Understanding variations in catastrophic health expenditure, its underlying determinants and impoverishment in Sub-Saharan African countries: a scoping review

Background To assess the financial burden due to out of pocket (OOP) payments, two mutually exclusive approaches have been used: catastrophic health expenditure (CHE) and impoverishment. Sub-Saharan African (SSA) countries primarily rely on OOP and are thus challenged with providing financial protection to the populations. To understand the variations in CHE and impoverishment in SSA, and the underlying determinants of CHE, a scoping review of the existing evidence was conducted. Methods This review is guided by Arksey and O’Malley scoping review framework. A search was conducted in several databases including PubMed, EBSCO (EconLit, PsychoInfo, CINAHL), Web of Science, Jstor and virtual libraries of the World Health Organizations (WHO) and the World Bank. The primary outcome of interest was catastrophic health expenditure/impoverishment, while the secondary outcome was the associated risk factors. Results Thirty-four (34) studies that met the inclusion criteria were fully assessed. CHE was higher amongst West African countries and amongst patients receiving treatment for HIV/ART, TB, malaria and chronic illnesses. Risk factors associated with CHE included household economic status, type of health provider, socio-demographic characteristics of household members, type of illness, social insurance schemes, geographical location and household size/composition. The proportion of households that are impoverished has increased over time across countries and also within the countries. Conclusion This review demonstrated that CHE/impoverishment is pervasive in SSA, and the magnitude varies across and within countries and over time. Socio-economic factors are seen to drive CHE with the poor being the most affected, and they vary across countries. This calls for intensifying health policies and financing structures in SSA, to provide equitable access to all populations especially the most poor and vulnerable. There is a need to innovate and draw lessons from the ‘informal’ social networks/schemes as they are reported to be more effective in cushioning the financial burden. Electronic supplementary material The online version of this article (10.1186/s13643-018-0799-1) contains supplementary material, which is available to authorized users.


Background
Financial barriers are a key limitation to access health services in low-and middle-income countries (LMICs) [1,2]. Financial barriers are usually related to out of pocket patient payments and their impact on household budget [3]. Two main approaches are used to assess the financial barriers: catastrophic health expenditure that occurs when out of pocket (OOP) payment equals or exceeds a pre-specified threshold of household expenditure or capacity to pay [4,5], and impoverishment that occurs when the average household consumption after health care payment is below the pre-specified international or national poverty line [6].
The incidence of catastrophic payments is reported to be higher in low-income countries that rely on OOP, and lower in countries that have some prepayment mechanisms [5,7]. While impoverishment is usually reported in LMICs, catastrophic payments also exist in high-income countries, and are slightly concentrated amongst the less well-off [5,8]. This trend is also observed in African countries. Studies in Sub-Saharan African (SSA) countries have shown that inequities in access exist as a result of income differences and the level of OOP within the country. The proportion of households facing catastrophic health care payments has been shown to vary widely between countries [9][10][11].The World Health organization (WHO) argues that when people suffer financial hardship due to OOP, it is impossible to get closer to universal health coverage (UHC) due to the high risk of catastrophe and impoverishment [12]. Moreover, UHC aims to ensure that health care benefits are distributed on the basis of need for care and not on ability to pay [13]. The burden of OOP payments has encouraged SSA countries to use different financial arrangements to prevent catastrophic payments [14]. One of them is introduction of insurance systems with universal population coverage [15], and another is removal of user fees. There is also a trend by governments to move out of the OOP payments that are considered to impoverish those who are already poor [16]. Given the over reliance on OOP payments in most of SSA countries, and with most countries having inadequate social insurance schemes, there is a strong need to evaluate systematically the existing evidence on financial inequity in access to healthcare. Furthermore, the effectiveness of the health financing system is seen through protecting people against the risk of becoming poor, while enabling them to make use of services [17]. In addition, WHO underscores health care financing as one of the crucial components of the broader efforts to ensure social protections [12].
Systematic reviews synthesise evidence given their clearly formulated structure and the methodological rigour [18]. Several studies have reviewed the variations in health indicators in SSA; however, most of these have focused on health status, service coverage and utilisation indicators like mortality rates and incidence/prevalence of diseases [19,20]. There have been few reviews that focus on unequal access to health care due to financial barriers [21]. Lack of systematic reviews in this area is perceived by policy makers as a limitation in decision making and developing new strategies [22]. While there are several systematic reviews on catastrophic payments and impoverishment in LMICs [23,24], very few have incorporated literature from SSA, and those that have done so have included only one or two countries from the SSA region [5,25]. A few other reviews that have been conducted are disease specific [26][27][28] and do not review the CHE risk factors. To our knowledge, there is currently no systematic and/or scoping review that examines the scale and variations of CHE and impoverishment across SSA countries. To understand the scope and nature of the studies conducted in SSA, we apply a scoping review approach. Scoping reviews are considered appropriate in that they not only bring together the available evidence but also provide broader synthesis of the evidence [29]. The aim of this study is to provide an overview of the magnitude and distribution of catastrophic health expenditure and impoverishment due to OOP for healthcare across SSA countries. Furthermore, we also look into the determinants of CHE that have been identified across SSA countries. This will not only highlight the scale of the problem but also identify any gaps that could potentially strengthen future research in CHE/impoverishment in SSA countries. The findings will help in developing effective health policies [30], that are more targeted, prioritise the vulnerable populations and address key risk factors. In addition, the findings could help to inform strategic health financing priorities of SSA member states by development partners/regional blocs like the African union (AU), WHO and World bank amongst others that invest in health initiatives in the region.
This study therefore responds to the research question: what variations exist in the distribution of CHE and/or impoverishment and the associated risk factors across SSA countries? The paper continues with the methods section (searching strategy and study selection) followed by the results section and discussion with conclusions.

