Impact Evaluation of Malaria Prevention and Treatment

1. Impact Evaluation of Malaria Prevention and Treatment Jed Friedman (DECRG, Economist) Edit V. Velenyi (DECRG/AIEI, MIEP Coordinator)

2. [Impact] Evaluating Health Programs is Different • Norm in medicine to use randomized control trials to figure out what works • What we know less about: • What are the socioeconomic effects of health interventions? • How to get people to utilize prevention / treatment services? • What is the most cost effective mode of prevention / treatment?

3. Some Differences • HIV/AIDS: Largely behaviorally driven • Malaria: Vector-born disease …But the behavioral aspects are important Treat Affected- Early Diagnosis & T- Monotherapy v. Non-ACT v. ACTProblems- Access- Compliance- Cost-Effectiveness- Long-Term Effect Parasite Control- Kill Asexual FormPrevents disease progression- Kill Sexual FormPrevent spread to mosquito- Full TreatmentProblem- Drug Resistance Vector Control- Prevent Breeding (DDT)- Prevent Entry (Proofing)- Prevent Bite (ITN, Spray) Problems- Resistance to Insecticides- Compliance Protect Unaffected- ITN, LLIN- Mosquito ProofingProblems- Valuation - Compliance

4. Dynamic Aspects of Malaria Control • Epidemiological / Entomological / Behavioral • Transmission Pattern / Time Path of Infection • Mosquito Resistance to Insecticide • Parasite Resistance to Drugs • Economic Rationale - Cost Effectiveness of Prevention & Treatment Regimes • Political Economy • Policy Making

5. Malaria Control Boundaries and Perspectives “Widespread use of ITNs and state-of-the-art drugs succeeded in cutting malaria deaths half in 2 countries most heavily affected by the disease, Rwanda and Kenya.” Washington Post (01/31/08) “This is a genuinely historic achievement. This is not theoretical. We do not have to wait for a vaccine or new drugs. If we implement today’s technologies aggressively on a national scale we will have a big impact.” Richard Feachem, former Director of the Global Fund “With the resources for research, we should be able to eradicate malaria before I hang up my lab coat.” Peter Agre, Malaria Research Institute, JHU

6. Impact of Policy Change on Malaria Prevalence: South Africa National Geographic 07/07 • Aggravating Exogenous Factors • - Refugee Flow • - Heavy Rains • - Worsening Resistance Policy Regime Switches • Pesticide Resistance - DDT Stops: 1996- Spraying Resumes: 2000 • Drug Resistance- Multidrug Therapy: 2000 In Kwazulu-Natal the combined effect of switching from SP to the fixed combination of AL and IRS with DDT was associated with a decrease in cases of 78% and an increase in cure rate of 87% (Barnes K, unpublished data; in Yeung et al. 2004).

7. Case 1: DDT - IndiaQuasi Experiment (Cutler et al. 2007) • Long-term Effects of Malaria Eradication • Outcome Measure: Educational Gains 1. Literacy Rate (LR); 2.School Completion Rate (SCR) • Method: Quasi Experiment using Diff-in-Diff Sample: 300,000 Households, 1 million Individual Observations (NSS) - Control: Pre-Eradication Cohorts (C0): 1912-1952 - Treatment: Post-Eradication Cohorts (C1): 1962-1972 - Omitted: Eradication Cohorts: 1953-1961 Outcome: Yicd = β0 + β1 (Post)c * (High)d + γd + αc + Xβ2 1947 Pre-Eradication- Population: 334 million- Cases: 75 million (annual)- Prevalence: 22% (annual)- Mortality: 800,000 (annual)- Mortality: 10% of total deaths 1953 Intervention1st: 1953 NMCP LaunchedDDT Spraying - 2 Rounds per Year - 125 Malaria Control Units2nd: 1958 NMEP Launched Gains12% ↑ in LR, SCR- Malaria explains half of these gains- Income ↓ through malaria 7-10%

