Health Systems & Reform ISSN: 2328-8604 (Print) 2328-8620 (Online) Journal homepage: http://www.tandfonline.com/loi/khsr20 Utilization of Health Care and Burden of Out-of- Pocket Health Expenditure in Zimbabwe: Results from a National Household Survey Wu Zeng, Laurence Lannes & Ronald Mutasa To cite this article: Wu Zeng, Laurence Lannes & Ronald Mutasa (2018) Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe: Results from a National Household Survey, Health Systems & Reform, 4:4, 300-312, DOI: 10.1080/23288604.2018.1513264 To link to this article: https://doi.org/10.1080/23288604.2018.1513264 Published with license by Taylor & Francis Group, LLC© 2018 International Bank for Reconstruction and Development / The World Bank Accepted author version posted online: 24 Sep 2018. Published online: 06 Nov 2018. Submit your article to this journal Article views: 354 View Crossmark data Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=khsr20 Health Systems & Reform, 4(4):300–312, 2018 Published with license by Taylor & Francis Group, LLC ISSN: 2328-8604 print / 2328-8620 online DOI: 10.1080/23288604.2018.1513264 Research Article Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe: Results from a National Household Survey Wu Zeng*,1, Laurence Lannes 2 and Ronald Mutasa2 1 Schneider Institutes for Health Policy, The Heller School for Social Policy and Management, Brandeis University, Waltham, MA, USA 2 Health, Nutrition and Population Global Practice, The World Bank, Washington, DC, USA CONTENTS Abstract—In the last decade, Zimbabwe has undertaken substan- Introduction tial changes and implemented new initiatives to improve health Methods system performance and services delivery, including results-based Results financing in rural health facilities. This study aims to examine the Discussion utilization of health services and level of financial risk protection of Conclusions Zimbabwe’s health system. Using a multistage sampling approach, 7,135 households with a total of 32,294 individuals were surveyed References in early 2016 on utilization of health services, out-of-pocket (OOP) health expenditure, and household consumption (as a measure of living standards) in 2015. The study found that the outpatient visits were favorable to the poor but the poorest had less access to inpatient care. In 2015, household OOP expenditure accounted for about one quarter of total health expenditure in Zimbabwe and 7.6% of households incurred catastrophic health expenditure (CHE). The incidence of CHE was 13.4% in the poorest quintile in comparison with 2.8% in the richest. Additionally, 1.29% of households fell into poverty due to health care–related expendi- tures. The study suggests that there are inequalities in utilization of health services among different population groups. The poor seek- ing inpatient care are the most vulnerable to CHE. INTRODUCTION Keywords: catastrophic health expenditure, health financing, out-of-pocket Addressing inequality in access to health services and pro- expenditure, poverty, Zimbabwe viding financial risk protection are two major, critical tasks Received 13 April 2018; revised 12 August 2018; accepted 12 August 2018. to establish an effective, efficient, and functional health *Correspondence to: Wu Zeng; Email: wuzengcn@brandeis.edu Color versions of one or more of the figures in the article can be found system. As part of the Sustainable Development Goals online at www.tandfonline.com/khsr. (SDGs) and universal health coverage (UHC) agendas,1 © 2018 International Bank for Reconstruction and Development / The World Bank many low- and middle-income countries have implemented This is an Open Access article distributed under the terms of the Creative health initiatives to improve access to basic health services Commons Attribution License (http://creativecommons.org/licenses/by/4.0/), for all and, additionally, provide financial risk protection to which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. reduce out-of-pocket (OOP) health expenditure. 300 Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 301 During the 1980s and most of the 1990s, Zimbabwe’s than Zimbabwe. Despite comparable income levels, economy was on the path to middle-income status. The Mozambique, Madagascar, Zambia, Tanzania, Lesotho, and government of Zimbabwe invested in primary and preventive Malawi spend comparatively more on health than Zimbabwe, health care and rolled out primary health care services to reflecting their governments’ commitment to the sector.4 within ten kilometers of at least 80% of the population. In 2015, 24% of health expenditures came from house- However, the deteriorating economic situation in the 1990s hold OOP payments in Zimbabwe. External assistance was and the decline in government financing led public and not- the biggest financing source (25%), whereas government for-profit health providers to introduce various forms of user spending comprised only 21% of all health spending.5 fees. From 2009 onwards, when the Zimbabwean economy Prior studies suggest the existence of inequality in began to improve after the 2008 crisis, the recovery in social resources, access to and use of health services, as well as sectors showed a mixed picture with promising trends in health outcomes between the rich and the poor. Zimbabwe some areas like infant/child mortality, but causes for concern has a large network of health facilities, with a higher per in others, such as the high incidence of poverty and the low capita distribution of facilities in provinces with higher pov- quality of reproductive health services and education. erty incidence.6 However, key health personnel, including Between 2001 and 2010, the share of household OOP expen- medical doctors, nurses, and midwives, are more concen- ditures in total health expenditure increased, and government trated in areas with lower poverty incidence. The inequitable spending fell to low levels. distribution