Cash, care and HIV-prevention for adolescent girls Latest evidence from southern africa L Cluver, M Orkin, M Boyes, L Sherr, F Meinck, A Yakubovich, E Toska, R Hodes, with the team C Ward, L Button, S Hoeksma, A Redfern, S De Stone, N Salah, M Isaacsohn, J Lachman, M Pantelic, J Steinert, J Sandelson, R Herrero Romero, Y Shenderovich, L Parmley. STRIVE, March 2015
COLLABORATIVE RESEARCH FOR POLICY Childfocused research Universities: Oxford, UCT, Wits, Curtin, UKZN
HIV-AFFECTED ADOLESCENTS World s first community survey of HIV-positive adolescents World s largest study of children in AIDS-affected families Free child abuse prevention programmes for families in the developing world
NATIONAL LONGITUDINAL STUDY OF ADOLESCENTS 6,850 adolescents, 2,500 adult caregivers, 2008-2012 Longitudinal national survey Main study: N=6000 (age: 10-18) 3 provinces South Africa; 6 sites >30% prevalence Stratified random sampling of census EAs Every household with a child aged 10-17 Urban/rural, 1 year follow-up in 2 provinces n=3401, 97% follow-up Measures Standardised scales, national surveys Ethics Approved by Universities of Cape Town, Oxford, KwaZulu-Natal, 6 Provincial Health & Education Departments Social & health service referrals Controlling for prior HIV risk
EFFECTS OF ABUSE, POVERTY & PARENTAL AIDS ON FEMALE ADOLESCENT RISK OF TRANSACTIONAL SEX 57% Cluver, Orkin, Boyes, Meinck, Makhasi (2011). JAIDS 1% 7% 13% Healthy family AIDS-sick parent Abused & hungry AIDS-sick parent, abused, hungry
STRIVE: STRUCTURAL DRIVERS OF HIV
PROOF OF CONCEPT: CASH INCENTIVES
UNCONDITIONAL CASH TRANSFERS: RANDOMISED TRIAL
NATIONAL PROGRAMMING IN SOUTH AFRICA: CHILD GRANT REDUCES INCIDENCE OF TRANSACTIONAL SEX AND AGE- DISPARATE SEX FOR GIRLS 8 7 6 5 4 3 2 1 0 % Incidence of transactional sex (OR.49 CI.26-.93*) 12-14 years 15-17 years No cash transfer Child cash transfer % Incidence of age-disparate sex (OR.29 CI.13-.67**) 12-14 years 15-17 years Cluver, Boyes, Orkin, Pantelic, Molwena, Sherr (2013). The Lancet Global Health.
KENYA NATIONAL PROGRAMMING: OVC CASH TRANSFER DELAYS SEXUAL DEBUT & REDUCES PREGNANCY FOR ADOLESCENTS (ODDS RATIOS) 2.4 2.2 All Females 2 1.8 1.6 1.4 1.2 1 Delayed Debut Never pregnant No depressive Symptoms Hope Scale<Median School Enrollment Handa, Halpern, Pettifor, Thirmurthy (2014) PLOS One.
Impacts on educational outcomes and indirect effects on prevention Direct impacts reducing a range of risky behaviours - Drug abuse - Alcohol use - Youth sexual activity Samson et al (2012) UNICEF and DSD.