Methods
This scoping review is based on the framework proposed by Arksey and O'Malley [31] and incorporates recommendations proposed by Levac [32]. In addition, the Preferred Reporting Items for Systematic Reviews and Meta-Analyses [PRISMA] (See Additional file 1) that provides key items considered to be essential and minimum components of a systematic review or meta-analysis protocol [33], was applied to guide the screening and eligibility of the studies.

Search strategy and inclusion criteria Inclusion and exclusion criteria
This review included studies that focused on all population groups including vulnerable groups like people living with disability, the elderly or children in both rural and urban settings. Studies with the primary aim of assessing catastrophic health expenditure and household impoverishment due to out of pocket payments in health care were included. We particularly look at the incidence of CHE and impoverishment, defined as the proportion of households whose out of pocket spending on health care is catastrophic or drives them into poverty. The intensity of CHE or impoverishment defined as the extent to which the household expenditure exceeds the set threshold or poverty line was also included. In addition, we reviewed studies that assessed the risk factors associated with the observed levels of incidence in catastrophic health expenditure.
We considered studies that assessed CHE and/or impoverishment due to seeking any type of health care service including HIV/AIDS, TB, chronic illnesses, malaria and maternal health services. The review was restricted to studies undertaken in any of the 45 Sub-Saharan African countries with coverage of either part of the country, the entire country or multiple countries. Articles were considered for inclusion if they were observational studies including cross-sectional studies, case-control, comparative or longitudinal studies. We excluded articles that were discussion papers or general literature review on CHE or impoverishment, qualitative studies that discussed CHE and those that addressed methodological issues and global macro analysis on CHE. These articles do not provide outcome measures that are relevant for our study such as the incidence or intensity of CHE.