of skilled health personnel compounds inequities In 2008, public health services collapsed as a result of the in health outcomes. Poorer households typically rely on financial crisis and Zimbabwe faced challenges in meeting lower-quality, low-level facilities when seeking care, whereas some key health-related Millennium Development Goals richer people are more likely to use provincial or central (MDGs) as well as its own broad set of national health hospitals as well as private services.6 targets. Life expectancy at birth in 2015 was only 60.3 years,2 Since 2010, the government of Zimbabwe has initiated with 65% of annual deaths attributed to communicable, health system reforms to improve access to health services maternal, perinatal, and nutritional illnesses in 2012.3 and financial risk protection. Recent reforms aim at tackling Under-five mortality dropped from 84 in 2010–2011 to 69/ core underlying health service delivery gaps and barriers that 1,000 live births in 2015 and infant mortality decreased from hinder utilization of health services by the poorest house- 57 to 50/1,000 live births during the same period, short of the holds. Results-based financing (RBF), combined with free MDG targets. In particular, Zimbabwe’s high maternal mor- health care for maternal and child health services, is one of tality ratio, estimated at 651 deaths per 100,000 live births in the key reforms being implemented in Zimbabwe at the 2015, in contrast to an MDG target of 174 per 100,000 live national scale to (1) increase demand for and utilization of births, remains an urgent concern despite improvements priority maternal and child health services by poor house- from the 2010 estimate of 960 deaths per 100,000 live holds by removing financial barriers to accessing health births.2 Furthermore, Zimbabwe has the fifth highest HIV services; (2) strengthen performance of health facilities, prevalence in sub-Saharan Africa, bringing with it a perni- including through quality improvement; and (3) rebuild cious and costly lifelong HIV treatment burden. The basic services that had collapsed in past years. After being Zimbabwean overextended health system is further com- piloted in the country, the RBF program was rolled out in all pounded by accelerated burden of noncommunicable dis- rural districts; it also benefits targeted low-income urban eases (NCDs); in 2012, approximately 31% of total deaths populations. A rigorous impact evaluation of RBF revealed in Zimbabwe were caused by NCDs.3 key gains in selected health coverage and quality indicators, Total health expenditure per capita in Zimbabwe (103.8 as well as improvement in equity in the use of health United States Dollars [USD] in 2015) compares favorably services.7 Notably, the in-facility delivery rate increased by with the sub-Saharan Africa average (84 USD), but spending 14 percentage points.7 is potentially regressive due to a high burden of household To better understand utilization of health care services in OOP expenditures at point of care. Zimbabwe’s public health Zimbabwe and the degree of financial protection in the care spending per capita is one of the lowest among coun- country, we used a national household survey to (1) examine tries in the subregion and amounted to 8.72% in 2015. In the the utilization of both inpatient and outpatient services, (2) Southern African Development Community group, only the estimate the incidence of catastrophic health spending, and Democratic Republic of the Congo allocates fewer public (3) investigate the impoverishing impact of OOP health resources for health as a share of total government spending spending. 302 Health Systems & Reform, Vol. 4 (2018), No. 4 METHODS Measurement of Utilization of Health Care and OOP per Visit or Admission In January 2016, the Zimbabwe National Statistics Agency (ZIMSTAT) conducted a national cross-sectional household For outpatient visits, the recall period was four weeks. The survey, with the main purpose of investigating utilization of number of outpatient visits, detailed information on the outpatient and inpatient health services and estimating type of care, disease categories associated with the visit, related OOP spending. and OOP spending on registration and consultation, drugs, medical investigation (e.g., lab tests, x-ray), food, and transportation, as well as in-kind payment for a maximum Sampling and Sample Size of four outpatient visits in the last four weeks were Originally, a total of 7,450 households were sampled for the recorded. When a household member could not remember household survey. All provinces in Zimbabwe are repre- detailed spending by category, a lump sum estimate was sented in the sample and the sample size for each province used. OOP spending for each reported visit up to four was proportional to the square root of the share of house- visits in the last four weeks was then calculated by sum- holds in each province to the total number of households in ming up all spending categories, and the average OOP Zimbabwe (see Table A1). A two-stage sampling process spending per visit was then estimated. OOP spending for was conducted in each of the provinces, with the first stage outpatient visits was estimated by multiplying the average being implemented at the enumerate area (EA) level and the OOP spending per visit by the number of outpatient visits second stage at the household level. EAs were selected using over four weeks. The result was then extrapolated to stratified random sampling and households within EAs were 52 weeks to obtain the annual OOP for outpatient visits. selected using systematic sampling. EAs were the primary The estimation of OOP spending for inpatient admissions sampling units and households were the secondary sampling used an approach similar to that for outpatient visits units. The final sample consisted of 7,135 households with a described above, except that a recall period of one year