View - Economic Research Forum
  • No tags were found...

View - Economic Research Forum

Income and the use of health care:an empirical study of Egypt and LebanonHeba ElgazzarOctober 27 th 2007Correspondence:LSE Health and Social CareLondon School of Economics and Political ScienceCowdray House, Portugal StreetLondon WC2A 2AETel: +44 (0) 20 7955 6617Fax: +44 (0) 20 7955 6131Email: and acknowledgement:This paper uses data from the World Health Organization’s Multi-Country Survey Study. The author isgrateful to Martin Knapp for valuable help and suggestions and to Nirmala Naidoo for facilitating useof the dataset. The views expressed are the author’s and do not reflect those of any institution.1

AbstractIn middle-income Arab countries such as Egypt and Lebanon, income-associated equity inhealth care remains an elusive policy objective due to a relatively high reliance on out-ofpocketpayments in financing care. This paper examines the effect of income on the use ofoutpatient and inpatient health care services in Egypt and Lebanon using econometric analysisof cross-sectional data from the World Health Organization. In light of noticeable differencesin income and public financing levels, these two countries serve as interesting case studies.Results suggest that Egyptian respondents were more likely to use health services, and inparticular outpatient services, than their Lebanese counterparts. Greater income-associatedinequity in out-of-pocket payments for health care was found in Lebanon, with paymentsexacerbating poverty levels substantially in both countries. Multivariate regression resultsindicate that income and insurance affect outpatient use more so than inpatient use and thatthis effect is more pronounced in Lebanon. Economic barriers to the use of health servicesare discussed within the context of health financing policy reforms aimed at improving equityin access to care.2

een observed in maternal and mental health (Keeler, 1992; Klavus and Häkkinen, 1996;Palmer et al, 2004; Ensor et al, 2005; Knapp et al, 2006). Furthermore, this effect appears tobe disproportionately greater for those at lower income levels, where out-of-pocket paymentscan dramatically impinge on overall living standards (Gertler et al, 1984; Litvack and Bodart,1993). Payments that account for at least five percent of household income have been shownto push people deeper into poverty throughout Asia (Grogan, 1995; O’Donnell et al, 2007; Luet al, 2007; van Doorslaer et al, 2007). Notably, insurance tends to reduce the share of thesepayments substantially while increasing utilization, particularly for poorer groups as seen inVietnam and Jordan (Wagstaff and Van Doorslaer, 2001; Jowett et al, 2003; Ekman, 2007).In general, it can be said that the higher the reliance on private financing, the more likely foreconomic barriers to induce horizontal inequity in utilization.The aim of this paper is to assess and compare the impact of income on utilization in Egyptand Lebanon. The central hypotheses that are tested empirically include: (1) whether thedegree of income-associated inequality in health and in out-of-pocket payments differsbetween countries, (2) whether the degree of impoverishment arising from out-of-pockets isrelegated to certain socioeconomic groups, and (3) whether the impact of income relative toinsurance differs for outpatient and inpatient utilization, controlling for various factors. Aquantitative approach was adopted by applying economic models for the demand for healthcare, using data from the 2001 Multi-Country Survey Study conducted by the World HealthOrganization. While income-associated equity has been examined in other world regions, thispaper contributes comparative evidence on economic barriers to the use of health servicesfrom two Arab countries.The paper is organized as follows. Section 2 presents the background on the regional context.Section 3 describes the methodology including the theoretical framework adopted, datasources, and analytic approach. Section 4 presents results of descriptive and bivariateanalyses of income-associated health inequality and impoverishment, and multivariateanalyses of the demand for health care. The paper concludes in section 5 with a discussion ofresults and policy implications towards improved equity.2. Regional contextEgypt and Lebanon have been chosen as case studies due to overall social similarities,noticeable differences in public financing and income, and the availability of reliable datafrom a single source. Appendix I displays key macroeconomic and health system indicatorsfor both countries as background. Overall, the epidemiologic profile and health status4

indicators are similar for both countries as shown in table 2 (Akala and El-Saharty, 2006),despite a higher literacy rate, per capita income, density of health care providers, andproportion of gross domestic product spent on health care in Lebanon (table 3). Bothcountries are also marked by a generally limited political and economic discourse on healthpolicy, with health care dominated by curative, specialist care (Karshenas and Moghadam,2005; Jabbour et al, 2006; Maziak, 2006). However, the nature of the welfare state and healthfinancing differ to some extent between both countries, as described in the following sections.The Egyptian state finances and provides health care through a citizenship-based welfarestate, owing to a constitutional commitment to social solidarity and universal coverage(Constitution, 1923/1952). At the same time, the system is a labyrinth of parallel operatingsystems, overseen by the Ministry of Health and Population (MOHP). Funding derives fromfive main sources, including the MOHP, the General Authority for Health Insurance/HealthInsurance Organization (HIO), specific ministries such as Transport and Defense, nongovernmentalorganizations, and households. Private health insurance covers only onepercent of the population (UNDP, 2005). In 1995, the majority of total health expenditurewas out-of-pocket (55%), followed by general revenue sources (25%) and social healthinsurance-based (SHI) financing (20%) (Rannan-Eliya et al, 2000).The MOHP is a general revenue-based service provider offering free care to any citizen,covering all inpatient and outpatient care. However, quality is generally perceived poorly,particularly for outpatient services (Rannan-Eliya et al, 2000; Salem, 2002). The HIO is asocial health insurance scheme financed by premiums and employer contributions, operatingunder the auspices of the MOHP. Eligibility is open to those employed in the formal sector,pensioners and enrolled students, excluding dependents, with half of the population enrolledin total (Abd El Fattah et al, 1997; Nandakumar et al, 1999; Salem, 2002). Services areprovided both through HIO facilities and external contracts, with coverage benefits varyingacross beneficiary categories.Previous research in Egypt suggests that there exists income-associated inequality in serviceutilization. Although 95% of the population is within five kilometers of a medical facility,most public outpatient services are inadequately equipped and staff poorly motivated (Salem,2002; El-Zanaty and Way, 2006). As a result, seventy percent of outpatient care is obtainedprivately, largely by the better-off, while poorer groups who cannot afford private care rely onthe MOHP (Ellis et al, 1994). Overall, the wealthy report up to four times as many outpatientvisits and one and a half times the length of inpatient stays as the poor (Rannan-Eliya et al,2000).5

different determinants was examined quantitatively by applying Grossman’s (1972) model forthe demand for health, elaborated below. Anderson’s framework posits the choice to usehealth services as a function of three groups of factors, including ‘predisposing factors’, ordemographic characteristics, social structure and health beliefs; ‘enabling factors’, or theability to secure resources; and ‘medical need’, perceived or evaluated. Importantly, theframework also models the relative importance of factors as a function of how discretionarythe service is perceived to be. For services generally considered more vital such as curative,inpatient care, medical need is typically the prime determinant. By contrast, servicesperceived as more discretionary such as preventative, primary outpatient care are moreheavily influenced by enabling factors, like income and insurance (Anderson and Newman,1973).Grossman’s economic model of the demand for health and derivations thereof are based on ahuman capital approach, where health is a matter of investment (Mushkin, 1962; VanDoorslaer, 1987). The model stems from classical production theory where health (output) isa function of investment (labor) and an initial stock of health (capital) (Estrin and Laidler,1995). The choice to invest in health is posited within a utility maximizing function underconstraints, where an individual’s total utility is a function of the utility derived from theirinherited initial ‘stock of health’, the ‘stock of health’ over time, and the consumption of othercommodities. The stock of health in the ith time (H i ) is subject to the amount of investment inrelated goods in the period immediately preceding it, expressed as:H i+1 - H i = I i - δ i H i , (1)I i = I i (M i , TH i ; E i ) (2)where I i is gross investment in health, δ i is the rate of depreciation during the ith period, M i ismedical care, TH i is time input, and E i is the stock of human capital (Grossman, 1972).In effect, the demand for health is only observable through the demand for health care givenits latent properties (Culyer, 1976; Williams, 1978). Following this logic, the demand forhealth care, M i , can be expressed as a function of social determinants in line with Anderson’sframework:M i = β i X i + c i H i + v, (3)where X is a vector of observed characteristics, i.e., predisposing and enabling factors, H irepresents stock of health/medical need, β and c represent parameter estimates, and v7