Search strategy
We commenced with a general search on Google Scholar, and then searched in several databases namely PubMed, EBSCO (EconLit, PsychoInfo, CINAHL), Web of Science and Jstor. We also searched through the grey literature of relevant organisations virtual libraries such as World Health Organization (WHO) and the World Bank. In addition, a forward search of authors mentioned in selected articles was also conducted. The search terms included 'Catastrophic' , 'Impoverishment' , 'Financial burden' , 'Economic burden' , and under PubMed search, we included the MESH terms for health expenditure, health care costs and Sub-Saharan African countries. These words were used for all the other database searches. The detailed search chain for PubMed is provided in Additional file 2. Only studies published in English language in the last 10 years (2006-May 2017) were included for review.

Data extraction and analysis
The main reviewer extracted and analysed data from all articles in consultation with the other authors. Information extracted from the publications included context of the study (country and year of publication), characteristics of the included population, methodology (design of the study, data source, sample size, type of analysis), primary (incidence and intensity of CHE/Impoverishment) and secondary outcomes (determinants of CHE). Studies were grouped by the outcome measures. As a primary outcome measure, we use the incidence and intensity of catastrophic expenditure and impoverishment. To measure the impact of OOP on household expenditure; varying thresholds were applied which varied from 5 to 40% as a ratio of household expenditure or non-food expenditure. Information on the determinants of CHE and impoverishment were reported as secondary outcomes. Articles were also classified according to four major SSA regions (West Africa, East Africa, South Africa and Central Africa).

Quality and risk of bias assessment
Although quality assessments are not a standard requirement in scoping reviews, it has been argued that the lack of it could minimise the rigour and challenge the interpretation of the findings [32]. In light of this, quality assessment of the studies was conducted by the main reviewer in consultation with the other authors. The Quality Assessment Tool for Observational Cohort and Cross-Sectional Studies was applied to evaluate the quality of the studies included. The tool is recommended by the NIH and has been used in several systematic reviews to assess internal validity [34] (See Additional file 3). For each question, studies are given scores on a Yes (1) or No (0), and others which include CD, cannot determine; NA, not applicable and NR, not reported.
All the studies included in this review were assessed for quality using the criteria that fits the respective studies, and for those studies which some elements of the criteria did not apply, these were marked as not applicable. The assessment of exposure measures was only done for those studies that focused on the risk factors associated with CHE. On average, all the studies met the quality criteria, apart from six studies [11,13,24,29,31,34] that assessed exposure factors that vary by levels that did not examine the different levels of exposure. In addition, one multi-country study [18] did not report on the sample size; thus, the three criteria related to the study sample could not be determined. Figure 1 shows the ratio of studies that met the respective criteria.

Study selection
The initial search identified a total of 512 articles from the main journals and another 33 articles from the additional databases of the WHO and the World Bank. Once duplicates were removed, a total of 501 articles remained. Using title and abstracts, one reviewer screened all the identified articles based on an agreed inclusion criteria with the other two authors. A total of 445 articles were excluded largely due to being non-SSA specific, or for having a general focus on national health expenditure instead of CHE and impoverishment. A total of 56 articles remained that were fully assessed for eligibility; a second reviewer went through these selected articles and provided recommendations. The three reviewers had concurrence to include 34 articles in the final review analysis. The main reason for dropping 22 studies included the fact that the outcome was level of OOP and not the proportion that was catastrophic. Also, these articles do not provide information that allow us to calculate the proportion of OOP that is/was catastrophic for households, a global analysis of studies that included one or two SSA countries, discussion papers or general literature review that provide a general understanding of CHE, qualitative studies that discussed CHE and methodological studies. Figure 2 represents the PRISMA flow chart for the studies selection process.

