The effectiveness of interventions to reduce the household economic burden of illness and injury: a systematic review

Abstract Objective To determine the nature, scope and effectiveness of interventions to reduce the household economic burden of illness or injury. Methods We systematically reviewed reports published on or before 31 January 2014 that we found in the CENTRAL, CINAHL, Econlit, Embase, MEDLINE, PreMEDLINE and PsycINFO databases. We extracted data from prospective controlled trials and assessed the risk of bias. We narratively synthesized evidence. Findings Nine of the 4330 studies checked met our inclusion criteria – seven had evaluated changes to existing health-insurance programmes and two had evaluated different modes of delivering information. The only interventions found to reduce out-of-pocket expenditure significantly were those that eliminated or substantially reduced co-payments for a given patient population. However, the reductions only represented marginal changes in the total expenditures of patients. We found no studies that had been effective in addressing broader household economic impacts – such as catastrophic health expenditure – in the disease populations investigated. Conclusion In general, interventions designed to reduce the complex household economic burden of illness and injury appear to have had little impact on household economies. We only found a few relevant studies using rigorous study designs that were conducted in defined patient populations. The studies were limited in the range of interventions tested and they evaluated only a narrow range of household economic outcomes. There is a need for method development to advance the measurement of the household economic consequences of illness and injury and facilitate the development of innovative interventions to supplement the strategies based on health insurance.


Introduction
Each year, globally, around 150 million people struggle to meet the costs of accessing and using health care and approximately 100 million people are driven below the poverty line by such costs. 1 Many people delay or avoid health care because it is -or, at least, is perceived to be -unaffordable. [2][3][4] Most of those who struggle to meet the out-of-pocket costs of health care live in low-income countries that have poorly funded health systems and inadequate measures to ensure the financial protection of households against high health-care expenditure. However, the problem is not limited to such countries. In 2007, for example, 62% of the personal bankruptcies recorded in the United States of America (USA) were attributed to medical debt 5 and 11% of the individuals found insolvent in Australia cited ill-health or lack of health insurance as the primary reason for their insolvency. 6 Substantial and unpredictable one-off health-care payments and a steady flow of unbudgeted medical bills can lead many households -particularly those already marginalized by socioeconomic disadvantage -towards catastrophic health-care expenditure. 7 The economic burden of illness in a household is only partly explained by out-of-pocket expenditure. The full evaluation of such burden requires a multidimensional framework -to move beyond absolute spending to incorporate measures that examine the broader impacts of illness or injury on the household economy -e.g. loss of employment -as well as the affordability of care, a household's response to an injury or illness and the consequences of those responses for the household. 8,9 Most research in this area has been observational and has demonstrated that households will employ several strategies -to deal with unbudgeted costs of medical care and unplanned departures from the workforce -when coping with the onset of an illness or injury, especially in the main income earner. Such coping strategies include drawing on available social resources and networks, cutting back on essential living expenses, drawing on savings, selling assets, borrowing money, entering into formal or informal loan agreements, increasing credit or debt and even moving house. 3,4 Although these strategies may help leverage the resources needed to pay for care, they can also have adverse effects on treatment-seeking behaviour and the long-term economic well-being and resilience of the household. 3,4,7 The provision of adequate financial protection -from the costs of seeking and using medical care -is a critical marker of the effectiveness of a health-care system. 10 The World Health Organization has encouraged its Member States to provide universal health coverage in some form and the United Nations has recently passed a declaration that calls for universal access to health care that does not cause financial hardship. 11 Such a goal -like other post-2015 development goals aimed at alleviating poverty -is unlikely to be achieved without further development and implementation of national health-insurance schemes. There is considerable evidence, most notably from the RAND Health Insurance Experiments, 12 that indicates how health insurance can protect the finances of households affected by illness or injury, by restricting individual healthcare expenditure. However, although such insurance is one of the most important population-based policy interventions to mitigate the economic burden of injury or illness, it is not sufficient, on its own, to provide full protection from catastrophic health expenditure. 13,14 The effectiveness of health insurance in protecting individuals who are intense users of medical caree.g. those with chronic illness or long-term injuries -has yet Systematic reviews Reducing the economic burden of illness and injury Beverley M Essue et al.
to be elucidated. Furthermore, limited coverage of services and high levels of co-payment can often mean that households with health insurance remain at risk of catastrophic health-care expenditures and economic hardship. 14,15 Evidence of the effectiveness of simple education and support interventions, in both clinic-and community-based settings, has highlighted the potential value of more targeted and patientfocused strategies in reducing the household economic burden of illness. Interventions that help patients and caregivers to navigate through health and social-welfare support systems 16,17 and informal loan and microcredit schemes [18][19][20] have the potential to buffer those with illness and injury against financial hardship. As the evidence of the effectiveness and cost-effectiveness of such interventions becomes more robust, opportunities for the development and scale-up of such interventions need to be explored.
There have been few systematic reviews of interventions to reduce the household economic burden of illness or injury. The reviews that have been conducted have tended to take a population-based approach -e.g. they have examined the impact of healthinsurance programmes on entire populations -and have often been based on studies that involved retrospective comparisons of before and after data. Furthermore, they have focused either on specific types of interventions -e.g. programmes for the management of chronic illness 21 or health-insurance schemes 22-24 -or have focused, narrowly, on out-of-pocket payments, as the sole measure of the economic impact of illness. 24 We decided to conduct a systematic review to try to determine the nature, scope and effectiveness of all interventions that have been designed to reduce the household economic burden of illness or injury.