epresents unobserved disturbance terms (Van de Ven and Van der Gaag, 1982; Wolfe andBehrman, 1984). Depending on the role of enabling factors such as ability to pay, asindicated by income, the relative importance of socioeconomic status in securing healthservices and overall equity implications can thus be evaluated.3.2. Data sources and variable specificationTable 4 in appendix II summarizes the data and variables used in the analysis. Data for Egyptand Lebanon were obtained from the 2001 Multi-Country Survey Study on health systemresponsiveness, expenditure, and utilization, conducted by the World Health Organization(Üstün et al, 2003). The survey has been applied to research on bias and disparities in selfreportedhealth (Bago et al 2006) and catastrophic expenditures and poverty (Van Doorslaeret al 2006; Van Doorslaer et al, 2007; Jones et al 2007). It was placed in over sixty countriesand is cross-sectional in nature, with approximately 3,500 – 5,000 respondents per country.The sample size for Egypt is n = 4,490 and for Lebanon, n= 3,246.The survey was conducted entirely by face-to-face, telephone, or computer-assistedinterviews using a standardized questionnaire. Samples consist of men and women aged 18years or older, non-institutionalized and living in private households. Multi-stage, stratifiedrandom cluster sampling procedures were used to identify eligible respondents, with samplesde facto representative of the target populations. One respondent per household wasrandomly selected using Kish tables by which an equal probability of selection is given toeach eligible individual within a household (Üstün et al, 2003). To ensure the samplesrepresented the populations from which they were drawn, age and sex distributions werecompared to United Nations population data. Deviations were measured using the SamplePopulation Deviation Index and adjusted where necessary by applying sample and poststratificationweights to the data prior to release for analysis (Üstün et al, 2003).The survey is a module-based instrument which captures a combination of household- andindividual-level information. Household-level information includes setting (i.e., urban orrural dwelling), roster, income, and expenditure. Individual-level information includesdemographic characteristics, health insurance, health state descriptions, and the prevalence ofchronic health conditions. Demographic characteristics include age, gender, marital status,education, employment status, and employment type.Health status was determined by assessing the proportion of respondents who were diagnosedor had suffered from a chronic health condition(s) over the past year, in addition to8

espondents’ ratings on self-assessed health status. Self-assessed health represents a usefulpolicy tool as it is a measure that encompasses functionality and perceptions, and has beencorrelated to future changes in health and mortality (Zimmer et al, 2000). The self-assessedhealth module is based on the WHO International Classification of Functioning, Disabilityand Health (ICF) (WHO, 2002). The health states that were measured are global healthstatus, pain, mobility, self-care, cognition, interpersonal activities, vision, sleep, and affect.For each state, respondents were asked to rate their health on an ordinal scale, withdescriptions shown in appendix III for indicators used in this analysis.The proportion of respondents who had used outpatient and inpatient health services over thepast year was collected. Inpatient care was defined as an overnight stay at a health care centreor hospital. Outpatient care was defined as ‘any place outside your home where you did notstay overnight’ such as a doctor’s consulting room, clinic, or hospital outpatient unit. Otherinformation that was collected includes whether access to care was refused due to lack ofaffordability and whether the respondent did not seek care due to lack of affordability.Expenditure data collected includes household consumption for selected categories andincome. Measures of household expenditure are the amount spent on accommodation andfood over the last month and on health care over the last year. The amounts spent oninsurance, medications, visits to doctors, or ‘other’ expenditures were captured. Whereavailable, the use of asset information is generally considered preferable over the use ofreported income as a proxy for households’ living standards, due to measurement error andsystematic reporting biases associated with the latter (Ferguson et al, 2003; Roy and Howard,2007). However, as insufficient information on permanent income indicators was available,this analysis uses annual household earnings income as a proxy for household livingstandards.Missing data on income was detected for approximately 17% of the observations from Egyptand 30% of those from Lebanon. In order to preserve the most observations as possible forthe analysis, missing values were imputed using multivariate imputation for a single variableof interest, recognized as a method to address missing data on income of a similar magnitude(Schenker et al, 2006; Horton and Kleinman, 2007).In order to allow for cross-national comparisons, expenditure and income data were convertedto international dollars using 2001 purchasing power parities and adjusted for householdcomposition (UN Stats, 2007). Adjustments for household composition and size were madeby applying an adult equivalency (AE) scale, defined as: AE = (A + αK) θ , where A represents9

the number of adults in the household, K the number of children, α the cost of childrenrelative to adults, and θ the degree of economies of scale for an average household, rangingfrom 0 to 1 (Deaton and Zaidi, 2002). For developing countries, values for α typically varybetween 0.3-0.5, with values for θ approaching 1 since food is a relatively large share ofoverall consumption (Deaton and Zaidi, 2002; Wagstaff et al, 2003). In this analysis, valuesused in assessing health equity in Asia were adopted, i.e., α = 0.5 and θ = 0.75 (Deaton, 1997;Equitap, 2002).3.3. Descriptive analysesSample characteristics were evaluated for all main variables of interest using two-tailed testsof significance where appropriate. Bivariate analyses were used to examine the relationshipbetween income and health status, insurance coverage, and out-of-pocket payments. Incomeassociatedinequity in the incidence of catastrophic expenditures was assessed using methodsdeveloped by van Doorslaer and Wagstaff (1992) and others (Wagstaff et al, 1991; Wagstaffet al, 2003). To quantify inequity, a concentration index for each country was constructed.The concentration index for catastrophic expenditures represents twice the difference betweenthe line of equality and the concentration curve of catastrophic expenditures, defined as thecumulative rate of the incidence of catastrophic expenditures against the cumulative levels ofincome. Values equal to zero suggest an income-neutral distribution, values less than zerodenote a pro-poor distribution, and values greater than zero denote a pro-rich distribution.The threshold for catastrophic health expenditures is typically set at a value between 5-25% oftotal household consumption, expenditure, or income (Wagstaff et al, 2003). For purposes ofthis analysis, the threshold was set at ten percent, where any amount of out-of-pocketpayments as a share of income that exceeds this threshold is considered catastrophic.In order to assess the impact of out-of-pocket payments on poverty, the poverty headcountwas estimated for each country before and after accounting for out-of-pocket payments forhealth care (Wagstaff et al, 2003). Measures of poverty do not typically take into accounthealth care-related expenditures (Van Doorslaer et al, 2006). The poverty headcount indicatesthe percent of households that fall below the poverty line, defined in relative or absoluteterms. Three analyses were conducted using a relative poverty line and two absolute povertylines. The relative poverty line was set equal to one-third the mean national income per capita(Wagstaff et al, 2003). Absolute poverty lines were set according to international povertylines estimated by the World Bank, although other approaches by which to measure povertyexist as described elsewhere (Coudouel et al, 2002; Wagstaff et al, 2003). Analysis wasconducted using poverty lines of 1993 US$1.08 and US$2.15 per capita per day inflated to10

2001 values using average annual consumer price indices for each country (UN Stats, 2007).Generally, the lower poverty line is used for low- or lower-middle income countries such asEgypt, with the higher line used for upper-middle income countries such as Lebanon(Coudouel et al, 2002). Using the dollar-a-day poverty line, conversions produce an Egyptianpoverty line of US$49.22 per capita per month and a Lebanese poverty line of US$10.23 percapita per month at 2001 prices.3.4. Multivariate analysis and econometric considerationsIn order to assess the determinants of health care utilization, general probit and recursivebivariate probit regression models were applied with variables described in appendix II, table4. The use of regression analysis is a recognized tool for evaluating the correlation betweenan individual factor on an outcome controlling for other factors, and the relative relationshipsbetween those factors (Dougherty, 2007). In conjunction with other econometric techniques,it proves valuable in helping to draw causal inferences from cross-sectional data (Winship andMorgan, 1999). Multivariate regression models have been used to address similar questionswithin the field of health services research (Zimmer et al, 2000; Erbsland et al, 2002;Hotchkiss et al, 2007; Roy and Howard, 2007).In particular, binary regression models are useful in estimating the probability of an eventoccurring by assuming either a logistic or normal distribution of the error terms, giving rise toa logit or probit model, respectively (Greene, 1997). The choice between the two is generallyarbitrary as both tend to yield equally satisfactory results in most cases (Amemiya, 1985).The binary regression relationship is denoted bywherey i * = β i ′ x i + u i , i = 1, 2, …, n, (4)y = 1 if y i * > 0 andy = 0 if otherwise.To ensure the reliability of parameter estimates, it is important to test for possible endogeneityamong explanatory factors. Endogeneity is said to occur if the error term of the independentvariables determine the value of the dependent variable, biasing the estimated effects of theindependent variables on the outcome (Waters, 1999; Dougherty, 2007). In this case, healthinsurance was suspected of being endogenous to the model since the likelihood of havinginsurance may be influenced by unobservable factors, such as the decision of where to live,the type of employment sought, or behavioral attitudes towards risk (Waters, 1999; Ekman,11