Characteristics of the included studies and quality of data
Of the 34 studies assessed, half were from the West African region (11 from Nigeria), eight from the East African region (4 from Kenya), seven from the South African region (2 from South Africa and 2 that were comparative of South Africa with Lesotho and Mozambique respectively), one from the Central African region and one covered three SSA regions (East, West and South Africa). One could argue that literature from the Central African region was missing because the region is largely francophone, while the review focused only on English studies. However, there were several studies included from other French speaking countries including DRC, Burkina Faso, Mali, Benin, Senegal and Côte d'Ivoire. All the studies identified were observational, of which 27 were crosssectional studies, 3 were cross-sectional comparative across countries in the regions, 2 were modelled longitudinal, 1 was a case control and another a prospective observational study.  Half of the studies covered a sub-national population within the respective country, while 14 studies (41%) had a national coverage, and 3 studies were multi-county in that they focused on more than one country. The national studies utilised data from various national household surveys including the National Living Standard, Social Economic Survey, Poverty Monitoring Survey, Health Expenditure and Utilization Survey, while the sub national studies sampled the respective regions or specific target population. A few other studies utilised hospital data to gather data on expenses paid for the various services provided; the limitation was the small sample sizes. See Appendix 1 for all studies included in the review by various characteristics.

Incidence and intensity of catastrophic health expenditure in SSA countries
The large majority of studies focused only on the incidence of CHE (n = 23), while some focused on both incidence and intensity (n = 11), and a set of others focused on the determinants of CHE (n = 18). Catastrophic health expenditure varied greatly between countries. However, cross-country comparisons are difficult because of the different thresholds, sample sizes and data sets used in the various studies. Given these variations, we shall discuss the magnitude and distribution of catastrophic payments based on the most commonly used thresholds; that is 10% of household income [35] and 40% of non-food expenditure [9]. Table 1 below summarizes the incidence and intensity of CHE as repoted in the articles reviewed.
The proportion of households facing catastrophic payments varied widely by the threshold applied. In most of the studies, the incidence of catastrophic expenditure was seen to be lower when higher thresholds were applied [36,37], at 10% the average incidence of CHE was 23% while at 40% the average was 17%. Generally, we noted that CHE was highest when a specific diagnostic service was assessed. Amongst the various diagnostics, HIV/ART and malaria had the highest incidence. Inpatient HIV patients in Nigeria had the highest incidence of CHE (100%) at 10% household expenditure, while at 40% non-food expenditure, the incidence was reduced to 94.3% [38]. There was a high incidence of CHE at 40% non-food expenditure in the Democratic republic of Congo amongst hospitalised children with severe malaria which was at 81.1% capacity to pay [39]. Both studies with a high incidence targeted specific groups of patients and thus were not national representative surveys. TB patients also incurred a high incidence of CHE in Benin at 71.8% at the threshold of 10% of household expenditure [40].
Variations are also observed within countries, for instance, two national studies in Nigeria that focused on CHE at 40% of non-food expenditure, one reported CHE of 1.7% [41], while the other [42] reported ten times more at 17.2%. However, in the same studies, at 10% of household expenditure, CHE was closer in range at 22.7% and 25.7% respectively. Figure 3 shows the variations in the level of incidence of CHE in various Sub-Saharan countries, with many countries still experiencing high CHE over time.
The intensity of catastrophic health expenditure, which is an indication of how much expenditure exceeds the thresholds ranged from 0.1 to 25% when the 40% proportion of non-food household expenditure threshold is applied, and 1-11% when the 10% proportion of household expenditure is applied. There was relatively little difference between the intensity reported for general health care and that reported for specific diagnostics like HIV/ART, TB, malaria and chronic illnesses. Most studies on diagnostic care did not report on intensity; thus, we have less evidence to discuss the severity of CHE due to the use of diagnostic services and treatment. The intensity of CHE was found to be lowest amongst South African countries, while in East Africa and West Africa, the intensity was within the same range.

Determinants of catastrophic health expenditure
Eighteen (18) studies assessed the determinants of CHE in the respective countries for both general health care and specific diagnoses. The articles assessed various determinants; thus, this review will discuss the overarching determinants reported in the majority of studies. These are summarised in Table 2. See Appendix 2 for determinants reported in each of the study.