total of 32,294 individuals from 373 EAs. More details on was used and the detailed spending was recorded up to two the sampling process were provided elsewhere.5 admissions. The annual OOP spending for inpatient admis- sions was estimated by multiplying the average OOP per admission by the number of admissions in the last Measurement of Household Characteristics, Income, and 12 months. Consumption The total annual OOP health expenditure was estimated A survey was designed to collect required information to by summing annual OOP for outpatient visits and inpatient estimate OOP spending, as well as characteristics of house- admissions for each individual. Data were aggregated at the holds and household members. The questions for estimating household and national levels, weighted by the probability of income and consumption were adopted from the survey used a household withdrawing from the national population. The for Zambia’s national health accounts. The household char- probability of sampling was estimated from Table A1. acteristics included location of households (province) and household size. To assess income, the household head or most informative household member was asked about 13 Measurement of Burden of OOP: Catastrophic Health income sources (including public and parastatal salaries, Spending and Headcount Index pensions, and sale of crops) and the amount received the The incidence of catastrophic health expenditure (CHE) was month preceding the survey. Household consumption was estimated as the share of households that spent more than estimated based on a one-month recall period for spending 25% of total consumption (OOP expenditure included) on on 37 items (e.g., food, education, transport, rental, etc.). OOP expenditure on health. In the last decades, there have Monthly income and consumption were extrapolated to been several cutoff values used for defining CHE, including obtain annual household income and consumption. We using OOP as a share of nonfood consumption expenditure9 used household consumption to measure living standard for and OOP as a share of total consumption.10 The cutoff the equity analysis, because consumption is less constrained values vary significantly when using the share of nonfood than income for households with limited resources.8 consumption, from 5% up to 40%. We primarily used the Individual characteristics included gender, age, relationship cutoff of 25% of total consumption to reflect recent devel- to head of household, education, and employment of house- opments in measuring catastrophic health spending.9 hold members. Additionally, we provided the incidence of CHE using a Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 303 cutoff of 10% of total consumption because this cutoff is predictors of total OOP expenditure once it occurred using a used for tracking the progress toward SDGs. regression model. The simplified models are expressed in (2) The poverty headcount index (incidence) is used to mea- and (3). sure the impact of health spending on poverty. The incidence Part I: of poverty was measured under two scenarios: with OOP   payments and without OOP payments. The difference in Poop In ¼ YX (2) poverty head count measures under the two scenarios cap- 1 À Poop tures the impact of OOP health expenditure on poverty. We use the poverty line obtained from ZIMSTAT, which is equal Part II: to 96.6 USD of consumption per month per capita.11 After removing 23 potential outliers (ratio of household consump- lnððOOPÞjOOPi0Þ ¼ ΛX (3) tion after excluding OOP health expenditure to the poverty line less than 2), we first estimated the incidence of poverty where POOP is the probability of a household incurring OOP by including OOP health expenditure as household con- expenditure, X is the same independent variable matrix as in sumption; second, we estimated the incidence of poverty by (1), and Υ and Λ are associated coefficient matrices. All data excluding OOP health expenditure from consumption. We analyses were conducted using STATA 15 (StataCorp LLC, then calculated the difference between the two incidences as College Station, TX). an indicator of the impoverishing impact of OOP health expenditure. RESULTS A total of 32,294 individuals (7,135 households) were Analysis sampled and included in the analysis. The average household size was 4.53, with a standard deviation of 2.16. Among We first calculated the utilization of inpatient and outpatient individuals, 48.3% were males and 51.7% were females; services by age, gender, and expenditure quintile. We then 21.1% of the population was urban and the rest lived in calculated total OOP health expenditure, incidence of cata- rural areas. The average age was 23.6 years old strophic health spending, and poverty headcount. (SD = 19.3). The share of individuals having preschool, We conducted a logistic regression (logit model) to esti- primary, secondary, and tertiary school education was mate the determinants of catastrophic health spending. The 13.2%, 41.0%, 35.6%, and 5.3%, respectively; another dependent variable was whether a household incurred CHE 4.7% of individuals were attending school at the time of (1 = yes, 0 = no), and the independent variables included the survey, and 0.2% individuals answered “do not know.” characteristics of household head (e.g., age, gender, educa- tion, employment, and marriage, as well as household con- sumption). The full logistic regression model presented in Outpatient Care (1) is During the four weeks preceding the interview, 18.5% of indi-   viduals had sought outpatient care. A total of 5,633 outpatient PCHE Ln ¼ β0 þ β1 gender þ β2 age þ β3 age2 visits were reported among the 32,294 individuals, with 0.174 1 À pCHE visits per capita over the four weeks. This was equivalent to þ β4 household size 2.26 visits per person per year. Table 1 shows the utilization of þ β5 education þ β6 marriage outpatient care by living standards (ranging from quintile one þ β7 employment [poorest] to quintile five [richest] based on the consumption). þ β8 consumption þ β9 urban Overall, there is no consistent pattern of utilization of out- þ β10 inpatent care (1) patient care across consumption groups. Among children under five, the third consumption quintile had the highest where PCHE is the probability of a household incurring CHE, number of outpatient visit (0.267) and the next richest had and β is the coefficient or coefficient matrix for the asso- the lowest (0.205). However, among the oldest age group (ages ciated variable or variable matrix. 