Tuberculosis and poverty: the contribution of patient costs in sub-Saharan Africa – a systematic review

Background Tuberculosis (TB) is known to disproportionately affect the most economically disadvantaged strata of society. Many studies have assessed the association between poverty and TB, but only a few have assessed the direct financial burden TB treatment and care can place on households. Patient costs can be particularly burdensome for TB-affected households in sub-Saharan Africa where poverty levels are high; these costs include the direct costs of medical and non-medical expenditures and the indirect costs of time utilizing healthcare or lost wages. In order to comprehensively assess the existing evidence on the costs that TB patients incur, we undertook a systematic review of the literature. Methods PubMed, EMBASE, Science Citation Index, Social Science Citation Index, EconLit, Dissertation Abstracts, CINAHL, and Sociological Abstracts databases were searched, and 5,114 articles were identified. Articles were included in the final review if they contained a quantitative measure of direct or indirect patient costs for treatment or care for pulmonary TB in sub-Saharan Africa and were published from January 1, 1994 to Dec 31, 2010. Cost data were extracted from each study and converted to 2010 international dollars (I$). Results Thirty articles met all of the inclusion criteria. Twenty-one studies reported both direct and indirect costs; eight studies reported only direct costs; and one study reported only indirect costs. Depending on type of costs, costs varied from less than I$1 to almost I$600 or from a small fraction of mean monthly income for average annual income earners to over 10 times average annual income for income earners in the income-poorest 20% of the population. Out of the eleven types of TB patient costs identified in this review, the costs for hospitalization, medication, transportation, and care in the private sector were largest. Conclusion TB patients and households in sub-Saharan Africa often incurred high costs when utilizing TB treatment and care, both within and outside of Directly Observed Therapy Short-course (DOTS) programs. For many households, TB treatment and care-related costs were considered to be catastrophic because the patient costs incurred commonly amounted to 10% or more of per capita incomes in the countries where the primary studies included in this review were conducted. Our results suggest that policies to decrease direct and indirect TB patient costs are urgently needed to prevent poverty due to TB treatment and care for those affected by the disease.


Background
In 2009, tuberculosis (TB) was the world's 7th leading cause of death, resulting in 1.7 million deaths worldwide, more than 9.4 million new infections and 14 million prevalent cases [1]. TB is often known as "a disease of the poor" because the burden of TB follows a strong socioeconomic gradient both between and within countries, and also within the poorest communities of countries with high TB incidence [2]. Some studies have shown a strong association between poverty and TB and have demonstrated that poor and vulnerable groups are at an increased risk of TB infection, have a higher prevalence of disease, have worse outcomes (including mortality), and display worse TB care-seeking behaviors [3][4][5][6][7][8]. Risk factors for these TB-related outcomes include structures, behaviors and other diseases commonly associated with poverty -overcrowded living or working conditions, poor nutrition, smoking, alcoholism, diabetes, exposure to indoor air pollution and HIV [2,[7][8][9][10].
It is also well-known that TB can contribute to poverty by reducing patients' physical strength and ability to work [8,[11][12][13]. However, another pathway through which TB can affect households' economic situation, the costs patients incur when utilizing TB care, has been less studied. These costs include both direct out-of-pocket costs incurred when seeking treatment and care and the indirect, or time costs, associated with utilizing healthcare. While most countries with high TB burden provide free sputum smear microscopy for patients with suspected pulmonary TB, more than half of these 22 countries charge for other TB-related diagnostic tests such as radiography, sputum culture, and drugsusceptibility testing [14]. Under Directly Observed Therapy Short-course (DOTS) programs, all high burden TB countries provide free first line anti-TB medication, but many patients purchase anti-TB drugs in private pharmacies (some without prescriptions), which can be costly [14,15]. In high TB burden countries, 60% of overall health expenditure is in the private sector, and a large proportion of these expenditures are paid out-of-pocket by patients [14].
A number of previous studies have documented the downstream consequences of the direct and indirect costs that TB patients incur. More than 50% of TB patients have been reported to experience financial difficulties due to TB [16], and these costs can be "catastrophic" in that they amount to more than 10% of patients' or households' annual income [17][18][19]. TB patient costs have been shown to lead to reduced food consumption, diversion of resources from other types of healthcare, taking children out of school, and borrowing or selling assets [17,[19][20][21]. Furthermore, financial constraints have been shown to predict non-adherence to TB medication [16]. In general, the World Health Organization (WHO) estimates that 100 million people every year fall into poverty from paying for health services [22].
One earlier review reports on the overall costs TBpatients in Africa face during the pre-and postdiagnosis phases of TB treatment and care as well as coping mechanisms for catastrophic costs [23]. In this study, we expand on this previous assessment by broadening the evidence base on TB patient costs in sub-Saharan Africa through screening of additional databases and broadening the study design inclusion criteria, systematically identifying the particular types of TB patient costs (both direct and indirect), systematically reviewing the evidence on the cost quantities for each cost type, and providing benchmarks for the magnitude of cost burdens on TB patients and households.