Household economic/income status
Households' income level is the most consistent determinant of catastrophic health expenditure with higherincome groups being less likely to incur CHE relative to middle income-and lower-income groups [39,43,44]. This is also observed amongst HIV/ART related studies. Lower-income groups had a higher likelihood of incurring CHE on ART services, given that they are more likely to use their savings on food and other routine household expenditures [45]. Besides the type of disease, the power of association varied by country of study. For instance, amongst studies conducted on CHE due to tuberculosis (TB), it was found that the power of association (CHE-lower-income groups) was higher in Benin [40] than in Nigeria [46].

Type of health care provider
In case health services are provided by public hospitals, a higher CHE is observed. This is especially observed for inpatient services [47,48]. Within the public health care system, seeking services at the primary health care level like health centres and posts had a reducing effect on CHE [30,49]. In some countries, seeking care from a private health facility is associated with increased CHE relative to seeking care at public facility [50], while in others, seeking care from public and private is associated with higher CHE compared to confessional structures [39]. Seeking services from traditional healers due to cultural beliefs on various illnesses is associated with high CHE [51].

Type of illness
The presence of a household member with a chronic disease increases the likelihood of experiencing CHE [44,51,52]. While the number of illness episodes amongst adults significantly increases the odds of CHE; the average number of illness episodes amongst children in a household has no effect on CHE [44]. Simple illness like coughs did not increase the risk of CHE [53]. In cases of TB care, households with an HIV patient are more likely to incur CHE than those not affected by HIV [50]. Contrary to expectation, having a disability has no effect on CHE [44]. However, occurrence of adverse events such as accidents or injury increases the likelihood of CHE [30,53].

Characteristics of household members
Characteristics of the household head and members were mentioned in the majority of studies. However, different studies focused on different parameters including age of the household member, employment status, education level and female/male headed households. Households with older heads and older main income earners, lower education or with unemployed heads are more likely to incur CHE [52,53]. Full-time employment is protective against CHE, especially amongst couples where the women has a full-time job [45]. Also, employment status and occupation are associated with CHE, for instance, having a household head who is a manual labourer increases the likelihood of CHE [16,51,52].
There are studies with different results, for instance, a study in Zambia shows that the education and employment status of the household head is not significantly associated with the likelihood of incurring CHE [49]. Also, a study in Nigeria finds counter intuitive evidence that more educated households are more likely to incur CHE than the less educated household [43].
There are mixed results about the probability of incurring CHE and gender. Female-headed households have a higher probability of facing CHE [47]. On the contrary, a study in Botswana observed that female-headed households are less likely to incur CHE [16]. In another study in Nigeria, households with a male patient are more likely to experience CHE [50], whereas in another study in Côte d'Ivoire, households of HIV-infected women have a higher risk of incurring CHE [54].

Geographical location and distance to health facility
Location of residence is seen as an important predictor of CHE. However, this varies by the location of the study. In Kenya, for instance, households located in marginalised counties have higher odds of incurring CHE [52], while in Benin, a study on TB patients shows the odds of CHE are higher for patients residing in urban areas, but when confounded with education, the effects disappear [40]. A study in Nigeria conducted amongst patients at a rural hospital shows that urban residents incur higher rates of catastrophic payments; this is due to transportation costs to the rural hospital [50]. Generally, living in urban areas is protective of CHE [30, 48, 55,56]. However, it is found to be protective for non-poor, but not for the poor [47]. Living far from the nearest health care centre is associated with increased CHE [49,55].
Social insurance/welfare scheme Informal financing mechanisms through mutual organisations, informal groups and merry go rounds unlike formal health insurance is observed to reduce the risk of CHE [43]. Thus, patients with a poor social network are more likely to incur CHE [40].
Households that are enrolled in health insurance are engaged in mutual health organisations, or an informal social safety net (such as membership in a merry go round) have a reduced risk of catastrophic spending [30,53].
In certain cases, health insurance is not a significant determinant, for instance, in Kenya, because it only covers a small proportion of households and only inpatient services [48]. Health insurance is observed not to protect households from CHE due to HIV/ART services. A study in Côte d'Ivoire observed no association between CHE and households having health insurance. This is because households continue to cope with HIV-related costs over time, thus, the financial burden increases [54].