2007). To test for endogeneity of the health insurance dummy variable, the recursivebivariate probit model was adopted given its utility in similar analyses elsewhere (Waters,1999; Fabbri et al, 2004; Jones et al, 2004; Jones, 2007). Elaborated in the literatureregarding simultaneous equations, the general two-equations model takes the form:y 1i * = β 1i ′ y 2i + γ 1i ′ x 1i + u 1i (5)y 2i * = β 2i ′ y 1i + γ 2i ′ x 2i + u 2i , (6)where the main outcome of interest, y 1i *,and the ‘treatment’ or endogenous variable, y 2i *, aresimultaneously determined (Maddala, 1983). In this analysis, the main outcome of interest,y 1i *, represents the likelihood of utilization, while the endogenous variable, y 2i *, representswhether or not the individual is covered by health insurance. The effect of insurance alongincome level was explored further by including an interaction term for insurance-income.A description of the dependent and independent variables for the main utilization model isshown in appendix II, table 3. Separate models were developed for outpatient (model I) andinpatient service utilization (model II) for each country. Separate versions for each type ofservice were run by excluding (version a) or including (version b) the insurance-incomeinteraction term. The criteria used to select variables for inclusion in the model were basedon the distribution of responses, previous knowledge of the predictive value of a variable, andthe correlation between variables. Regression models were developed using a forwardstrategy based on the likelihood ratio test (criterion 0.05). Estimates include probit indices,robust standard errors, p-values and 95% confidence intervals. All analyses employed twotailedsignificance levels of p < 0.05 and were conducted using Stata 8 SE (StataCorp LP,College Station, Texas, USA). The probit indices represent the extent to which covariatessuch as income and insurance determine utilization. These covariates signify potentialenabling factors and policy levers for health care reform.4. ResultsResults are presented in appendix III for three sets of quantitative analyses. The first sectionpresents sample characteristics for each country sample. The second section presents findingsfrom bivariate analysis between socioeconomic status and key variables. Finally, the thirdsection presents the econometric estimates from multivariate regression models for serviceutilization.12

4.1. Sample characteristicsTable 5 presents descriptive statistics for each sample, with tests of significance shown whereapplicable. Mean age and gender distribution were similar between countries, with anaverage age of 39-42 years and males constituting half of all respondents. More respondentsin Lebanon had completed secondary education than had in Egypt (53% versus 40%,respectively). The Lebanese sample reported twice the mean per capita income as theEgyptian sample ($2989 versus $1579, respectively) (table 8). These figures are relativelysimilar to published 2001 estimates for gross national income per capita of $4010 and $1530,respectively (World Bank, 2001). The employment rate in both samples was only 43-44%,which includes formal and informal employment. Half of Egyptian respondents lived inurban settings, with no information available on setting for the Lebanese sample.Regarding health status, the Lebanese sample reported a relatively lower health status onaverage than did the Egyptian sample. For example, twice as many respondents in Lebanonas in Egypt reported a ‘bad’ or ‘very bad’ self-assessed health status (8.4% versus. 4.5%,respectively). The same was found for ‘severe’ or ‘extreme’ pain (8.3% versus 4.7%,respectively). Similarly, although the aggregate incidence of chronic health conditions wassimilar between samples, with 57% and 58% reporting at least one chronic health condition inEgypt and Lebanon, respectively, a closer look reveals important differences. Table 6 showsthe incidence of disaggregate chronic health conditions. The incidence of several individualconditions was higher in the Lebanese sample, such as heart disease, high blood pressure,asthma, back problems, depression, and vision problems. At the same time, the incidence ofarthritis and ‘other’ conditions was higher among Egyptian respondents. Conditionsclassified as ‘other’ were specified in the Lebanese sample only and include conditions suchas allergy, anemia, bone disorders, diabetes, thyroid problems, kidney disease,uterine/prostate problems, and rheumatism.Regarding health insurance coverage and out-of-pocket payments, the majority of respondentsin both samples were not covered by any health insurance, as shown in table 7. 47% ofLebanese respondents had some form of insurance, compared to only 32% in the Egyptiansample. It is important to note that in the Arabic-administered version of the questionnaire,health insurance was distinguished from free care, such that these figures should captureenrolment in social or private health insurance exclusively. These figures are relativelysimilar to published data showing that social health insurance covers 45.9% of the Lebanesepopulation and 45% of the Egyptian population (WHO, 2006a; WHO, 2006b). At the sametime, Lebanese respondents reported higher average out-of-pocket payments than Egyptian13

espondents as shown in table 9, which may reflect both the degree of financial risk as well asinherent differences in the price of health care.Importantly, more respondents in Egypt than in Lebanon utilized health services as shown intable 10 (61% versus 42%, respectively). Of those who sought care, outpatient services weremore frequently utilized in Egypt than in Lebanon (56% versus 35%, respectively), while lessinpatient care was sought in Egypt than in Lebanon (6% versus 13%, respectively) (table 5).More respondents in Lebanon reported foregoing care due to affordability (25% versus 12%,respectively) (table 10).4.2. Results of bivariate analysis: need, out-of-pocket, and poverty levels bysocioeconomic statusOverall, the distribution of self-assessed health, pain, and chronic conditions bysocioeconomic status appears more pro-poor amongst Lebanese respondents than theirEgyptian counterparts, shown in tables 11 through 15. This finding suggests that healthinequality is more pronounced in the Lebanese sample. In general, the richest income quintileis three times more likely to be covered by health insurance than the poorest quintile in bothcountries (tables 16 and 17). Similarly, the distribution of out-of-pocket expenditure as ashare of income appears more regressive in the Lebanese sample than in the Egyptian one, asdoes the incidence of catastrophic payments (table 18). The concentration index for theincidence of catastrophic payments is more negative for Lebanon than for Egypt, affirming arelatively higher degree of regressivity in out-of-pocket payments for Lebanon.As far as the change in poverty headcount after accounting for out-of-pocket payments, tables19 and 20 show that the post-payment incidence of poverty increases in both countries,regardless of which poverty line is used. Before accounting for out-of-pocket payments,poverty headcount at a dollar-a-day was estimated as 7.5% and 0.4% for Egypt and Lebanon,respectively, which are relatively similar to published data. Available data for Egypt from theWorld Bank (2007) suggests the poverty headcount was 2.58% in 1995 and 3.08% in 2000,while Jolliffe et al (2004) estimate the headcount was 7.6% in 1997. At a poverty line of $1per day for Egypt, twice as many households fall below the poverty line after accounting forout-of-pocket payments; at a poverty line of $2 per day for Lebanon, approximately fourtimes as many households do.In addition, while this effect is largely driven by the poorest income quintile in Egypt, somehouseholds in higher income groups are also driven into poverty in Lebanon. Using the14

elative poverty line for Lebanon, 2%, 10%, and 41% more households in the middle, secondpoorest, and poorest income quintiles, respectively, fall below the poverty line after paymentsare accounted for. Overall, validating poverty rates presents as a challenge due to the relativepaucity of data. However, this analysis suggests that the relative impact of out-of-pocketpayments on living standards can be substantial, and that this effect is largely relegated to thepoorest groups but can be felt to some extent across the socioeconomic spectrum.4.3. Results of multivariate analysis: outpatient and inpatient utilizationEstimation results from multivariate regression analyses for inpatient and outpatientutilization are presented in tables 21 through 24. The full output for the recursive bivariateprobit model is displayed only for the base outpatient model for brevity. Parameter estimatesfor utilization, but not for insurance, are shown for the remainder of the models. In testing thesuspected endogeneity of health insurance, as in table 21 for example, results of the Wald testshow that the chi-squared statistic in the Egyptian model does not differ significantly fromzero, indicating that insurance is not endogenous to the model. By contrast, insurance wasendogenous in the Lebanese model. These results were consistent across all country-specificmodels. As such, the standard probit model was run with results compared to the recursivebivariate probit models. Parameter estimates for the standard probit Egyptian models and therecursive Lebanese models are displayed in the tables. Interestingly, the determinants andtheir relative importance for both outpatient and inpatient insurance were similar in bothcountries. The most important factors determining insurance were a higher educational leveland being employed, followed by income and gender, with age playing a relatively minorrole. Health status covariates were tested but were not found to be significant determinants(results not shown).Table 21 presents the estimation results of the base model for the probability of outpatientutilization in each country. Overall, health status is the most important determinant ofoutpatient use, as indicated by the variables for self-assessed health, pain, and the presence ofchronic health conditions. Insurance significantly increases the probability of visitingoutpatient providers in both settings, although this effect is greater in Lebanon as indicated bythe larger probit index, as compared to the Egyptian model. Being married is positivelyassociated with utilization in both settings. While people seek more care as they grow olderin Egypt, they tend to use less in Lebanon. This could reflect differences in social healthinsurance coverage policies since pensioners in Lebanon are typically not eligible forcoverage. In addition, while being employed does not have an effect on utilization in Egypt,it decreases the probability of using services in Lebanon.15

