Economic burden of malaria in rural Tanzania: variations by socioeconomic status and season


Corresponding Author Masha F. Somi, Australian Centre for Economic Research on Health, Australian National University, Cnr Mills and Eggleston Roads, Canberra ACT 0200, SA, Australia. Tel.: +61 2 6125 3688; Fax: +61 2 6125 9123; E-mail:


Objective  To determine the economic burden of malaria in a rural Tanzanian setting and identify any differences by socioeconomic status and season.

Methods  Interviews of 557 households in south eastern Tanzania between May and December 2004, on consumption and malaria-related costs.

Results  Malaria-related expenses were significantly higher in the dry, non-malarious season than in the rainy season. Households sought treatment more frequently and from more expensive service providers in the dry season, when they have more money. Malaria expenses did not vary significantly across socioeconomic status quintiles, but poorer households spent a higher proportion of their consumption in both seasons.

Conclusion  Poorer households bear a greater economic burden from malaria relative to their consumption than better-off households. Households are particularly vulnerable to malaria in the rainy season, when malaria prevalence is highest but liquidity is lower. Alternative strategies to assist households to cope with seasonal liquidity issues, including insurance, should be investigated.


Objectifs  Déterminer la charge économique de la malaria dans une zone rurale tanzanienne et identifier toutes les différences selon le statut socio-économique et la saison.

Méthodes  Entretiens dans 557 ménages dans le sud-est de la Tanzanie entre mai et le décembre 2004, concernant les coûts pour la malaria.

Résultats  Les dépenses des ménages liées au contrôle de la malaria (prévention, coûts directs et indirects) variaient sensiblement à travers les saisons (plus élevées durant la saison sèche et non-malarique que durant la saison des pluies). Les ménages recherchaient un traitement plus fréquemment et chez des pourvoyeurs de service plus coûteux durant la saison sèche, quand la liquidité disponible des ménages est plus élevée. Les dépenses des ménages liées à la malaria ne variaient pas de manière significative à travers les quantiles de statut socio-économique, mais les ménages plus pauvres consacrent une proportion plus élevée de leurs dépenses à la malaria durant tous les deux saisons.

Discussion  Les ménages pauvres portent un plus grand fardeau économique de la malaria relatif à leur consommation que ceux de statut socio-économique élevé. Les ménages sont particulièrement vulnérables à la malaria durant la saison des pluies où la prévalence de la malaria est la plus élevée mais l’accès aux ressources est plus limité. Des stratégies alternatives d’aide aux ménages pour faire face aux problèmes de liquidité saisonnière, y compris une assurance, devraient être étudiées.


Objetivos  Determinar la carga económica de malaria en una localidad rural de Tanzania e identificar cualquier diferencia por estatus socioeconómico y estación.

Métodos  Los datos sobre costes relacionados con el consumo y la enfermedad por malaria fueron recolectados en 557 hogares del sureste de Tanzania, entre Mayo y Diciembre del 2004.

Resultados  Los costes asociados con malaria estaban mayores en la estación seca, sin malaria, que en la estación lluviosa. La liquidez de los hogares (mayor en la estación seca) es un determinante importante en esta zona de Tanzania a la hora de gastar en malaria. Mientras que los costes asociados con malaria no variaron significativamente a lo largo de los diferentes quintiles de nivel socioeconómico, si varió la cantidad comprometida a la hora del consumo: los hogares con un nivel socioeconómico bajo dedicaron una proporción más alta de su consumo a la malaria en ambas estaciones.

Conclusión  Los hogares con bajo nivel socioeconómico tienen una mayor carga económica por malaria con respecto a su consumo, que el que tienen hogares con un mayor nivel socioeconómico. Los hogares son particularmente vulnerables a la malaria durante la estación de lluvias, durante la cual la prevalencia de malaria es mayor pero el acceso a recursos es menor. Se deberían estudiar estrategias alternativas para ayudar a los hogares a sobrellevar los problemas de liquidez relacionados con la estacionalidad, incluyendo la opción de un seguro.


In African settings households face high economic burdens from diseases such as malaria. These burdens comprise three components: preventive expenditures, direct and indirect costs of illness. This study had two aims: to explore all three components of the economic burden of malaria in rural Tanzania and to determine if socioeconomic differences exist in any or all three; and to see whether seasonal differences exist in the economic burden of malaria.

