Enhancing the rational use of antimalarials: The cost-effectiveness of rapid immunochromatographic dipsticks in sub-Saharan Africa Chantal Morel, Sam Shillcutt, Paul Coleman, Catherine Goodman & Anne Mills Health Economics and Financing Programme, Public Health & Policy Department Disease Control & Vector Biology Unit, Infectious and Tropical Diseases Department The London School of Hygiene & Tropical Medicine
Abstract Problem Statement: The massive burden of malaria, along with a severe scarcity of economic resources, makes efficiency in antimalarial drug programs a critical issue in sub-Saharan Africa. Parasite resistance has developed to currently-used first line therapies, to which artemisinin-based combination therapies (ACTs) provide a cost-effective alternative. Rapid immunochromatographic dipsticks may be an efficient method in certain settings to allocate these more-expensive but more-effective drugs. Objectives: This study evaluates the cost-effectiveness of using dipsticks to diagnose malaria in sub-Saharan Africa relative to presumptive antimalarial treatment for all people presenting to a clinic with fever. To set an upper limit for how much a decision maker should be willing to pay to reduce parameter-uncertainty within the model, the expected value of perfect information (EVPI) is calculated. Design: A theoretical decision-analytic model is used to determine the probability that dipsticks are cost-effective across a spectrum of possible prevalence levels. Drug savings are measured according to unnecessary treatments avoided through improved accuracy of diagnosis. Given the uncertainty surrounding cost-effectiveness estimates, the per-person EVPI is calculated. Setting: Sub-Saharan Africa. Population: A hypothetical population presenting with fever to a clinic for treatment. Intervention: Immunochromatographic dipsticks may be used to diagnose malaria in approximately 20 minutes. After blood and buffer are mixed in a sample well, immersion of an antibody-covered strip will indicate the presence of malaria parasites. Outcome Measures: Incremental Cost per Disability Adjusted Life Years (DALYs). Results: At a ceiling ratio of US$150/DALY averted, it is 95% certain that dipsticks are cost-effective where fewer than 15% of febrile patients have parasitemia, and not cost-effective above 55%. The EVPI is greatest between 15% to 55% prevalence with a peak at 33%, the point at which uncertainty around the cost-effectiveness of dipsticks is at its maximum. Conclusions: Based on criteria of economic efficiency, dipsticks should be used in areas where the proportion of febrile illnesses caused by malaria is low. The simplicity and clarity of this diagnostic strategy is likely to provide incentives to encourage people to seek treatment, encourage more rational use of ACTs, and impede the development of resistance to ACTs.
Study Questions • At what levels of malaria prevalence is dipstick diagnosis cost-effective relative to presumptive treatment? • How much should a decision-maker be willing to pay to eliminate uncertainty about model parameters before making a decision?
Introduction Inappropriate diagnosis of febrile illness is a common problem in sub-Saharan Africa. Presumptive treatment for malaria dominates where malaria is prevalent, which leads to excessive prescription of antimalarials and inappropriate treatment of non-malarial fevers. These fevers may become severe with delayed treatment. With the introduction of artemisinin-based combination therapies, presumptive treatment may no longer be affordable. Rapid dipstick tests are an inexpensive and simple diagnostic tool, and are currently being developed for use in endemic areas. This paper examines the cost-effectiveness of using dipsticks to diagnose malaria, given treatment with ACTs, across the possible range of malaria prevalence in low-income countries of sub-Saharan Africa. Beyond the scope of the cost-effectiveness analysis, it is important to consider the value of collecting additional information about parameter values. Both deciding to implement an intervention and deciding to obtain further information involve potential opportunity costs – choosing a sub-optimal intervention, or spending money to confirm an existing recommendation. An EVPI analysis may be used to evaluate the maximum value that further information could add to the model. While EVPI does not determine the value of information given by studies with finite sample sizes, it provides a threshold above which the option to sample further can be rejected. This study estimates the EVPI for the overall model.