Table 22 shows the same model with the addition of the insurance-income interaction term.The probit index for this variable is significant only in Egypt, where the negative valuesuggests that as income increases insurance becomes gradually less important in explainingoutpatient use. In both countries, including the interaction term also increases the values ofthe probit indices for insurance, showing that the base model underestimates its effect onseeking care. The effects of most of the other explanatory variables remain largely the same.Estimation results from the base model for the probability of using inpatient care arepresented in table 23. Health status continues to be the most important factor positivelydetermining inpatient use. The effect of insurance on inpatient use is less than its effect onoutpatient use in both countries. In Egypt, insurance is not a significant determinant ofinpatient use, while in Lebanon the absolute value of the probit index is smaller than that inthe outpatient model. However, its negative value suggests that having insurance decreasesthe probability of seeking care in Lebanon. By the same token, income positively increasesthe chances of seeking inpatient care there, but has no effect in Egypt. It is important to notethat most Lebanese social and private health insurance schemes do not cover major inpatientcare, instead referring patients to the MOHP.After including the insurance-income interaction term as shown in table 24, results from themodel indicate that the effect of insurance in Lebanon disappears. However, incomeelasticityof demand for inpatient care increases, as shown by the larger probit index ascompared to that in the base model. Estimates for the other explanatory variables do notchange substantially.Overall, results from multivariate analysis of utilization demonstrate that the demand foroutpatient care is generally more influenced by insurance and other socioeconomic variablesthan is the demand for inpatient care, as expected. Secondly, insurance-elasticity of demandis greater in Lebanon than in Egypt for both types of health care. In addition, while incomewas not a direct determinant in the outpatient models or in the Egyptian inpatient models,income-elasticity of demand was significant in Lebanon in both the base and interactionmodels. Finally, while insurance strongly influences the probability of using services, the factthat income and employment determine the likelihood of being covered by insurance areimportant findings.16

5. Discussion and Policy ImplicationsThe aim of this paper was to evaluate the impact of income on access to different healthservices in Egypt and Lebanon. First, bivariate methods were used to explore the relationshipbetween income and health status, insurance coverage, and out-of-pocket payments. Next,income-associated equity in out-of-pocket payments and the effect of payments on povertywere quantified. Finally, multivariate econometric methods were employed to robustly assessthe determinants of utilization while controlling for important factors. Results suggest thatthe degree of income-associated inequity in health care differs between Egypt and Lebanon,as does the role of income and insurance in explaining utilization. A discussion of results andtheir relevance to policy formation follows.5.1. Overall findingsResults from descriptive analyses reveal interesting yet seemingly paradoxical results. In thisanalysis, the Egyptian sample was on average relatively less wealthy, somewhat lesseducated, and reported less ill health, but was still more likely to access health services ascompared to the Lebanese sample. A greater degree of inequity was observed in Lebanonwith regard to the distribution of ill health and out-of-of pocket payments. Importantly, morehouseholds in Lebanon were found to be at risk of facing catastrophic payments. Herein liesthe paradox; that the distribution of health and payments in a relatively wealthier society canbe more heavily skewed pro-poor than that in a relatively poorer one, and not the reverse.These results however are consistent with those found in the US, European, Latin American,and Asian contexts.Furthermore, the percentage changes in poverty levels for Egypt and Lebanon appearrelatively large as compared to findings from other countries, suggesting that the impact ofout-of-pocket payments on living standards is particularly high. For example, the percentagechange in headcount appears to be lower in Vietnam (pre-payment headcount of 3.6% versuspost-payment headcount of 4.7%, respectively; change of 30%), Indonesia (7.9% versus8.6%, respectively; change of 8.7%), and Sri Lanka (3.8% versus 4.1%, respectively; changeof 8.3%) as reported by Van Doorslaer et al (2006). These findings are likely due todifferences in health financing policy and utilization practices, among other considerations.Interestingly, the observation that more respondents in Egypt utilized services was apparentlydriven by their greater propensity to use outpatient care than their Lebanese counterparts. Theincreased rate of inpatient use amongst the Lebanese sample was not high enough to17

compensate for their lower rate of outpatient use. The supply of services may not fullyexplain this result; if anything, there appear to be more health care providers per capita overallin Lebanon than in Egypt (table 2).Upon closer investigation through multivariate regression analysis, income-elasticity andinsurance-elasticity of demand for outpatient services were greater in Lebanon than that inEgypt. This finding may be explained by country differences regarding the nature ofinsurance coverage and incentives to seek certain types of care over others. While theEgyptian state health system provides free outpatient and inpatient care, outpatient care isoften excluded from coverage or is associated with high co-payment rates in Lebanon. Theseempirical results suggest that the incentive to seek outpatient care in Lebanon is relativelylow. Instead, most individuals seek curative, inpatient care which is generally covered by theMOHP, possibly even in cases where such care is neither medically necessary nor particularlycost-effective.Coupled with findings discussed above, these results imply two general conclusions. First, ahigher income alone does not necessarily increase the use of health services, but that someform of risk-pooling or insurance is important in mitigating against the effect of economicbarriers. Secondly, improving income-associated equity in outpatient utilization can go a longway towards improving overall equity in access, assuming that this care meets medical needs.5.2. Methodological considerationsThe methodological issues and limitations of this study which may have impacted the resultsinclude the following. While a variety of methods to measure living standards and povertyexist, a full exploration of these methods was outside the scope of this paper. It is possiblethat the approaches taken for this analysis may have impacted absolute figures to an extent,although not necessarily the relative differences between them. This area represents animportant avenue for future research. In addition, as ratings for self-assessed health statusare subject to differences in interpretation across countries (King et al, 2003), the ratings ineach country may not necessarily be comparable, such that ‘bad’ in one setting may mean‘moderate’ in the other, for example. However, in light of the relatively similar culturalframework and notions regarding the sick role (Gallagher, 2001; Adib, 2004), potential biasarising from this limitation was considered relatively low.Despite limitations in drawing causal inferences from cross-sectional data, the use ofrecognized econometric techniques in this analysis has helped to overcome such limitations to18

a great extent. Although debate is found regarding the handling of sample selection bias andmissing data, standard techniques were adopted for purposes of this analysis. Notably, thesuspected endogeneity of the health insurance variable was tested using the recursive bivariateprobit regression model together with robust estimates of standard errors. Results indicatedthat endogeneity was present only in the Lebanese models, suggesting that standard probitestimates were reliable for the Egyptian data. While the relationship between factors maydiffer from country to country which may have biased results (Nord, 2002), when comparingtwo reasonably similar contexts this issue was considered less of a confounding factor.Future research may help to illuminate further methodological explorations.5.3. ConclusionsIn the face of an increasing demand for health services due to epidemiological transitions,governments face the question of how best to design health financing policies in order toensure access to care and protect those most financially at risk. As Deaton (2002) points out,the question of whether, for example, income redistribution versus better policies are likely toimprove health is a prime question. While economic and political constraints limit some ofthe options that might otherwise be viable in middle-income countries like Egypt andLebanon, this analysis highlights at least three equity-related policy issues that should beconsidered in evaluating options for health financing policy reform.First, the greater degree of income-associated inequity observed in Lebanon as compared toEgypt should be addressed through the lens of health financing policy. This finding isconsistent with the general observation from other studies that wealthier countries that relylargely on market-based financing are associated with higher degrees of systemic unfairnessthan poorer countries that have at least some form of a comprehensive safety net for healthcare. Hence, income redistribution alone may not be sufficient in improving equity. Second,the nature of health insurance can lead to dramatic differences in its effect on using services.Benefits included and co-payment rates are likely to be critical factors that influenceutilization. Therefore, demand- and supply-side incentives to use outpatient care in particularshould be considered carefully in policy-setting. Third, the importance of insurance appearsto be greater for those at the lower ends of the income ladder, especially as these groups tendto face a higher burden of out-of-pocket payments and are more likely to be pushed furtherinto poverty. Yet insurance coverage tends to be more concentrated among the well-off,which only exacerbates inequity. Hence, eligibility criteria for risk-pooling schemesrepresent an important determinant of equity; although social values play a part in shaping19

criteria, nonetheless, the most vulnerable should at the very least be relieved of economicbarriers to health care.While a number of other factors should be considered in improving equity such as quality ofcare, the nature and distribution of supply, and social beliefs about curative versuspreventative care, this paper demonstrates that well-planned health financing policy is centralto the process of ensuring and sustaining equity. Providing protection from unforeseenadverse economic consequences of using services should reduce the likelihood that peopleforego medically-necessary care, otherwise plunging them into a vicious cycle of ill healthand poverty. In the future, more research is needed to distil whether the effect of income onthe use of services pertains to specific types of health insurance, regions, or providers, be theypublic or private. Overall, alleviating economic barriers to access to care through soundhealth financing policy reforms is likely to bring about substantial gains in equity throughouthealth systems in middle-income countries such as Egypt and Lebanon.20