Data and methods

Study area

Over 96% of Tanzania’s population lives in areas of malaria risk (Roll Back Malaria 2003), and it is the most commonly reported health complaint in the country (Government of Tanzania 2001). This study took place in 25 villages in Kilombero and Ulanga Districts, Morogoro Region, south eastern Tanzania, under the umbrella of the Ifakara Demographic Surveillance Site. The area is described in detail in an INDEPTH monograph (INDEPTH 2002). The most common occupations are subsistence farming, fishing and small-scale trading. Rice and maize are the predominant food crops. The site has no paved roads, and some villages are cut off for parts of the year as a result of flooding. Most families have a second house at a farm, where they stay during the planting and harvesting seasons. Malaria transmission at the study site is intense and year round, with a mean entomological inoculation rate of 395 infective bites per person per year (G. Killeen, personal communication).

Data collection and analysis

Informed consent was obtained for all participating adults, and from parents or guardians of children under the age of 11 years. The institutional review boards of the Australian National University, Ifakara Health Research and Development Centre and the Tanzanian Medical Research Coordinating Committee approved the study.

Household surveys were conducted over two Rounds. In the first household survey (conducted between May and August 2004 and called Round 1), information was collected on: the amount the household had spent on malaria prevention in the month before interview; whether any member had experienced malaria or fever (used as a proxy for malaria in the absence of microscopic examination); whether individuals with malaria/fever had sought treatment; and the sources and costs of treatment. The first interviews coincided with the rainy season when malaria is more common. Data collection in Round 1 was undertaken in collaboration with the Interdisciplinary Monitoring Project for Antimalarial Combination Therapy in Tanzania (IMPACT-Tz)1. For the requirements of the IMPACT-Tz project, 1591 households from the Ifakara Demographic Surveillance Site were randomly selected to participate in the study; 1295 of these households participated in Round 1.

Of the 1295 households which participated in Round 1, 600 were randomly selected to participate in Round 2 data collection, and 557 of these provided sufficient information to be included in the analysis. These households were reinterviewed three to four months after their initial interview (between September and December 2004) on consumption patterns to determine socioeconomic status. We used three recall periods: 1 year for expensive and infrequently purchased items such as furniture, 1 month for items such as clothes and 1 week for frequently purchased items such as food. The value of consumption of items produced within the home was imputed and included in the aggregate. Households were also asked about ownership of houses and durable items. Data analysis for the consumption survey followed the method described by Deaton and Zaidi (1999). Total household consumption comprised imputed rental from all houses the household owned, user fees from all durable items owned and reported household consumption. Total consumption was deflated by a price index (calculated for each household) and adjusted for household size, to give household consumption per equivalent adult in the household with the effects of inter-household differences in prices removed. Health-related spending and tax costs were not included in the total, as neither increases welfare. Households were classified on the basis of their consumption in socioeconomic quintiles.

During Round 2 data collection, households were asked only about preventive spending, illness bouts and treatment seeking for the reported illnesses, not about the direct and escort costs associated with treatment seeking at each service provider. The cost per treatment at each source, including home medication, was calculated from Round 1 data, and applied to the treatment seeking patterns reported in Round 2. Because of the data collection method, individual costs of treatment seeking are only reported for Round 1, and seasonal comparisons of the costs of treatment are only made at the household level. There were two reasons for not collecting data on the costs of malarial illness (particularly the direct costs) in Round 2. First, this limited the responder burden on households, particularly the time costs. Given that the collection of malarial illness costs and consumption data were the largest component of the data collection, we decided to split these segments between the two interview Rounds. Secondly, budgetary constraints limited the number of staff that could be employed by the project, so collecting of malarial illness costs in Round 2 would have reduced the number of households that could be interviewed.

The significance of differences in household consumption and malaria expenditure results was determined using anova for cross quintile comparisons, and Satterthwaite’s t tests (Casella & Berger 1990) for comparing the highest and lowest spending quintiles. Cuzick’s test of trend (Cuzick 1985) was used to determine if there was a trend across socioeconomic status quintiles in the costs associated with malaria. Pearson’s chi-squared and Fisher’s exact t tests were applied to compare the proportion of households from each socioeconomic status quintile: using different prevention methods; reporting the presence of a malarious/feverish individual; and seeking malaria treatment. Two sample t tests were used to compare spending variables across the two data collection Rounds.


