Introduction

Improvements in healthcare, hygiene and sanitation have increased the possibility to live until older age. Together with a growing global population, this has meant that non-communicable diseases (NCDs), including coronary heart disease (CHD), stroke, chronic obstructive pulmonary disorder (COPD), cancer, type 2 diabetes mellitus (DM) and chronic kidney disease (CKD), are now the leading causes of morbidity and mortality worldwide. The burden exerted by NCDs extends beyond morbidity and mortality and generates an enormous societal impact, including on households and impoverishment [1–5].

Limited insurance coverage and lack of social security nets can force households of NCD patients to spend large amounts of money out-of-pocket (OOP). NCDs reduce family income, savings and consumption of non-health items, and prompt early retirement [6, 7]. The impact of NCDs on households is likely to be especially severe in low- and middle-income countries (LMIC) where low-income populations, many of whom already experience extreme absolute poverty and precarious living conditions, are especially vulnerable to impoverishment due to any degree of healthcare spending [1, 8–10]. With some exceptions, such vulnerable groups suffer a double burden of chronic and infectious diseases [2, 10–13]. The interplay between exposure to disease and financial vulnerability among low-income households can drive families and societies into deeper poverty.

Despite greater appreciation on the likely deleterious role of NCDs on households and impoverishment, the extent of this impact in various geographical regions, is unclear. While several studies have addressed the issue, they have not been systematically evaluated in a single comprehensive investigation. Therefore, we report a systematic review to investigate the economic consequences of the major NCDs on the micro-economic indicators (1) at the level of households (such as consumption choices, coping strategies, OOP, direct and indirect costs) and (2) of poverty (such as financial burden, catastrophic spending, impoverishment, poverty line and financial vulnerability), across various global regions.

Methods

Conceptual framework

To guide the systematic review of the literature regarding the household impact of NCDs, a conceptual framework was adopted. This theory, previously described by McIntyre and colleagues, focuses on the economic consequences of illness and paying for health care [14]. The economic consequences that NCDs incur on the household level are preceded by levels of perceived illness and the resulting treatment seeking behaviour. Seeking care can lead to economic consequences in the form of direct (e.g. costs for hospitalization, medicines, transportation) and indirect costs (e.g. time costs of informal caregivers, time costs of the ill). The indirect costs associated with not seeking care can exert a similar burden on the microeconomic level. Economic consequences in combination with divergent coping strategies (e.g. household labour substitution, use of savings, changing consumption choices) can result in poverty.

Although the importance of the first two steps (perceived illness and treatment seeking behaviour) is conclusive, the focus of this review was on economic consequences, coping strategies and poverty.

Search strategy and inclusion criteria

We conducted a systematic search of electronic medical databases (Medline, Embase and Google Scholar) from inception to November 6th 2014 to identify scientific articles assessing the economic consequences of NCDs on households and on impoverishment. Given their large burden in populations worldwide, the following NCDs were selected: CHD, stroke, COPD, DM, cancers (lung, colon, breast, and cervical) and CKD [1]. The step-wise inclusion and exclusion procedure outlined in Fig. 1 was followed. Eligible study designs included randomized controlled trials (RCTs), systematic reviews (used to identify further references), cohort, case–control, cross-sectional, ecological studies and modeling studies. Studies were included evaluating the impact of at least one of the selected NCDs and on at least one of the measures of interest: expenditure on medication, transport, co-morbidities, OOP or other indirect costs; consumption choices, coping strategies, impoverishment, poverty line and catastrophic spending; the household or individual financial cost. Only studies carried out in adults (>18 years old) were included and no language or date restrictions were considered. The search strategy in “Appendix 1” was applied.