Household size and composition
Households size is associated with CHE, with larger households (of more than five members) having a higher risk of incurring CHE [16,51,52,54].
It is observed that having an elderly member (above 65 years) in the household imposes a higher risk of CHE for the household, meaning that elderly people are more vulnerable [39,47,52]. If the household has an elderly patient, (older than 40 years) CHE is likely to be high [40,50]. In Nigeria, there seems to be a positive but not significant elderly effect [43]. Despite children being vulnerable to diseases, a study in Kenya showed that having a member aged under five decreased the odds of CHE [39].

Household impoverishment in Sub Saharan-Africa countries
Household impoverishments due to catastrophic health expenditure are measured using different poverty lines in the different studies including the subsistence poverty line, the national poverty line (NPL) and the international poverty line (IPL). A study in Uganda that used both the national ($1.31) and the international ($1.25) poverty line [57] finds that the percentage poverty head count after health payment is higher when IPL is applied at 18.1%, while that of the NPL was 17.1%. Only one study in Malawi assesses impoverishment due to CHE for chronic illness [58]; all the other assessed impoverished due to CHE related to general health care.
The percentage of households that is impoverished ranges from 1.4 to 4.5% in the various countries. On average, 2% of households are impoverished due to health payments across all countries with Nigeria and Uganda having the largest proportion of household impoverished, 4.1% and 4.5% respectively. We note that the proportion of household impoverished in some instances increased and also decreased over time across countries and also within the countries. This is, for example, the case of Kenya in 2003, percentage of households impoverished was 1.5%, which increased in 2007 to 2.7% and decreased in 2013 to 1.6%. In Nigeria, a study using data collected in 1999 showed 2.5% households being impoverished and 4.1% in 2009, and in Malawi, households being impoverished were 0.9% in 2011 and 1.7% in 2012. Out of pocket payments induced a further 5.6% (ranging from 2-7%) of households on average into poverty with Uganda being the highest and an outlier at 18%. We observe no regional variations, but within regions there are variations, for instance, in West Africa, a study in 2009 [36] found that 4.1% of households in Nigeria were impoverished due to OOP relative to 1.4% of households observed in 2011 in Senegal [30]. In East Africa, similar variations were observed, a study in Uganda showed that 4.5% of households were impoverished [57] compared to neighbouring Kenya where 2.7% [59] and 1.6% [52] households were impoverished in 2007 and 2013 respectively Fig. 4. Table 3 summarises the pre-and post-poverty head count after health care payment and the associated poverty incidence.