Data sources and search strategies
We used eight electronic databases to identify papers reporting on patient costs for TB care in sub-Saharan Africa available by January 3-4, 2011: PubMed, Embase, Science Citation Index, Social Science Citation Index, EconLit, Dissertation Abstracts, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Sociological Abstracts. Each search strategy comprised a Boolean operator of "and" with two elements: tuberculosis and cost/economic aspects. For the PubMed search, MeSH and "all fields" terms comprising tuberculosis and OR fields for cost estimates such as "employment," "out of pocket," "patient costs" and MeSH terms for "costs and cost analysis" were used. Similar search strategies were employed for the other 7 databases (see Additional file 1 for the precise search algorithms for each database). Each database was searched from the earliest referenced publication date through January 1, 2011. Studies were included regardless of language. Two reviewers independently screened articles identified from the initial search of the databases by title and/or abstract.
To identify additional articles, conference abstracts written in English from 1994-2010 from the International Union Against Tuberculosis and Lung Disease (IUATLD) annual conference were searched. Furthermore, we performed a secondary search of reference lists of articles identified through the database search, including both the primary studies included in our synthesis and review studies.
Articles were considered for inclusion if they contained a quantitative measure of a direct or indirect patient-incurred cost (including time costs) relating to TB treatment or care for adult pulmonary tuberculosis. Following Rajeswari et al. [24] and Jackson et al. [25] we defined costs as follows: Direct costs included both medical expenditures (such as consultation fees or costs of medication or diagnostic tests) and non-medical expenditures (such as money spent on travel, lodging, and food for both patients and caregivers). Indirect costs were defined as time costs associated with utilizing healthcare, or time costs converted into monetary units based on loss of wages for both patients and caregivers or decreased earning ability [26].
We excluded the following articles: i) published before 1994 when DOTS was officially launched as a framework for a TB control strategy recommended by the WHO [27]; ii) not taking place in Sub-Saharan Africa; iii) not pertaining to TB; iv) not involving human subjects; v) on MDR-TB, HIV/TB co-infection, latent or pediatric TB; vi) focusing solely on diagnostic tests, screening tests, or vaccinations; vii ) not containing any primary data on cost estimates or economic analysis; or viii) not relating to individual patient costs.
Articles were assessed for study quality. In particular, we examined the studies to ensure that we included only those in our final review that had clearly defined objectives, clearly defined study populations, and a quantitative measure of patient-costs. Additionally, articles were categorized by study type and whether costs were incurred pre-or post-diagnosis. We adhered to the PRISMA guidelines [28].