Malaria is defined as ‘self report of malaria or fever in the 2 weeks before interview.’ This definition is supported by authors such as Mugisha et al. (2002) who argue that it is the perception of illness that determines whether or not an individual seeks treatment, and that the resulting expenditure depends to some extent on the perceived type and severity of illness. Malaria prevention expenditure includes household spending in the month before interview on mosquito nets, insecticide treatment for mosquito nets, insect spray, mosquito coils, mosquito repellent, burning organic materials such as rice husks and malaria prophylaxis. The direct costs of malarial illness comprise medicines, prescription drugs, food, transport and accommodation. The indirect costs of illness represent the value of production lost because of the illness, and equal time lost from work multiplied by the wage rate. The amount of time lost to work is the sum of the number of days lost to illness by those reporting malaria/fever and the hours any adults spent accompanying ill individuals who were seeking treatment. Escort costs were calculated for accompanying individuals over the age of 12 years (assuming that <12 years do not work). The wage rate used for calculating the indirect costs of illness is based on the Tanzanian minimum wage of 48000 Tsh/month. The full daily wage rate of 2087 Tsh/day was assigned to adults (≥18 years) and half (1044 Tsh) to young adults (15–18 years) and older persons (>60 years). The economic burden of malaria to households is the sum of prevention, direct and indirect costs arising from preventing and treating malaria.

In both Rounds of data collection some individuals had ongoing malaria bouts at the time of interview (23% in Round 1 and 33% in Round 2). Estimates of projected length of illness remaining for those people were calculated using conditional probability curves, by plotting the duration of illness for each completed bout of malaria/fever in Microsoft Excel. On these charts illness duration (days) was recorded on the x-axis, and the proportion of individuals still ill on that day on the y-axis. These curves provided an estimation of the probability that an individual who was ill today would still be ill the next day. The curves are downward sloping because only individuals reporting malaria/fever were included, all of whom reported being sick for at least 1 day. For Round 1, two outliers were removed before the curves were developed (55 and 89 days illness duration). The expected duration of incomplete bouts is the number of days already ill plus the sum of the conditional probability of still being ill on the following day multiplied by 1 day (which also equals the area under the curve for the days which the individual has not yet been ill). For those individuals still ill on the day of interview, the expected duration of illness was used in calculating the economic burden of malaria. For indirect costs, the cost of lost production was calculated using the expected duration of illness rather than the actual duration at time of interview (for individuals with incomplete episodes). For the direct costs, a daily cost of illness was calculated (by dividing the sum of direct costs by the number of days the individual was ill) and then applied to each additional day the individual was expected to be ill. This same method was used to calculate the expected costs of escorting individuals to treatment. The costs reported in this paper reflect the total costs of seeking treatment incurred by households and include the projected indirect and direct costs predicted for incomplete episodes of illness.


Table 1 shows the projected number of days of illness remaining and the expected durations of bouts of malarial/fever illness using conditional probability curves for individuals with incomplete bouts of illness, for each of the two Rounds of data collection. Whilst the proportion of individuals reporting malaria/fever in the 2 weeks before interview was similar in the two Rounds, there was a difference in the proportion of sick individuals seeking treatment and self medicating using medicines from home or from the neighbourhood (Table 2). Direct costs varied by service provider (Table 3).