Fig. 1
figure 1

Flowchart of Studies for the Global Economic Impact of NCDs on Households and Impoverishment. aThis exclusion criterion includes letters, abstracts and conference proceedings

Full size image

Study selection

Two independent reviewers reviewed the abstracts and selected eligible studies. Any disagreements between the two reviewers were resolved through consensus or consultation of a third reviewer. To ensure consistent application of the inclusion criteria, a sample of the full texts was reviewed by a third reviewer. The references of the retrieved studies were scanned to identify additional relevant publications that were missed by the initial search. Authors of included studies were contacted to retrieve missing full texts and to identify any missing studies.

Data extraction

A data collection form was prepared to extract the relevant information from the included full texts, including study design, World Health Organization (WHO) region, characteristics of study participants, and characteristics of the NCDs evaluated and measures included. Local currencies were converted to US dollars (USD) to enhance comparability between the eligible studies, preferably using exchange rates given by the studies, if used. If no exchange rate was given, a conversion rate of the publication year of the study was used. All USD were converted to dollars of 2013 using the consumer price index conversion factors [15].

Quality evaluation

To evaluate the quality of all studies included, the Newcastle–Ottawa Scale (NOS) was applied [16]. NOS scale assesses the quality of the articles in three domains of selection, comparability and exposure. Within the selection category, four items are assessed and maximum one star can be awarded to each item. Two stars can be awarded to the one item within the comparability category. Finally, one star can be awarded to each of the three items in the exposure category. A score was made by adding up the number of stars and therefore, NOS scale can have maximum nine stars for the highest quality. For cross-sectional and descriptive studies, an adapted version of NOS scale was used (“Appendix 2”).

Statistical methods

Heterogeneity permitting, we sought to pool the results using a random effects meta-analysis model. If pooled, results were expressed as the pooled estimate and the corresponding 95 % confidence intervals. All costs presented are converted in USD 2013.

Results

From 3,241 references initially identified, 64 studies met the inclusion criteria (Fig. 1; Table 1) [17–80]. The eligible studies were published between 1999 and 2014, and included more than 835 million individuals.

Table 1 General characteristics of the studies included in this review
Full size table

General characteristics of the included studies

Of these 64 studies, three studies focused on multiple WHO regions, 20 studies originated from the WHO Western Pacific region and 25 from the WHO region of the Americas [22 from Canada or the United States of America (USA)]. Thirteen studies were from South-East Asia (eight from India); five studies from Europe and the African region contributed four studies. We found three studies from the Eastern Mediterranean region.

Fifty-seven studies had an observational design, of which twelve were prospective cohort studies, one was retrospective and 44 cross-sectional. One study presented a retrospective analysis of a randomized clinical trial and six were economic modeling studies. Most of the studies (51) used solely self-reported NCDs and economic measures data. Eligible participants were mostly sampled from hospitals, from disease registries or the general population. The remaining thirteen studies used data from regional, national and international databases and insurance data. In less than half of the studies, a control group was present; this was either a sample of the general population or sometimes sought within the same environment as the patients (e.g. same insurance company, same registry).

Sixteen studies focused on the impact of more than one NCD on households and impoverishment. The most frequently studied diseases were breast cancer and DM. Of the studies reporting on cancers, breast cancer was included in 21 studies, followed by colon cancer (eleven studies), lung cancer (eight studies) and cervical cancer (four studies). Two studies mentioned cancer, without specifying cancer types. DM was the NCD of interest in 21 studies, stroke in ten, CVD in eight and CKD in five studies. Three studies focused on COPD and three on NCDs in general terms.

Quality of the included studies

A quality score was appointed to all except 2 of the 64 included studies (Tables 2, 3, 4, 5, 6, 7, 8). In these two studies quality assessment was unfeasible due to their methodology and design. The median quality score over all the studies was 4.5 out of 9 (interquartile range 3–6). Two thirds of the eligible studies scored 5 points or less, indicating that the majority of the studies were of low or moderate quality.

