Economic Costs of Malaria in Children in the Sub-Sahara
Economic Costs of Malaria in Children in the Sub-Sahara
Three countries were selected to provide estimates for different epidemiological settings within the SSA region: Ghana, Tanzania and Kenya. Ghana is a West African country with about 23 million inhabitants, presenting high malaria endemicity, with 100% of the population living in high transmission areas. Tanzania, with almost 42 million inhabitants including Zanzibar, is an East African country with moderate malaria endemicity. Almost three quarters (73%) of the population in Tanzania live in high transmission areas and approximately one quarter in areas of low transmission. Kenya is another East African country with relatively low malaria endemicity. Kenya has more than 39 million inhabitants with 36% of the population living in high transmission areas, 40% in low transmission areas, and 24% in malaria-free zones.
The costs of treating uncomplicated (outpatients without co-morbidities) and hospitalized cases (all cases requiring parenteral treatment, despite WHO case definitions for severe malaria) were based in each of the three countries on earlier studies that evaluated the economics of intermittent preventive treatment of malaria in infants (IPTi) and in children (IPTc). In IPTi study, data were collected at different health facilities representing the three levels of health care i.e. primary, secondary and tertiary care in each country. Household costs were collected through surveys from a sample of carers of approximately 300 children after an outpatient visit or at discharge (150 outpatients and 150 inpatients) in Kenya and Tanzania. Data collection in Ghana included 207 outpatients and 10 inpatients cases interviewed at home. Household costs were divided into direct and indirect. Direct costs were then divided into the cost of the visit or hospitalization (including facilities and personnel) and the cost of the resources used for treatment (tests and medications). Indirect costs included the carers' reported productivity loss for the entire episode of malaria. Health provider treatment costs included both recurrent and capital costs attributable to malaria care in U5 children.
The breakdown of costs collected during the IPTi study were updated to 2009 rates using the Consumer Price Index of the USA and combined treatment costs for co-morbidities, such as anemia, cerebral malaria and neurological sequelae. In addition, costs were modified according to the new first-line treatment for uncomplicated malaria introduced in recent years [artemisinin-based combination therapy (ACT)]. ACT costs incurred by the households were taken from a recent report. International drug supplier prices were used and augmented by 15% to include shipment costs when drug costs were entirely borne by the health system.
Standards of care and associated costs of co-morbidities and complications were estimated based on interviews with clinicians, health workers and managers of the malaria control programme in the three countries.
In the current study, incremental costs associated with treatment and care of co-morbidities and medium-term consequences not included in the base estimates were considered, for both the health system and the household. Health system additional costs were considered in terms of incremental personnel effort and other resources, such as the extra costs associated with the administration of parenteral treatment compared to oral therapies. Household additional costs were considered in terms of incremental direct (user fees, transportation) and indirect costs (additional value of time lost). Drug costs were imputed to the health system or households depending on national or local policies. Specific costs associated with severity of disease and the presence of co-morbidities was: cost of blood transfusion (severe anemia), cost of anti-seizure/anticonvulsant therapies (cerebral malaria) and rehabilitation costs post-discharge (neurological sequelae).
A live chicken was assumed to be the payment for traditional treatment for one episode of malaria. Institutional local market prices were used to estimate the monetary value of such a payment. Total costs for treating a malaria episode were estimated for the following categories: uncomplicated malaria, malaria hospitalization, malaria hospitalization + severe anemia, cerebral malaria, and cerebral malaria + neurological sequelae. Malaria hospitalization refers to all inpatient cases, regardless of being severe cases according to WHO definition. These categories were based on the perceptions of clinicians and health workers interviewed rather than on institutional definitions. 'Uncomplicated malaria' included all malaria cases (usually laboratory confirmed) where no hospitalization was required. 'Cerebral malaria' was generally referred to as malaria hospitalization of children in deep coma. Total costs were calculated by adding health system and household costs and subtracting user fees paid by the households for consultation or admission at health facility.
The human capital approach was applied to estimate the potential life-long productivity losses due to death. This cost was represented, in each country, by the present value of an annuity with instalment equal to the institutional minimum wage in force, for the period defined by adulthood (from 15 years) and life expectancy. The present value at the time of childhood death, of future potential earnings for an individual (onset of work at 15 years of age) was calculated using the following formula:
R is the annual earning; n is the time (in years) between 15 years and life expectancy; i is the discount rate (assumed to be 3%); m represents the number of years between childhood death and 15 years of age. In the model, death was assumed to occur either at 0–1 or 1–4 years of age. Life expectancies differ between these two age groups.
Models were developed to estimate the expected treatment cost per malaria episode per child by severity and presence of co-morbidities and clinical complications, from the household and health system perspectives. Therefore, the result of each model is the expected value of treatment cost per episode per child, including household and health system costs.