Methods
We searched electronic databases, using a predefined search strategy and confining the search to reports published on or before 31 January 2014 (Box 1). The reference lists of retrieved articles were screened to identify additional studies, and investigators known to be carrying out relevant research were contacted for unpublished data. Non-English articles were translated where necessary.
To be included in our review, a study (i) had to be a prospective controlled trial of one or more interventions -i.e. a randomized or nonrandomized controlled trial, an interrupted time series study with control, or a controlled before-and-after study; (ii) involve a study population with any, chronic or acute, communicable or noncommunicable disease or injury; and (iii) use a study outcome that was a measure of the household economic burden of illness or injury -e.g. out-of-pocket expenditure or level of economic hardship.
Interventions directed at the individual, household or population and delivered in any setting were eligible for inclusion. Studies that were primarily treatment or medical interventions -e.g. cataract surgery or chemotherapy -were excluded even if they included economic measures as additional outcomes.
Two authors carried out the literature search and screened titles and abstracts using a standardized eligibility assessment form based on our inclusion criteria. The full texts of articles of potential interest were reviewed by two authors and a final decision on which studies to include was confirmed by consensus. A third author provided arbitration if consensus was not reached.
One author used a predefined form 25,26 to extract data from each included study. The data extraction was verified by a second author. Authors of included studies were contacted for any missing information or data. Where possible, effect estimates were calculated as standardized mean differences between the intervention and control groups, with 95% confidence intervals. 27 Where reported, data on the impact of the interventions on health-service-utilization -e.g. numbers of hospital admissions or medical appointments -and medication adherence were also collected.
The risk of bias in each of the included studies was assessed by one author -using the criteria suggested for Effective Practice and Organisation of Care reviews 28 -and verified by a second author.
Quantitative analysis of the data was deemed inappropriate because of the heterogeneity in the collected data, designs and settings of the included studies.

Results
The initial literature search identified 4330 citations. There were 90 articles of potential interest and, after examination of the full texts, nine articles described studies that met all of our inclusion criteria ( Fig. 1). Each of the nine articlesseven conducted in the USA, [29][30][31][32][33][34][35] one in Finland 36 and one in China 37 -described a single study. Most of the included studies had investigated adult urban patients with noncommunicable disease (6/9) and had involved data from more than 1000 participants (7/9; Table 1). Illness and injury inclusion criteria had been assessed using diagnostic codes, the health-service use reported in insurance claims, clinical presentations or self-reporting.
Seven of our included studies had evaluated policy interventions that involved health-insurance schemes ( Table 2). Of these, three had involved the reduction or elimination of co-payments for disease-specific medications or outpatient care. 31,32,37 Another three studies had evaluated the effectiveness of a similar intervention -that offered parity in service coverage for mental health and substance use disorders -in different subgroups. 29,30,34 One study had investigated the extension of coverage of an existing health-insurance scheme to a new patient population. 33