ReferencesAbd El Fattah H, Saleh I, Ezzat E et al (1997). The health insurance organization of Egypt: ananalytical review and strategy for reform. Technical report no. 43. Bethesda, MD:Partnerships for Health Reform Project, Abt Associates Inc.Adeyi O, Smith O and Robles S (2007). Public policy and the challenge of chronicnoncommunicable diseases. Human Development Network. Washington DC: TheWorld Bank.Adib SM (2004). From the biomedical model to the Islamic alternative: a brief overview ofmedical practices in the contemporary Arab world. Social Science & Medicine 58:697-702.Akala AF and El-Saharty F (2006). Public health challenges in the Middle East and NorthAfrica. The Lancet 367:961-964.Amemiya T (1985). Advanced Econometrics. Oxford: Basil Blackwell, Ltd., pp.267-359.Ammar W (2003). Health system and reform in Lebanon. Cairo: Regional Office for theEastern Mediterranean, World Health Organization, and Beirut: EntrepriseUniversitaire d’Etudes et de Publications.Andersen R and Newman JF (1973). Societal and individual determinants of medical careutilization in the United States. The Milbank Memorial Fund Quarterly. Health andSociety 51(1):95-124.Anderson R (1975). ‘Introduction’, in Anderson R and Kravits J (eds.), Equity in healthservices. Cambridge: Ballinger Publishing Company.Anderson R and Benham L (1970). ‘Factors affecting the relationship between family incomeand medical care consumption,’ in Klarman H (ed.), Empirical studies in healtheconomics, Baltimore: Johns Hopkins Press.Arrow K (1963). Uncertainty and the welfare economics of medical care. The AmericanEconomic Review. 53(5):941-973.Bago d’Uva T, van Doorslaer E, Lindeboom M, O’Donnell O and Chatterji S (2006). Doesreporting heterogeneity bias the measurement of health disparities? University ofYork: Health Econometrics and Data Group (HEDG) Working Paper 05/02.Barnett JR (1984). Equity, access and resource allocation: planning hospital services in NewZealand. Social Science & Medicine 9(11):981-989.Barr NA (2004). Economics of the welfare state. Oxford: Oxford University Press.Bloom G (2001). Equity in health in unequal societies: meeting health needs in contexts ofsocial change. Health Policy 57:205–224Boutayeb A and Serghini M (2006). Health indicators and human development in the Arabregion. International Journal of Health Geographics 5:61. Available from: [Accessed 21 January 2007]Chernichovsky D (1995). What can developing economies learn from health system reformsof developed economies? Health Policy 32:79-91.Cockerham WC (2001). ‘Medical sociology and sociological theory’, in Cockerham WC(ed.), The Blackwell companion to medical sociology, Malden: Blackwell PublishersInc.Constitution of the Arab Republic of Egypt (1923/1952). Cairo: People’s Assembly.Available from: [Accessed 7 August 2007]Coudouel A Hentschel JS and Wodon QT (2002). ‘Poverty measurement and analysis’, InKlugman J (ed.), A sourcebook for poverty reduction strategies, Washington DC: TheWorld Bank.Culyer AJ (1976). Need and the national health service. London: Martin Robertson & Co.Ltd.Deaton A (1997). The analysis of household surveys. Baltimore: Johns Hopkins UniversityPress.Deaton A (2002). Policy implications of the gradient of health and wealth. An economistasks, would redistributing income improve population health? Health Affairs 21(2):13–30.21

Deaton A and Zaidi S (2002). Guidelines for constructing consumption aggregates. LSMSWorking Paper 135. Washington DC: The World Bank.Dougherty C (2007). Introduction to econometrics. 3 rd ed. Oxford: Oxford University Press,pp. 291-321.Egypt (2007). The World Factbook. Washington DC: US Central Intelligence Agency.Available from: [accessed 8 October 2007]Ekman B (2007). The impact of health insurance on outpatient utilization and expenditure:evidence from one middle-income country using national household survey data.Health Research Policy and Systems 5:6. Available from[Accessed 1 August 2007]Ellis RP, McInnes DK, and Stephenson EH (1994). Inpatient and outpatient health caredemand in Cairo, Egypt. Health Economics 3(3):183-200.El-Zanaty F and Way A (2006). Egypt Demographic and Health Survey 2005. Cairo: Ministryof Health and Population, National Population Council, El-Zanaty and Associates, andORC Macro.Ensor T and Ronoh J (2005). Effective financing of maternal health services: A review of theliterature. Health Policy 75:49–58.Equitap (2002). Equity in Asia-Pacific Health Systems. Empirical guidelines 1. Rotterdam:Erasmus University, and Colombo: Institute of Policy Studies.Erbsland M, Ried W and Ulrich V (2002). ‘Health, health care, and the environment:econometric evidence from German micro data’, in Jones AM and O’Donnell O (eds.),Econometric analysis of health data, Chichester: John Wiley & Sons, Ltd.Estrin S and Laidler D (1995). Introduction to microeconomics, 4 th ed. Hemel Hempstead:Prentice Hall/Harvester Wheatsheaf.Fabbri D, Monfardini C and Radice R (2004). Testing exogeneity in the bivariate probitmodel: Monte Carlo evidence and an application to health economics. Department ofEconomics, working paper no. 514. Bologna: University of Bologna.Ferguson BD, Tandon A, Gakidou E, and Murray CJ (2003). ‘Estimating permanent incomeusing indicator variables,’ in Murray CJ and Evans DB (eds.,), Health systemsperformance assessment: debates, methods and empiricism. Geneva: World HealthOrganization.Gallagher E (2001). Health, health care, and medical education in the Arab world, inCockerham WC (ed.), The Blackwell companion to medical sociology, Malden:Blackwell Publishers Inc.Gertler P, Locay L, Sanderson W (1987). Are user fees regressive? The welfare implicationsof health care financing proposals in Peru. Journal of Econometrics 36:67-88.Goodin, R (1988). ‘Reasons for welfare: economic, sociological, and political – butultimately moral’, in Moon DJ (ed.), Responsibility, rights, and welfare: the theory ofthe welfare state. Boulder and London: Westview Press.Greene W (1997). Econometric analysis. 3 rd ed. Upper Saddle River: Prentice-Hall, pp.871-894, 418-449.Grogan (1995). Urban economic reform and access to health care coverage in the People’sRepublic of China. Social Science & Medicine 41(8):1073-1084.Grossman M (1972). The demand for health: a theoretical and empirical investigation.Occasional paper 119. New York: National Bureau of Economic Research.Gulliford M et al (2002). What does `access to health care’ mean? Journal of HealthServices Research & Policy 7(3):186–188.Horton NJ and Kleinman KP (2007). Much ado about nothing: a comparison of missing datamethods and software to fit incomplete data regression models. The AmericanStatistician 61(1):79-90.Hotchkiss DR, Hutchinson PL, Malaj A and Berruti AA (2007). Out-of-pocket payments andutilization of health care services in Albania: evidence from three districts. HealthPolicy 75:18-39.Jabbour S, El-Zein A, Nuwayhid I and Giacaman R (2006). Can action on health achievepolitical and social reform? British Medical Journal 333(7573):837-839.22