65+) there was a more consistent pattern of demand of health We used the same set of independent variables and con- services: the poorest quintile had the lowest number of visits ducted a two-part model analysis to examine determinants of with an average of 0.326 visits per person in four weeks and the occurrence of OOP expenditure using a logit model and the fourth quintile had 0.462 visits per person in four weeks. 304 Health Systems & Reform, Vol. 4 (2018), No. 4 Poorest Next Poorest Middle Next Richest Richest Age Group (Quintile 1) (Quintile 2) (Quintile 3) (Quintile 4) (Quintile 5) Total F Value <5 0.233 0.245 0.267 0.205 0.241 0.238 1.70 5–19 0.130 0.125 0.102 0.085 0.085 0.107 7.39*** 20–29 0.145 0.142 0.117 0.130 0.131 0.132 0.61 30–39 0.217 0.194 0.197 0.151 0.146 0.177 3.33** 40–49 0.195 0.275 0.278 0.218 0.201 0.232 2.54* 50–64 0.320 0.243 0.293 0.309 0.322 0.296 1.28 65+ 0.326 0.341 0.431 0.462 0.387 0.380 2.08 Total 0.188 0.183 0.181 0.161 0.155 0.174 5.71*** *P < 0.05. **P < 0.01. ***P < 0.001. TABLE 1. Utilization of Outpatient Care per Capita per Four Weeks by Expenditure Quintile Across all age groups, outpatient visits were more frequently admissions per person. The richest used more inpatient ser- used by the poorest quintile (0.188 visits per person in four vices than did the poorest (Table 2). The poorest quintile, on weeks in comparison with 0.155 in the richest quintile). The average, had 0.021 admissions per person per year compared overall concentration index for outpatient visits was −0.029, to 0.036 (71.4% higher) for the richest. The population in the suggesting that the utilization of outpatient care was pro-poor. third consumption quintile had the second highest number of admissions per capita of 0.034. Inpatient Care Among the 32,294 individuals, 2.48% (783 individuals) Out-of-Pocket Spending reported that they were hospitalized during the 12 months Total OOP was estimated at 343.7 million USD, equivalent to preceding the survey, with 0.03 admissions per capita per 24.90 USD per capita per year. As shown in Table A2, 73.59% of year (Table 2). The use of inpatient care increases with age, OOP expenditure was used for curative care and 9.91% was for except for children under five, who had higher average hospi- long-term care. Expenditures on rehabilitative care, ancillary tal admissions than those aged 5–19 (0.021 versus 0.012 services, and medical goods were relatively low, accounting for admissions per person per year). The oldest group (ages 65+) 3.78%, 2.74%, and 1.77% of total OOP expenditure, respectively. had an average admission of 0.174 per person per year. About 88% of OOP expenditure was used for outpatient care and Although the concentration index for inpatient care was the remaining 12% was for inpatient care. estimated to be −0.015, suggesting an overall pro-poor pat- Table A3 shows OOP by diseases/conditions. Conditions tern, the utilization of inpatient care for the poorest remained with the highest OOP were hypertensive diseases; respiratory the most striking, with the lowest average number of system diseases; accidents, poisoning, and injuries; and Poorest Next Poorest Middle Next Richest Richest Age Group (Quintile 1) (Quintile 2) (Quintile 3) (Quintile 4) (Quintile 5) Total F Value <5 0.016 0.017 0.028 0.020 0.027 0.021 1.03 5–19 0.006 0.013 0.010 0.013 0.020 0.012 3.23* 20–29 0.026 0.026 0.040 0.033 0.042 0.034 1.03 30–39 0.045 0.056 0.045 0.044 0.043 0.046 0.29 40–49 0.021 0.054 0.048 0.037 0.036 0.040 1.30 50–64 0.056 0.043 0.079 0.053 0.082 0.062 1.10 65+ 0.037 0.096 0.105 0.100 0.102 0.082 1.11 Total 0.021 0.029 0.034 0.030 0.036 0.030 3.72** *P < 0.05. **P < 0.01. TABLE 2. Utilization of Inpatient Care per Capita per Year by Income Quintile Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 305 Consumption Total Consumption OOP Expenditure Total Consumption Percentage of OOP Expenditure as a Quintile (USD) (USD) Excluding Households Percentage of OOP (USD) Incurring OOP Total Consumption 1 199.8 ± 3.0 34.0 ± 6.1 165.9 ± 6.7 17.3 16.5 2 593.2 ± 3.5 75.5 ± 10.0 517.7 ± 10.5 22.4 12.7 3 1,212.1 ± 6.8 128.0 ± 16.1 1,084.1 ± 17.4 28.6 10.7 4 2,377.7 ± 11.6 115.2 ± 13.1 2,262.5 ± 17.1 26.7 5.0 5 7,341.9 ± 266.7 203.8 ± 16.3 7,138.1 ± 266.0 34.8 2.8 Average 2,404.4 ± 64.3 112.5 ± 5.9 2,291.8 ± 64.0 26.1 9.4 a Total consumption, OOP expenditure, and total consumption excluding OOP were measured per household per year. TABLE 3. Consumption and Out-of-Pocket (OOP) Expenditure by Quintilea intestinal infectious diseases, with shares of 11.93%, 9.27%, households having CHE, in contrast with 2.8% of house- 9.87%, and 6.14%, respectively. HIV/AIDS accounts for holds in the richest quintile (Table 5). When the cutoff of 2.23% of total health OOP expenditure. Because 31% of 10% of total consumption was used, the overall incidence of expenditure was not categorized according to disease, we CHE was estimated to be 12.6% in Zimbabwe, with 16.7% could not include the category of disease as a predictor of and 9.4% in the poorest and wealthiest quintiles, CHE. respectively. Table 3 shows a summary of indicators for consumption Figure 1 shows that the OOP drove 1.29% of households and OOP expenditure. Total yearly consumption per house- into poverty, with a 95% confidence interval of (1.02%, hold was highly skewed, with a mean