Discussion
We observed some limitations that should be considered when interpreting the findings. First, the studies utilised different survey data including national household surveys, targeted population surveys and hospital data. Secondly, there were variations in the measurement of expenditure with some studies including only direct medical costs while others assessed both direct and indirect medical costs [52]. Also, there were variations of the thresholds applied across the different studies to measure catastrophic health expenditure, which makes it challenging to draw direct country comparisons. Furthermore, the proportion of households that experience CHE is dependent on the threshold used to define it [25].
In addition, impoverishment was measured using different poverty lines including subsistence, national and international poverty line. Given the main aim was to focus on CHE studies, the articles assessed on impoverishment were not exhaustive of the available literature on the same in SSA countries, but a representation of those that assessed both CHE and impoverishment.
We note that several studies rely on data collected several years (up to 10 years) back before the article was published, thus not providing a true reflection of the current context. Furthermore, use of alternative data means that the data was not solely collected for the purpose of this type of analysis, thus could bias the results.
The search and selection process was mainly conducted by the lead author. This could lead to a  limitation or bias in the information retrieved from the articles selected for final review. However, all authors were involved in deciding the key search words, and the search string was discussed and agreed upon by the three reviewers. At every stage of the selection process, the three authors held frequent discussions to analyse the output(s). In addition, the use of MESH terms for key journal searches like PubMed ensured that all possible words were included in the search. We note that this has no substantial impact on the findings given the final articles that were reviewed represented 17 Sub-Saharan African countries from across all the different regions and a range of health areas. West African countries incurred higher CHE relative to the other regions. This could be because most studies utilised convenient sampling of pre-selected vulnerable groups with small sample sizes rather than national representative household economic surveys, which were largely used in other regions. Furthermore, it has been argued that the use of convenience sampling is likely to bias results and conclusions; thus, interpretation should be done with caution [28]. For instance, of the 11 studies conducted in Nigeria, only two [41,42] utilised national representative household surveys, and we note that most studies reported different incidences of CHE.
Patients with HIV/ART, TB and malaria experience the highest incidence of catastrophic expenditure. This could largely be due to the fact that individuals with HIV continue to incur health expenses throughout the time of their illness, while those with TB are in continuous medication for about 6 months or more, and Malaria could have several repeat episodes within a family.
Studies have found that affordability of treatment in LMICs is low as large proportions of population are pushed into poverty due to medicine procurement, hence the need for subsidies [60]. However, this review revealed that non-medical related costs like transportation costs which are invariably greater for the poor living far from the health facilities, food related costs, non-routine tests and inadequate care (due to shortages of drugs and medical services) in public primary health care facilities largely influence CHE which is consistent with other studies [54,55,61,62]. This therefore means that on the contrary, subsidising the cost of drugs or removal of user fees alone may not necessarily protect households from CHE. It is revealing that non-medical expenditures are much higher than medical expenditures, with food and transport being the two most significant expenditure components [38]. We note that where user fees are abolished, CHE declines for the non-poor but surprisingly remains the same for the poor, thus not encouraging the poor to seek care [47].
All study findings are consistent that the poor have a higher incidence and are more likely to incur CHE than the well-off. Furthermore, studies show that the poor are more burdened with out of pocket payments and catastrophic expenditure [59,[63][64][65]. This is largely due to the fact that for households with a low income, even a small amount of health care costs can be catastrophic [49]. This is contrary to studies in low-and middle-income countries elsewhere like Asia, whereby the well-off are seen to have a higher incidence of CHE given their likelihood to spend more on health care unlike the poor [8,66,67]. This demonstrates that there is significantly less financial protection going to the poorest sections of the population in Sub-Saharan African countries.
Surprisingly, seeking services from the public sector increases the risk of CHE, despite no or modest charges for public sector [68]. This is possibly because most people who seek services from public service providers are from lower-income quintiles. In most SSA countries, people who seek care in the private sector are more likely to be well-off, hence have the capacity to pay. Nevertheless, this is not the case in many other countries that show CHE to be higher amongst people seeking care in private hospitals. This could be another factor that may explain the relatively high incidence of CHE even where user fees have been removed, given the inadequate quality of services in public facilities (due to shortages of drugs and medical services), individuals are compelled to seek better care elsewhere [49].
The review underscores the role of the type of illness in CHE. Consistent with other studies that have shown the impact of non-communicable chronic illnesses [25,69], this review also notes that chronic illnesses contribute to a high risk of CHE [70][71][72]. Putting into consideration that infectious disease like HIV, TB and malaria are highly prevalent in Africa and have the highest incidence of CHE. This potentially poses a double burden on the households that are affected by both, thus driving the incidence of CHE further up. We observe that the time on ART decreases the risk of CHE; meaning that, patients who can access continuous ART treatment can be more financially secure [45]. However, if the main income earner is the one affected, time on ART increases the risk of CHE [54]. Contrary to the notion of collaborative HIV/TB services, we note that in case of TB care, households with HIV patients are more likely to incur CHE than those not affected by HIV because of the double disease burden [50].
Unlike in developed countries where health insurance is protective of CHE, this review emphasises informal social networks and mutual organisation common in the African setup, which help households to cope with costs. However, the review is inconclusive about the effect of formal health insurance in reducing CHE in SSA. There are nuances on the size of the household as a predictor of CHE. Although a larger household size is associated with higher CHE, households with more working adults are less likely to incur CHE [53] perhaps supporting the economies of scale argument [43]. Furthermore, it is also noted that smaller households have an increased risk of CHE which reflects a smaller support network from which financial assistance can be sort [45]. Elderly members in the household are seen to increase the risk of CHE [30] unlike children less than 5 years despite both being vulnerable to illnesses. This could be due to the fact that the elderly tends to be also income earners, thus when ill, there is dual burden unlike children who are under care of an elder. This is consistent with findings in other studies [73].
Distance to the health facility is associated with an increased likelihood of CHE, highlighting the significance of distance in increasing cost of access to health care [49]. Households in rural areas are also seen to experience higher CHE relative to those in urban areas excluding slum dwellers. Similar findings are observed in studies in Vietnam, Thailand and Serbia [6,73,74].
We observe that women are more likely to incur CHE due to their low financial status [43,54]. In addition, we note that domestic violence against women increases the likelihood of experiencing CHE [51], given women's welfare is vital to the household and injustice against them affects their income contribution, health and well-being [51].
Contrary to the notion that health payments have a higher impact in countries where poverty is high [3], we observe variations in the level of impoverishments in relation to the poverty head count before health payments. For instance, in Uganda, the level of poverty before health care payment was low, but the proportion of households impoverished as a result of health payment was higher than in all other countries. The proportion of households impoverished was seen to increase over time across the various countries, with the rapidly increasing population in Africa where the majority live below the poverty line; more people could be pushed into poverty if the right financial protection measures are not put in place. It is inconclusive if impoverishment due to health care payments was permanent or transitory as no study in in this review provided for that. An answer to this could however be given using panel data which are only limited available at a national scale in SSA.