Table 2 Results of the included studies investigating the impact of cardiovascular disease on households and impoverishment
Full size table
Table 3 Results of the included studies on the impact of stroke on households and impoverishment
Full size table
Table 4 Results of the included studies on the impact of cancers on households and impoverishment
Full size table
Table 5 Results of the included studies on the impact of chronic obstructive pulmonary disease on households and impoverishment
Full size table
Table 6 Results of the included studies on the impact of chronic kidney disease on households and impoverishment
Full size table
Table 7 Results of the included studies on the impact of Type 2 diabetes mellitus on households and impoverishment
Full size table
Table 8 Results of the included studies investigating the impact of non-communicable diseases on households and impoverishment
Full size table

Measures of economic impact on households and impoverishment

There was substantial heterogeneity among the studies in the measurement methods of the economic impact of NCDs on households and impoverishment. Therefore, pooling the outcomes of the included studies was not feasible.

For economic consequences (e.g. direct and indirect costs), OOP cost was the most common measure evaluated and was reported either as absolute costs or as a percentage of varying income proxies (e.g. individual income, family income, monthly non-food expenditure or household capacity to pay). Different OOP definitions were applied and could include the following expense types: cost of treatment or hospitalization (direct medical costs) and, among others, costs for transportation, food and lodging (referred to as direct non-medical costs or indirect costs). For catastrophic spending, mostly defined as a scenario in which OOP costs exceed a certain percentage of household income, different thresholds ranging from 10 to 40 % were used. Studies applying higher thresholds (e.g. 40 %) did not necessarily find lower percentages of households that experience financial catastrophe when compared to studies using lower thresholds (e.g. 10 %). Two other frequently reported measures of micro-economic burden were income loss and perceived financial hardship (e.g. worries about or change for the worse in financial situation), the latter capturing a different, more subjective perspective of the economic impact of NCDs on individuals and households.

Of the 64 eligible studies, five reported on the impact of NCDs on coping strategies, wherein the applied definitions differed between studies. Impoverishment was reported in three studies and was expressed as the percentage of people dropping below the 1, 1.25 or 2 USD per day poverty line due to the economic burden of treatment.

Impact of cardiovascular disease

Huffman et al. (Table 2) reported that 14.3 % of high-income families in China experienced some form of household income loss due to cardiovascular disease (CVD) hospitalization, rising to 26.3 % in India, to 63.5 % in Tanzania, and to 67.5 % in Argentina. This impact was patterned by socio-economic position, as greater household CVD-attributable income losses were reported for lower income groups [47]. In the USA, 10.4 % of CHD patients reported that OOP spending was more than 20 % of the family income [69]. CVD patients in India spent 30 % of their annual family income on direct CVD health care, where mean OOP per hospitalization increased from 364 USD in 1995–575 USD in 2004 [30, 59]. In CVD-affected households in India, >30 % borrowed or sold assets to pay for inpatient treatment, compared to 12 % in matched control households [78]. Also in India, the risk of impoverishment due to CVD was 37 % greater than for communicable diseases [95 % confidence interval (CI) 1.2–1.5] [59].

Impact of stroke

The average OOP burden as a percentage of income in Japan ranged between 5.1 and 17.2 % (Table 3) [33]. In China, OOP costs in the first 3 months after diagnosis of stroke was 158 % greater than the annual income. Catastrophic spending (e.g. OOP spending >30 % of annual income) was experienced by 71 %, pushing an estimated 23 % of insured and 62 % of uninsured stroke patients below the 1 USD per day poverty line [49]. In the USA, 27.8 % of stroke patients reported OOP spending at >20 % of the family income [69]. Among Australian stroke survivors, an estimated 473 USD were spent in the first year after diagnosis and 61 % perceived financial hardship after 12 months [57, 61].