of 2,404 USD and 1.56%), which is equivalent to 179,868 individuals who fell median of 1,186 USD. Yearly OOP expenditure per house- into poverty in 2015 in Zimbabwe due to health expenditure. holds averaged 112 USD, accounting for 9.4% of yearly total As a result, the poverty rate increased from 55.39% to consumption. Although households with higher total con- 56.69%. If outliers were included, OOP would increase the sumption tended to incur OOP, OOP expenditure as a per- poverty rate by an additional 1.36%. centage of total consumption declined, from 16.5% among the poorest to 2.8% among the wealthiest. Determinants of Catastrophic Health Spending Table 6 shows the determinants of CHE. We found that Determinants of Out-of-Pocket Spending household size, consumption, residing in urban areas, and Table 4 shows the results from the two-part model on deter- having inpatient care were the major determinants of CHE. minants of incurring OOP expenditure and predictors for the One more member in the household was associated with amount of OOP if it occurred. Larger household size, house- an 8.4% increase in odds of incurring CHE (P < 0.05). holds in urban areas, and a higher consumption level were Compared to the poorest households, all other groups of associated with a higher chance of incurring OOP and higher households had a lower odd of incurring CHE amount of OOP expenditure. For example, compared to the (P < 0.001). The odds decreased as the households poorest quintile group, the second poorest quintile group had became wealthier. All other things being equal, living in 26% (exp(0.236)−1) higher odds of incurring OOP spending urban areas was associated with a 48.5% higher odds of and 37.2% higher OOP expenditure occurred. Age and edu- incurring CHE (P < 0.05). Having inpatient care was cation were not associated with the occurrence of OOP but associated with 6.03 times higher risk of incurring CHE were associated with the amount of OOP when it occurred. (P < 0.001). Education, marriage, and employment were not statistically significant at a significance level of 0.05. Catastrophic Expenditure and Impoverishing Effect of OOP Health Expenditure DISCUSSION Using the cutoff of 25% of total consumption, we estimated This study provides evidence on utilization of care, burden that 7.6% of households in Zimbabwe incurred CHE in of OOP health spending in terms of CHE, and impoverish- 2015. The poorest suffered the most, with 13.4% of ment due to OOP health expenditure in Zimbabwe. 306 Health Systems & Reform, Vol. 4 (2018), No. 4 Determinants of Risk of OOP (Logit Model) Coefficient SE t P>t 95% Confidence Interval Female 0.107 0.093 1.150 0.251 −0.076 0.290 Age 0.001 0.013 0.080 0.940 −0.025 0.026 Age2 0.000 0.000 0.710 0.477 0.000 0.000 Household size 0.108 0.016 6.760 0.000 0.077 0.140 Preschool Reference Primary school 0.220 0.276 0.800 0.426 −0.322 0.762 Secondary 0.316 0.283 1.120 0.264 −0.239 0.871 Tertiary 0.296 0.299 0.990 0.322 −0.290 0.882 Never married Reference Married 0.205 0.169 1.210 0.227 −0.127 0.537 Divorced/separated 0.035 0.189 0.180 0.855 −0.336 0.405 Paid employee Own account workers −0.074 0.084 −0.880 0.378 −0.238 0.090 Not employed −0.037 0.105 −0.350 0.727 −0.243 0.169 Quintile 1 (consumption) Reference Quintile 2 0.236 0.118 2.000 0.045 0.005 0.467 Quintile 3 0.536 0.116 4.610 0.000 0.308 0.764 Quintile 4 0.366 0.124 2.960 0.003 0.124 0.609 Quintile 5 0.622 0.135 4.630 0.000 0.359 0.886 Urban 0.343 0.086 3.990 0.000 0.175 0.512 Inpatient care 0.000 (omitted) Constant −2.946 0.421 −6.990 0.000 −3.772 −2.120 ln(OOP)|OOP > 0 (regression model) Female 0.019 0.093 0.200 0.842 −0.164 0.201 Age −0.028 0.013 −2.270 0.023 −0.053 −0.004 Age2 0.000 0.000 3.570 0.000 0.000 0.001 Household size 0.060 0.017 3.530 0.000 0.027 0.093 Preschool Reference Primary school −0.511 0.251 −2.030 0.042 −1.003 −0.018 Secondary −0.388 0.255 −1.520 0.128 −0.887 0.111 Tertiary 0.069 0.267 0.260 0.795 −0.454 0.593 Never married Reference Married −0.170 0.141 −1.210 0.227 −0.447 0.106 Divorced/separated −0.235 0.162 −1.450 0.147 −0.553 0.083 Paid employee Reference Own account workers −0.032 0.071 −0.460 0.648 −0.171 0.107 Not employed 0.043 0.094 0.450 0.651 −0.142 0.227 Quintile 1 (consumption) Reference Quintile 2 0.372 0.118 3.150 0.002 0.140 0.604 Quintile 3 0.611 0.111 5.500 0.000 0.393 0.828 Quintile 4 0.611 0.113 5.420 0.000 0.390 0.832 Quintile 5 1.001 0.117 8.520 0.000 0.770 1.231 Urban 0.328 0.070 4.690 0.000 0.191 0.465 Inpatient care −0.029 0.041 −0.720 0.469 −0.109 0.050 Constant 4.699 0.387 12.150 0.000 3.941 5.458 TABLE 4. Determinants of Out-of-Pocket (OOP) Health Expenditure (Two-Part Model) Utilization of outpatient and inpatient care is not distributed more inpatient admissions than the poorest. In total, OOP equally among different populations. Those in the poorest spending was estimated at 25 USD per person per year, and quintile seek slightly more outpatient care than those in the OOP health spending results in an additional 1.29% of richest quintile. By contrast, the richest quintile had 71.4% households falling into poverty. Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 307 Poorest Next Poorest Middle Next Richest Richest Total F Value (Quintile 1) (Quintile 2) (Quintile 3) (Quintile 4) (Quintile 5) Incidence of CHE (25% of 13.38% 8.68% 8.37% 5.20% 2.77% 7.64% 28.07*** total consumption) Incidence of CHE (10% of 16.67% 12.98% 14.10% 10.20% 9.36% 12.63% 10.76*** total consumption) ***P < 0.001. TABLE 5. Incidence of Catastrophic Health Expenditure (CHE) by Consumption Quintile 20 Household consumption as multiple of poverty line 18 HH consumption 16 HH consumption without OOP 14 Poverty line 12 10 8 6 4 2 0 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% -2 Cumulative proportion of household ranked by consumption -4 expenditure FIGURE 1. The Impact of Out-of-Pocket (OOP) Health Expenditure on Poverty The study shows that there are substantial inequalities in relatively low utilization of inpatient care among the poor utilization of inpatient care: The poorest utilize inpatient may suggest a potential high incidence of foregone care for care to a much lesser extent than the richest, although they inpatient services due to financial barriers. A study has tend to use slightly more outpatient care. The findings of shown that OOP could be one of the major reasons why lower use of inpatient care by the poor are consistent with patients forgo health care.6 Unfortunately, this study did not that found in Afghansitan,12 China,13 and Ethiopia.14 In allow us to estimate the incidence of forgone care due to the Zimbabwe, the exemption of user fees for maternal and lack of relevant questions in the survey. This is the major child health services, coupled with RBF in all primary health limitation of this study, preventing us from making more care facilities and some district hospitals, removed the finan- meaningful comparisons of utilization of care among differ- cial barrier to access to health care services, thus explaining ent expenditure groups. the equitable distribution of utilization of outpatient health The emerging NCD disease burden is confirmed by our services. This is in contrast with inpatient care, because findings, which show that OOP spending has been concen- hospitals charge user fees. This is consistent with earlier trated on diseases such as hypertension.5 The OOP of house- studies in Zimbabwe that found that the poorest households hold members with NCDs could result in CHE. A study in relied on primary health care facilities and the richest house- Nepal showed that a household affected by chronic illness holds had greater hospital utilization, where the quality of such as diabetes, heart disease, asthma, or hypertension is care provided is better and range of services provided is more likely to incur CHE.15 Zimbabwe is also facing an much wider.6 Additionally, high hospital treatment costs increasing burden from NCDs.16 Cerebrovascular disease is could deter the poor from using inpatient care, and the one of the leading causes of disability-adjusted life years.17 308 Health Systems & Reform, Vol. 4 (2018), No. 4 Determinants of 95% utilization of health care, particularly for the poor who Risk of CHE Odds Confidence use inpatient care, which incurs higher costs. (Logit Model) Ratio SE t P>t Interval Inpatient care is significantly associated with CHE, with 6.03 times higher odds of incurring CHE. OOP expenditure Female 1.023 0.154 0.150 0.878 0.762 1.375 per inpatient admission was estimated to be 103 USD, com- Age 0.998 0.019 −0.110 0.911 0.962 1.036 pared to 9.64 USD per outpatient visit. OOP expenditure is Age2 1.000 0.000 1.380 0.166 1.000 1.001 much higher for inpatient care. OOP expenditure per inpa- Household size 1.084 0.027 3.220 0.001 1.032 1.139 tient admission is equivalent to 51.4% of average monthly Preschool Reference consumption per household. Additionally, a hospitalized Primary school 0.736 0.203 −1.110 0.266 0.428 1.264 patient is absent from work, and forgone earnings may Secondary 0.866 0.253 −0.490 0.623 0.489 1.536 reduce consumption due to financial constraints, leading to Tertiary 0.965 0.353 −0.100 0.923 0.472 1.976 a higher chance of incurring CHE. As mentioned previously, Never married Reference chronic illness is the major reason for hospitalization. Focusing on preventive care and prevention of chronic dis- Married 0.963 0.260 −0.140 0.889 0.568 1.634 eases (e.g., behavior change and lifestyle modification) and Divorced/ 0.805 0.242 −0.720 0.471 0.447 1.450 developing mechanisms to reduce OOP for expensive inpa- separated tient care would help alleviate the concern of CHE. Paid employee Reference The poor suffer most from the high OOP expenditure, Workers in the 0.874 0.122 −0.960 0.335 0.666 1.149 with a much higher incidence of CHE. Although the weal- informal sector thier group tends to have a higher chance of spending OOP, Not employed 1.017 0.165 0.110 0.916 0.740 1.398 the amount of OOP expenditure as a percentage of total Quintile 1 Reference consumption is lower that that among the poor. In fact, (consumption) given that some poor households may not seek care due to Quintile 2 0.589 0.083 −3.760 0.000 0.448 0.776 high OOP spending, the financial burden could be even Quintile 3 0.497 0.075 −4.650 0.000 0.371 0.668 higher for the poor if this factor is accounted for. Quintile 4 0.255 0.046 −7.570 0.000 0.179 0.363 Developing financial protection for the poor is thus critical. Quintile 5 0.099 0.026 −8.860 0.000 0.059 0.165 Prior to 2012, mechanisms to protect the poor against finan- Urban 1.485 0.239 2.460 0.014 1.084 2.035 cial loss in the event of illness were limited in Zimbabwe: Inpatient care 7.029 0.931 14.730 0.000 5.422 9.112 according to the 2014 Labor Survey, only 9% of the total population, primarily the rich, were covered under any form Constant 0.077 0.044 −4.530 0.000 0.025 0.234 of health insurance.20 Accordingly, direct user fees remain TABLE 6. Determinants of Catastrophic Health Spending (Logit an important source of funding for district, mission, central, Model) and local government health facilities. In a country like Zimbabwe where almost 90% of the population is in the informal sector, expanding employer- Given the increasing burden of chronic diseases in based health insurance to increase financial protection Zimbabwe,16 a focus on preventive care to avert high treat- cannot be a short- to medium-term option to reduce the ment costs of NCDs and a more balanced budget with more financial burden of health care spending on the poorest funds provided for preventive services could help avoid high households. Given the limited financial resources avail- OOP and CHE.18 able, in 2012, Zimbabwe started by offering a limited This study also found that OOP spending plays a sig- package of free basic health services to reduce