Jolliffe D, Datt G and Sharma M (2004). Robust poverty and inequality measurement inEgypt: correcting for spatial-price variation and sample design effects. Review ofDevelopment Economics 8(4):557–572Jones AM (2007). Applied econometrics for health economists: a practical guide. 2 nd ed.Oxford: Radcliffe Medical Publishing.Jones AM, Koolman X, and van Doorslaer E (2004). The impact of supplementary privatehealth insurance on the use of specialists in European countries. ECuity III workingpaper no. 21. Rotterdam: Erasmus University.Jones AM, Rice N, Bago d’Uva T and Balia S (2007). ‘Bias in self-reported health’, in JonesAM, Rice N, Bago d’Uva T and Balia S (eds.), Applied health economics. London andNew York: Routledge.Jowett M, Contoyannisb P and Vinh ND (2003). The impact of public voluntary healthinsurance on private health expenditures in Vietnam. Social Science & Medicine56:333–342Jurjus AR (1995). Health financing in Lebanon: cost and insurance coverage. Beirut:Ministry of Public Health.Karshenas M and Moghadam VM (eds) (2005). Social Policy in the Middle East: Economic,Political and Gender Dynamics, Basingstoke: Palgrave Macmillan.Keeler EB (1992). Effects of cost sharing on use of medical care and health. Journal ofMedical Practice Management 8:317-321.King G, Murray CJ, Salomon JA and Tandon A (2004). Enhancing the validity and crosscultural comparability of measurement in survey research. American Political ScienceReview 98(1):191-207.Klavus J and Häkkinen U (1996). Health care and income distribution in Finland. HealthPolicy 38(1):31-43.Klein R and Millar J (1995). Do-it-yourself social policy: searching for a new paradigm?Social Policy & Administration 29(4):303-316.Knapp M, Funk M, Curran C, Prince M, Grigg M and McDaid, D (2006). Economic barriersto better mental health practice and policy. Health Policy and Planning 21(3):157-170.Kutzin J (2001). A descriptive framework for country-level analysis of health care financingarrangements. Health Policy 56:171–204.Le Grand J (1991). Quasi-markets and social policy. The Economic Journal 101: 1256-1267.Le Grand J, Robinson R and Propper C (1984). The Economics of Social Problems. London:Macmillan.Lebanon (2007). The World Factbook. Washington, DC: U.S. Central Intelligence Agency.Available from: [accessed 8 October 2007]Lerberghe WV, Ammar W, El Rashidi R et al (1997). Reform follows failure: I. unregulatedprivate care in Lebanon. Health Policy and Planning 12(4):296-311.Livack JI and Bodart C (1993). User fees plus quality equals improved access to health care:results of a field experiment in Cameroon. Social Science & Medicine 37(3):369-383.Lu JF, Leung GM, Kwon S, Tin KY, Van Doorslaer E and O'Donnell O (2007). Horizontalequity in health care utilization evidence from three high-income Asian economies.Social Science & Medicine 64(1):199-212.Maddala GS (1983). Limited-dependent and qualitative variables in econometrics.Cambridge: Cambridge University Press, pp. 22-32, 94-147.Marmot M (2005). Social determinants of health inequalities. Lancet 365(9464):1099-104.Maziak W (2006). Health in the Middle East. British Medical Journal 333(7573):815-816.Muennig P, Franks P, Haomiao J, Lubetkin E and Gold M (2005). The income-associatedburden of disease in the United States. Social Science & Medicine 61:2018-2026.Mushkin SJ (1962). Health as an investment. Journal of Political Economy 70(5):129-157.Nandakumar AK, Nada K, Ibrahim A et al (1999). Health expenditure review: Alexandria,Egypt. Technical report no. 35. Bethesda, MD: Partnerships for Health Reform Project,Abt Associates Inc.Nord E (2002). Measures of goal attainment and performance in the World Health Report2000: a brief, critical consumer guide. Health Policy 59:183-191.23

O’Donnell O, E van Doorslaer E, Rannan-Eliya R et al (2007). The incidence of publicspending on health care: comparative evidence from Asia. World Bank EconomicReview. Washington DC: The World Bank.Palmer N, Mueller DH, Gilson L, Mills A and Haines A (2004). Health financing to promoteaccess in low income settings--how much do we know? Lancet 364(9442):1365-1370.Phelps CE (1975). ‘Effects of insurance on demand for medical care’, in Anderson R, KravitsJ and Anderson W (eds.), Equity in health services: empirical analyses in social policy,Cambridge: Ballinger Publishing.Rannan-Eliya RP, Blanco-Vidal C and Nandakumar AK (2000). The distribution of healthcare resources in Egypt: implications for equity. Boston: Harvard School of PublicHealth, Data for Decision Making Project.Roy K and Howard DH (2007). Equity in out-of-pocket payments for hospital care: evidencefrom India. Health Policy 80:297-307.Salem MA (2002). Policy research in Egypt’s health sector reform. Geneva: The Alliancefor Health Policy, World Health Organization.Schenker N, Raghunathan TE, Chiu P-L et al (2007). Multiple imputation of missing incomedata in the National Health Interview Survey. Journal of the American StatisticalAssociation 101(475):924-933.Sen A (2001). Development as freedom. Oxford: Oxford University Press.Titmuss RM (1956/2000). ‘The social division of welfare: some reflections on the search forequity’, in Goodin RE and Mitchell D (eds.), The foundations of the welfare state,Volume I, Cheltenham: Edward Elgar Publishing.United Nations Development Program (2005). Egypt human development report. New York:United Nations Development Program and the Institute of National Planning, Egypt.United Nations Statistics Division (2007). Common Database (UNCDB), International LaborOrganization estimates for consumer price index (CPI). New York: United NationsStatistics Division. Available from [Accessed 22 Sept 2007]Üstün TB, Chatterji S, Villanueva M, Benib L, Celik C, Sadana R, Valentine NB, Ortiz JP,Tandon A, Salomon J, Yang C, Xie XJ, Ozaltin E, Mather CD and Murray CJL (2003).‘WHO multi-country survey study on health and responsiveness 2001-2’, in MurrayCJL and Evans DB (eds), Health systems performance assessment: debates, methodsand empiricism. Geneva: World Health Organization.Van de Ven WP and Van der Gaag J (1982). Health as an unobservable: a MIMIC-model ofdemand for health care. Journal of Health Economics 1:157-183.Van Doorslaer E and Wagstaff A (1992). Equity in the delivery of health care: someinternational comparisons. Journal of Health Economics 11:389-411.Van Doorslaer E, Masseria C and the OECD Health Equity Research Group Members (2004).Income-related inequality in the use of medical care in 21 OECD countries. OECDHealth working paper no. 14. Paris: Organisation for Economic Cooperation andDevelopment.Van Doorslaer E, O'Donnell O, Rannan-Eliya RP et al (2006). Effect of payments for healthcare on poverty estimates in 11 countries in Asia: an analysis of household survey data.Lancet 368(9544):1357-64.Van Doorslaer E, O'donnell O, Rannan-Eliya RP, et al (2007). Catastrophic payments forhealth care in Asia. Health Economics 16(11):1159-1184.Van Doorslaer EK (1987). Health, knowledge and the demand for medical care. Assen: VanGorcum.Wagstaff A and van Doorslaer E (2001). Paying for health care: quantifying fairness,catastrophe, and impoverishment, with applications to Vietnam 1993-98. PolicyResearch Working Paper No. 2715. Washington DC: The World Bank.Wagstaff A, van Doorslaer E, and Paci P (1991). On the measurement of horizontal inequityin the delivery of health care. Journal of Health Economics 10:169-205.Wagstaff A, van Doorslaer E, O’Donnell O and Lindelow M (2003). Quantitative techniquesfor health equity analysis: a series of 20 technical notes. Washington DC: The WorldBank.24

Walker GK (1986). Reforming Medicare: the limited framework of political discourse onequity and economy. Social Science & Medicine 13(12):1137-1130. 1986Ware JE, Rogers WH, Davies AR et al (1986). Comparison of health outcomes at a healthmaintenance organization with those of fee-for-service care. Lancet 327(8488):1017-1022.Waters HR (1999). Measuring the impact of health insurance with a correction for selectionbias – a case study of Ecuador. Health Economics 8:473–483Weber, M (1978). ‘Class and status’, in Roth G and Wittich C (eds.), Economy and Society:an outline of interpretive sociology, Vol.2. Berkeley: University of California Press.Williams A (1978/1991). ‘Need – an economic exegesis’, in Culyer AJ (ed.), The economicsof health, volume I. Aldershot: Edward Elgar Publishing.Winship C and Morgan SL (1999). The estimation of causal effects from observational data.Annual Review of Sociology 25:659-706.Wolfe BL and Behrman JR (1984). Determinants of women's health status and health-careutilization in a developing country: a latent variable approach. The Review ofEconomics and Statistics 66(4):696-703.World Bank (2001). World Development Indicators. Washington DC: The World BankGroup. Available from: [Accessed 27 July 2007].World Bank (2005). Country Assistance Strategy for the Republic of Lebanon. InternationalBank for Reconstruction and Development, report no. 34463-LB. Washington DC:The World Bank.World Bank (2007). World Development Indicators. Washington DC: The World BankGroup. Available from: [Accessed 27 July 2007].World Health Organization (2002). Towards a common language for functioning, disabilityand health: ICF, the International Classification of Functioning, Disability and Health.Geneva: World Health Organization. Available from: [Accessed 13 March 2007]World Health Organization (2005). Action on the social determinants of health: learningfrom previous experiences. Geneva: World Health Organization. Available from: [Accessed 15 May 2007]World Health Organization (2006a). Health system profile: Egypt. Cairo: Regional Officefor the Eastern Mediterranean, World Health Organization.World Health Organization (2006b). Health system profile: Lebanon. Cairo: RegionalOffice for the Eastern Mediterranean, World Health Organization.World Health Organization (2007). National Health Accounts. Geneva: World HealthOrganization. Available from: [Accessed 8 Oct 2007]Zimmer Z, Natividad J, Lin H-S and Chayovan N (2000). A cross-national examination ofthe determinants of self-assessed health. Journal of Health and Social Behavior41:465-481.25