Probabilities of incurring a malaria episode were taken from the results of previous clinical trials (Table 1) where health outcomes were measured at health facilities rather than within the community. Therefore, such data may be biased towards more intensive users of health services. In order to assess if treatment-seeking behaviour would impact the results, two different types of models were constructed. The two models are identical in structure except that the treatment-seeking behaviour for uncomplicated malaria was considered in model type 1 but not in model type 2 (Figure 1). The models start with the probability of experiencing at least one episode of malaria, consider the probability of such an episode becoming severe and conclude with the probability of the severe episodes (with and without co-morbidities) resulting in sequelae or death. In model type 1, episodes of non-complicated malaria are associated with different types of costs depending on the type of treatment sought. No cost was applied when malaria treatment was not sought.
(Enlarge Image)
Figure 1.
Models to estimate expected malaria treatment costs per episode per child (U5) and with or without treatment seeking behaviour (TSB) component.
The burden of malaria in children varies by age. Therefore, several versions of the two model types were developed by age groups for which clinical incidence data were available. In order to account for different epidemiological contexts, the models were used for cost estimation in Ghana, Tanzania and Kenya separately. To represent intra-country epidemiological heterogeneity, data from different areas were used. The two types of model were estimated for five age groups in Ghana, three age groups in Kenya and four age groups in Tanzania. Therefore, a total of 24 scenarios were constructed.
Table 1 summarizes parameters used to populate the models (Figure 1), with their relative sources. The age-specific probabilities of experiencing at least one episode of malaria, the probability of hospitalization and of co-morbidities or complications/sequelae were taken from several different sources, including clinical trials for IPTi and of Intermittent Preventive Treatment of malaria in children (IPTc). Probabilities included in the models were obtained by rates published as outcomes of children in the control group of each trial considered, translated into yearly probabilities. Case fatality rates (CFR) were calculated as average ratios of the number of malaria deaths in U5 children to the number of malaria cases in U5 children reported in World Malaria Reports (between 2000 and 2009), apart from the case of Ghana (2–15 and 16–24 months children) for which malaria death was included among trial outcomes.
Information on treatment-seeking behaviour for malaria was taken from Demographic and Health Survey (DHS) data bases (Standard DHS, Standard AIS – AIDS indicator surveys - in the case of Tanzania). For model type 1, Chi-square tests were performed to test the presence of a statistically significant association between the age of the child and treatment choices.
Most of the variables used to populate the models were taken from studies providing local information, with both cost and epidemiological data being derived from small rural areas in each country ( Table 1 ). To test the uncertainty around estimated mean values, sensitivity analyses were conducted. More specifically, input variables were assigned a range of possible values, to generate a probability distribution. Triangular or uniform distributions were constructed with the estimates of the current study being the most likely value, and the minimum or the maximum being used as the comparator value.
Values used as comparators for health system costs for malaria treatment were WHO-choice cost estimates. For each country, comparator household costs for malaria treatment were derived from different sources. For Ghana, household costs, comparator costs were taken from Asante et al. Household costs in Tanzania were taken from Hutton et al. For Kenya, estimates from Chuma et al. were used as comparators for uncomplicated case costs; Ayieko et al. cost data were used to estimate the costs of complicated cases.
Epidemiological data were also inserted as a probability distribution by comparing estimates with values taken from the World Malaria Report 2009. Monte Carlo simulations were performed within the constructed ranges, (N = 1000 iterations). All analyses were performed using TreeAge Software. 2008. TreeAge Pro 2008 (Tree Age Software, Inc., Williamstown, MA, USA).
As no age breakdown is reported in the World Malaria Report 2010, the total number of malaria cases occurring in U5 children during the year 2009 in Ghana and Tanzania was estimated by assuming the same proportion between U5 cases and all-age cases published in the World Malaria Report 2009. For Kenya, the number of malaria cases occurring in U5 children was assumed to be 40% of cases occurring across all ages. U5 malaria cases were grouped according to severity, using the same clinical/epidemiological data mentioned above (Table 1). Each unit cost per episode, for households and the health system, was multiplied by the number of cases grouped by severity. Death was included in the household indirect cost calculation and the value of one death was represented by the net present value of future potential earnings. Total annual costs were presented from both the households and the health system perspectives. The proportion attributable to U5 children of total annual costs for prevention, from the health system perspective (bed nets and indoor residual spraying) and for the same year in each country, was added to household and health system costs to yield total annual costs. The proportion of prevention costs imputable to U5 children was calculated by multiplying the total cost by the proportion of the total population accounted for U5 children in each country (28%, 18% and 17% for Ghana, Tanzania and Kenya, respectively). Average treatment costs, including household and health system, were calculated by dividing total costs (excluding prevention) by the total number of cases.