OOP nificant role in financing the health system in Zimbabwe. through removal of user fees at primary health care facil- OOP spending amounts to 25 USD per person per year, ities, combined with RBF for maternal and child health accounting for 24% of Zimbabwe’s total health services in rural areas7 or a voucher system for maternal expenditure.5 The share of OOP in total health expenditure services in urban areas. RBF schemes covered 3.5 million in Zimbabwe is similar to that in Tanzania (23.31%) and people in 2012, representing 23.7% of the national popu- Angola (23.96%) but is lower than that in the Democratic lation. As of March 2016, the voucher schemes had helped Republic of the Congo (38.77%) and Madagascar more than 2,500 mothers access needed health services, (41.36%).19 The high share of OOP spending affects although coverage remains low.21 Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 309 OOP spending on health drove 1.29% households into the poor is a major concern and should remain at the heart poverty, increasing the poverty rate from 55.39% to of Zimbabwe’s health financing reforms. Although OOP 56.69%. This may still be in a lower bound of the estimate. expenditure and CHE remain concerns in Zimbabwe, as If households expect OOP expenditures, they may reduce many low- and middle-income countries are moving toward expenditures in other categories and thus reduce total con- UHC and aiming to achieve SDGs by 2030, initiatives (e.g., sumption. This could drive some households into poverty RBF and the voucher scheme) that Zimbabwe has taken to before OOP expenditure incurs and underestimate the impact address financial risks among populations, particular the poor, of OOP expenditure on impoverishment. The 1.29% is simi- provide valuable lessons for other countries to design and lar to that reported in Ghana: 1.95% among the rural popula- implement tailored financing mechanisms to reduce OOP tion and 1.01% among the urban population.22 Poverty and expenditure and CHE. utilization of health services affect each other.10 Making users of health services pay OOP impoverishes some house- holds that choose to seek services. ACKNOWLEDGMENTS Such impoverishment, in turn, has an impact on a popula- The authors are indebted to Clare L. Hurley at Brandeis tion’s health and affects utilization of health care services. University for editorial assistance and to Paolo Belli, Son This points to the need to develop innovative ways to reduce Nam Nguyen, and Pia Schneider at the World Bank for the financial burden on households, particularly for the reviewing the article. poorest, to improve equity in access to health care services and improve financial protection. In the short term, this includes improving resource allocation across provinces DISCLOSURE OF POTENTIAL CONFLICTS OF and user fee exemptions for some population groups and INTEREST or/services. However, the country needs to explore ways to None of the authors have any conflicts of interest to report. expand prepayment mechanisms that eliminate user fees from the point of care for priority diseases beyond maternal and child health services. becoming increasingly necessary, ORCID because the accelerating incidence of NCDs requires a more comprehensive policy response to address the financial bur- Laurence Lannes http://orcid.org/0000-0001-8275-7135 den that households are experiencing and to improve the provision of NCD services to all segments of the population. REFERENCES 1. World Health Organization. Health systems financing: the path CONCLUSIONS to universal coverage. Geneva (Switzerland): World Health Organization; 2010. This study revealed inequality in utilization of health care 2. World Bank. World Bank Development Indicators [database on services in Zimbabwe. The poorest use much less inpatient the Internet]. World Bank. 2017 [accessed 2017 Dec 10]. http://databank.worldbank.org/data/reports.aspx?source= care than the richest do, and the poor are provided with limited world-development-indicators. financial protection against CHE. Though OOP expenditure 3. World Health Organization. Noncommunicable diseases coun- remains an important mechanism to finance the health system try profiles, 2014: Zimbabwe. Geneva (Switzerland): World in Zimbabwe, it also increases the risk of household impover- Health Organization; 2014 [accessed 2017 Dec 4]. http://www. ishment. For Zimbabwe to achieve UHC, policy makers and who.int/nmh/countries/zwe_en.pdf development partners should carefully explore options for 4. World Bank. Analyzing fiscal space options for health in Zimbabwe. Washington (DC): World Bank; 2017. financing health care that minimizes user fees at the point of 5. ZIMSTAT. Ministry of Health & Child Care, ZIMREF. care, promote equitable access to inpatient care, and expand National Health Accounts 2015: estimates for Zimbabwe. access to preventive care. Some of these financing approaches Harare (Zimbabwe): Ministry of Health & Child Care; 2017. are well articulated in the government of Zimbabwe’s 2018 6. World Bank. Health public expenditure review Zimbabwe. health financing strategy (e.g., RBF coupled with user fee Washington (DC): World Bank; 2015. 7. World Bank. Rewarding provider performance to improve removal, health insurance, etc.), the implementation of quality and coverage of maternal and child health outcomes: which will be critical to Zimbabwe’s progress toward key Zimbabwe results-based financing pilot program evidence to national and global SDG targets. As is the case for other inform policy and management decisions. Washington (DC): low- and middle-income countries, financial protection for World Bank; 2016 [accessed 2017 Nov 21]. http://www. 310 Health Systems & Reform, Vol. 4 (2018), No. 4 rbfhealth.org/sites/rbf/files/Zimbabwe-Impact-Evaluation- Nepal: a cross-sectional survey. Bull World Health Organ. Overview-Report.pdf 2014;92(10):760–767. doi:10.2471/BLT.13.126615. 