Introduction

Fair financing is a critical issue in health systems. Efficient financial mechanisms in health systems can protect households from facing catastrophic health expenditures (CHE) (Murray & Frenk, 2000). According to the World Health Organization (WHO) definition, CHE occurs when a household's total out-of-pocket expenditures on health are equal to or exceed 40 percent of households’ capacity to pay or non-subsistence spending. The households’ capacity to pay is defined as the income left after all basic subsistence needs are met (Xu et al., 2003). Globally, more than 150 million people face CHE every year because of health-care spending (Wagstaff et al., 2018). A study reported the wide range of prevalence of CHE (1 to 25 percent) in Latin America and the Caribbean countries and based on that study health insurance was one of the important factors in preventing households from CHE in these countries (Knaul et al., 2011). In general, the proportion of households facing CHE in different countries is extensively varied, and developed countries have a much lower incidence of catastrophic payments than developing ones (Xu et al., 2003). The World Health Organization (WHO) considers health insurance as an important instrument to provide financial protection and to achieve universal health coverage. UHC was defined as ensuring that everyone within a country can access the health services they need, which should be of sufficient quality to be effective, and provide all with financial protection from the costs of using health services (Aryeetey et al., 2016). It has been shown that health insurance schemes can decrease the incidence of CHE (Xu et al., 2007). Asian countries have been achieved different levels of successful financial protection from households and the incidence of CHE varies between the countries (Van Doorslaer et al., 2007).

Evidence shows that the health insurance schemes for the poor can result in a significant reduction in CHE and financial hardship (Galárraga, Sosa-Rubí, Salinas-Rodríguez, & Sesma-Vázquez, 2010). However, some health insurances are weak in coverage of services and their related costs and therefore, insured people have to pay a considerable fraction of health expenditures in form of out-of-pocket (OOP) payment (Preker et al., 2002). Based on the report of the national health accounts (NHA), the proportion of different health insurances in total health expenditures is very low and OOP payments consist of the highest share of total health expenditures in Iran(Oliaie Manesh et al., 2008). This type of payment usually prevents the poor from using the health-care services that they need (Preker et al., 2002).

There are four main social health insurance funds in Iran. The most important health insurance organizations in Iran are the Iran Health Insurance Organization (IHIO) and Social Security Organization (SSO). IHIO is a governmental organization and has four sub-funds that provide health insurance for government employees and their families, rural residents, the self-employed and their dependents, and other sectors (such as students, some professional associations, and so on). SSO is a non-governmental company that covers all the individuals employed in the formal private sector and their families. The Armed Forces Medical Services Insurance Organization (AFMSI) provides health insurance for military personnel and their dependents. The fourth is the Imam Khomeini Relief Foundation (IKRF), a governmental fund providing health insurance coverage for the poor. In addition, there are about 17 smaller institutional health insurance funds such as those offered by banks, the National Broadcasting Organization, the Tehran Municipality, the Petroleum Industry Health Organization and so on which provide health insurance coverage for their own employees and families (Bazyar et al., 2016). Although about 96 percent of the Iranian population is covered by health insurance, the pooled estimation of the prevalence of CHE in the Iranian population showed that on average, about 7 percent of the population suffers from CHE (Rezaei, Woldemichael, Hajizadeh, & Kazemi Karyani, 2019).

The effect of health insurance on the financial protection of Iranian households has been investigated in several studies (Mobaraki et al., 2018; Moradi et al., 2017; Piroozi, Moradi, Nouri, Mohamadi Bolbanabad, & Safari, 2016; Rezapour et al., 2017). However, different and conflicting results have been reported by them. Most of the studies indicated that having health insurance reduced CHE (Amery, Jafari, & Panahi, 2013; Kavosi et al., 2012; Mehrara & Fazaeli, 2010; Piroozi et al., 2016; Sabermahani et al., 2014; Soofi et al., 2013). On the other hand, in some studies having health insurance did not have a significant effect on reducing CHE (Amery et al., 2013; Kavosi et al., 2012; Moghadam et al., 2012; Razavi et al., 2005). Therefore, using all accessible evidence, we conducted a meta-analysis to assess the protective effect of health insurance on incurring CHE in Iran. The results can help to evaluate the effects of health insurance coverage on providing financial protection against CHEs in Iran's health system.