Methods
Three countries were selected to provide estimates for different epidemiological settings within the SSA region: Ghana, Tanzania and Kenya. Ghana is a West African country with about 23 million inhabitants, presenting high malaria endemicity, with 100% of the population living in high transmission areas. Tanzania, with almost 42 million inhabitants including Zanzibar, is an East African country with moderate malaria endemicity. Almost three quarters (73%) of the population in Tanzania live in high transmission areas and approximately one quarter in areas of low transmission. Kenya is another East African country with relatively low malaria endemicity. Kenya has more than 39 million inhabitants with 36% of the population living in high transmission areas, 40% in low transmission areas, and 24% in malaria-free zones.
Cost Estimates
The costs of treating uncomplicated (outpatients without co-morbidities) and hospitalized cases (all cases requiring parenteral treatment, despite WHO case definitions for severe malaria) were based in each of the three countries on earlier studies that evaluated the economics of intermittent preventive treatment of malaria in infants (IPTi) and in children (IPTc). In IPTi study, data were collected at different health facilities representing the three levels of health care i.e. primary, secondary and tertiary care in each country. Household costs were collected through surveys from a sample of carers of approximately 300 children after an outpatient visit or at discharge (150 outpatients and 150 inpatients) in Kenya and Tanzania. Data collection in Ghana included 207 outpatients and 10 inpatients cases interviewed at home. Household costs were divided into direct and indirect. Direct costs were then divided into the cost of the visit or hospitalization (including facilities and personnel) and the cost of the resources used for treatment (tests and medications). Indirect costs included the carers' reported productivity loss for the entire episode of malaria. Health provider treatment costs included both recurrent and capital costs attributable to malaria care in U5 children.
The breakdown of costs collected during the IPTi study were updated to 2009 rates using the Consumer Price Index of the USA and combined treatment costs for co-morbidities, such as anemia, cerebral malaria and neurological sequelae. In addition, costs were modified according to the new first-line treatment for uncomplicated malaria introduced in recent years [artemisinin-based combination therapy (ACT)]. ACT costs incurred by the households were taken from a recent report. International drug supplier prices were used and augmented by 15% to include shipment costs when drug costs were entirely borne by the health system.
Standards of care and associated costs of co-morbidities and complications were estimated based on interviews with clinicians, health workers and managers of the malaria control programme in the three countries.
In the current study, incremental costs associated with treatment and care of co-morbidities and medium-term consequences not included in the base estimates were considered, for both the health system and the household. Health system additional costs were considered in terms of incremental personnel effort and other resources, such as the extra costs associated with the administration of parenteral treatment compared to oral therapies. Household additional costs were considered in terms of incremental direct (user fees, transportation) and indirect costs (additional value of time lost). Drug costs were imputed to the health system or households depending on national or local policies. Specific costs associated with severity of disease and the presence of co-morbidities was: cost of blood transfusion (severe anemia), cost of anti-seizure/anticonvulsant therapies (cerebral malaria) and rehabilitation costs post-discharge (neurological sequelae).
A live chicken was assumed to be the payment for traditional treatment for one episode of malaria. Institutional local market prices were used to estimate the monetary value of such a payment. Total costs for treating a malaria episode were estimated for the following categories: uncomplicated malaria, malaria hospitalization, malaria hospitalization + severe anemia, cerebral malaria, and cerebral malaria + neurological sequelae. Malaria hospitalization refers to all inpatient cases, regardless of being severe cases according to WHO definition. These categories were based on the perceptions of clinicians and health workers interviewed rather than on institutional definitions. 'Uncomplicated malaria' included all malaria cases (usually laboratory confirmed) where no hospitalization was required. 'Cerebral malaria' was generally referred to as malaria hospitalization of children in deep coma. Total costs were calculated by adding health system and household costs and subtracting user fees paid by the households for consultation or admission at health facility.
The human capital approach was applied to estimate the potential life-long productivity losses due to death. This cost was represented, in each country, by the present value of an annuity with instalment equal to the institutional minimum wage in force, for the period defined by adulthood (from 15 years) and life expectancy. The present value at the time of childhood death, of future potential earnings for an individual (onset of work at 15 years of age) was calculated using the following formula:
R is the annual earning; n is the time (in years) between 15 years and life expectancy; i is the discount rate (assumed to be 3%); m represents the number of years between childhood death and 15 years of age. In the model, death was assumed to occur either at 0–1 or 1–4 years of age. Life expectancies differ between these two age groups.
Description of the Models
Models were developed to estimate the expected treatment cost per malaria episode per child by severity and presence of co-morbidities and clinical complications, from the household and health system perspectives. Therefore, the result of each model is the expected value of treatment cost per episode per child, including household and health system costs.
