OBJECTIVE To assess the costs of tuberculosis at household level in Dar es Salaam and to compare them with the provider costs of the national tuberculosis control programme.
DESIGN Tuberculosis patients were found by active case searching within a routine census in three areas of Dar es Salaam, and by examining records for residents already receiving treatment. Costs at household level were evaluated through a cross-sectional household survey.
RESULTS One hundred and ninety-one tuberculosis cases were included in the survey. With treatment periods of 8 to 12 months, extrapolated average costs of a period of illness to patients and their families were as follows: US $2 for examination and laboratory costs, between US $17 and US $50 for consultation and drugs, less than US $1 for hospitalization and between US $13 and US $20 for transport. The analysis revealed high costs due to inability to work, ranging from US $154 to US $1384. These data were compared with the operation costs of the tuberculosis programme and proved to comprise 68% to 94% of total costs.
CONCLUSIONS For patients and their families, tuberculosis implies three main types of cost: drugs, transportation and, most importantly, financial loss due to inability to work. They represent around two thirds of total cost and are a high economic burden for households, in particular those with a low-income. While assessing tuberculosis control strategies such as direct case finding at home, it is therefore important to also include costs incurred at household level.
Over the last several years, tuberculosis has been increasing in many sub-Saharan countries. One reason is its association with Human Immunodeficiency Virus (HIV) seropositivity. Tuberculosis is now a major source of morbidity and mortality in Africa with an estimated incidence of smear-positive cases of 100–450 per 100 000 per year ( Murray et al. 1993 ; Snider 1994; Murray & Lopez 1996; Dye et al. 1999 ). The tuberculosis control programme in Tanzania is based on the health services and mostly confined to governmental services ( Ipuge et al. 1997 ). Tanzania was one of the first countries to introduce short-course therapy, achieving effective cure rates of 86% of patients enrolled in chemotherapy ( Murray et al. 1991 ; De Jonghe et al. 1994 ). It is seen as one of the most successful control programmes. As the incidence and prevalence of tuberculosis is also increasing in Dar es Salaam ( Chum et al. 1991 ), and in the context of the current health sector reform in Tanzania, attempts are underway to identify new ways of organizing control and treatment. Elements being discussed and implemented include the transfer of tuberculosis control to peripheral levels in addition to active case-finding at home ( Ministry of Health 1999).
As costs of tuberculosis at household level may place a heavy economic burden on the individual and society, this research assessed the direct and indirect costs of pulmonary tuberculosis to households, and compared these expenditures to provider costs of the national tuberculosis programme, to give an assessment of the total cost of tuberculosis in Tanzania.
A cost-effectiveness assessment of chemotherapy for pulmonary tuberculosis in Tanzania ( Murray et al. 1991 ; De Jonghe et al. 1994 ) estimated operation costs of the control programme to be US $174 (1989) per person for short-course therapy with hospital admission, and US $72 for ambulatory short course chemotherapy. The World Bank has reviewed the cost-effectiveness of nearly all known health interventions applicable on a mass scale in developing countries. The costs per death averted and per year of life saved mean that chemotherapy for smear-positive tuberculosis is one of the cheapest strategies along with immunization for measles and tetanus, oral rehydration therapy for diarrhoea, blood bank screening for HIV, and BCG vaccination ( Jamison et al. 1993 ).
The current standards for cost-effectiveness analysis advocate the use of a societal perspective ( Gold et al. 1996 ; Weinstein et al. 1996 ). It is recommended for conducting such an analysis to include all costs and all health effects, regardless of who incurs costs and who is affected. For assessing costs related to tuberculosis treatment, this means that while comparing different control strategies, costs at household level should be considered, as these may be substantial ( Saunderson 1995; Sawert et al. 1997 ; Kamolrantanakul et al. 1999 ).