Methods

We used the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement for conducting and reporting the study (Liberati et al., 2009).

Inclusion Criteria

We considered studies conducted on estimating CHE in the Iranian setting. Studies in which health insurance was used as a predictor of CHE were included. We considered primary peer-reviewed studies, including cross-sectional studies at both the regional and national levels. Different thresholds have been used to estimate exposure to CHEs in various studies (Rashidian et al., 2018). In this study, the studies that used the WHO methodology to measure CHE were included. WHO methodology proposed 40 percent capacity to pay as a threshold for the calculation of CHE (Xu et al., 2003).

Exclusion Criteria

We excluded studies that lacked quantitative data on CHE or insurance status and full-text not accessible.

Information Sources and Search Strategy

We searched the English-language medical literature published until January 30, 2018, in four databases: Scopus, PubMed, Web of Science, and Google Scholar. In addition, a systematic search of Persian-language medical literature was conducted in three databases: Scientific Information Systems (SID), Iranedex, and Magiran. Also, we hand-searched all the references of the included studies to maximize the sensitivity of our search and contact with study authors to identify additional studies or missing information. We used a combination of Medical Subject Headings (MeSH) terms and free terms in the database search. The search was based on the title and abstract with the appropriate keywords and terms such as CHE, health insurance, and Iran. Our search was limited to English and Farsi language studies.

Study Selection and Data Extraction

After completing the search, all records imported to Endnote v8 and the duplicates were removed. Two reviewers (HA and MAZ) independently screened the records based on the title, abstract, and the full-text according to the eligibility criteria. Any discrepancy was resolved through consensus with a third reviewer (MS).

Two independent reviewers (HA and MAZ) used a data collection form to extract the relevant information from the selected studies. The data collection form included items such as author(s) name, published year of study, study design, target population, sample size, sampling method, geographical origin, mean age of participants, data collection period, type of questionnaire, type of insurance and proportion of exposure to CHE. The extracted data were rechecked by a third reviewer (MS) to prevent the occurrence of any error.

Quality Appraisal

The JBI checklist for cross-sectional studies was used for appraising the quality of included studies (The Joanna Briggs Institute, 2017). Two reviewers (HA and MAZ) independently assessed the included studies. Potential discrepancies were resolved through consensus with a third reviewer (MS). The studies are divided into three categories according to quality. As this checklist has eight items, studies with a score of less than three were weak, between three and five were moderate and more than six were strong.

Data Analysis

In this meta-analysis, the odds ratio (OR) was considered as the common measure of the association between health insurance with CHE. We used the Cochran Q test and I 2 statistics for quantifying heterogeneity between the included studies (Higgins & Thompson, 2002). Due to the high heterogeneity in studies, we performed a random effect model using the Mantel–Haenszel method in our analysis (Riley, Higgins, & Deeks, 2011). Subgroup analyzes were performed based on the type of insurance and target population. All data synthesis and analysis were performed using RevMan 5.3 (RevMan5.3; The Cochrane Collaboration, Oxford, UK).

Results Study Selection

Initially, a total of 1,444 records were screened and 102 full publications were retrieved, yielding 13 studies that met inclusion criteria (Ghiasvand et al., 2010; Amery et al. 2014; Karami et al., 2009; Kavosi et al., 2009, 2012; Mobaraki et al., 2018; Moghadam et al., 2012; Moradi et al., 2017; Piroozi et al., 2016; Rezapour et al., 2017; Soofi et al., 2013). Figure 1 shows the study selection process and the reasons for exclusion (Figure 1).

Flow of Selection for Studies Through Review.

Study Characteristics

The included studies were published between 2001 and 2018, and the data collection was done between 2001 and 2017. All papers used in our analysis were published in English (61.5 percent) and Farsi (38.5 percent). The sample size ranged from 189 to 39,008. The main results are based on information from 13 studies, which included 54,128 participants. The WHO questionnaire (77 percent) was the most used questionnaire. Table 1 summarizes the characteristics of the included studies (Table 1).