Impact of cancer

All but five of the 28 studies reporting on cancer originated from high-income countries (Table 4). OOP spending as a percentage of annual income was estimated by two different studies at 9.7 and 44 % for breast cancer in the USA [32, 72]. In Canada, the percentage was 2.3 % [41]. In these countries, perceived financial hardship (e.g. worries about, or change for the worse in, financial situation) for breast cancer was reported by 1–92 % of women [40, 41, 52]. This perception of financial burden was experienced by 70 % of breast cancer patients in a study from Pakistan [18]. When comparing early to late expenditures for cervical cancer in Nigeria, the costs rose from 240 to 558 USD [34]. Among Norwegian women, income loss for cervical, breast, colon and lung cancer was experienced by 3.8, 5.7, 6.2 and 21.1 %, respectively. A loss in income due to cervical cancer was reported by 39 % of Argentinean women [71]. When comparing cancer to communicable diseases in India, the risk of catastrophic spending, defined as OOP costs exceeding 40 % of household income, and the risk of impoverishment was 2.7 times (95 % CI 2.1–3.1) and 2.3 times (95 % CI 1.9–2.9) higher [59].

Of the five studies focusing on coping strategies, all except one did so for the assessment of the impact of cancer [27, 64, 68, 77]. The results of a study by Chirikos and colleagues suggested that losses incurred by breast cancer patients were compensated by other individuals in the household [64]. Income and savings were used to pay for health care in up to 80 % of breast cancer patients, 10 % increased credit card debt, 7 % borrowed from friends or family and 5 % left some medical bills unpaid [77].

Impact of chronic obstructive pulmonary disorder

In Australia, financial hardship (e.g. worries about, or change for the worse in, financial situation) was felt by 36–78 % of COPD patients (Table 5) [46, 58]. Financial catastrophe, at a 10 % income threshold, was experienced by 46 % of COPD patients. In absolute terms, annual OOP expenditure among COPD sufferers was 2048 USD [58].

Impact of chronic kidney disease

57 % of Australian CKD patients reported financial hardship (Table 6). Using the same income threshold of 10 %, financial catastrophe was experienced by 71 % of CKD patients, which is equivalent in absolute terms to annual OOP expenditure of 3,755 USD [56]. In Japan, mean annual OOP expenditure was 2,604 USD [48]. OOP expenses due to CKD increased by 60 % between 2002 and 2005, and 32.6 % of CKD patients spent more than 10 % on income OOP [67, 69].

Impact of Type 2 diabetes mellitus

From the 21 studies focusing on DM, eight originated from India and showed a consistent impact on households (Table 7). Mean OOP expenditure per in-patient hospital stay for DM increased from 134 USD to 211 USD between 1995 and 2004 and direct total OOP spending per year was estimated at 262–280 USD [29, 50, 59]. The percent wise household consumption spent OOP ranged between 7.7 and 17.5 % [26, 30]. In Japan, the average OOP burden for DM, as a percentage of household income, ranged from 4.8 to 11.3 % [33].

In the USA, the mean annual OOP diabetes care cost was 1,237 USD and increased by 23 % from 2002 to 2005 [28, 67]. Nearly 40 % of DM cases in the USA experienced catastrophic spending (using the >10 % threshold); 13 % experienced catastrophic spending even above the 20 % threshold [69]. A cross-country analysis, performed by Niens et al., quantified the impoverishing effects of purchasing medicines for different diseases, including DM. Buying lowest price generic or originator brand glibenclamide would plunge either 2 million (5 %) or 3 million (10 %) chronic patients below the 1.25 USD/day poverty line, respectively. When stratifying across the 16 countries, these percentages ranged between 0 and 58 % [35].

Impact of NCDs in general terms

The proportion spent OOP on NCDs increased from 31.6 to 47.3 % between 1995 and 2004 in India (Table 8) [59]. In Japan, the average OOP burden was 2.1 % of available income [33]. The threshold for what is considered ‘catastrophic spending’ has a large impact on the proportion of households who experience it. For example, in Burkina Faso, the proportion of households experiencing catastrophic spending gradually increased from 4.5 to 10.6 % (and in absolute numbers from 79 to 108 USD annually) as the catastrophic threshold lowered, stepwise, from >60 to >40 %, >30, and >20 % [24]. The mean NCD expenditure as a proportion of household capacity to pay in Vietnam was 27.7 %. When using different catastrophic spending thresholds, nearly 60 % of the participants spent between 20 and 30 % of their income on NCDs [21].