Methods A simple decision tree, restricted to patients that present with fever to a public health facility, was developed to calculate incremental cost-effectiveness. The decision tree in Figure 1 follows an individual patient entering the system through to being cured, dying, or surviving with neurological sequelae, according to the sensitivity and specificity of each diagnostic strategy and level of malaria prevalence. Evidence on the progression of non-malarial illnesses is lacking, and it was assumed that their consequences would be similar to untreated malaria. All parameter values, their associated uncertainty, were abstracted from a variety of sources and sub-Saharan African (SSA) countries. A population structure including 50% adults and 50% children was assumed in the model. Costs, in 2002 US dollars, were calculated using the ingredients approach. Only direct costs of medical diagnosis and care were included in this analysis. A range of ACTs were considered, including artesunate-sulfadoxine-pyrimethemine, artemether-lumefantrine (Coartem™), and artesunate-mefloquine1. Drugs recommended by the Integrated Management of Childhood Illness (IMCI) protocol for febrile illness were considered for negative diagnoses: paracetamol, amoxicillin, and chloramphenicol2. 1. Bloland (2001) WHO 2. WHO (1999) IMCI Information Package
Methods Health outcomes were measured in terms of DALYs averted, calculated according to standard methods. Full compliance with diagnosis and treatment was assumed on the part of the patient and the health worker. Parameter uncertainty was quantified using probabilistic sensitivity analysis, and incremental cost-effectiveness ratios (ICERs) were determined. The probability dipsticks are cost-effective was evaluated using a ceiling ratio equal to US$150/DALY averted (λ)1. ICERs were converted to net-benefits using the following formula. Net Benefit = Effects * λ – Costs Expected Value of Perfect Information (EVPI) was calculated according to methods shown in Figure 22. The average net-benefit of the optimal strategy at each iteration was used to approximate the expected value of making a decision under complete certainty. The difference between this and expected net benefit with current uncertainty is the EVPI. 1. WHO (1996) Investing in Health Research and Development, Report of the Ad Hoc Committee on Health Research 2. Fenwick (2000) York Discussion Paper
Figure 1: Simple decision tree model sensitivity, a True positive malaria, p 1-a Falsenegative Suspected malaria specificity, b True negative 1-p 1-b False positive • Give all suspected malaria ACTs: a=1 and b =0 • Use dipstick before giving ACTs: a0.95 and b 0.95
Discussion of Results Rapid dipstick tests will introduce a tradeoff between reducing the prescription of antimalarials with reducing the sensitivity of diagnosis. Our model indicates that where 15% or fewer fevers are caused by malaria, dipsticks are the dominant strategy, and should be used in public health care clinics to diagnose malaria. This result is most sensitive to malaria prevalence and the cost and accuracy of dipsticks. When prevalence is high, the probability that a person will return for treatment if symptoms become severe is important. Reducing amounts of antimalarials prescribed may affect drug pressure on parasites, which would impact the growth of drug resistance. However, improved diagnosis may increase compliance to ACTs (around 40%)1, and use of the public health care system among people receiving antimalarials (around 50%)2. Thus, the net effect on drug pressure is unclear. Improved information on prevalence may help health planners more effectively target preventive and treatment measures towards people who need them most, both within the context of malaria and across disease areas. 1. Depoortere (2004) TMIH 2. Foster (1991) WHO Bull
Limitations and Further Work This analysis is limited in several respects. It assumes that patients and health workers will follow the mode of action suggested by the dipstick results, restricting drug treatment to those with positive tests. In reality, patients who test negative may be given antimalarials. Health workers may lack faith in test quality, and patients may demand drugs anyway. In areas of high transmission intensity, some patients may be immune to levels of malaria parasites that cause illness in others. The interaction of these factors pose complex questions for diagnostics that are not dealt with in this model. Further work is necessary to clarify the causes of treatable febrile illness in people who incorrectly receive antimalarials. Some evidence exists to suggest that pneumonia, salmonella, meningitis, and other illnesses are common. Treatments and outcomes for these diseases differ, and studies are needed to determine their relative contributions to misdiagnosis. Our EVPI estimate provides only a rough estimate of the maximum amount a decision-maker should be willing to pay for perfect information in the entire model. A more useful analysis would estimate the value of testing individual parameters according to the power associated with specific sample sizes. A Bayesian two-step Monte-Carlo simulation approach has recently been developed to make this analysis possible1. 1. Brennan (2004) J. Health Economics
Conclusion and Policy Implications Rapid dipstick tests are highly effective and simple tools for diagnosing malaria. Our model suggests that they should be used where 15% or fewer people that present to public health clinics with fever have malaria. These results should not be interpreted according to endemicity as transmission intensity and parasitemia are not linearly correlated. Further information to reduce uncertainty around model parameters may be useful between 15% and 55% prevalence. However, if these studies are projected to cost more than US$2.20 per person, decision-makers should proceed with the choice to adopt dipsticks. Dipsticks represent a significant investment, costing between US$0.50 and US$1.85 per test1, or about one-half as much as first-line treatment with ACTs2. Currently, 42% of malaria costs are borne by households in SSA, with 39% covered by donors3. The international community must contribute to this efficient use of resources to combat this disease. 1. Kindermans (2002) 2. Bloland (2001) 3. WHO (2003) Africa Malaria Report