Table 1.   Expected duration of incomplete bouts of malarial/fever illness
Actual length of illness at time of interviewRound 1 (Rainy season)Round 2 (Dry season)
Number of individualsProjected days of illness remainingExpected duration of illnessNumber of individualsProjected days of illness remainingExpected duration of illness
(1)(2)(3)(4) = (1) + (3)(5)(6)(7) = (1) + (6)
Table 2.   Illness reporting and treatments seeking in the sample
Illness and treatment seeking informationRound 1 (Rainy season)Round 2 (Dry season)
Number of individuals in the sample20332001
Number (%) of individuals reporting malaria/fever 208 (10.2) 193 (9.6)
Average duration of illness4.24.1
Number (%) of individuals who sought treatment 159 (70) 176 (91)
Number of treatments sought 173 183
Location of treatments (n and %):
 Health facility – government run 35(20)  4 (2)
 Health facility – non-government  run 17 (10) 76 (42)
 Medicine stores 80 (46) 62 (34)
 General stores 37 (22) 33 (18)
 Traditional healer  4 (2)  –
 Other (unspecified)  –  8 (4)
 Total 173 (100) 183 (100)
 Number of individuals  who self medicated 20 (13) 13 (7)
Table 3.   Average direct costs of treatment, by service type
Service typeAverage direct costs per service provided (Tsh)
Health facility – government run897
Health facility – non-government run1183
Medicine store558
General store89
Traditional healer195
All service types643

Household costs associated with malaria

Differences in the average amounts spent on malaria prevention by households of different socioeconomic status (Table 4) were not significant (anova, P > 0.05), and there was no trend across quintiles (Cuzick’s test, P > 0.05). Households spent more on malaria prevention in Round 2 than in Round 1 (with the exception of households in the Poor and Less poor quintile), but again these differences were not significant (t test, P > 0.05).

Table 4.   Average household spending on malaria prevention by socioeconomic status (SES) qunitile (Tsh)
SES quintileNumber of householdsAverage spending on malaria prevention (Tsh)
Round 1 (Rainy season)Round 2 (Dry season)
1. Most poor112149179
2. More poor111113185
3. Poor112251144
4. Less poor111196196
5. Least poor11170188
All households557156178

Whilst a large proportion of households used mosquito nets in the month before interview, few treated those nets with insecticide, and almost no households used alternative methods such as insect spray or mosquito repellent (results not shown). The only statistically significant difference across socioeconomic status quintiles in the use of preventive measures in both Rounds was in mosquito net use, with better-off households being more likely to use them (Pearson’s chi-squared, P = 0.03 in Round 1 and P = 0.002 in Round 2), and the trend across quintiles is positive and significant (Cuzick’s test, P = 0.002 in Round 1 and P < 0.001 in Round 2). In Round 1 the difference in use of insecticide treatment across socioeconomic status quintiles is significant (Pearson’s chi-squared, P = 0.04), though there is no specific trend (Cuzick’s test, P > 0.05). Grouping non-mosquito net and insecticide treatment preventive measures to form an ‘other’ category made no difference in significance (Fisher's exact test, P = 1.00). Mosquito nets were used significantly more frequently in Round 1 than in Round 2 (t test, P < 0.001); however, the differences in use of the other preventive measures were not significant, even when grouped into one category.

Malarial illness costs

Whilst a higher proportion of households reported that at least one individual had malaria/fever in Round 1 than in Round 2 (Table 5), this difference was not significant (30% and 28%; t test, P > 0.05). Reporting of malaria or fever varied by the socioeconomic status of the households, though this difference was not significant (Pearson’s chi-squared, P > 0.05). Of the households with at least one ill individual, 83% (n = 140) reported seeking treatment in Round 1, against 90% (n = 141) in Round 2. Whilst the difference in the proportions of households seeking treatment did not vary significantly across households from different socioeconomic status (Pearson’s chi-squared, P > 0.05), the difference across the two Rounds is significant (t test, P < 0.001).

Table 5.   Households with individuals with malaria/fever and who sought treatment
Socioeconomic status quintileNumber of households in sampleRound 1 (Rainy season)Round 2 (Dry season)
Reporting malaria/fever (%)*Seeking treatment (%)†Reporting malaria/fever (%)Seeking treatment (%)
  1. *At least one individual in the household has reported malaria or fever in the 2 weeks before interview.

  2. †At least one individual in the household (only households reporting ill individuals included) has sought treatment (self medication or at a service provider).