8. Case 2: ITN – KenyaField Experiment (Cohen and Dupas 2008) • Explore tradeoffs between cost-sharing (CS) & free distribution for ITNs • Randomize price of ITNs (0 ≤ p <pPrevailingCS) in prenatal clinics in Kenya • Evaluate impact on pregnant women • Demand / Uptake - Cost-sharing (C/S) does considerablydampen demand. - Uptake Drop: i) by 75% from 0 to prevailing CS price; ii) by 20% for ↑10Ksh. • Usage - No evidence that C/S reduces wastageon those who do not use the net. - Free ITN owner is not less likely to use net than those who paid higher prices. - Coverage (Uptake + Usage): 63% (Free Net) v. 14% (40Ksh) CF≠ CPCS • Need (Health) - No evidence that C/S induces selectionof those who need net more. - Those paying higher prices appear no sicker (anemia) than control group. • Compare Cost Effectiveness (Externality Assumptions) - Number of child lives saved highest under free distribution. - Free distribution is more CE when externality threshold is medium level. Source: Cohen and Dupas 2008

9. 1) Demand for ITNs: Monthly Net Sales by ITN Price Source: Cohen and Dupas 2008

10. 2) ITN Usage Rates by Price: Share of “Takers” who Report Using ITN at Home Source: Cohen and Dupas 2008

11. Case 3: ITN – UgandaExperimental Evidence on ITNs from Rubagano and Kimuli Villages (Hoffman1) Do free goods stick to poor households? Design NHH=193 T1:Free Net: 71,T2:Cash Transfer: 72,C:Uncompensated: 50 Findings - Wealth and endowment effects result in very few HHs selling free net (FN). - Simulation: Only 6% of FN would be sold in frictionless market. Accounting for transaction cost would further reduce this number. - No significant gender gradient in average compensated valuation of ITNs. -Man have higher income elasticity of supply for ITNs. Men have higher ATP. Implications - Distributing FN to women ►less leakage. - Marketing among man ► more effective. Source: Hoffmann, Barrett, and Just (2007)

12. Case 3: ITN – UgandaExperimental Evidence on ITNs from Rubagano and Kimuli Villages (Hoffman2) Psychology, gender, intra-HH allocation of ITNs Design: NHH=143 T1:Free Net: 71,T2:Cash Transfer: 72 Findings - Free nets lead to greater number of children covered, even for HHs with ATP. - Net retention is higher for free nets (Endowment Effect): NFN > NCT - Women tend to cover larger proportion of HH with nets. - Intra-HH allocation of purchased nets (CT) depends on cost-benefit calculations, with income-earners, net purchasers, and people often suffering from malaria receiving it. - Accompanied with a message, in-kind nets (FN) induce allocation to children. Implications Beyond price, mode of allocation and communication are important. Source: Hoffmann (2007)

13. Case 4: IPT in Schools (CRCT) – Kenya (Clarke et al. 2008) • Malaria Control: Health & Education of School Children 5-18 • Cluster-Randomized Control Trial (CRCT) NSCHOOLS = 30 • Protocol: 3 Treatments of SP/AQ (Dual Placebo) / 4 Months • Significant ↓in prevalence of anaemia & asymptomatic P. falciparum parasitaemia Risk of anaemia halved for IPT recipients v. placebo. Protective Efficacy: 53% Effectiveness: 1 case of anaemia prevented for every 15 children treated with IPT. Adjusted difference in haemoglobin concentration (Hb) is +5.5g/L for the treated. Marked ↓ in prevalence of Pf. Protective Efficacy against Pf Infection: 91% Adjusted risk difference: -0.36, average of 1 infection averted for every 3 children treated. - Significant ↑ in sustained attention. Mean ↑ in test score: +6.74 • No impact was observed on educational achievement. Evidence • IPT’s role in improving the health and cognitive ability of schoolchildren. • Need integrated effective malaria interventions in school health programs. Clarke et al. (2004)

14. Policy Debate: ACT Subsidy Economics of Malaria Treatment (Arrow et al. 2004) • Institute of Medicine (2002) – Economics of Malaria Treatment • Strategy: Enable widespread use of effective drugs to fight malaria, while at the same time delay the emergence of resistance to ACT for as long as possible. • Saving Lives, Buying Time (Arrow et al. 2004): ACT v. AMT Concern - “If ACTs become as inexpensive to consumers as CQ is now, chances are they will be used as frequently as CQ, including over-use for febrile illnesses due to causes other than malaria. All overuse increases the probability that artemisinin-resistant parasites will arise and spread.” (IOM) Economic Case -The greater use of ACTs would delay resistance only if the subsidized ACT would crowd out the practice of artemisinin monotherapy (AMT) or monotherapy of a partner drug (PMT). If the benefits of crowding out monotherapies outweighed the resistance related costs of greater use of ACTs, a global subsidy of ACTs would contribute to the expansion of a global public good, to antimalarial efficacy. Arrow et al. (2004)