Appendix I: Development indicatorsEgypt LebanonTotal population (in thousands) 74 000 4 050 (‘06)GNI per capita, Atlas method $1,260 $6,320Unemployment rate 2 10.3% (‘06) 20% (‘06)Population below poverty line 3 20% 28% (‘99)Table 2 Health status and supply indicatorsEgypt LebanonHealth status 1 aboveLife expectancy at birth, total (years) 71 72.5Mortality rate, infant (per 1 000 live births) 28 30Literacy rate, adult total (% of people ages 15 and above) 2 71 87.4 (‘03)Supply of health providers 4Physicians (density per 1 000 population) 0.54 (‘03) 3.25 (‘01)Nurses (density per 1 000 population) 1.98 (‘04) 1.18 (‘01)Dentists (density per 1 000 population) 0.14 (‘04) 1.21 (‘01)Pharmacists (density per 1 000 population) 0.1 (‘04) 0.95 (‘01)Hospital beds (per 10 000 population) 22 36Egypt LebanonI. Expenditure ratiosTotal expenditure on health (THE) as % GDP 6.1 11.2General government expenditure on health (GGHE) as % THE 38.4 28.3Private expenditure on health (PvtHE) as % THE 61.6 71.7General government expenditure (GGHE) on health as % THE 7.2 10.4Social security expenditure on health as % GGHE 28.6 34.8Out-of-pocket spending on health as % PvtHE 94.6 82.1Private prepaid plans expenditure on health as % PvtHE 0.3 15.4Externally funded expenditure on health as % THE 0.9 1.7II. Per capita levels ($)THE per capita at exchange rate 77 693THE per capita at international dollar rate 276 840GGHE per capita at exchange rate 30 196GGHE per capita at international dollar rate 106 237III. Expenditure by category (‘02) (‘03)Expenditure on inpatient care as % THE 20.7 30.6Expenditure on preventative and public health care as % THE 9.6 1.91 World Bank, 2007. Values shown for 2005 unless otherwise indicated.2 Egypt, Lebanon, 2007, CIA World Factbook (see References)3 Rates are based on national poverty lines for Egypt and Lebanon.4 World Health Organization, 2007. Values shown for 2005 unless otherwise indicated.26

Appendix II: Variable definitionsTable 4 Descriptions of variables used in the analysisVariable Type DefinitionDependent variables 5 :Outpt Binary Dummy variable =1 if the individual has visitedoutpatient provider over the past 12 months.Inpt Binary Dummy variable =1 if the individual has visitedinpatient provider over the past 12 months.Independent variables:Policy variables:Op_ins Binary Dummy variable = 1 if the individual has insurancecoverage for outpatient care.Ip_ins Binary Dummy variable = 1 if the individual has insurancecoverage for inpatient care.Demographic variables:Age Continuous Years of age.Male Binary Dummy variable =1 if the individual is male.Married Binary Dummy variable =1 if the individual is married.Educlev Binary Dummy variable =1 if individual has completedsecondary education or above, 0 if less than secondaryeducation or no formal education.Emplyd Binary Dummy variable =1 if individual is employed.Logincome Continuous Log of reported income or imputed income over thepast 12 months.Urban_i Binary Dummy variable =1 if the individual lives in an urbansetting, rural if otherwise. Data is available only forEgyptian sample.Health need variables:SAH Categorical Self-assessed health, i.e., ‘In general, how would yourate your health today?’ 1 = Very good; 2 = Good; 3 =Moderate; 4 = Bad; 5 = Very bad.Pain Categorical ‘Overall in the last 30 days, how much pain ordiscomfort did you have?’ 1 = None; 2 = Mild; 3 =Moderate; 4 = Severe; 5 = Extreme.Chronic_1 Binary Dummy variable =1 if the individual has suffered orhas been diagnosed with at least one chronic healthcondition over the past 12 months.Interaction terms:Logincome X Op_ins Interaction term Log of reported income by outpatient insurance.Logincome X Ip_ins Interaction term Log of reported income by outpatient insurance.5 For binary variables, reference category = 0.27

Appendix III: Data TablesSample CharacteristicsTable 5 Descriptive summary statistics by country 6EgyptLebanonN Mean N MeanDependent variables 7 :Outpt* 4490 59 3246 39Inpt* 4490 6 3246 13Independent variables:Op_ins* 4352 27 3223 47Ip_ins* 4352 25 3223 35Age* 4480 39 3220 42Male* 4479 56 3245 51Married* 4484 74 3236 63Educlev* 4480 40 3211 53Emplyd 4478 44 3218 43Logincome 8 4342 7.14 2899 7.75Urban_i 4401 55 n/a 9 n/aSAH 4484 -- 3243 --Very good 37.8 33.2Good 34.8 33.2Moderate 22.9 25.3Bad 3.4 6.8Very Bad1.11.6Pain 4482 -- 3233 --None 58.2 57.4Mild 23.2 19.7Moderate 13.9 14.5Severe 4.1 7.5Extreme0.60.8Chronic_1 4416 58.2 3174 57.1*Difference is significant with a 95% confidence interval.6 A two-tailed t–test was used to test for significant differences with a 95% confidence level betweenmeans for continuous variables; z-statistic was used to test differences between proportions for binaryvariables.7 For binary variables, percentages for the non-reference category are shown.8 Tests of significance were not applied to income data.9 Data was not available.28

Table 6 Prevalence of chronic health conditions by type and countryEgyptLebanonN % N %Arthritis/arthrosis* 4451 25 3195 12Heart disease/coronary disease/heart attack* 4444 4 3192 7Asthma* 4441 6 3191 8Depression/anxiety* 4444 7 3189 17Diabetes 4445 5 3192 5High blood pressure/hypertension* 4447 12 3192 14Chronic bronchitis* 4442 2 3191 3Back pain/disc problems* 4448 19 3196 22Migraine 4443 13 3191 13Stroke* 4442 0.4 3187 1Sleep problems* 4444 6 3190 12Hearing problems* 4445 3 3190 5Vision problems* 4446 11 3191 18Gastritis/ulcer 4443 16 3194 15Tumor/cancer 4436 .5 3180 .3Other* 4124 20 2668 14*Difference is significant with a 95% confidence interval.Table 7 Proportion of respondents with insurance coverage for different services by countryEgypt(%)Lebanon(%)N 4352 3223Outpatient only* 5 0.8Inpatient only* 7 13Both outpatient and inpatient* 20 34None* 68 53*Difference is significant with a 95% confidence interval.Table 8 Income per adult equivalent (mean and distribution) and public employment by countryEgypt LebanonMean ($) Mean ($)N 3703 1960Annual income per adult equivalent, mean (observed) 1579 2989Richest 20% 3795 78632 nd richest 1608 2916Middle 1224 20902 nd poorest 931 1484Poorest 20% 581 838N 4342 2899Annual income per adult equivalent, mean (imputed) 1580 3480Richest 20% 3910 88522 nd richest 1836 3762Middle 1321 24132 nd poorest 928 1624Poorest 20% 362 801N 4478 3218Government employee* 41% 14%*Difference is significant with a 95% confidence interval. Tests of significance were not applied toincome data.29

Table 9 Amount of out-of-payments for health care by type and countryEgyptLebanonN Mean ($) N Mean ($)Insurance, mean 2012 14 1149 61Medications, mean 4086 71 2811 158Doctor visits, mean 3708 20 2478 72Other, mean 183 192 658 127Total, mean (including medications and doctor visits) 3680 94 2445 235Table 10 Overall rates of health service utilization and foregone care by countryEgyptLebanonN % N %Visited any health care provider in last 12 months* 4466 61 3246 42Was refused care because could not afford* 4490 12 3246 5Did not seek care because could not afford* 4490 12 3246 20*Difference is significant with a 95% confidence interval.Bivariate analysis: Distribution of need, out-of-pocket payments, and poverty by socioeconomicstatusTable 11 Income-associated distribution of self-assessed health, Egypt% Very good % Good % Moderate % Bad % Very badRichest 20% 44.5 34.8 17.9 2.1 0.72 nd richest 37.2 33.3 25.3 3.3 0.9Middle 39.4 34.8 20.8 4.0 1.02 nd poorest 37.6 33.7 24.9 3.0 0.9Poorest 20% 23.3 34.8 27.4 4.2 1.2Total 37.9 34.2 23.6 3.3 0.9Table 12 Income-associated distribution of self-assessed health, Lebanon% Very good % Good % Moderate % Bad % Very badRichest 20% 40.9 39.1 17.1 2.6 0.22 nd richest 40.9 30.7 23.9 4.0 0.6Middle 35.3 35.3 24.2 4.7 0.52 nd poorest 26.0 34.5 29.6 8.0 1.9Poorest 20% 21.2 25.9 38.5 11.2 3.3Total 32.6 33.0 26.9 6.2 1.3Table 13 Income-associated distribution of pain, Egypt% None % Mild % Moderate % Severe % ExtremeRichest 20% 70.0 17.0 9.6 3.1 0.32 nd richest 56.2 22.5 17.4 3.4 0.5Middle 56.6 24.8 13.1 5.2 0.32 nd poorest 56.8 25.4 13.2 4.4 0.1Poorest 20% 53.2 26.9 15.0 4.0 0.8Total 58.3 23.5 13.8 4.0 0.430