Most poor11227 (24)21 (78)32 (29)28 (88)
More poor11140 (36)35 (88)37 (33)33 (89)
Poor11241 (36)37 (93)31 (28)29 (94)
Less poor11134 (31)28 (82)35 (32)33 (94)
Least poor11126 (23)19 (53)21 (19)18 (88)
Total557168 (30)140 (83)156 (28)141 (90)

In Round 1 the Least poor households spent most on direct costs (Table 6) for malaria treatment (1036 Tsh), followed by the More poor (1007 Tsh) and the Less poor (738 Tsh). Whilst the difference across all socioeconomic status quintiles is not significant (anova, P > 0.05) and there is no trend across the quintiles (Cuzick’s test, P > 0.05), the difference between the highest (Least poor, 1036 Tsh) and the lowest (Poor, 536 Tsh) average direct costs was significant (t test, P = 0.04). In Round 2 there was a clear and significant gradient in direct spending (anova, P = 0.03; Cuzick’s test, P = 0.005), with the Most poor households spending the least (889 Tsh) and the less poor households spending the most (1354 Tsh). The difference in direct spending between the two Rounds was also significant (t test, P < 0.001), with spending averaging 767 Tsh in Round 1 and 1147 Tsh in Round 2.

Table 6.   Average direct and indirect costs of malarial illness per household, by socioeconomic status (SES) quintile (Tsh)
SES quintileDirect costsIndirect costs
Round 1 (Rainy season)Round 2 (Dry season)Round 1 (Rainy season)Round 2 (Dry season)
Most poor2156928889247331325913
More poor351007331099394009348815
Less poor28738331354282637346320
Least poor191036181348194519206347

In Round 1, the Most poor households faced the highest indirect costs (7331 Tsh) and the Less poor households the lowest (2637 Tsh). Whilst the difference across all quintiles was not significant (anova, P > 0.05) and there was no trend across the quintiles (Cuzick’s test, P > 0.05), the difference between the highest (Most poor) and lowest (Less poor) indirect costs was significant (t test, P < 0.001). In Round 2 the More poor households incurred the highest indirect costs (8815 Tsh) and the Poor households the least (5563 Tsh). Whilst the difference between all quintiles was not significant (anova, P > 0.05) and there was no trend across quintiles (Cuzick’s test, P > 0.05), the difference between the highest (More poor) and lowest (Poor) indirect costs was significant (t test, P = 0.03). The difference in indirect costs between the two Rounds was also significant (t test, P < 0.001), with indirect costs averaging 4477 Tsh in Round 1 and 6671 Tsh in Round 2.

Economic burden of malaria

Average total costs (sum of preventive, direct and indirect), or the economic burden of malaria, and the share of annual consumption they comprise are outlined in Table 7. In Round 1, the average economic burden of malaria was highest in the Most poor households (7242 Tsh), followed by the Least poor households (4815 Tsh). Households in the Less poor quintile experienced the lowest economic burden in Round 1 (3154 Tsh). Whilst the difference in total costs across all quintiles was not significant (anova, P > 0.05) and there was no trend across the quintiles (Cuzick’s test, P > 0.05), the difference between the highest (Most poor) and lowest (Less poor) quintiles was significant (t test, P = 0.002). In Round 2 spending on malaria was highest in the More poor households (10199 Tsh), followed by the Less poor households (7914 Tsh). Poor households bore the lowest economic burden (6813 Tsh). Again, the difference in malaria expenditure across the quintiles was not significant (anova, P > 0.05) and there was no trend across the quintiles (Cuzick’s test, P > 0.05), but the difference in spending between the highest (More poor) and lowest (Most poor) households was significant (t test, P = 0.04). The difference in the economic burden of malaria between the two Rounds was significant (t test, P < 0.001), averaging 4871 Tsh in Round 1 and 7970 Tsh in Round 2.

Table 7.   Economic burden of malaria (Tsh) and its share of consumption, by socioeconomic status (SES)
SES quintileEconomic burden of malariaEconomic burden as share of consumption (%)
Round 1 (Rainy season)Round 2 (Dry season)Round 1 (Rainy season)Round 2 (Dry season)
Most poor267242326813267.2326.4
More poor4048013410 199402.6345.6
Less poor313154347914310.9342.4
Least poor224815207785220.9201.6

Average economic burden as a share of consumption averaged 2.7% in Round 1 and 3.9% in Round 2. In both Rounds the share of consumption taken by malaria is largest for the Most poor (7.2% in Round 1 and 6.4% in Round 2), and smallest for the Least poor (0.9% in Round 1 and 1.6% in Round 2). The share of household consumption dedicated to the economic burden of malaria varied significantly by socioeconomic status in both Rounds (Cuzick’s test, P < 0.001 in both Rounds). The difference between highest and lowest shares of consumption spent on malaria (Most poor and Least poor) is significant in both Rounds (t test, P < 0.001 in both Rounds). The share of consumption spent on malaria varied significantly by season (t test, P < 0.0001).