8. Brewer M, O’Dea C. Measuring living standards with income 16. Mutowo MP, Owen AJ, Billah B, Lorgelly PK, Gumbie KE, and consumption: evidence from the UK. Colchester (UK): Mangwiro JC, Renzaho AM. Burden attributable to cardio- Institute for Fiscal Studies, University of Essex; 2012. metabolic diseases in Zimbabwe: a retrospective cross-sec- 9. World Health Organization, World Bank. Tracking universal tional study of national mortality data. BMC Public Health. health coverage: first global monitoring report. Geneva 2015;15:.1213–1220. doi:10.1186/s12889-015-2554-z. (Switzerland): WHO; 2015. 17. Institute for Health Metrics and Evaluation. Health data: 10. Xu K, Evans DB, Kawabata K, Zeramdini R, Klavus J, Murray Zimbabwe. 2016 [accessed 2018 Feb 21]. http://www.health CJ. Household catastrophic health expenditure: a multicountry data.org/zimbabwe. analysis. Lancet. 2003;362(9378):111–117. doi:10.1016/ 18. Chambers JD, Lord J, Cohen JT, Neumann PJ, Buxton MJ. S0140-6736(03)13861-5. Illustrating potential efficiency gains from using cost- 11. ZIMSTAT. Poverty analysis: poverty datum lines — December effectiveness evidence to reallocate Medicare expenditures. 2015. Harare (Zimbabwe): ZIMSTAT; 2016 [accessed 2017 Nov Value Health. 2013;16(4):629–638. doi:10.1016/j. 13]. http://www.zimstat.co.zw/sites/default/files/img/Poverty% jval.2013.02.011. 20Leaf%20%26%20LOGO%20December%202015.pdf 19. World Health Organization. Global health expenditure data- 12. Kim C, Saeed KM, Salehi AS, Zeng W. An equity analysis of base [database on the Internet]. World Health Organization. utilization of health services in Afghanistan using a national 2018 [accessed 2018 Feb 12]. http://apps.who.int/nha/data household survey. BMC Public Health. 2016;16(1):1226– base/Select/Indicators/en. 1236. doi:10.1186/s12889-016-3894-z. 20. ZIMSTAT. Labour force survey. Harare (Zimbabwe): 13. Zhu D, Guo N, Wang J, Nicholas S, Chen L. Socioeconomic ZIMSTAT; 2014. 2015. inequalities of outpatient and inpatient service utilization in 21. RBFHealth. Zimbabwe’s urban voucher programme. 2016 China: personal and regional perspectives. Int J Equity Health. [accessed 2018 Feb 21]. https://www.rbfhealth.org/resource/ 2017;16(1):210–219. doi:10.1186/s12939-017-0706-8. zimbabwes-urban-voucher-programme. 14. Memirie ST, Verguet S, Norheim OF, Levin C, Johansson KA. 22. Akazili J, Ataguba JE, Kanmiki EW, Gyapong J, Sankoh O, Inequalities in utilization of maternal and child health services in Oduro A, McIntyre D. Assessing the impoverishment effects Ethiopia: the role of primary health care. BMC Health Serv Res. of out-of-pocket healthcare payments prior to the uptake of 2016;16:.51–58. doi:10.1186/s12913-016-1296-7. the national health insurance scheme in Ghana. BMC Int 15. Saito E, Gilmour S, Rahman MM, Gautam GS, Shrestha PK, Health Hum Rights. 2017;17(1):13–20. doi:10.1186/s12914- Shibuya K. Catastrophic household expenditure on health in 017-0121-7. Zeng et al.: Utilization of Health Care and Burden of Out-of-Pocket Health Expenditure in Zimbabwe 311 APPENDIX Total Number of % of Planned Sample Actual Sample Province Households Households Square Root of the Share Size Size Bulawayo 165,345 0.05 0.23 558 533 Manicaland 410,082 0.13 0.37 879 850 Mashonaland 263,923 0.09 0.29 705 559 Central Mashonaland East 326,825 0.11 0.33 785 792 Mashonaland West 345,223 0.11 0.34 807 800 Matabeleland North 160,912 0.05 0.23 551 541 Matabeleland South 154,875 0.05 0.23 540 540 Midlands 359,572 0.12 0.34 823 819 Masvingo 338,153 0.11 0.33 798 723 Harare 534,106 0.17 0.42 1,003 978 Grand total 3,059,016 1.00 3.10 7,450 7,135 TABLE A1. Sampling of Households for the Survey Share of Outpatient OOP for Outpatient Share of Inpatient OOP for Inpatient % of Heath Care (%) Care (USD) Care (%) Care (USD) Total (USD) Spending Curative care 70.16 222,532,336 74.90 30,421,979 252,954,315 73.59 Rehabilitative 1.73 9,375,152 5.43 3,617,893 12,993,045 3.78 care Long-term 9.52 33,038,230 5.12 1,040,068 34,078,298 9.91 care Ancillary 1.01 9,218,738 0.56 196,751 9,415,489 2.74 services Medical 1.71 6,064,399 0.19 3,867 6,068,266 1.77 goods Preventative 9.78 2,466,154 0.76 86,780 2,552,934 0.74 care Other services 6.08 19,662,098 13.03 6,018,291 25,680,389 7.47 Total 100.00 302,357,107 100.00 41,385,629 343,742,736 100.00 TABLE A2. Out-of-Pocket (OOP) Health Expenditure by Function 312 Health Systems & Reform, Vol. 4 (2018), No. 4 % of OOP for % of OOP for Outpatient Outpatient Care Inpatient Inpatient Care Total OOP Disease/Condition Care (USD) Care (USD) (USD) % Tuberculosis 1.48 5,120,422 2.59 629,883 5,750,305 1.67 Malaria/fever 7.77 15,758,161 5.82 965,187 16,723,348 4.87 Intestinal infectious diseases 10.40 18,242,120 5.11 2,861,197 21,103,317 6.14 Human immunodeficiency virus infection and 6.07 7,450,080 2.76 222,917 7,672,997 2.23 acquired immune deficiency syndrome (HIV/ AIDS) Sexually transmitted infections (syphilis, etc.) 0.29 586,091 0.00 — 586,091 0.17 Diseases of neoplasms (tumors) 0.73 2,292,104 1.82 978,096 3,270,200 0.95 Diabetes 1.47 13,870,916 2.77 406,155 14,277,071 4.15 Nutritional diseases 0.41 2,443,046 0.36 16,213 2,459,259 0.72 Mental and behavioral disorders 0.94 3,425,010 1.52 551,079 3,976,089 1.16 Hypertensive disease 5.80 39,656,413 5.29 1,345,411 41,001,824 11.93 Heart disease 1.03 9,990,117 0.88 444,318 10,434,435 3.04 Respiratory disease 12.53 29,408,743 7.75 2,460,514 31,869,257 9.27 Digestive system disorders 3.40 16,895,942 4.07 1,865,647 18,761,589 5.46 Pregnancy, child birth, family planning 3.26 15,430,940 12.97 5,235,854 20,666,794 6.01 Skin diseases 4.04 9,571,294 3.16 627,024 10,198,318 2.97 Accidents, poisoning, and injuries 4.98 27,718,455 9.76 6,216,167 33,934,622 9.87 Eye diseases 3.50 15,415,836 2.36 726,154 16,141,990 4.70 Other 31.90 69,081,415 31.00 15,833,812 84,915,227 24.70 Total 100.00 302,357,105 99.99 41,385,628 343,742,733 100.00 TABLE A3. Out-of-Pocket (OOP) Health Expenditure by Disease