Discussion

This systematic review summarizes 64 studies published worldwide of the impact of the major NCDs (CHD, stroke, COPD, major cancers, DM and CKD) at the micro-economic level on households and impoverishment. The studies show a steady global increase in household expenditure on NCDs between 1999 and 2014. The importance of these trends in global health is further underlined by the ‘WHO Global Action Plan for the Prevention of non-communicable diseases 2013–2020’, which highlights the need for further research into NCDs and their impact at the micro-economic level [81].

There is evidence that a substantial number of people experience financial hardship due to NCDs, as income losses affect patients and their caregivers and OOP medical expenditure for NCDs drive households into financial catastrophe and impoverishment. This rising burden is directly related to the global rise of NCDs, particularly in LMIC, many of which have under-resourced healthcare systems that impose OOP payments on individuals and households as a means to supplement other sources of revenue [1]. As healthcare systems in LMIC often experience a dual burden of infectious and chronic disease, they are less able to allocate resources towards primary prevention of NCDs. Most eligible studies used OOP expenditure to quantify the magnitude of the economic impact of NCDs on households and for mapping the extent of financial catastrophe, in particular. OOP expenditure was self-reported in most of the studies, with some exceptions where studies used health insurance claim data. Relative to different income proxies, OOP expenditure ranged widely between 2 and 158 % across different NCDs and countries.

The threshold for what is considered ‘catastrophic spending’ has a large impact on the proportion of households who experience it; depending on the income threshold taken by the study, the global proportion of households suffering from financial catastrophe ranged from 6 to 84 %. Heterogeneity in the use of an income threshold in combination with differences in study samples (among others, related to insurance coverage levels) undermine comparability across the studies, although evidence does suggest that financial catastrophe due to NCDs is an important issue for all countries and across all income strata. This observation is in accordance with other reports that took a broader (chronic) illness perspective [8, 10, 14, 82]. Variations in OOP spending and financial catastrophe across and within countries depend a great deal on the triad of factors, described by Xu and colleagues, as poverty levels, healthcare service access and use, and the presence or absence of financial risk pooling mechanisms such as health insurance or taxed-based systems [9]. Although it was outside the scope of this study to review the impact of this triad on catastrophic spending, these factors are very likely to be key components of the (varying) relation between OOP spending and catastrophic spending. Therefore, although OOP spending and financial catastrophe are valuable methodological approaches to provide insights into the impact on households, these measures cannot be interpreted without being placed within the specific health system perspective from which the sample is drawn. Standardized definitions and thresholds would facilitate unbiased and cross-country comparisons.

A minority of the studies addressed the absolute impoverishing effects of NCDs. A large study by Niens et al., in 16 LMICs, showed that the purchase of lowest price generic medication rather than originator brand DM (and other) medication could reduce absolute impoverishment, at the 1.25 USD/day poverty line, from 11 to 6 % of the total population. This finding reinforces the need to improve availability of low-priced generics, which for NCDs receives comparatively little attention compared to infectious disease treatment [83, 84].

The extent to which NCDs drive households into relative poverty were more difficult to estimate from the eligible studies, partly due to the fact that relative poverty is more difficult to measure and the definitions are less clear. We observed that some eligible studies used income losses to estimate the relative impoverishing influence of NCDs. For instance, for Norwegian women suffering from cervical, breast or lung cancer, the percent-wise income deviation compared to healthy women was 3.8, 5.7 and 20 % respectively [22]. Household income losses after CVD diagnosis were 67.5, 14.3, 26.3 and 63.5 % in high-income groups in Argentina, China, India and Tanzania respectively, and were even higher in the lower income groups [47]. These findings are consistent with similar studies, which showed that poor households are less able to cope with healthcare costs compared to more affluent households [9, 85, 86]. Solely five eligible studies provided insights in the coping strategies adopted by households to cope with a family member suffering from NCDs. The paucity of evidence regarding coping strategies, together with the significant role that illness perceiving and absence of health care seeking due to financial reasons play, are likely to reflect a considerable underestimation of the true extent to which NCDs impact households.