Table 1. Summary Characteristics of Included Studies Author (year) Study Design Sample Size Sampling Method Data Collection Year Region/Scope (Name of Place) Type of Questionnaire Target Population Type of Insurance Mean Age Male (%) Language Piroozi et al. (2015) Cross-sectional 646 Cluster 2015 Sanandaj WHO questioner Household Basic health insurancea 43 92.0 English Supplementary Health insurance Rezapour et al. (2015) Cross-sectional 772 Proportional stratified random 2014 Tehran WHO questioner Chronic diseaseb Supplementary Health insurance NS 88.86 English Moradi et al. (2015) Cross-sectional 385 Stratified random sampling 2015 Kurdistan WHO questioner Households with members with special diseases Supplementary Health insurance NS NS English Kavosi, Keshtkaran, Hayati, Ravangard, and Khammarnia (2014) Cross-sectional 245 Random sampling 2014 Shiraz WHO questioner Chronic diseasec Basic health insurance 49.4 90.4 English Kavosi et al. (2013) Cross-sectional 376 Multi-stage cluster sampling 2013 Shiraz WHO questioner Patients admitted Complimentary insurance 52.13 46.4 English Karami et al. (2008) Cross-sectional 189 Systematic Random 2008 Kermanshah WHO questioner Household Basic Health insurance 48.96 87.3 English Mobaraki et al. (2018) Cross-sectional 553 Random 2017 Tehran WHO questioner Old adults Supplementary Health insurance 70.6 90 English Nekoei Moghadam et al. (2008) Cross-sectional 39,008 Convenience 2008 National statistical Centre of Iran (SCI) questionnaire Household Basic Health insurance NS 88.9 English Amery et al. (2011) Cross-sectional 384 Multi-stage cluster sampling 2011 Yazd WHO questioner Household Basic Health insurance 27 93.8 Persian Amery et al. (2012) Cross-sectional 400 Cluster sampling 2012 Torbat Heydariyeh WHO questioner Household Basic Health insurance 49.87 51.9 Persian Soofi et al. (2001) Cross-sectional 10300 Multi-stage cluster sampling 2001 National WHO questioner Household Basic Health insurance 52.4 51 Persian Ghiasvand et al. (2009) Cross-sectional 340 Random 2009 Tehran Researcher-made questioner (RM) Patients admitted Basic Health insurance NS 80 Persian Supplementary Health insurance Kavosi et al. (2004) Cross-sectional 579 Multi-stage cluster sampling 2004 Tehran WHO questioner Household Basic Health insurance NS NS Persian Kavosi et al. (2008) Cross-sectional 592 Multi-stage cluster sampling 2008 Tehran WHO questioner Household Basic Health insurance NS NS Persian Quality Assessment

The JBI tool for quality assessment in cross-sectional studies yielded scores ranging from 2 to 5. Of 13 included studies, eleven studies were scored as moderate, and two were scored as weak. The overall agreement between the two reviewers was judged as good (Cohen's κ = 0.85). The agreement rate for each question ranged from 0.76 to 1.00) (Table 2).

Table 2. The Results of the Inter-Rater Reliability Test (Cohen's κ Coefficient) 1 2 3 4 5 6 7 8 9 10 11 12 13 Questions Piroozi et al. (2015) Rezapour et al. (2015) Kavosi et al. (2014) Kavosi et al. (2013) Karami et al. (2008) Mobaraki et al. (2018) Moradi et al. (2015) Nekoei Moghadam et al. (2008) Amery et al. (2011) Amery et al. (2012) Soofi et al. (2001) Ghiasvand et al. (2009) Kavosi et al. (2008) Cohen's κ Were the criteria for inclusion in the sample clearly defined? Y N N N Y Y Y Y Y Y Y Y Y 0.84 Were the study subjects and the setting described in detail? Y N Y N Y Y Y N Y Y N N Y 0.92 Was the exposure measured in a valid and reliable way? Y Y Y Y Y Y Y Y Y Y Y Y Y 0.84 Were objective, standard criteria used for measurement of the condition? N N N N N N N N N N N N N 0.76 Were confounding factors identified? N N N N N N N N N N N N N 0.76 Were strategies to deal with confounding factors stated? N N N N N N N N N N N N N 0.84 Were the outcomes measured in a valid and reliable way? Y Y Y Y Y Y Y Y Y Y Y N Y 1.00 Was appropriate statistical analysis used? Y N Y Y N Y Y N Y Y N N Y 0.84 Overall appraisal 5 2 4 3 4 5 5 3 5 5 3 2 5 0.85 Note: H, high risk of bias (0 to <3); L, low risk of bias (6–8); M, moderate risk of bias (3 to <6); N, no; U, unclear; Y, yes. Synthesis of the Results