In health care, economic costs can be categorized as provider (programme) and household costs, or as direct and indirect costs ( Table 1) ( Drummond et al. 1988 ; Patrick & Erickson 1993; Creese & Parker 1994). Provider costs are those associated with developing and operating a health care programme or service. They include staff costs (e.g. medical assistants, nurses, etc.) as well as costs of supplies and equipment (e.g. microscope, reagents, cars, etc.) needed for making diagnoses and providing treatment and rehabilitation. Personal or household costs are those borne by patients and their families. They include both direct and indirect costs. Direct costs include medical costs, i.e. payments made to investigate and treat symptoms and cover expenditure for hospitalization, consultation fees or drugs. People suffering from tuberculosis can also incur direct nonmedical costs as a result of the illness or as part of treatment. Such costs may include modification of accommodation at home and transport to obtain care. In addition to direct costs related to health care, patients and their families often incur indirect costs. Some authors ( Drummond et al. 1988 ; Gold et al. 1996 ) have drawn attention to potential confusion about the term ‘indirect’ costs due to accounting practices, where the expression is used to describe overheads of fixed costs of production. However, indirect costs here refer to the value of resources lost, including reduced levels of work output and loss of productivity resulting from inability to work or from a change of employment. The cost of care provided by relatives and friends may be direct if they are reimbursed, or indirect in the form of time spent by household members on care rather than on other work.
Cost classification used by this study
Study population and methods
Dar es Salaam is the biggest city in Tanzania and the economic centre of the country. At the time of the study (1996) there were around 2.5 million inhabitants. Three areas were identified within the city by the Dar es Salaam Urban Health Project (DUHP) and the Adult Morbidity and Mortality Project (AMMP) in 1991 as a surveillance area, in order to set up and maintain collections of records of births, deaths, immigration and emigrations as well as of mortality and other health-related data. Details of the ongoing surveillance system have been described elsewhere ( Kitange et al. 1996 ; United Republic of Tanzania 1997). The three areas covered by the sentinel surveillance system are Ilala, Keko and Mtoni. Data on the cost of tuberculosis incurred to households were collected in these three surveillance sentinel areas of the city, covering approximately 70 000 people.
Two methods were used to identify tuberculosis cases in the study area: screening through the routine census in the three sentinel surveillance areas and active search for tuberculosis cases by field workers. During the yearly census in the sentinel surveillance area, undertaken between April and June 1996, all those who complained of coughing for more than 3 weeks were counselled to attend a nearby health service offering laboratory services including sputum examination (Temeke and Ilala district hospitals and Mnazi Mmoja health centre). Sputum was then checked three times for AFB. Those who were detected to be positive began treatment and were enrolled in the study. In addition to screening through the census, an active search for potential tuberculosis cases in the surveillance area took place between August and September 1996. This was done in two ways. Firstly, consultation books of health services in Dar es Salaam offering tuberculosis treatment were examined for patients living in the surveillance area. Secondly, eight field workers of both sexes conducted an exploration at household level with the assistance of community leaders to trace known patients in the study area. The identification card used by the national tuberculosis programme was used to check whether people were currently enrolled in tuberculosis treatment. If so, they were included in the study.
People identified as having tuberculosis were interviewed with a questionnaire. Data collection on economic costs at household level was undertaken from October to December 1996 and was carried out by the same eight field workers. They had a nonmedical background and were well known in the sentinel surveillance area through the regular census. They went from house to house, visiting each household of a given area until each of the previously identified persons could be directly interviewed. For persons under 15 years, an adult member of the household acted as proxy for that individual (in most cases the mother). The interviewers were supervised daily by a field research co-ordinator responsible for co-ordination of the work and control of the data collection. The completed questionnaires were verified by the field research co-ordinator and, if necessary, interviewers were asked to return to the household to collect any missing data. EpiInfo version 6.0 was used for data entry and analysis.