This study aimed to calculate the economic burden of malaria in rural Tanzania and to determine if differences exist in the burden by socioeconomic status and by season. There are no clear differences or trends across the socioeconomic status quintiles for the three components of economic burden of malaria (prevention, direct and indirect costs), with the exception of the direct costs of treatment in Round 2. These findings are generally consistent with others undertaken in Africa, and those differences that have been found reflect differences in the study samples being investigated. Whilst Ettling et al. (1994) and Wiseman et al. (2006) found that high socioeconomic status households spent significantly more on malaria preventive activities in Malawi and The Gambia respectively, the vast majority of household spending was on consumables such as insecticide sprays, and mosquito coils and repellents; in our Tanzanian sample, most malaria preventive spending was on mosquito nets, which last for long periods of time. In the Ifakara DSS, the vast majority of households (including 94% of the Least poor households) already have access to mosquito nets. It is likely that the relatively high spending in the lower socioeconomic status quintiles reflects a catch up period, given that the highest socioeconomic status quintiles have already nearly reached a point of saturation in mosquito net ownership. These findings indicate that higher socioeconomic status households do not need to spend significantly more on malaria prevention to gain the reduced malaria parasitaemia rates reported by Njau et al. (2006), which implies that they gain benefits from previous (longer run) expenditure on items such as mosquito nets.

In Malawi, Ettling et al. (1994) similarly found that the direct costs of malaria did not vary by socioeconomic status: US$19.13 amongst very low income households and US$19.94 amongst low to high income households. Whilst the results are consistent with those from our study, a direct comparison is difficult because of the groupings of socioeconomic status selected; Ettling et al.’s (1994) study compared two, we compared five. Njau et al. (2006) found that the direct costs of malaria were significantly higher amongst the better-off third than amongst the worse-off two-thirds. This different finding may be due an important difference in design: Njau et al. (2006) collected data from both Ifakara DSS and Rufiji DSS (a site close to Ifakara DSS but not contiguous, and with a population with differing characteristics), we only collected data from Ifakara DSS. The differences in findings may reflect real life differences between the two sites. The demarcation of households into socioeconomic status groups may have also contributed, as Njau et al. (2006) formed thirds and we formed quintiles. The Least poor households do in fact have the highest direct costs associated with malaria and the Less poor the third highest costs (Table 6). How similar the results of the two studies may have been had similar groupings been used for socioeconomic status was not explored.

Ettling et al. (1994) found that high socioeconomic status households bore larger burdens of indirect costs; we saw no difference across the quintiles. Ettling et al. (1994) costed the time of high socioeconomic status individuals at a higher rate than the low socioeconomic status individuals; thus very low income households in the Malawi sample lost an average 28 days to malaria over 1 year costed at US$2.13 whereas high income households lost 22 days costed at US$20.61.

The pattern of costs across socioeconomic status quintiles is quite volatile. This result, which reflects the large variation in household costs of malaria in the sample (ranging from 0 up to almost 28 900 Tsh), is unlikely to be due to sampling error or the particular measures of socioeconomic status adopted in this analysis. The same uneven pattern of household spending on health was found in another analysis of this sample investigating catastrophic health expenditures. It may be, however, that Cusick’s test of trend (which uses the five socioeconomic status quintiles as data points) is not sufficiently sensitive to detect a subtle socioeconomic gradient in costs within the uneven pattern observed across the quintiles.

Households from all quintiles face similar preventive, direct and indirect costs in relation to malaria, but these costs comprise significantly different proportions of their consumption. This finding is consistent with two other African studies: in Malawi the share of income that malaria spending comprised varied significantly by income (Ettling et al. 1994); in Kenya the share of expenditure spent on malaria each month varied across socioeconomic status quintiles (Chuma et al. (2006).