CAN CASH + CARE REDUCE HIV RISK BEHAVIOR? BEHAVIORAL HIV-RISK INCIDENCE: CASH Transactional sex Age-disparate sex CARE Sex using alcohol/drugs Multiple partners Unprotected sex
THE WILSON VALIDITY CHECK HIV-risk behavior Link to pregnancy Link to HIV-incidence Transactional sex OR 11.0 CI5.6-21.5 HIV+ OR 19.2 Meta-analysis Watts et al 2015 Age-disparate sex OR 5.7 CI 2.8-11.6 HIV+ OR 7.8 Meta-analysis (abstract) Akileswaran 2006 [Tanser 2011 no link] Sex using alcohol/drugs OR 5.1 CI 2.3-11.6 Meta-analysis Woolfe-King et al 2013 Multiple partners OR 3.8 CI 2.1-6.9 HIV+ OR 14.8 Tanser 2011 yes, but not concurrency Unprotected sex OR 11.6 CI 7.3-18 HIV+ OR 17.9 Meta-analysis Drake 2014
Child-focused grant Food garden Positive parenting Regular food parcels School counsellor Free school meals Teacher support
% ADOLESCENTS INCIDENCE OF 1+ HIV RISK BEHAVIOR: CASH PLUS CARE = HALVED RISK 60 50 41% Cash alone: OR.63 Cash plus care: OR.55 42% Cash alone: no effect Cash plus care: OR.50 40 28% 30 25% 20 15% 17% 10 0 no support cash cash plus care Cluver, Orkin, Boyes, Sherr (2014). AIDS. no support cash cash plus care Controlling for: family HIV/AIDS, informal/formal housing, age of child, poverty levels, number of moves of home, baseline HIV risk behaviour
PREDICTORS OF ADOLESCENT HIV-RISKS 2011 Structural deprivation Hunger Community violence Parental HIV/AIDS Informal settlement all p<.001 2012 HIV-risk behavior incidence Transactional sex Age-disparate sex Sex using substances Multiple partners Unprotected sex Pregnancy Cluver, Orkin, Meinck, Boyes, Sherr controlling for: baseline HIV-risk, age, gender
school dropout child abuse conduct problems psychological distress drug/alcohol use Structural deprivation Psychosocial problems p<.004 HIV-risk behavior incidence controlling for: baseline HIV-risk, age, gender
GIRLS: STARVING SEX INCIDENCE (longitudinal, 60% of HIV-risk behavior explained) CARE (pos parenting, monitoring, teacher support CASH (child grant, pension, food garden) High vulnerability (AIDS-affected, informal housing, domestic violence in home etc) Psychosocial risks (depression, suicidality, abuse) Starving Sex (transactional, age-disparate) Cluver, Orkin, Meinck, Boyes, Sherr CLASSROOM (free school & books, school feeding) controlling for: age, baseline HIV-risk
GIRLS: CARELESS SEX INCIDENCE (longitudinal, 44% of HIV-risk behavior explained) CARE (pos parenting, monitoring, teacher support CASH (child grant, pension, food garden) Psychosocial risks (depression, suicidality, abuse) High vulnerability (AIDS-affected, informal housing, violent areas etc) Stupid Sex (unprotected, multiple partners, alcohol/drugs) Cluver, Orkin, Meinck, Boyes, Sherr CLASSROOM (free school & books, school feeding) controlling for: age, baseline HIV-risk
WHAT DO POLICY-MAKERS WANT TO KNOW? Which types of social protection work best? Can a social protection that doesn t work alone become effective when combined with another? (integrated effects) Does adding two or more social protection interventions work better than one? (cumulative effects) How much will it cost and where will we find the money?
SPECIFIC COMBINATIONS FOR HIV-PREVENTION with odds ratios, in multivariate logistic regression Males Careless sex Females CommodityCareless sex sex Commodity sex Cash Child Grants 0.48 Classroom Care School feeding 0.61 Free school & books 0.67 0.41 Monitoring 0.49 0.09 0.63 Teacher support Cluver, Orkin, Yakubovich
COMBINATIONS: ADDITIVE EFFECTS % probability 30 MALES: % PROBABILITY INCIDENCE OF CARELESS SEX (modelled marginal effects using multivariate logistic regression coefficients) 25 20 15 10 5 0 No interventions Free school Good monitoring Teacher support Free school + monitoring Free school + teacher support Monitoring + teacher support All interventions Cluver, Orkin, Yakubovich
COMBINATIONS: ADDITIVE EFFECTS % probability FEMALES: % PROBABILITY OF INCIDENCE OF CARELESS SEX HIV-RISKS 25 (modeled marginal effects using multivariate logistic regression coefficients) 20 15 10 5 0 No intervention Good monitoring School feeding Both interventions Cluver, Orkin, Yakubovich