The costs which apply are those of 1996. The exchange rate applicable at that time was 665 Tanzanian Shillings to US $1. In order to estimate expenditures at household level for a whole tuberculosis treatment period, expenses had to be extrapolated. To do so, some assumptions were necessary ( Table 4). They are discussed in detail in the results and discussion sections in the part on the ‘extrapolation of household costs’.
Assumptions for the extrapolation of household costs
Results and discussion
In total, 191 tuberculosis cases were found in the sentinel surveillance area. According to the census information, 71 225 people are resident in this area, therefore the point prevalence was 268 per 100 000 people. Among the tuberculosis cases identified, 87 were female (46%) and 104 (54%) were male. The age distribution of cases was as follows: less than 15 years, 9%; 15–24 years, 22%; 25–34 years, 36%; 35–44 years, 20%; 45–54 years, 8%; 55–64 years, 6%.
Direct costs at household level
Tuberculosis patients were asked whether they had undergone a sputum smear analysis, X-rays and other examinations, and how much they had spent on these tests since the onset of the disease. One hundred and forty-one (74%) said that they had undergone a sputum smear analysis, 124 (65%) an X-ray examination and 30 (16%) other examinations such as biopsies or aspirations. Average expenditures of the whole sample since the onset of disease were US $0.6 for laboratory tests, US $1.4 for X-ray and US $0.1 for other examinations adding up to a total of US $2.1 ( Table 2). Although laboratory and X-ray services are theoretically provided free by the Tanzanian tuberculosis control programme, the patients reported that they had incurred costs. Different explanations are possible: it may take some time for tuberculosis to be diagnosed and during this time people may have to pay for examinations and laboratory tests. In order to have an X-ray examination, patients may have to buy films themselves. Patients often also have to make informal payments in order to receive services.
Direct costs to households related to tuberculosis
Health care delivery in Dar es Salaam is provided through several sectors and multiple consultation is frequent, especially among people suffering chronic illness ( Wyss et al. 1996 ). The cross-sectional household survey showed that tuberculosis patients had used different types of treatment up to the moment of diagnosis and after they had been enrolled in the national control programme. The most popular choices for health care since the onset of disease were stated to be governmental health services (93%), followed by private services (58%), self-treatment mainly by purchasing drugs at pharmacies (52%) and ‘traditional’ therapists (28%), including a wide range of possible treatments such as religious practitioners, herbal medicine and divining. The average total expenditure for consultations and drugs was stated to be US $16.6 since the onset of the illness. Tuberculosis patients and their families spent an average of US $8.6 for governmental health services and US $ 6.8 for private services ( Table 2). These out-of pocket payments included expenses from the onset of illness until diagnosis in addition to costs during treatment.
Tuberculosis patients were asked whether they had been admitted to hospital during the previous week and how much they had spent for transportation to the treatment place(s) ( Table 2). Of 191 respondents, five (3%) stated that they had been hospitalized. Average monthly costs were US $0.1. Average monthly expenditure for public buses, taxis and other means of transport were US $1.3 for the patient and US $0.3 for the people accompanying the patient.
Indirect costs at household level
During the interview patients were asked about their principal activity, how much time they usually spent on it (when healthy), and how much time they could allocate to it at the time. This made it possible to calculate the average decrease in the principal daily activity. Of the women, about half (41 out of 87) stated that they worked as housewives. Around a quarter (23 of 87) had (or had had before their illness) small businesses, 14 of 87 were unemployed and the remaining nine had other occupations. Among men, about 20% were civil servants (10 of 104) or employed in the private sector (11 of 104), about a quarter had small businesses (25 of 104) and more than half (58 of 104) had other or no occupations (e.g. students, military, unemployed).