Box 1. Basic literature search strategy for systematic review of interventions to reduce the household economic burden of ill health
The following databases were searched: CENTRAL, CINAHL, Econlit, Embase, MEDLINE, PreMEDLINE and PsycINFO Search terms: 1. "intervention" OR "program" OR "programme" OR "policy" OR "scheme" 2. "catastrophic" AND "finance OR cost OR medical OR expenditure" 3. "finance OR economic" AND "hardship OR strain OR stress OR well-being" 4. "burden" AND "household financial OR household economic" 5. "household" AND "economic impact" 6. "out-of-pocket" AND "cost OR expenditure OR spend OR payment OR catastrophic" A detailed search strategy for each database is available from the authors.
The other two studies trialled different models of delivering patient-focused education and support -e.g. by web-or telephone-based communication or inperson. 35,36 Out-of-pocket expenditure had been the primary outcome in six of our included studies -including one post-hoc analysis -and a supplementary outcome in another two ( Table 1). The researchers involved in most of the studies had ascertained out-of-pocket expenditures from databases of insurance claims. Household economic burden had also been measured in terms of the likelihood of a household paying any out-of-pocket costs for care, the prevalence of catastrophic health expenditure -i.e. out-of-pocket costs that were greater than 40% of the maximum amount that a household could payand the prevalence of cost-related delays in seeking care. None of the studies had evaluated the effectiveness of an intervention in reducing economic hardship.
Six of the studies had also investigated the effectiveness of an intervention on clinical and health-system outcomes, health-service use, adherence to pharmaceuticals, direct costs to private health insurers or the indirect costs to patients and household caregivers in terms of the time spent seeking health care.
There was a high or unclear risk of bias in the randomized and nonrandomized controlled trials and controlled before-and-after studies ( Fig. 2; available from: http://www.who.int/ bulletin/volumes/93/2/14-139287). In these studies, inadequate allocationsequence generation and concealment could have resulted in an overestimate of the effects of an intervention on the household economic burden -particularly since absolute out-of-pocket expenditure was often the main outcome and such expenditure was selfreported in three studies. [35][36][37] Attrition bias due to incomplete reporting of outcome data -which may also lead to overestimates of an intervention -was potentially an issue in three studies. [35][36][37] There was also a high risk of reporting bias in two of the studies. 35,36 The data we reviewed from interrupted time series studies (3/9) had a generally low risk of bias ( Fig. 3; available from: http://www.who.int/ bulletin/volumes/93/2/14-139287). However, in such studies, there is some risk that the intervention effect may not have occurred independently of other changes occurring over time and that the outcome observed may have been influenced by confounding factors. These two issues may have resulted in an overestimate of the effect of the intervention. Attrition bias may also be an issue in these studies since there is unclear bias introduced by the incomplete reporting of outcome data.
The outcomes of the interventions investigated in all of our included studies are summarized in Table 3.
Two studies conducted in the USA evaluated the effectiveness of reducing or eliminating co-payments and found statistically significant reductions in out-of-pocket costs for cardiovascular pharmaceuticals and medical services. 31,32 Another three studies conducted in the USA evaluated the effectiveness of parity in service coverage for mental health problems and substance use disorders. 29,30,34 In these three studies, statistically significant reductions in outof-pocket expenditure were reported for the whole study population, 34 among children with high expenditure 29 and in specific disease groups. 30 For example, the reported mean annual reductions in out-of-pocket costs per patient were 148, United States dollars (US$) for bipolar disease, US$ 100 for major depression and US$ 68 for adjustment disorder. 30 A sixth study in the USA found a statistically significant association between the expansion of health-insurance coverage and the proportion of people who had moderate out-of-pocket costs of US$ 1-2000 per person. 33 In rural China, the implementation of a voluntary community-based insurance programme that offered higher reimbursement for outpatient services for a poor population was not found to reduce the prevalence of catastrophic health expenditure significantly. 37 In Finland, the web-based delivery of information to patients was not associated with any change in out-ofpocket expenditure. 36 In the USA, an intervention that targeted information at caregivers was found to increase the care-associated spending of the caregivers and had no significant effect on total  Reducing the economic burden of illness and injury Beverley M Essue et al.
out-of-pocket expenditure on health for the patients. 35 Outcomes other than out-ofpocket expenditure were assessed in several studies (Table 4; available from: http://www.who.int/bulletin/volumes/93/2/14-139287). Two insurance interventions were adequately powered to measure their effect on clinical and health-service outcomes. One study found significant reductions in the rates of total major vascular events or revascularization. 31,32 None of the other seven studies we included in our systematic review appeared to show a significant impact on the clinical or health-service outcomes assessed -probably because they were underpowered to assess the effect.