COMBINATIONS: ADDITIVE EFFECTS % probability FEMALES: % PROBABILITY OF INCIDENCE OF ECONOMIC SEX HIV- RISKS (modeled marginal effects using multivariate logistic regression coefficients) 16 14 12 10 8 6 4 2 0 No intervention Child grant Free school Both interventions Cluver, Orkin, Yakubovich
COMBINATIONS: ADDITIVE EFFECTS % probability 9 8 FEMALES: % PROBABILITY OF INCIDENCE OF PREGNANCY (modeled marginal effects using multivariate logistic regression coefficients) 7 6 5 4 3 2 1 0 No interventions Monitoring Free school School feeding Monitoring + free school Monitoring + school feeding Free school + school feeding All interventions Cluver, Orkin, Yakubovich
ADOLESCENT ART-ADHERENCE AND HIV-PREVENTION Eastern Cape South Africa, 39 state health facilities N=700 (+200) HIV+ adolescents Every teen who ever started on ART Community tracing: including lost-to-follow-up Standardized interviews; Biomarkers (VL/CD4) PAST-WEEK NON-ADHERENCE TO ART: 36% (SELF-REPORT) ADHERENT NON-ADHERENCT Symptom validation: significantly associated* Shingles (p<.03) Bodily sores (p<.01) Mouth ulcers (p<.01) Diarrhea (p<.001) TB symptoms (p<.03) *MV logistic regression 0 20 40 60 80 100 Cluver, Orkin, Sherr, Meinck
PREDICTORS OF NON-ADHERENCE (SEPARATE LOGISTIC REGRESSION CONTROLLING FOR SOCIODEMOGRAPHICS, THEN MODELLED PERCENTAGE PROBABILITIES) Medication side-effects OR (95% CI) 2.1 (1.6-2.9) Hunger 2.2 (1.5-3.4) Abuse 2.1 (1.3-3.1) Domestic violence 2.1 (1.3-3.2) Depression 1.7 (1.2-2.4) Attention deficit 1.6 (1.2-2.2) Behavioral problems 1.5 (1.0-2.2) Cluver, Orkin, Sherr, Meinck 50 45 40 35 30 25 20 15 10 5 0 13 no risk factors % adolescents non-adherent 23 34 hunger hunger + side-effects 44 hunger, sideeffects, attention deficit27
CUMULATIVE ADHERENCE-PROMOTING INPUTS (SEPARATE LOGISTIC REGRESSION CONTROLLING FOR SOCIODEMOGRAPHICS, THEN MODELLED PERCENTAGE PROBABILITIES) Free school meals OR (95% CI).65 (.43-.98) 35 31 % adolescents non-adherent: Parental monitoring.63 (.42-.34) 30 25 24 22 High social support.63 (.45-.09) 20 15 17 10 5 0 Cluver, Orkin, Sherr, Meinck No intervention School feeding Social Support School Feeding and social support 28
FREE, NGO-LED, EVIDENCE-BASED PARENTING PROGRAMS Sinovuyo Teen/Parenting for Lifelong Health Aims: Reduce child abuse, improve parenting and monitoring Local NGO staff, no materials Free: Creative Commons WHO and UNICEF: scale-up to other countries 2012 Qualitative 2013 Pre-post test N=60 2014 Pre-post test N=224 2015-16 Cluster RCT N=1000, 1 year follow-up
Physical abuse p<.001 Emotional Abuse p<.001/.002 2.5 2 1.5 1 0.5 0 Pre Post 3.5 3 2.5 2 1.5 1 0.5 0 Pre Post Caregiver Teen Positive Parenting p<.001/<.01 Poor monitoring p<.002/<.04) 22.5 14 22 13 21.5 12 21 20.5 20 19.5 19 18.5 11 10 9 8 7 Pre Post Caregiver Teen
STRIVE: REMME, WATTS et al: CROSS-SECTORAL CBA Net intervention costs US$ 74,284 Implementation costs US$ 110,250 HIV treatment savings US$ 35,966 Net intervention benefits US$ 478,373 HIV infections and DALYs averted US$ 83,600 Long-term benefits to education and health (excl. HIV) US$ 384,773 Benefit-cost ratio (overall) 6.4 HIV only 1.3 Health and education only 3.5 Net Benefit US$ 404,088 Financing decision: Worth funding
REMME, WATTS et al: CO-FINANCING AND AFFORDABILITY IN MALAWI (Sub-) Sector National scale (million US$) National sector budget (million US$) 2011/12 HIV 0.93 78 Health 1.92 222 Donor disbursements (million US$) 2010/11 Average size of donor projects (million US$) 2010/11 298.2 2.6 Education 1.82 312 167.7 4.1 Total 3.2 (national programme) 1,980 (national budget) 1,022 (overall) 2.3 (overall) Relative contributions for a national-scale scheme appear quite affordable (1.2% HIV, 0.9% health, 0.6% education national budgets) (Source: GoM Budget Statement 2011/12, MoF Aid Atlas 2010/11) (Remme et al, AIDS, 2014)
Funders: thank you.
EVIDENCE SUMMARY Unconditional cash reduces adolescent HIV risks Cash + care + classroom = greater effects Cash + care + classroom buffer structural risk Cumulative impacts of 2+ interventions Effective in real-world sub-saharan Africa
I would like to advice the government to issue grants and support. And see how the impact it could create in the life of many youngsters. How it can better their decision and destiny. Take it from me I am the living difference. Noxolo, 19 yrs. UNAIDS PCB.