The average decrease of the principal daily activity was calculated by dividing the declared usual allocation of time to work by the current one. On average, the loss was 74% of one person’s working capacity ( Table 3). In urban Zambia, care-givers incurred costs equivalent to about a third of the monthly income ( Needham et al. 1998 ). In addition to the patient’s own work capacity loss, members of the family may have to look after the patient. According to the responses of tuberculosis patients, for nearly two thirds (65%) no provision of assistance by household members that involved their abandoning of their daily activities or occupations was necessary. The remaining third of cases ‘benefitted’ to various degrees from the care of someone else. In the whole sample the average time lost by household members was two days per week. Projected to a week, this corresponds to a loss of 29% of the working time of one person per patient. Overall, the average productivity reductions per case were situated at a loss of 1.02 person work force.
Average time lost in principal daily activity
Extrapolation of household costs
The data set (191 cases) from the sentinel surveillance area was a mixture of smear-positive cases, extrapulmonary tuberculosis cases and relapses, all requiring different treatment and cure periods. For smear-positive cases the current standard in Dar es Salaam is ambulatory short-course chemotherapy for 8 months, whereas extrapulmonary tuberculosis requires longer treatment periods – in most cases at least one year of ambulatory care. Also, some patients need longer treatment periods than 8 months due to tubercle bacilli that are isozianid resistant and due to other (opportunistic) diseases. Other patients were not immediately enrolled in the services of the national control programme after onset of the illness. In Thailand and Tanzania, the average delay from the onset of illness until diagnosis was around 70 and 185 days, respectively ( Kamolrantanakul et al. 1999 ; Wandwalo & Morkove 2000). During this time patients pay for services. In order to include these variable time periods during which costs are incurred at household level, low, middle and high treatment times were used for the extrapolation of costs at household level.
The extrapolation of household costs relied on three scenarios: The low and best case estimation was based on the assumption that tuberculosis is detected shortly after the onset of the disease and requires a short-course, ambulatory treatment of 8 months ( Table 4). The high case estimation assumed an ambulatory treatment period of 12 months corresponding to standard Tanzanian smear-negative treatment. For examination and laboratory costs it was assumed for the best, middle and worst case that tuberculosis patients incur these costs just once, until the diagnosis of the disease is established. For consultation and drug costs it was assumed that tuberculosis cases mentioned expenses since the onset of disease, which covered a period of 8 months (best case), 6 months (middle case) or 4 months (worst case). For costs related to hospitalization it was assumed that the average monthly expenditure would be incurred for 2 months (best case), 2.5 months (middle) and 3 months (worst case). Weekly transportation costs as revealed by the questionnaire were extrapolated to an 8-month period (best case), 10 months (middle case) and 12 months (worst case), respectively. Costs for travelling and waiting times were assumed to be contained in costs related to productivity losses. For costs related to productivity losses, and after discussions with experts and the consultation of various documents such as World Bank, UNDP, and other statistics, the average foregone monthly earnings in 1996 were estimated to be US $38 (TSH 25 000), US $53 (middle; TSH 35 000) and US $113 (high; TSH 75 000). These income estimations do not reflect high, middle or low income groups, but reflect the lack of precise statistics on this issue and the subsequent uncertainty ruling over monthly average earnings in urban Tanzania. Furthermore, it was assumed that in the best case the ability to work is reduced over 4 months (low), 8 (middle) and 12 months (high). Intangible costs including pain, anxiety, and loss of well-being associated with tuberculosis were not included in the cost estimates. However, in a cost-effectiveness assessment of tuberculosis control, the reduction or increase in well-being has to be computed on the effectiveness side.
Using these assumptions, costs of tuberculosis at household level for a complete treatment period were estimated to be in the range of US $186 to US $1457 ( Table 5). Three categories of expenses were important: costs for consultation and drugs, transportation costs, and productivity losses.