Data extraction and analysis
In addition to study-specific variables (authors, study type, year, location, setting, period of observation, population under study, and study objectives), the two reviewers extracted patient-borne quantitative direct and indirect cost measurements. We extracted only costs measured empirically in the reviewed studies. Costs that were extrapolated or projected in mathematical models were not included in the analysis. Data were categorized into costs related to health insurance, prepayment, consultation or provider fees, hospitalization, medication, and diagnostic test costs, traditional healer and food costs, travel costs, time costs, reported impact on income, reported direct, indirect, and total costs, caregiver costs and catastrophic costs.
To compare costs expressed in different currencies and measured in different years, we converted all cost 180 estimates into 2010 international dollars (I$). We rounded all cost estimates to the nearest integer except for costs less than 1I$, which we rounded to the second decimal place. For costs that were presented in US$, costs were first converted to respective local currency units using OANDA currency conversions [29] based on exchange rates at the commencement of the study period. For studies that did not specify the year or period of currency estimates, January 1 st of the beginning of the study year was used as a standard conversion date except for one study in Botswana [30], for which conversion rates were only available beginning November 1, 1993 instead of January 1, 1993, and for one study in Uganda [31], for which conversion rates were available beginning January 1, 1996 instead of January 1, 1992. Next, local currency units were adjusted to 2010 rates using the International Monetary Fund's database on average consumer price inflation over time [32]. Finally, costs were adjusted to 2010 international dollars based on the World Bank's purchasing power parity (PPP) conversion factors (in local currency units per international dollar) [33].
To compare costs across studies, we report travel costs as single visit costs and hospitalization costs over the entire treatment period. To provide a benchmark for the magnitude of cost burdens of TB care on patients, we expressed the expenditures as percentage of per-capita annual (or monthly) GDP (in I$) of the country and in the year when the study, which generated the cost estimates, was conducted. The per-capita GDP figures are taken from the World Development Indicators published by the World Bank [33]. Per-capita GDP, i.e., the average income, is one benchmark that is meaningful to understand and commonly used for such purposes, and we thus use it here. However, since TB is a disease that predominantly affects poorer populations, we also express the cost estimates as a percentage of an alternative income benchmark-the percapita income of the income-poorest 20% of the population, calculated according to the following equation: We chose these two income benchmarks, rather than study population-specific incomes, because very few of the studies included in our review reported the study population income. We also report whether the patient costs of TB treatment are "catastrophic" for the person of average income or the person of average income amongst the income-poorest 20% of the population, classifying costs as "catastrophic" when they were at least 10% of average annual income for the respective population. While definitions of catastrophic expenditures commonly relate to household income, [18,[34][35][36][37] we chose to use 10% of annual individual income as a benchmark for catastrophic costs because for most studies we lack householdlevel income data as well as the household-level TB data that would be necessary to judge whether the financial burdens of TB care and treatment is "catastrophic" or not. Without the latter data, householdlevel income data is not an appropriate indicator since TB tends to cluster in households [38][39][40].
Total GDP Ã Income share of the incomeÀpoorest 20% of the population Total population size Ã 0:2 Results 5,114 articles were identified from the initial search of the eight databases. After excluding 1,112 duplicate articles, 1,510 were excluded because they did not include TB as a major subject heading, 777 did not include cost estimates, 632 were published before 1994, 427 focused solely on diagnostic tests, screening tests or vaccinations, 331 did not include patient cost estimates, and 95 did not involve human subjects. Reviewing the full text of the remaining 230 articles, we found that 55 did not include a quantitative cost measurement; 14 were reviews, commentaries, letters or editorials that did not include primary data; 7 were on pediatric TB, 13 on MDR-TB, 6 on latent TB, 4 on HIV/TB, 105 studies did not take place in sub-Saharan Africa; and 2 studies were by the same authors and included the same data (leading to the exclusion of 1 of the 2 studies). This selection process resulted in 25 relevant studies; 5 additional relevant publications were identified through the search of the conference database and reference lists, so that a total of 30 articles were included in the final synthesis of our review (see Figure 1). Table 1 describes each of the 30 studies in terms of their study populations, main objectives, types of cost estimates, and time period in which costs were incurred (pre-vs. post-diagnosis). Eight studies reported direct costs, one study reported indirect costs, and twenty-one studies reported both indirect and direct costs. Table 2 describes the cost categories, including definitions for each cost type, whether the costs are considered direct or indirect, whether the cost are incurred pre-or postdiagnosis, the number of studies reporting a particular cost type, and the cost range and median among all studies reporting the cost type.
Health insurance, prepayment fees, consultation and private provider fees Two studies reported health insurance fees that ranged from I$2 to I$3 in Zambia [41,59], and four studies reported consultation or prepayment fees that ranged from I$2 in Ethiopia [55] and Botswana [30] to I$7 in Zambia [58]. Patients who did not seek care from the public sector paid fees for care in the private sector. Four studies noted such fees for private services, which ranged from I$24 (median I$10) in Zambia [58] to I$141 in Uganda [47] (see Table 3). Additionally, one study from Uganda reported the practice of "tipping" healthcare providers in the range of I$5-I$40 [46].