Probabilities of incurring a malaria episode were taken from the results of previous clinical trials (Table 1) where health outcomes were measured at health facilities rather than within the community. Therefore, such data may be biased towards more intensive users of health services. In order to assess if treatment-seeking behaviour would impact the results, two different types of models were constructed. The two models are identical in structure except that the treatment-seeking behaviour for uncomplicated malaria was considered in model type 1 but not in model type 2 (Figure 1). The models start with the probability of experiencing at least one episode of malaria, consider the probability of such an episode becoming severe and conclude with the probability of the severe episodes (with and without co-morbidities) resulting in sequelae or death. In model type 1, episodes of non-complicated malaria are associated with different types of costs depending on the type of treatment sought. No cost was applied when malaria treatment was not sought.
(Enlarge Image)
Figure 1.
Models to estimate expected malaria treatment costs per episode per child (U5) and with or without treatment seeking behaviour (TSB) component.
Scenarios
The burden of malaria in children varies by age. Therefore, several versions of the two model types were developed by age groups for which clinical incidence data were available. In order to account for different epidemiological contexts, the models were used for cost estimation in Ghana, Tanzania and Kenya separately. To represent intra-country epidemiological heterogeneity, data from different areas were used. The two types of model were estimated for five age groups in Ghana, three age groups in Kenya and four age groups in Tanzania. Therefore, a total of 24 scenarios were constructed.
Model Inputs and Sources
Table 1 summarizes parameters used to populate the models (Figure 1), with their relative sources. The age-specific probabilities of experiencing at least one episode of malaria, the probability of hospitalization and of co-morbidities or complications/sequelae were taken from several different sources, including clinical trials for IPTi and of Intermittent Preventive Treatment of malaria in children (IPTc). Probabilities included in the models were obtained by rates published as outcomes of children in the control group of each trial considered, translated into yearly probabilities. Case fatality rates (CFR) were calculated as average ratios of the number of malaria deaths in U5 children to the number of malaria cases in U5 children reported in World Malaria Reports (between 2000 and 2009), apart from the case of Ghana (2–15 and 16–24 months children) for which malaria death was included among trial outcomes.
Information on treatment-seeking behaviour for malaria was taken from Demographic and Health Survey (DHS) data bases (Standard DHS, Standard AIS – AIDS indicator surveys - in the case of Tanzania). For model type 1, Chi-square tests were performed to test the presence of a statistically significant association between the age of the child and treatment choices.
Sensitivity Analyses
Most of the variables used to populate the models were taken from studies providing local information, with both cost and epidemiological data being derived from small rural areas in each country ( Table 1 ). To test the uncertainty around estimated mean values, sensitivity analyses were conducted. More specifically, input variables were assigned a range of possible values, to generate a probability distribution. Triangular or uniform distributions were constructed with the estimates of the current study being the most likely value, and the minimum or the maximum being used as the comparator value.
Values used as comparators for health system costs for malaria treatment were WHO-choice cost estimates. For each country, comparator household costs for malaria treatment were derived from different sources. For Ghana, household costs, comparator costs were taken from Asante et al. Household costs in Tanzania were taken from Hutton et al. For Kenya, estimates from Chuma et al. were used as comparators for uncomplicated case costs; Ayieko et al. cost data were used to estimate the costs of complicated cases.
Epidemiological data were also inserted as a probability distribution by comparing estimates with values taken from the World Malaria Report 2009. Monte Carlo simulations were performed within the constructed ranges, (N = 1000 iterations). All analyses were performed using TreeAge Software. 2008. TreeAge Pro 2008 (Tree Age Software, Inc., Williamstown, MA, USA).
Annual Cost Estimates
As no age breakdown is reported in the World Malaria Report 2010, the total number of malaria cases occurring in U5 children during the year 2009 in Ghana and Tanzania was estimated by assuming the same proportion between U5 cases and all-age cases published in the World Malaria Report 2009. For Kenya, the number of malaria cases occurring in U5 children was assumed to be 40% of cases occurring across all ages. U5 malaria cases were grouped according to severity, using the same clinical/epidemiological data mentioned above (Table 1). Each unit cost per episode, for households and the health system, was multiplied by the number of cases grouped by severity. Death was included in the household indirect cost calculation and the value of one death was represented by the net present value of future potential earnings. Total annual costs were presented from both the households and the health system perspectives. The proportion attributable to U5 children of total annual costs for prevention, from the health system perspective (bed nets and indoor residual spraying) and for the same year in each country, was added to household and health system costs to yield total annual costs. The proportion of prevention costs imputable to U5 children was calculated by multiplying the total cost by the proportion of the total population accounted for U5 children in each country (28%, 18% and 17% for Ghana, Tanzania and Kenya, respectively). Average treatment costs, including household and health system, were calculated by dividing total costs (excluding prevention) by the total number of cases.
Source...