Extrapolation of household costs for a complete treatment period
Cost estimates indicate that expenses for consultation and drugs incurred by tuberculosis patients and their families over the whole illness period can be as high as US $50. An important proportion of these expenditures is incurred before patients are actually diagnosed, between the onset of disease and actual diagnoses. Considering the time gap prior to enrolment into the national tuberculosis control programme, when patients should receive at least the drugs used for chemotherapy free, and that during the chemotherapy additional remedies may be necessary, this estimate seems to be reasonable. However, in some instances, as also confirmed by other studies ( Kamolrantanakul et al. 1999 ; McPake et al. 1999 ), the Tanzanian policy that tuberculosis services are free is probably not always respected by health services staff.
Transportation costs were estimated to be between US $13 in the best case and US $20 in the worst case. During the intensive 2-month phase of an ambulatory short-course therapy, patients have to visit services six times a week. Therefore the costs estimated by the household survey seem to be reasonable considering prices for the cheapest means of transport in Dar es Salaam, the so called Dala Dala (private buses). In 1996 the price for a single one-way trip was US $0.23. As people usually have to take two or more buses to make a one-way trip from home to a tuberculosis treatment service, they easily can spend US $1 for a single consultation at a health service.
The extrapolation of household costs revealed very high costs related to the loss of work ability, between US $154 and US $1384 in the worst case estimate. These numbers are based on the assumptions that per patient an average of 1.02 care giver is not able to follow his or her usual occupation and that the monthly foregone earnings are between US $38 and US $113. To attribute to every adult the same potential earning regardless of age, sex or occupation, can be justified on the basis of equity (giving all members of society the same value). Clearly, however, these calculations are based on crude assumptions. In future, it is important to consider how to assess exactly the amount of work productivity lost due to tuberculosis: how long patients are unable to work, to what degree they have to decrease their activities, how important home production (value of services provided by family members) is, etc. However, as tuberculosis patients are predominantly very ill adults who cannot work, high costs for the lost of work force have to be expected. Costs of productivity losses such as those calculated by this study may even be higher, as relapses are frequent and tuberculosis cure rates in eastern Africa often under 85% ( Glynn et al. 1998 ). In other words, some tuberculosis patients will endure longer treatment periods than 8 or 12 months, resulting in higher mean costs of productivity losses.
Provider costs versus household costs
In the framework of a cost-effectiveness exercise, provider costs for different tuberculosis control strategies were calculated for the national tuberculosis programmes in Malawi, Tanzania and Mozambique ( Murray et al. 1991 ; De Jonghe et al. 1994 ). Short-course ambulatory chemotherapy proved to be most cost-effective, and provider costs inflated to US $ (1996) rates were as follows ( Table 6): programme management costs US $31, laboratory costs US $9, drug costs US $40 and costs for ambulatory care US $10, adding up to a total cost of US $90 per treated case. This assessment was done from the viewpoint of the Ministry of Health. The implicit question for the purpose of this analysis was ‘how can the provider best use a limited budget to achieve the maximum health gain?’.
Estimation of average costs for short-course ambulatory chemotherapy in Dar es Salaam
The current standards for cost-effectiveness analysis recommend a societal perspective ( Gold et al. 1996 ). Thus, not only the provider costs of short-course ambulatory chemotherapy should be considered, but also costs incurred at household level, to give the total cost of treatment ( WHO, 1996; Sawert et al. 1997 ; Kamolrantanakul et al. 1999 ).
In the present study, in the best case extrapolation, household costs accounted for 68% of total costs and for 94% in the worst case extrapolation ( Table 6). In other words, costs on household level account for more than two thirds of total costs. Productivity losses constitute an important percentage of cost, and this is not apparent if only provider costs are measured.