Hospitalization, medication, and diagnostic tests costs
Ten studies reported hospitalization expenses. Costs ranged from I$4 in Uganda [61] to over I$530 in Kenya [60]. Some patients were required to pay hospital admission fees. Patients in Freetown, Sierra Leone paid an average of I$1 at a missionary hospital and I$47 at a government hospital, which included the cost of food [52]. Five studies reported medication costs that ranged from I$20 in Uganda [46] to I$548 in Nigeria [42] (see Table 4). Moreover, one study from Kenya reported that patients paid I$46 monthly for syringes and needles for streptomycin treatment (not including streptomycin itself ) [43]. Three studies reported diagnostic test costs other than sputum smears which ranged from I$7 for chest radiographs [41] to I$10 for examination, laboratory, and X-ray fees in Tanzania [68].

Traditional healer and food costs
Five studies reported that patients paid between I$3 in Malawi [44] to I$563 in Uganda [47] to see traditional healers, and four studies reported the cost of food, which ranged from I$4 (interquartile range (IQR) I$1-I$7) in Zambia to I$36 in Ethiopia and Zambia (median I$19) for "special food" [41,48,58] (see Table 5).

Travel costs
Eighteen studies reported travel costs for patients, families, or guardians for single visits or for multiple visits during treatment. Costs ranged from less than I$1 in South Africa for a single health clinic visit [51,63] to I$70 in Ethiopia for pretreatment transportation costs [65] (see Table 6). Travel time also varied from 48 minutes in Cape Town, South Africa [63] and Kampala, Uganda [46] to almost 70 hours in Ethiopia under a health-facility based DOTS system for a single visit [48]. One study from Zambia distinguished between pre and post diagnosis travel costs: pre-diagnosis travel costs were I$3 (IQR I$1-I$7) while post-diagnosis costs were I$11 (IQR I$4-I$29) [41]. In addition to travel costs, one study reported accommodation costs (in Ethiopia) [55].

Time costs
Twenty-one studies reported time costs. Clinic visit wait time varied from 30 minutes in Limpopo Province, South Africa [49] to 111 minutes in Kampala, Uganda [46]. One study from Uganda reported that patients spent on average 22 minutes for a volunteer-supervised outpatient DOTS visit or I$0.23 (95% CI I$0.00-I$0.42) in lost income, and an average of 110 minutes for a health-facility visit, or I$1 (95% CI I$1-I$2) in lost income [61].
Lost work time varied by treatment system. For hospitalized patients in South Africa, each hospital day led to an average of 402 minutes of lost work time (I$7 in lost income) compared to 128 minutes (I$2) for a health clinic visit, 50 minutes (I$0.85) for a DOTS visit with a community health worker chosen as a treatment supervisor, and 4 minutes (I$0.51) with another type of health worker chosen as a supervisor [51]. In Malawi patients lost an average of 22 workdays resulting in an average income loss of I$68 [54]. Two studies in Zambia found that patients missed an average of 18 workdays before being diagnosed with TB [58] and 48 days of missed work in total [59]. Foregone earnings reported for any type of care-seeking activity ranged from I$3 in South Africa [63] to I$169 in Tanzania [68].

