Background: Monitoring financial protection is a key component in achieving Universal Health Coverage, even for health systems that grant their citizens access to care free-of-charge. Our study investigated out-of-pocket expenditure (OOPE) on curative healthcare services and their determinants in rural Malawi, a country that has consistently aimed at providing free healthcare services. Methods: Our study used data from two consecutive rounds of a household survey conducted in 2012 and 2013 among 1639 households in three districts in rural Malawi. Given our explicit focus on OOPE for curative healthcare services, we relied on a Heckman selection model to account for the fact that relevant OOPE could only be observed for those who had sought care in the first place. Results: Our sample included a total of 2740 illness episodes. Among the 1884 (68.75%) that had made use of curative healthcare services, 494 (26.22%) had incurred a positive healthcare expenditure, whose mean amounted to 678.45 MWK (equivalent to 2.72 USD). Our analysis revealed a significant positive association between the magnitude of OOPE and age 15-39 years (p = 0.022), household head (p = 0.037), suffering from a chronic illness (p = 0.019), illness duration (p = 0.014), hospitalization (p = 0.002), number of accompanying persons (p = 0.019), wealth quartiles (p 2 = 0.018; p 3 = 0.001; p 4 = 0.002), and urban residency (p = 0.001). Conclusion: Our findings indicate that a formal policy commitment to providing free healthcare services is not sufficient to guarantee widespread financial protection and that additional measures are needed to protect particularly vulnerable population groups.
With a gross national per capita income of 1064 PPP USD [34], the landlocked sub-Saharan African (SSA) country Malawi is ranked 171th out of 189 countries on the 2017 Human Development Index [34, 36]. 71.4% of the population live under the poverty line of 1.90 PPP USD per day [35]. Nearly 80% of the population live in rural areas and rely primarily on subsistence farming [36]. The country is affected by a high degree of morbidity and mortality, mainly due to malnutrition and infectious diseases such as HIV/AIDS, malaria, acute respiratory infections, stroke and diarrhea [37, 38]. In addition, chronic conditions such as asthma, cancer, high blood pressure, cardiovascular diseases, and diabetes are on the rise, imposing an additional challenge to effective service provision in an already strained healthcare system [36, 39–41]. Healthcare provision is organized in a three-tier system, with health centers, community hospitals, dispensaries, and maternity units at the primary level serving as the first point of contact for most patients; district hospitals equipped with basic surgical facilities providing secondary care; and central hospitals located in the four largest cities of the country providing tertiary care. As described earlier, government-owned health facilities and selected private facilities (for example, those of the Christian Health Association (CHAM)) contracted by the Ministry of Health via Service Level Agreements (SLAs) are expected to provide EHP services with no fees at point of use. In principle, the EHP includes a wide range of cost-effective services for the prevention and treatment of communicable and non-communicable diseases, malnutrition, and maternal and perinatal conditions [42]. Approximately 60% of all health facilities in Malawi belong to the government, 36% to the CHAM, and the remaining 4% belong to private for-profit providers [11, 37, 43]. In 2014, total per capita health expenditure amounted to 93 PPP USD per year, equivalent to 11.4% of GDP [38]. Of this amount, donor funding accounted for 74%, domestic funding accounted for 19%, and OOPE for 7% [44]. This study used data from the first (August to October 2012) and the second (March to May 2013) round of a household survey conducted in three districts in rural Malawi (Chiradzulu, Thyolo, Mulanje) and was initially set up to evaluate the impact of a micro-health insurance scheme planned, but never implemented, by the largest Malawian micro-finance organization, MUSCCO. Details of the survey have been described before [45]. In brief, data were collected on a total sample of 1639 households selected across 114 villages using a two-stage sampling procedure. Approval was granted by the relevant ethical committees. The questionnaire collected information on households’ demographic and socio-economic profiles as well as on individual’s acute and chronic illness reporting, health care seeking behavior, including use of both formal (i.e., Western facility-based care) and informal (i.e., traditional healers, community health workers, pharmacies) health services, and related OOPE (including transport). We categorized pharmacies, community health workers and community nurses as non-formal care because, at the time of data collection, these two categories were not part of any formal curative healthcare provision program. Information from individual household members were collected by trained research assistants using a digitalized data entry system. Mothers or primary caretakers