Conclusions
Overall, we observe that CHE and impoverishment are pervasive across all Sub-Saharan African countries, and the magnitude varies across and within countries and over time. The factors that keep CHE higher vary across the countries and are seen to cut across various socio-economic and demographic characteristics including economic status, type of health care provider, type of disease, household size, geographical location and social support schemes/network.

Implication for research
This review underscores the importance of studies that assess CHE in SSA, and we notice the increased interest in this area given the rise in number of studies over time. However, we observe that majority of the studies were cross sectional, thus not sufficient for overtime analysis. Further research in SSA would be more beneficial if panel data were utilised to facilitate continuous monitoring of trends and robust over-time analysis on CHE and impoverishment.

Implications for policy
Social protection interventions in Africa have primarily focused on the supply side through subsidising drugs, removal of user fees or provision of free health care, and most recently, expansion of social insurance schemes. However, this review has shown that most of these do not necessarily protect households from CHE due to other related non-medical costs like transport and food. The review emphasised on the role of informal social networks which are common in Africa like merry go round/mutual organisations and, hence, the need to explore policy innovations through these social networks, like insurance packages for informal/mutual groups. This review further highlights specific illnesses that drive CHE. In light of this, it is paramount for SSA countries to consider comprehensive and integrated health financing policies that cut across diseases, as this could help to draw synergies and efficiencies across disease areas and deal with possible dual disease burden. In addition, this review has paid specific attention to groups that are not financially autonomous. The fact that CHE was seen to be higher amongst the poor is an indication that the measures put in place have not been effective in protecting the poor. Given the context, there is a need to strengthen the social protection policies such that they are more holistic and effective in protecting the most vulnerable of population from catastrophic and impoverishing effects of health payments.