15. A Case for ACT Subsidy Ento-Epi Mathematical Modeling (Laxminarayan, Over, and Smith 2005) Objective Explore whether ACT subsidy saves lives and reduces malaria compared to the counterfactuals AMT and PMT, and, if affirmative, at what cost? Findings • Regardless of the degree of responsiveness of antimalarial consumption to price, a subsidy to ACT would save lives even considering that a subsidy could hasten the arrival of parasite resistance to artemisinin-based drugs. • For moderate to high elasticities of demand a larger subsidy is not necessarily more cost-effective than a smaller one. • A delay (even by two years) in instituting a subsidy for ACTs would exacerbate resistance engendered by use of AMT and PMT • A global subsidy for two or more ACTs is likely to be more effective in delaying the onset of resistance and saving lives than reliance on a single or even a limited number of combinations. Laxminarayan, Over, and Smith (2005)

16. A Case for ACT Subsidy (Cont.)Ento-Epi Mathematical Modeling (Laxminarayan, Over, and Smith 2005) Policy / Research Implications • Achieving a high level of subsidy for ACTs should not stand in the way of instituting a more modest subsidy immediately. • The introduction of price subsidies should be accompanied by studies of demand response and of the quality of malaria service delivery, which would enable a more precise determination of the optimal subsidy level. • A subsidy to ACTs should be implemented sooner rather than later. Postponing action is likely to result in diminished CE of future treatment using ACTs. • The global subsidy program should support locally appropriate combination and treatment strategies, while keeping in mind the importance of trans-boundary spillovers of actions/decisions undertaken by any single country. Laxminarayan, Over, and Smith (2005)

17. Socio-Economic EffectsIs there a Malaria Trap? (1) • “Malaria is truly a disease of poverty. It afflicts primarily the poor, who tend to live in malaria-prone areas in dwellings that offer few, if any, barriers against mosquitoes.” (UNICEF 2005). • “As a general rule of thumb, where malaria prospers most, human societies have prospered least…. The extent of the correlation suggests that malaria and poverty are intimately related.” (Sachs and Malaney 2002) Causality - Direction: Does malaria cause poverty? Or does poverty cause malaria? - Spurious: Correlation may be caused by some other direct connection between climate and geography with growth rates or income levels. Gollin and Zimmermann (2007)

18. Socio-Economic EffectsMacro Data Studies (2) Two Cross Country Regressions with Instruments and Controls 1. McCarthy, Wolf, and Wu (1999) - Malaria prevalence is negatively related to growth of per capita income. - Malaria morbidity is linked to climatic differences across countries. • SSA countries experience a reduction in income growth of 0.55% annually 2. Gallup and Sachs (2001) • Countries with “intensive” malaria experience a reduction in per capita income growth of 1.3% annually. • Ceteris paribus, a country experiencing intensive malaria would have its long-term level of income per capita reduced by one-third, compared with the same country in the absence of malaria. Gollin and Zimmermann (2007)

19. Socio-Economic EffectsMicro Empirical Studies (3) SKEPTICAL of the Growth Effects of Malaria • Acemoglu and Johnson (2006); Weil (2007); Cutler, Fung, Kremer and Singhal (2007) EVIDENCE on the Socio-Economic Impact of Eradication Campaigns Endemic & Epidemic Malaria (MMR, Ecology Index, etc.) • Bleakley (2007): US, Brazil, Colombia, and Mexico. Uses MMR & Ecology Index to identify prevalence. Malaria Exposure: ↓ Labor Productivity, and ↓ Adult Income. • Lucas (2005): Women in Paraguay, Sri Lanka, Trinidad 1940-60. Eradication leads to increases in female education and literacy rates. Epidemic Malaria (Weather condition as IV for malaria exposure) • Hong (2007): Malaria risk leads to adverse long-run health outcomes, lower labor force participation, and lower wealth. • Barecca (2007): Utero malaria exposure leads to lower educational attainment. Difficulties • Paucity of reliable data • Inherent difficulty of identification

20. Conclusions • Impact evaluation studies have already contributed much information for understanding the efficacy and cost-effectiveness of various malaria prevention options • Recent impact evaluation studies will help policy makers work through certain malaria control decisions such as free ITN distribution vs. cost-recovery • However many critical questions remain. For example, how do we affect people's behavior to ensure adoption and proper usage of nets? How do we increase the proportion of fever cases seeking treatment at facilities with adequate diagnostic and curative care? • This week we will discuss methods through which we can answer these question.