Reported impact on income
Five studies surveyed patients on their salaries and reported the impact of TB patient costs on household incomes [41,55,58,59,65]; one study used average household income estimates from an external source to calculate the impact of TB patient costs on household incomes [54]. In Malawi, patients spent between 129% and 244% of their mean monthly income (MMI) on TB diagnosis [54].
In Zambia, patients spent 16% of their MMI on transportation costs and 66% of their MMI on food [59]. Direct medical expenditures ranged from between 10% of MMI for men and 132% of MMI for women in Zambia [41] to 31% for all patients in Ethiopia [55], while non-medical expenditures ranged from 42% in Ethiopia [55] to 55% of MMI in Zambia [58]. In Ethiopia, 48% and 35% of annual household income was lost due to TB treatment and pretreatment costs, respectively [65].

Reported direct, indirect and total costs
Eight studies reported aggregated overall direct costs, and six studies reported aggregated overall indirect costs incurred by patients (although authors defined direct and indirect costs differently). Reported direct costs    [41] to I$145 in Ethiopia [55]. Thirteen studies reported overall total costs (direct and indirect), which ranged from I$2 in South Africa [63] to I$584 in Uganda [31] (see Table 7). Five studies reported the percentage of all costs that patients paid out-of-pocket. In South Africa, out-of-pocket expenses varied by district in which patients were responsible for paying between 13% and 34% of all costs [51]. Similarly, in Tanzania patients paid between 13% and 30% of total costs in community-based DOTS and health facility-based DOTS programs, respectively [66]. In Malawi and Ethiopia, patients paid close to 50% of total costs of their care.

Caregiver and guardian costs
Eight studies reported both direct and indirect costs incurred by TB patients' guardians or caregivers. Direct costs included transportation costs that ranged from less than I$1 (standard deviation [SD] I$4) under a community-based DOTS program in Ethiopia [48] to I $27 in Botswana [56]. The amount of time spent traveling for one care-related visit ranged from 20 minutes in Botswana [56] to 17 hours (median 0) in Ethiopia [55]. Other direct costs for caregivers included food costs that ranged from I$3 (standard deviation [SD] I$5) in Ethiopia [48] to I$1,209 in Botswana [56] and time spent providing care-giving activities, which ranged from 1 hour each day in Botswana [59] to 6 days (median 1) for hospitalized patients in Ethiopia [55]. Indirect costs included foregone earnings for caregivers which ranged from I$19 (median $10) in Zambia [58] to I$89 in Ethiopia [55]. Total reported caregiver cost ranged from I$24 (median I$12) in Zambia [58] to I $1510 under a home-based care strategy in Botswana [56].

Catastrophic costs
Twenty studies reported costs that were found to be catastrophic for those with average income, and twentyfive studies had costs that were catastrophic for the lowest income earners (see Table 8). Catastrophic costs constituted between 11% of average annual income in The Gambia for private providers [53] and almost three times average annual income in Ethiopia for total pretreatment costs [65]. For those in the income-poorest 20% of a country's population, catastrophic costs constituted between 10% of annual income for traditional healers in Tanzania [68] to roughly ten times annual income for hospitalization costs in Malawi [50]. Two aspects of our extracted data are important to note in this context: First, we extracted data for a range of different cost categories (Tables 1, 2, 3, 4, 5). For all cost categories, there are at least a few studies reporting catastrophic costs according to our definition both for those with average income and those with the average income among the income-poorest 20% of the population: traditional healer, food, travel, private provider, medication, "tipping" providers, hospitalization, caregiver, and overall direct, indirect and total costs. Second, the extracted costs are presented in the tables in the units they were reported in the original papers because we did not have sufficient information, either from the papers or external sources, to allow translation into a common unit. However, catastrophic costs were found in all units of reported costs, including per-visit, per time period, and per treatment course and in both the pre-and post-diagnosis periods.