The proportion of exposure to CHE in the group with health insurance was 5.7 percent and in the group without health insurance was 8.6 percent. The 13 studies yielded a non-significant pooled OR of 0.93 (95% confidence interval [CI], 0.68–1.28), suggesting that the exposure to CHE was not significantly different between people with insurance and people without insurance (Figure 2).

Meta-Analysis of Exposure to Catastrophic Health Expenditures in People With Insurance and People Without Insurance.

The overall proportion of exposure to CHE in people with insurance and people without insurance based on the type of insurance was not different between two groups of insurance and the protective effect of these types of insurances was statistically insignificant (OR: 0.97; 95% CI, 0.7–1.34, p < .83 for basic insurance, and 0.78; 95% CI, 0.37–16.5, p < .52 for complementary insurance) (Figure 3).

Meta-Analysis of Exposure to Catastrophic Health Expenditures Based on Type of Insurance.

The target population for nine studies was households, for two studies was hospitalized patients and for three studies was patients with chronic diseases. The subgroup analyses indicated that the protective effect of health insurance was significant for the “hospitalized patients,” suggesting that hospitalized patients with health insurance are less likely to experience CHE (OR: 0.52; 95% CI, 0.37–0.75, p < .0004) subgroup but not for the “household” (OR: 0.9; 95% CI, 0.64–1.25, p < .52) and “patients with chronic diseases” (OR: 1.54; 95% CI, 0.26–9.2, p < .64) subgroups (Figure 4).

Meta-Analysis of Exposure to Catastrophic Health Expenditures Based on Target Population.

Discussion

In this study, we aimed to estimate a pooled estimate of the effect of health insurance in providing financial protection against CHE in Iran. This study is the first meta-analysis, to the best of our knowledge, to investigate the role of health insurance in protecting households/individuals from CHE in Iran. The findings from the meta-analysis indicated that the role of health insurance in protecting individuals from CHEs in Iran is limited. The pooled OR suggested that there was no significant difference between insured and uninsured people in facing catastrophic health costs, and therefore, it seems that individuals incurred financial hardship due to health care usage regardless of the existence of health insurance coverage.

A possible explanation for this is the way in which the Iran health system is financed. The main financing methods include general budgeting, social insurance, and household out-of-pocket expenses in Iran. It should be noted that the majority of health care is financed through OOP payments, which account for 55 percent of the total health-care costs. This form of financing can impose CHE on households, particularly among the poor (McIntyre, Thiede, Dahlgren, & Whitehead, 2006). In addition, the fee-for-service payment mechanism and high tariffs in Iran exacerbate financial problems stemming from OOP payments. A systematic review also showed that in some low- and middle-income countries such as Kenya, Bangladesh, Myanmar, Tanzania, and India the health-care financing system remains dependent on OOP payments. Such dependency represents an unsustainable way of financing health-care systems and is thus inadequate to ensure that disadvantaged users of the health-care services are protected financially (Azzani, Roslani, & Su, 2019).

Some previous studies in low and middle countries also have found partial, or no impact of health insurance on out-of-pocket and CHE depending on the structure and services offered by the scheme (Wagstaff et al., 2009; Buigut, Ettarh, & Amendah, 2015; Archarya et al., 2013 Chankova, Sulzbach, & Diop, 2008; Choi et al., 2016; Devadasan, Criel, Van Damme, Ranson, & Van der Stuyft,