mental health 10

11. Our sample N = 1848 Average age: 11.5 years 55% female 11

12. Who are CBOs reaching? Odds Ratio (95% CI)p-Value Female gender0.674 (0.514–0.884).004 Age0.532 (0.471–0.599)<.001 Informal housing0.461 (0.321–0.662)<.001 Household employment0.321 (0.241–0.429)<.001 Orphanhood6.243 (4.719–8.259)<.001 School non-enrolment1.545 (0.299–7.992).604 HIV+ caregiver0.576 (0.404–0.823).002 Seen someone attacked2.061 (1.553–2.736)<.001 Care for younger children3.704 (2.776–4.943)<.001 Household size1.216 (1.149–1.287)<.001 Note. Multivariable binary logistic regression with CBO attendance as the outcome. 12

13. Who are CBOs reaching? Odds Ratio (95% CI)p-Value Female gender0.674 (0.514–0.884).004 Age0.532 (0.471–0.599)<.001 Informal housing0.461 (0.321–0.662)<.001 Household employment0.321 (0.241–0.429)<.001 Orphanhood6.243 (4.719–8.259)<.001 School non-enrolment1.545 (0.299–7.992).604 HIV+ caregiver0.576 (0.404–0.823).002 Seen someone attacked2.061 (1.553–2.736)<.001 Care for younger children3.704 (2.776–4.943)<.001 Household size1.216 (1.149–1.287)<.001 Note. Multivariable binary logistic regression with CBO attendance as the outcome. 13

14. Who are CBOs reaching? Odds Ratio (95% CI)p-Value Female gender0.674 (0.514–0.884).004 Age0.532 (0.471–0.599)<.001 Informal housing0.461 (0.321–0.662)<.001 Household employment0.321 (0.241–0.429)<.001 Orphanhood6.243 (4.719–8.259)<.001 School non-enrolment1.545 (0.299–7.992).604 HIV+ caregiver0.576 (0.404–0.823).002 Seen someone attacked2.061 (1.553–2.736)<.001 Care for younger children3.704 (2.776–4.943)<.001 Household size1.216 (1.149–1.287)<.001 Note. Multivariable binary logistic regression with CBO attendance as the outcome. 14

15. What outcomes are associated with CBOs? Binary OutcomeOdds Ratio (95% CI)p-Value Weekly physical abuse0.065 (0.015–0.276)<.001 Weekly emotional abuse0.223 (0.109–0.459)<.001 Regular parental praise2.363 (1.733–3.221)<.001 Suicidal ideation0.764 (0.373–1.564).461 Weekly domestic conflict0.124 (0.077–0.199)<.001 Weekly domestic violence0.161 (0.048–0.545).003 Continuous OutcomeUnstandardized B (95% CI)p-Value Depressive symptoms−0.325 (−0.554, −0.096).005 Post-traumatic symptoms0.034 (−0.092, 0.160).597 Note. Separate logistic/linear regressions for each outcome. Predictor is CBO attendance, covariates are all variables in the previous table. 15

16. What outcomes are associated with CBOs? Binary OutcomeOdds Ratio (95% CI)p-Value Weekly physical abuse0.065 (0.015–0.276)<.001 Weekly emotional abuse0.223 (0.109–0.459)<.001 Regular parental praise2.363 (1.733–3.221)<.001 Suicidal ideation0.764 (0.373–1.564).461 Weekly domestic conflict0.124 (0.077–0.199)<.001 Weekly domestic violence0.161 (0.048–0.545).003 Continuous OutcomeUnstandardized B (95% CI)p-Value Depressive symptoms−0.325 (−0.554, −0.096).005 Post-traumatic symptoms0.034 (−0.092, 0.160).597 Note. Separate logistic/linear regressions for each outcome. Predictor is CBO attendance, covariates are all variables in the previous table. 16

17. How are CBOs associated with positive child mental health? Process: CBO Family strengthening Positive child mental health 17

18. How are CBOs associated with positive child mental health? CBO attendance Less physical or emotional abuse More parental praise Less domestic conflict or violence Less depressive symptoms B = -0.07 [95% CI -0.11, -0.04] B = -0.03 [95% CI -0.04, -0.02] B = -0.03 [95% CI -0.06, -0.01] Note. Multiple mediation model run using 5000 bootstrap samples, controlling for all variables in slide 12. 18

19. Understanding CBO reach Results suggest CBOs in SA are reaching more vulnerable children  Unemployment, overcrowding, orphaned, high caregiving responsibility, community violence Some vulnerabilities missing:  Girls, younger age, informal housing, HIV+ caregivers 19

20. Understanding outcomes and pathways associated with CBOs CBO attendance associated with more positive home environments and in turn better child wellbeing Areas for improvement:  Post-trauma symptoms and suicidal ideation  Better referral or improved services? 20

21. Limitations Two data sources: may have driven some of the differences observed Data are cross-sectional and CBO attendance was not randomized 21

22. Conclusions First known, large-scale study of reach, psychosocial correlates, and potential mechanisms of CBOs in SA ‘Real-world’ investigation that highlights creative data solutions Suggests good targeting and positive conditions among CBO attenders 22

23. Acknowledgments CBO Recruitment partners: World Vision, Bernard van Leer Foundation, Firelight Foundation, Save the Children, REPSSI, Stop AIDS Now, AIDS Alliance, The Diana Memorial Fund, Comic Relief, HelpAge, UNICEF 23

24. CBO services for children 24

25. CBO services for caregivers 25

26. Detailed measures: Reach VariableMeasure Age, gender Census items Informal housing Children indicate type of housing. Shack = informal; all else = formal Household size & employment Children counted how many people live in their house and indicated who had a job School enrolment Children indicated whether they attend school. YC: corroborated with school registers; CCC: corroborated with caregiver report Community violence Adapted Child Exposure to Community Violence Checklist (Richters & Martinez, 1993) Provision of care YC: Adapted Multidimensional Assessment of Caring Activities questionnaire (Joseph et al., 2009). CCC: ‘Do you help look after younger kids in your home?’ ‘Have you ever helped unwell people in your home?’ HIV status of caregiver YC: Child report of the verbal autopsy (Lopman et al., 2006); CCC: caregiver self-report of HIV 26

27. Detailed measures: Outcomes VariableMeasure Depression Summation of short-form 9 item Child Depression Inventory (Kovacs, 1992) Suicidal ideation ‘Have you thought about killing yourself?’ Post-traumatic symptoms YC: 28 item Child PTSD Checklist (Amaya et al., 1995); CCC: 10-item Trauma Symptom Checklist for Children (Briere, 1996) Scale scores standardized and combined into single continuous scale Physical abuse Child report of carers using stick or belt to child or slapping or punching child at least weekly (UNICEF items for sub-Saharan Africa, Snider & Dawes, 2006) Emotional abuse Carers threatened to send child away, invoked ghosts or harm upon the child, insulted the child at least weekly (UNICEF for SSA, Snider & Dawes, 2006). YC: child report; CCC: caregiver report. Domestic conflict & violence Conflict: Adults shouting at each other in the home; Violence: Adults hitting each other in the home. CC: 0 = never, 1 = at least once this year, 2 = monthly, 3 = weekly; YC: How many times in the past week? Combined into dichotomous variable: 1 = at least weekly, 0 = all else Parental praise Child received praise for behaving or doing something well. 1 = regular praise (YC: always or often; CCC: often), 0 = irregular praise (YC: sometimes, almost never, never; CCC: rarely or never) 27