Discussion
In expanding on a previous review of TB patient costs in sub-Saharan Africa [23], we extended the evidence base on TB patient costs in sub-Saharan Africa through screening of additional databases and broadening the study design inclusion criteria. We have further added to the literature by systematically identifying the particular types of costs TB patient incur and by systematically reviewing the evidence on the cost quantities for each cost type. Our review furthermore provides benchmarks for the magnitude of these cost burdens by comparing them to average income earners and the average income of the income-poorest 20% of the population.  The data reviewed here demonstrate that direct and indirect patient costs for TB patients and their households can be substantial and often "catastrophic" for average income earners and, in particular, for those in the income-poorest 20% of the population-the proportion of the population most at-risk of acquiring TB. The data we extracted from the literature thus suggest that expenditures for TB treatment and care can cause or exacerbate poverty. TB patients in sub-Saharan Africa incur both substantial direct and indirect costs before, during, and after a TB diagnosis. The largest costs these patients incur are for hospitalization, medication, transportation, and treatment or care in the private sector. In addition, caregivers incur substantial indirect, or time costs, of providing care or support for TB patients.
Results also show that total TB treatment and care costs vary greatly between studies: from only I$2 in   South Africa [63] to almost $600 in Uganda [31] in total estimated costs. Additionally, the types of costs that patients often pay are numerous: eleven main categories of costs were captured in this review, and nine of the thirty studies reported costs in at least five of these cost categories. Though patients in sub-Saharan Africa often do not incur all of these fees, merely paying for one or some of them can have substantial impacts on their economic circumstances. Patient costs may negatively impact health-seeking behaviors, leading to delays in hospital presentation, further worsening of disease and increasing risk of disease spread [16]. Patients adopt several mechanisms to cope with these costs ranging from asset selling, borrowing, and diversifying income-generating activities [19,20]. In order to alleviate the financial burden borne by TB patients, policy makers should consider incorporating policies to support patients receiving TB treatment into general financing and risk-pooling strategies, such as tax-based or social insurance systems as used by many developed and, increasingly, developing economies. While in some settings strategies aimed at reducing patient costs incurred when utilizing healthcare may be feared to lead to increases in demand for healthcare exceeding the underlying need, in many developing countries like those in sub-Saharan Africa, healthcare demand is currently far below need, including for the priority diseases TB and HIV, so that financial and non-financial support for healthcare seeking is likely to contribute substantially to improving population health [70,71]. In addition to the direct benefits to the treated patient, TB treatment also reduces onward transmission of the disease in the community. This positive externality needs to be taken into account when considering public investments to decrease the costs patients bare when utilizing TB care.
Comparable types and extents of costs incurred by patients in sub-Saharan have been found in studies assessing the household financial burden of other infectious diseases including malaria and HIV/AIDS. Although some health districts in sub-Saharan Africa provide free care for HIV/AIDS and malaria patients, studies assessing the burden of expenditures for malaria and HIV/AIDS-affected households have found that costs are commonly "catastrophic," a general finding that conforms with our results for the case of TB. Moreover, the main types of costs that patients and households incurred based on these studies were similar to those found in our review: direct costs of transportation for patients and caregivers, medications, diagnostic tests, hospitalization, food, medical consultations, and traditional healers, as well as the indirect costs such as loss of time at work or school [34,[72][73][74][75][76].
More research needs to be done in order to assess the burden of costs to TB patients in other geographical areas, including in the 36 other countries in sub-Saharan Africa for which our systematic review could not identify any empirical estimates of TB treatment and care costs. While our results broadly support calls for health policy changes to alleviate patients' financial burdens of TB treatment, it is unclear which particular interventions will be most effective and cost-effective. Examples of interventions that can reduce the direct patient expenditures for TB treatment include free transport between communities and TB treatment facilities, food vouchers given to TB patients when visiting a facility, and systematic elimination of user fees. Interventions that can reduce indirect patient expenditures for TB treatment include improved health infrastructure, such as, additional TB treatment facilities, better transport infrastructure (to decrease the time spent travelling from a household to a healthcare facility), as well as improved patient scheduling systems (to decrease patient wait times at healthcare facilities).     Finally, since we find that patients in some countries spend substantial amounts of money for TB treatment and care in the private sector despite public-sector TB programs that offer some services free of charge, interventions to decrease the demand for private-sector TB treatment and care could reduce the financial burden to TB patients. Such interventions could improve media campaigns to ensure that patients are aware of where and when they can access public-sector TB treatment and health sector reforms leading to increased quality of care in the public sector. Rigorous evaluation studies, accompanying the implementation of such interventions, will contribute to our understanding of their performance and impact on patient healthcare expenditures and population health outcomes.