Table 14 Income-associated distribution of pain, Lebanon% None % Mild % Moderate % Severe % ExtremeRichest 20% 66.6 18.6 12.2 2.1 0.52 nd richest 62.2 18.2 13.1 6.0 0.6Middle 59.0 18.2 15.4 7.2 0.22 nd poorest 52.2 20.7 17.4 9.1 0.6Poorest 20% 42.0 21.4 19.2 15.7 1.6Total 56.1 19.4 15.6 8.2 0.7Table 15 Income-associated distribution of chronic health conditions by countryEgypt(%)Lebanon(%)Richest 20% 48.1 50.52 nd richest 58.6 51.4Middle 61.8 56.22 nd poorest 60.2 60.8Poorest 20% 63.8 70.7Total 58.7 58.2Table 16 Income-associated distribution of insurance coverage, EgyptInpatient Coverage (%) Outpatient coverage (%) Any coverage (%)Richest 20% 41.8 46.4 54.72 nd richest 33.5 25.2 37.1Middle 27.4 21.8 30.72 nd poorest 22.2 18.6 26.0Poorest 20% 17.6 15.8 20.4Total 28.1 25.2 33.3Table 17 Income-associated distribution of insurance coverage, LebanonInpatient Coverage (%) Outpatient coverage (%) Any coverage (%)Richest 20% 65.7 49.3 66.82 nd richest 63.5 52.0 64.7Middle 47.4 40.0 47.92 nd poorest 34.2 29.2 34.5Poorest 20% 24.5 18.9 24.7Total 46.4 37.3 47.031

Table 18 Income-associated distribution of out-of-pocket payments and catastrophic payments bycountryEgyptLebanonMean out-of-pocket payments per adult equivalent ($)Richest 20% 134 2852 nd richest 109 236Middle 87 2342 nd poorest 81 225Poorest 20% 59 213Total 93 238Out-of-pocket payments as a share of income (%)Richest 20% 4.7 5.62 nd richest 6.8 8.2Middle 7.2 11.32 nd poorest 8.6 15.3Poorest 20% 13.8 43.9Total 8.5 16.9Incidence of catastrophic payments 10 (%)Richest 20% 9.7 12.62 nd richest 17.2 23.7Middle 16.6 40.52 nd poorest 22.8 48.8Poorest 20% 34.0 64.7Total 20.9 38.3Rank-weighted incidence of catastrophic paymentsConcentration index of incidence -.0242 -.052910 Percent of households facing catastrophic payments for health care, or an amount of out-of-pocketpayments as share of income that is at least ten percent.32

Table 19 Poverty headcount: effect of accounting for out-of-pocket payments for health care, Egypt (%)Richest 20% 2 nd richest Middle 2 nd poorest Poorest 20% TotalPoverty line of $1.08 per dayPre-payment headcount 0 0 0 0 32.8 7.5Post-payment headcount 0.2 0.2 0.4 2.3 63.1 14.9Percentage change +0.2 +0.2 +0.4 +2.3 +92.4 +100.0Poverty line of $2.15 per dayPre-payment headcount 0 0 7.9 1 1 45.9Post-payment headcount 0.2 3.5 68.5 1 1 56.1Percentage change +0.2 +3.5 +767.0 0 0 +22.2Relative poverty linePre-payment headcount 0 0 0 0 28.3 6.5Post-payment headcount 0.2 0.2 0.4 1.5 40.8 9.7Percentage change +0.2 +0.2 +0.4 +1.5 +44.2 +49.2Table 20 Poverty headcount: effect of accounting for out-of-pocket payments for health care, Lebanon (%)Richest 20% 2 nd richest Middle 2 nd poorest Poorest 20% TotalPoverty line of $1.08 per dayPre-payment headcount 0 0 0 0 1.6 0.4Post-payment headcount 0 0 0.9 1.0 8.1 2.0Percentage change 0 0 +0.9 +1.0 +400.0 +400.0Poverty line of $2.15 per dayPre-payment headcount 0 0 0 0 3.0 0.7Post-payment headcount 0 0 0.9 1.0 11.0 2.6Percentage change 0 0 +0.9 +1.0 +267.0 +271.0Relative poverty linePre-payment headcount 0 0 0 0 60.7 13.3Post-payment headcount 0.3 0.3 2.1 9.6 85.7 19.6Percentage change +0.3 +0.3 +2.1 +9.6 +41.2 +47.433

Multivariate analysis: Outpatient and Inpatient UtilizationTable 21 Estimation of model I-a: Probability of outpatient use (base model)EgyptLebanonDependent variable: Visited outpatient provider in last 12 monthsOutpt Probits** Robust SE Probits Robust SEOp_ins 0.120* 0.059 1.030* 0.234Age 0.005* 0.002 -0.007* 0.002Male 0.317* 0.061 0.057 0.067Married 0.166* 0.053 0.180* 0.052Educlev 0.186* 0.055 -0.147* 0.061Emplyd 0.057 0.061 -0.221* 0.064Logincome -0.024 0.033 -0.019 0.033SAH 0.219* 0.037 0.248* 0.041Pain 0.1852* 0.041 0.298* 0.038Chronic_1 1.402* 0.051 1 .400* 0.065Constant -1.657* 0.254 -1.454* 0.319Op_insAge 0.019* 0.002 0.005* 0.002Male -0.232* 0.059 0.209* 0.060Educlev 0.997* 0.053 0.299* 0.056Employed 0.408* 0.061 0.247* 0.061Logincome 0.257* 0.035 0.141* 0.033Constant -3.841* 0.266 -2.069* 0.265/athrho 0.185 0.328 -0.781* 0.247rho 0.183 0.317 -0.653 0.141Wald test of rho=0: chi2(1)=0.319 Prob>chi2=0.573 chi2(1)=10.040 Prob>chi2=0.002N 4133 2752Prob > chi2 (overall model) 0.000 0.000*Significant with a 95% confidence interval.**Probits for top half for Egypt shown for general probit model since OP_INS is exogenous asindicated by Wald chi-squared test.Table 22 Estimation of model I-b: Probability of outpatient use (with interaction effect)EgyptLebanonDependent variable: Visited outpatient provider in last 12 monthsOutpt Probits Robust SE Probits Robust SEOp_ins 1.243* 0.560 2.020* 0.517Age 0.005* 0.002 -0.007* 0.002Male 0.319* 0.066 0.030 0.075Married 0.095 0.059 0.149* 0.054Educlev 0.146* 0.060 -0.104 0.066Emplyd 0.043 0.066 -0.273* 0.072Logincome 0.053 0.047 -0.079 0.057SAH 0.205* 0.040 0.201* 0.041Pain 0.231* 0.044 0.267* 0.038Chronic_1 1.409* 0.055 0.247* 0.069Logincome X Op_ins -0.152* 0.076 -0.091 0.060Constant -2.198* 0.345 -0.808* 0.473N 3526 1857Prob > chi2 0.000 0.000*Significant with a 95% confidence interval.34

Table 23 Estimation of model II-a: Probability of inpatient use (base model)EgyptLebanonDependent variable: Visited inpatient provider in last 12 monthsInpt Probits Robust SE Probits Robust SEIp_ins 0.094 0.081 -0.778* 0.277Age -0.002 0.003 0.004 0.002Male -0.094 0.081 0.065 0.077Married 0.050 0.078 0.232* 0.069Educlev -0.070 0.085 0.008 0.087Emplyd -0.061 0.086 -0.114 0.084Logincome -0.026 0.047 0.124* 0.043SAH 0.120* 0.050 0.125* 0.049Pain 0.238* 0.046 0.165* 0.037Chronic_1 0.285* 0.092 0.289* 0.081Constant -2.210* 0.361 -2.756* 0.332N 4133 2752Prob > chi2 0.000 0.000*Significant with a 95% confidence interval.Table 24 Estimation of model II-b: Probability of inpatient use (with interaction effect)EgyptLebanonDependent variable: Visited inpatient provider in last 12 monthsInpt Probits Robust SE Probits Robust SEIp_ins -0.808 0.890 -0.887 0.679Age -0.002 0.003 0.003 0.003Male -0.127 0.088 0.004 0.087Married 0.074 0.085 0.208* 0.077Educlev -0.078 0.091 0.032 0.094Emplyd -0.060 0.091 -0.119 0.094Logincome -0.042 0.066 0.252* 0.084SAH 0.122* 0.053 0.142* 0.053Pain 0.212* 0.050 0.120* 0.040Chronic_1 0.352* 0.099 0.256* 0.092Logincome X Ip_ins 0.121 0.123 -0.033 0.100Constant -2.061* 0.490 -3.322* 0.545N 3526 1857Prob > chi2 0.000 0.000*Significant with a 95% confidence interval.35

More magazines by this user
Similar magazines