Findings of this systematic review generally concur with and further extend previous reviews on this topic. Previous work was focused on specific types of NCDs, was focused in specific regions of the world or provided methodological commentaries [10, 87–100]. A recent narrative review emphasized the importance of standardized definitions for OOP spending, the use of larger sample sizes and prospective study designs and a better collecting of data on economic consequences of NCDs (e.g. direct and indirect costs) [89]. Kankeu and colleagues assessed financial burden of four domains of NCDs (cancers, CVD, COPD, and diabetes) but did not include CKD in their review. In addition and interestingly, they included only studies conducted in LMICs [91]. Mahal and colleagues summarized the economic impact of NCDs for India [94]. A second study, conducted by Engelgau et al. [10], non-systematically reviewed studies mostly conducted in India. Costs involved in cancer care, without stratifying for cancer type, were reviewed in three domains in a systematic review by Pearce and colleagues. The domains included cost-effectiveness and cancer treatment, the indirect cancer costs and human costs of cancer. Definite conclusions were missing due to conceptual and methodological limitations of the included studies. Nevertheless, the complexity of the costs attached to cancer care was observed [95]. Pisu et al. [96] reviewed OOP expenses in breast cancer patients only. Tong and colleagues thematically synthesized patient and caregiver perspectives in CKD. Out of 26 included studies in this review, one study from Thailand focused on economic consequences, and found a large economic strain due to forced early retirement [97]. Coping with OOP health payments was assessed in 15 African countries and showed that borrowing and selling assets was an important coping mechanism, its prevalence ranging from 23 to 68 %. Unfortunately a specification of the included diseases was not provided [93].

The strength and limitations of our work merit careful consideration. An important strength of this review is the exhaustive search for relevant articles. We used extensive, precise search terms and applied stringent inclusion criteria, specifically the exclusion of studies focusing solely on ‘chronic diseases’ or ‘illness’. We feel that this specific approach gave rise to a comprehensive undiluted perspective of the micro-economic impact of NCDs, since all available evidence was gathered via the initial search and was supplemented by an extensive screening of reference lists for possibly missed eligible studies. However, we do emphasize that precisely defining included chronic illnesses would greatly benefit future research and the disease specific policy implications this research could give rise to.

The methods used by the eligible studies to measure household impact and impoverishment were remarkably heterogeneous which, along with a broader disease burden perspective than NCDs, is a recurrent challenge in similar reviews and did not allow us to pool the reported estimates in a meta-analysis [14, 91]. Furthermore, in many studies convenience sampling was used to assemble study samples, and the overall quality of the included studies was moderate to low. Therefore, country-wide and disease-specific implications of the results must be interpreted with caution. Given the already wide scope of our systematic evaluation, we were unable to explore wider impacts associated with NCDs such as non-economic and indirect impacts including educational dropout among children, healthcare utilization and costs of premature death. Estimation of the number and experiences of marginalized and vulnerable people who do not seek care for NCDs for financial reasons is currently neglected and their inclusion could give a more comprehensive overview of the impact of NCDs on households and impoverishment.

Conclusions

NCDs impose a large and growing global impact on households and impoverishment, in all continents and levels of income. The true extent, however, remains difficult to determine due to heterogeneity across existing studies in terms of populations evaluated, outcomes reported and measures employed. The impact that NCDs exert on households and impoverishment is likely to be underestimated since important economic domains, such as coping strategies and the inclusion of marginalized and vulnerable people who do not seek health care due to financial reasons, are overlooked in literature. Given the scarcity of information on specific regions, further research is required to estimate impact of NCDs on households and impoverishment in LMIC, especially the Middle Eastern, African and Latin American regions.