The cost-effectiveness studies on different TB control strategies in East Africa compared ambulatory short-course chemotherapy, short-course chemotherapy with hospital admission and standard 12-month chemotherapy ( Murray et al. 1991 ; De Jonghe et al. 1994 ). The authors concluded that ambulatory short-course chemotherapy is preferable to standard 12-month chemotherapy. Differences in household costs between different strategies will not influence the effectiveness of an intervention and cure and death or transmission rates are not altered (the benefits of chemotherapy for smear-positive tuberculosis can be divided in direct benefits to the patient in terms of increase in well-being and reduction of morbidity and mortality, and indirect benefits to others through reduced transmission of tuberculosis). During a short-course therapy costs on household level are smaller and therefore costs higher during a 12-month chemotherapy, hence the principal conclusions of the cost-effectiveness studies are reinforced and expressed by a better cost-effectiveness ratio. In other words, due to high costs for transport, drugs and productivity losses the cost-effectiveness ratio of ambulatory short-course therapy will perform more favourably compared to the 12-month therapy.
Any new strategy under discussion for tuberculosis control, such as active case-finding, should be assessed in the light of total cost. Case-finding at home leading to prompt treatment could reduce cost to households. Considering high costs due to foregone workforce and that costs for providers of case-finding compared to ambulatory chemotherapy should only be higher for costs associated to field health workers, average unit, incremental and marginal cost of active case-finding might be, from a societal perspective, cheaper than the other strategies.
This research demonstrated that among 191 tuberculosis patients in Dar es Salaam, Tanzania the average costs of tuberculosis to patients and their families are considerable. Assuming treatment periods of 8 to 12 months and various costs for foregone productivity, extrapolated costs of a illness period to patients and their families were as follows: US $2.1 for examination and laboratory costs, between US $16.6 and US $49.9 for consultation and drugs, between US $0.3 and US $0.4 for hospitalization and between US $14.5 and US $21.8 for transport. The survey revealed that on average 1.02 persons in a patient’s household (patient and care givers) were not able to follow their usual occupations during the period of illness. The resulting loss of earnings were estimated to be between US $153.8 and US $1384.1. Costs of waiting and travelling time were included in this index. Expenses due to home modifications (buying of a new bed, etc.) were not included in the survey, because it was assumed that these costs were in general very low in the African urban context. Intangible costs (pain, anxiety, and loss of well-being) associated with tuberculosis are difficult to translate into dollars, and this was not attempted in this study. Even estimates of loss of productivity and income are inevitably approximate, especially for occupations such as housework where the product cannot easily be measured. It would be valuable to explore this question in more depth, for example, using time-and-motion studies. Furthermore, in a retrospective household survey, recall biases may influence the quality of information including cost estimates, for example for drugs and consultations, laboratory examinations and hospital expenses. Thus, future research on economic costs of tuberculosis at household level may be based on prospective studies following patients from diagnosis over an appropriate period in order to obtain solid estimates on these type of costs.
The three major types of costs identified, cost for drugs and consultation, transportation costs and in particular costs due to lost work force, put a high economic burden on patients and households. The sample of this research was a mixture of high-middle and low-income groups in Dar es Salaam. No differentiation was made between them. Thus, for poorer households with very limited financial resources, consequences of tuberculosis are particularly serious and potentially devastating.
Compared with provider costs as calculated on the basis of other studies ( Murray et al. 1991 ; De Jonghe et al. 1994 ), household costs make up between 68% and 94% of total costs. As the household costs are thus more than two-thirds of the total costs it is important when assessing tuberculosis control strategies that future cost-effectiveness analyses are undertaken not only from the viewpoint of Ministry of Health but also include a societal perspective. For example, patient-finding at community level, followed by prompt treatment, might reduce household costs and increase compliance without substantially increasing provider costs. When the societal perspective is considered, case finding could prove to have lower unit, incremental and marginal costs than other strategies, including ambulatory short-course chemotherapy.
We are very grateful to the people in the sentinel surveillance area of Dar es Salaam who kindly agreed to participate in the study. We would like to thank Dr Deo Mtasiwa and Mrs Jennifer Jenkins for advice and helpful comments during the writing of the document. The World Health Organization (WHO), the Swiss Agency for Development and Cooperation (SDC) and the Dar es Salaam Urban Health Project (DUHP) provided financial support for this study.