Limitations
There are several limitations of this review: First, we included studies that were conducted from 1994 to present. Earlier studies might not be generalizable to current strategies because TB control program policies may have changed since their publication. Second, although no official criteria exist for assessing studies on patient costs, other systematic reviews of patient costs have used various quality assessment criteria for evaluating study quality [23,[77][78][79]. Due to our broader study inclusion criteria, we were unable to adopt similar formal criteria for evaluating study quality beyond ensuring that studies had clearly defined objectives, clearly defined study populations, and a quantitative measure of patient-costs. This limitation may have led to a relatively higher weight placed on results from "low quality" studies in the interpretation and discussion of the summary findings from our systematic review than would have been the case if we could have more clearly distinguished between "low" and "high quality" studies.
Third, many studies assessed only a few specific types of costs, so comparing total cost burden to patients across studies is difficult and may be underestimated. Similarly, the studies that reported aggregate direct and indirect costs might have defined direct and indirect costs differently, so these reported measures might not be comparable. Additionally, since costs were measured differently and may reflect different time periods, (for instance some travel costs were measured as one-time costs, and others were measured over the entire course of treatment), comparing these cost burdens on patients' average annual or monthly income are difficult and must be interpreted with caution.
Moreover, only five studies directly surveyed patients on their incomes. In our estimation of the impact of TB patient costs on patients' economic circumstances, we thus had to rely on national income estimates. However, it is very likely that TB patients' average income deviates substantially from the national average, so that our relative measures of the out-of-pocket cost burden to patients and households may be biased.
Furthermore, since data on income shares of the income-poorest 20% of the population were missing for several country-years, we used estimates from the closest years for which data were available for a country, but this could potentially yield inaccurate estimates of per capita GDP for this portion of the population. Finally, evidence on TB patient costs was only available for 11 out of 47 countries in sub-Saharan Africa. Results from these studies cannot be extrapolated to the rest of the continent, and it is possible that patients in other parts of Africa might incur lower TB patient costs.

Conclusion
Tuberculosis can place considerable financial and economic burden on patients and households in sub-Saharan Africa. Here, we identified 30 relevant studies investigating TB patient costs in sub-Saharan Africa before, during and after diagnosis and treatment for TB. We found that patient costs vary considerably both by total amounts and according to the different types of costs incurred by patients in DOTS and non-DOTS settings. In many settings, patient costs were found to be "catastrophic" in that they amounted to 10% or more of average annual incomes. These costs have the potential to financially strain patients and their households, leading to detrimental effects such as delayed care seeking and increased default rates of TB patients, potentially fueling the spread of TB or increasing multi-drug resistant TB in sub-Saharan Africa. Future research needs to assess which interventions and health systems reforms are most effective and cost-effective in reducing the financial burdens that patients incur when seeking TB treatment and care.

Additional file
Additional file 1: Systematic review search algorithm.