Discussion
To the authors' knowledge this is the only systematic review to synthesize published evidence on the effectiveness of interventions that address the diverse ways that illness and injury adversely affect household economics. In the reviewed studies, the economic burden of illness at household level was measured predominantly in terms of out-of-pocket costs. The interventions that were found to be most effective at mitigating the burden of illness were implemented in the context of existing health-insurance schemes and involved reducing or eliminating co-payments for disease-specific treatments. Offering parity in the bene-fits for specific illnesses also significantly reduced out-of-pocket costs.
However, any reductions in outof-pocket expenditure should be interpreted in the context of total spending -by the individual and the household -for the management of an illness or injury. 30 One study reported that, although the 21% reduction in out-ofpocket expenditure found in their study was statistically significant, the absolute annual reduction -of US$ 100-148 per patient -was unlikely to confer protection from catastrophic expenditure. 30 Total household expenditure on health-related care -including the costs of transport, home assistance, medical equipment and accommodation -can be much greater than the direct out-ofpocket costs of medicines and surgery. 38 Moreover, such indirect costs of care are seldom covered by health-insurance schemes, particularly in low-income settings. Few of our included studies incorporated other categories of out-ofpocket expenditure beyond the direct costs of medical care. Interventions that solely reduce co-payments for specific aspects of care will only be effective if the care that is covered represents the main economic burden of the illness or injury at household level. Furthermore, many households may have more than one member with illness or injury. Therefore, interventions will need to move beyond targeting disease-specific aspects of treatment and, instead, take a holistic view of the multiple and diverse ways that illness and injury affect household economic circumstances.
Of the nine studies we reviewed, seven involved changes to -or extensions of -an existing package of healthinsurance benefits, with the sole aim of shifting the costs of care to the insurer and minimizing the costs to the patient. Only one of these health-insurance studies was conducted in a low-or middle-income country. Although most of the health-insurance interventions were associated with statistically significant effects within the study period, such interventions will not be put into widespread practice unless they can be shown to be economically viable. To the authors' knowledge, only one of the health-insurance studies was accompanied by a published cost-effectiveness investigation of the type needed to inform priority setting and resource planning for any sustainable intervention. In low-and middle-income countries, the financial sustainability of such measures is critical. If the post-2015 development goals relating to poverty reduction are to be achieved, good evidence is needed to inform the development of stronger and more financially sustainable health systems in these settings.
There is a general scarcity of evaluations of innovative interventions to address the economic burden of illness and injury. Such interventions have the potential to supplement existing health-   insurance policies, particularly those being rolled out to achieve universal health coverage in low-and middle-income settings. The interventions uncovered in this review tended to be health-insurance-based or, to a lesser extent, involve some form of patient education. If used in isolation, such interventions cannot resolve the fundamental issues of social disadvantage and poverty and overlook the multidimensional pathways in which illnesses or injuries are linked to economic outcomes. For instance, there appear to have been few attempts to examine the role of strategies such as income support or programmes to support household consumption in addressing the financial challenges of long-term chronic illness. This might be due to the narrow disciplinary perspectives of the relevant researchers. 39 This review highlights a need for method development in this field, to take account of the capacity of households to afford out-of-pocket expenditure and the impact of coping strategies on household economic outcomes. There is an interconnection and, potentially, a vicious cycle between poor economic circumstances and illness. 3,40 Social disadvantages can predispose individuals to a risk of illness. This, in turn, can predispose individuals and their households to illness-related poverty and economic hardship. These economic consequences can further perpetuate poor health, through impaired quality of life, depression and non-adherence to treatment. Interventions to address the economic burden of illness have the potential to break this nexus. However, research has been slow to adopt tools for measuring outcomes in this field beyond out-of-pocket expenditure, and the relevant studies that have been conducted have been of variable quality and rarely randomized controlled trials. There have also been inconsistencies in the measurement and reporting of outcomes such as out-of-pocket costs and catastrophic health expenditures. 41,42 Once a consistent approach to measuring outcomes has been developed, research in this area will allow for greater comparability between studies 8,9 and offer opportunities for the routine assessment of household expenditures within research on clinical interventions. 43,44 This review has limitations. First, the authors of excluded studies were not contacted to determine if they had collected data on relevant outcomes but not reported them. Second, the household economic burden of illness or injury was not the primary outcome in all of the included studies. It is possible that some included studies were not sufficiently powered to detect a change in this outcome. Third, this review was limited to studies published in the peerreviewed literature. Fourth, most of the included studies were conducted in the USA and so low-and middle-income settings were underrepresented. Finally, there were few randomized controlled trials included. As a result of the two latter issues, our findings are unlikely to be representative of all health systems.

Conclusion
Health-insurance programmes that reduce or eliminate co-payments for defined illness-specific treatments can effectively provide some financial protection, by reducing out-of-pocket expenditure. However, little is known about the cost-effectiveness of such programmes and about other forms of intervention that may provide relief from adverse economic outcomes to households. Given the multiple and diverse ways that illness and injury can affect the economic circumstances of households, this review highlights the need for method development in this field -above and beyond the limited focus on out-of-pocket expenditure. Additionally, especially in low-and middle-income countries, there is wide scope for research on the effectiveness of innovative non-insurance interventions that could provide low-cost and bettertargeted support. ■ Relative changes in out-of-pocket costs were adjusted for age, sex, income, race, coronary artery disease, congestive heart failure, diabetes, hypertension, Charlson comorbidity score, number of hospitalizations and prescription drugs on enrolment. c Difference in difference analysis of gender-, age-and area-adjusted change in annual out-of-pocket costs among those in at least the 90th percentile of MH/SUD treatment expenditure. The corresponding proportion of total MH/SUD costs fell significantly by 5% (P ≤ 0.05

Systematic reviews
Reducing the economic burden of illness and injury Beverley M Essue et al.
For each domain, a study was categorized as high, low or unclear risk, using the criteria suggested for Effective Practice and Organisation of Care reviews. 28 Note: Each domain refers to an area of potential bias that could affect the validity of the three studies.
For each domain, a study was categorized as high, low or unclear risk, using the criteria suggested for Effective Practice and Organisation of Care reviews. 28