There are significant seasonal differences in the economic burden of malaria in Ifakara DSS. Households spent more on each component of the economic burden of malaria in Round 2 (dry season) than in Round 1 (rainy season), despite reporting more malaria/fever in Round 1 (P > 0.05). The finding that malaria was not more prevalent in the rainy season was unexpected; Sauerborn et al. (1996) also found that illness reporting fell during the rainy season, when the opportunity cost of time away from work is higher. Perhaps households are less likely to report illness when they have pressing tasks. Higher spending on malaria in the dry season is likely to reflect greater liquidity because of rice planting and harvesting cycles. Liquidity seems an important determinant of household spending on malaria treatment, and socioeconomic differences become more pronounced as cash availability increases. Other studies have also found that household spending on health (Sauerborn et al. 1996) and, in particular, on malaria (Chuma et al. 2006) is higher in the season with greater cash availability.

Given that liquidity, and thus malaria spending, is seasonal, strategies need to be developed to assist households to cope in the months they are most vulnerable. Researchers investigating the economic burden of malaria, or in fact of any disease, need to consider seasonal variations in liquidity and spending patterns when designing their studies (Litvack & Bodart 1993; Chima et al. 2003). Whilst health insurance is generally regarded as the best strategy to deal with these fluctuations in household income, difficulties with implementing insurance schemes have been reported in Tanzania (Munishi 2003): low community participation, poor use of revenues collected, and inconsistent drug availability at health centres. Further research would help identify functional alternatives to health-centre based insurance.

A limitation with this Tanzanian study in comparing seasonal variations in the costs of treatment is that direct costs associated with malaria were only collected during Round 1, and extrapolated to Round 2. It is likely that direct spending on malaria treatment would have been higher in Round 2 than Round 1 given greater household liquidity. Furthermore, we did not measure severity of illness for the periods individuals reported illness. Estimates for the indirect costs of illness therefore incorporate the whole time the individual was ill, and do not differentiate between periods when an individual may have been working (even at partial capacity) and when they were unable to. In this same vein, no account was taken of time that adults may have spent taking care of ill children. Further, assumptions were made about the productivity of individuals, specifically that all individuals of a particular age group had similar productivity and hence were paid the same wage rate. While this wage rate was used to value all time losses associated with malarial illness regardless of whether the person was formally employed or not, it may give a distorted estimate of the true time cost as the actual wage rate is usually less than the minimum wage rate in rural Tanzanian settings. Various aspects of the data collection would have been subject to recall and social desirability biases, including consumption information, treatment seeking patterns (particularly about traditional healers) and their associated costs. It is likely, however, that the effects of these biases would have been random. A final limitation of the study is that the estimates of the economic burden of malaria do not accurately account for complications arising from malarial illness, such as anaemia and sequelae/disability (Snow et al. 2004) or for the social costs associated with illness (Jones & Williams 2004).


Poorer households bear a greater economic burden from malaria relative to their consumption than better-off households. Households are particularly vulnerable to malaria in the rainy season, when malaria prevalence is highest but liquidity is lower. Alternative strategies to assist households to cope with seasonal liquidity issues, including insurance, should be investigated. The seasonal variation in the economic burden of malaria has implications for the design and interpretation of studies.


  • 1

    The Interdisciplinary Monitoring Project for Antimalarial Combination Therapy in Tanzania (IMPACT-Tz) is a multiyear implementation research evaluation project that rests on a collaborative platform comprising the US Centers for Disease Control and Prevention (CDC), Ifakara Health Research and Development Centre, the National Institute for Medical Research, London School of Hygiene and Tropical Medicine (UK), and the Tanzanian Ministry of Health, including its National Malaria Control Programme, the Tanzania Essential Health Interventions Project, and the Council Health Management Teams of Rufiji, Morogoro, Mvomeru, Kilombero and Ulanga Districts. IMPACT-Tz is primarily supported by funding from CDC, the United States Agency for International Development and Wellcome Trust.


We thank the Ifakara Health Research and Development Centre for their support and assistance throughout the data collection. The field work would not have been possible without the help of Rashid Khatib, Jensen Charles and Chrisostom Mahutanga and Berty F. Elling. The IMPACT Tanzania project was also very supportive during data collection and shared information with this project. The findings and conclusions presented in this paper are those of the authors and do not necessarily represent those of the United States Public Health Service or the Centers for Disease Control and Prevention.