acted as proxy respondents for children below the age of 14, while households determined whether individuals aged 14 to 17 years should respond on their own or not. Table 1 contains a list of all variables included in this study, their measurement, and the expected sign of the association with OOPE. Statistics about health expenditures are listed in Table 3. Most of the variables listed in Table Table11 are self-explanatory. Variables, their measurements, and hypothesized sign of the association with out-of-pocket expenditure on medical care at formal healthcare facilities 0 = no care or informal care (incl. Self-care; community health worker or community nurse; traditional healer or herbalist) 1 = formal care (incl. Visit to either a public, a private or a not-for-profit Western healthcare facility) 0 = 0–4 years 1 = 5–14 years 2 = 15–39 years 3 = 39+ years 0 = male 1 = female 0 = no formal education 1 = any formal education 0 = other 1 = being household head 0 = no chronic illness reported 1 = chronic illness reported 0 = no perceived limitation on routine activities 1 = perceived limitation on routine activities 0 = no hospitalization 1 = hospitalization 1 = poorest (1st wealth quartile) 2 = poor (2nd wealth quartile) 3 = less poor (3rd wealth quartile) 4 = least poor (4th wealth quartile) 0 = 0–5 members 1 = 5+ members 0 = rural 1 = urban Individual out-of-pocket expenditureb on medical treatment among individuals seeking care at a healthcare facility (n = 1884; zeros excluded) SD Standard Deviation, MWK Malawian Kwacha aValues are expressed in MWK (249.11 MWK = 1 USD at time of data collection (World Bank, 2019)) bWinsorized direct costs (excluding zeros): Replacement of the right outliners with the 95th percentile (Facility healthcare expenditures) or the 95th percentile (transport expenses) cOut-of-pocket expenditures on medical treatment for formal care include all consultation and treatment expenses, including: laboratory tests (x-rays etc.), drugs (tablets, injections, infusions, topical preparations etc.), medical devices (crutches, glasses etc.), as well as informal fees In line with our objective to explore the extent to which direct payments at point of use persist within the framework of a free healthcare system, we defined OOPE (i.e., individual nonzero healthcare expenditures) for people seeking care at formal healthcare facilities as our primary outcome. In line with the abovementioned definition of OOPE, we defined having sought care at a formal healthcare facility as our outcome for the selection model. More specifically, we focused on the sub-sample of individuals reporting at least one acute illness episode over the course of the prior 4 weeks and distinguished individuals who sought formal care at a health facility (coded as 1) from individuals who visited a community health worker or community nurse, a traditional healer or herbalist, or who did not seek care at all (coded as 0). Our OOPE variable included only consultation and treatment expenses, such as expenses for laboratory tests (X-rays etc.), drugs (tablets, injections, infusions, topical preparations etc.), medical devices (crutches, glasses etc.), and any additional formal or informal fees paid. In spite of being aware that transportation expenses represent an important component of the financial burden imposed on households in rural African settings [46–48], we excluded them from the computation of our OOPE variable, since our objective was to estimate the financial protection granted specifically by the Malawian free healthcare policy, and at the time of study (and until today), there is no direct provision to include travel expenses in the EHP. Still, we report transportation expenses separately to provide a comprehensive picture of the financial burden illness episodes impose on Malawian households. A detailed listing of the different medical expenses for medication, laboratory etc. was not available. We adopted distance to the nearest healthcare facility, measured as a straight-line distance using GPS coordinates [23, 49], as the selection variable since prior studies in Malawi [21, 50] and elsewhere in Sub-Saharan Africa [51, 52] have identified it as a major determinant of healthcare seeking, but not of OOPEs on medical treatment. Accordingly, we did not include distance as an explanatory variable in our primary model, but only in our selection model. Explanatory variables were selected on the basis of prior evidence indicating their association with OOPE [21, 23, 53] and of pragmatic considerations regarding availability in the specific database at our disposal. We divided age into four groups in alignment with the defined ages rages for child labor [54] to facilitate the interpretation of our findings for policy purposes. We included a measure of education, assuming that better educated individuals are generally more empowered to make decisions on their own or their dependents’ health and may therefore face higher OOPE on medical treatment. We assigned the educational status of the household head to children under 14, since we assumed that they would be the ones mediating the decision to seek care for minors [16, 55]. We determined whether the individual reporting an illness was the household head himself/herself, because prior research indicates a higher propensity to seek formal care [56] and incur higher OOPE [20, 23, 57]. We included a measure of whether the individual reported any chronic illness, defined as any condition a person suffered from for more than 3 months and not restricting it only to non-communicable diseases as done in previous studies [23]. We postulated that individuals with an underlying chronic condition might be exposed to important co-morbidities and hence face higher OOPE. Similarly, we assumed that individuals might face higher OOPE for more severe conditions, and therefore included three different measures of illness severity: self-reported illness duration, resulting degree of limitation imposed on routine activities, and hospitalization. Further, we determined the number of accompanying persons who assisted an ill individual when seeking care, since, based on prior qualitative evidence also from Malawi [55], we postulated that individuals affected by more severe conditions may require additional support in seeking care and, as a consequence, may incur higher OOPE. Household wealth quartiles were used as proxy of household socio-economic status. We computed a wealth index by aggregating information on physical household infrastructure, durable assets, and owned animals using Multiple Correspondence Analysis [58, 59]. In line with prior literature [21, 22], we hypothesized that, given a higher capacity to pay, better off individuals would face higher OOPE than poorer individuals. We also included a measure of household size, since we assumed that decisions on intra-household resource allocation might be more complex for larger households resulting in lower ability to pay for the single individual member. We included the location of the household, since we assumed that in urban settings individuals might incur higher OOPE due to a greater availability of diagnosis and treatment options, in line with previous findings [21, 22, 60]. It ought to be noted, however, that in our study urban setting only refers to small district towns and not to cities such as Lilongwe or Blantyre. It also ought to be noted that we would have liked to differentiate OOPE for people seeking care at public vs. private (including not-for-profit) facilities, but unfortunately information on facility ownership was missing for over two-thirds of our sample, possibly suggesting that individuals may not be able to recall facility ownership. Prior to beginning our analysis, we pooled all illness episodes detected in the 2012 and in the 2013 survey rounds into a single sample. Then, using the pooled sample, we relied on descriptive statistics to identify a sample distribution for all variables included in our study. We calculated mean, standard deviation (SD), median, and range values for OOPE on medical treatment (our primary outcome) and for expenditure on transport. To account for the heavily right-skewed distribution of OOPE [61], we used boxplots to detect outliers [62, 63]. To handle the outliers, we used winsorization, since truncation or trimming can lead to substantial bias for resulting mean values [61]. Accordingly, we replaced the upper 5% of outliers with the respective highest value of the sub-sample without outliers (i.e., 95% percentile value) [64]. This approach allowed us to keep the entire sample while avoiding a situation where extreme outliers would distort the findings of the regression analysis [61, 62, 64]. Last, we relied on a Heckman selection model to identify the determinants of OOPEs on medical treatment conditional upon having sought formal care at a healthcare facility. We relied on a Heckman selection model rather than standard linear regression, since the outcome of interest, OOPE on medical treatment, could only be observed for individuals who sought care at a formal health facility in the first place [28, 30]. Through the application of a two-step statistical approach, the Heckman model offers a means of correcting for non-random samples. In the Heckman model, OOPE on medical treatment for individual i with the attributes xi is defined as (primary equation): under the condition that the individual sought care at a formal health facility (selection equation): where β and ∂ are the coefficients of the attributes in the primary and in the selection equation. The selection variable is zi (here the distance to the nearest health facility), γ its coefficient and the following applies: This means that if there is no self-selection effect (ρ = 0), the selection and the regression equation can be analyzed separately. In our case, however, we could not reject the null-hypothesis and could identify a selection effect, which could be effectively accounted for by the selection variable, i.e., distance to the nearest healthcare facility (Wald test of independence significant on level 1% and selection variable significant on level 5%). To take possible intra-correlation on individual level into account, robust standard errors (SE) were estimated. Data analysis was performed using STATA 14.
N/A