Correlates of out-of-pocket and catastrophic health expenditures in Tanzania: results from a national household survey

We employed a cross-sectional study design. This study uses secondary data from the first round of Tanzania National Panel Survey (TZNPS), implemented by National Bureau of Statistics of United Republic of Tanzania [24]. The first wave of the survey was conducted between October 2008 and October 2009. TZNPS employed multistage, stratified, random sampling to generate nationally representative samples. The first stage of sampling units comprised 410 clusters of enumeration areas in urban and rural areas. The second stage involved 3265 households. All members of the selected households were invited and those who provided verbal informed consent, were included in the study. The final sample included 8297 participants, of which 7279 were aged between 18 and 59 years and 1018 participants were above 60 years. The details of the participants’ sampling information are available elsewhere [24]. To obtain individual and household level data, trained field staffs conducted person-to-person interviews employing structured questionnaires and recorded anthropometric measures [24]. At the individual level, the data included socioeconomic characteristics, self-reported disabilities, health service visits and OOP health expenditures by type of health facilities. Data on household characteristics, availability of durable assets, household’s total health expenditures and basic facilities like access to safe drinking water and toilet facilities for the past 4 weeks were recorded. The OOP health expenditures included the utilization of private outpatient clinics, preventive health services, over-the counter medications and use of traditional healer services.

Based on Anderson’s model, the factors determining the health care demand were categorized into predisposing, enabling and need factors. We included age, gender, marital and occupation status variables as predisposing factors; education, asset index as proxy measures for wealth were considered as enabling factors to OOP health expenditures. The asset index was estimated using eighteen variables, which provided information on housing characteristics, type of water supply, sanitation in the household, and ownership of the household’s durable assets. Obesity, presence of visual defects, hearing defects, limb defects, self-reported functional disabilities and psychological morbidity were included as need variables.

We initially analyzed the study variables using descriptive statistics. We used the statistical technique, Principal Component Analysis (PCA) to construct the asset index [25]. We performed Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy. The obtained value of 0.80 was liable for conducting PCA. Scoring factors were derived from the first principal component, which was a linear index of variables with common information, and this score was used to construct the household’s asset index. We used Mann–Whitney U and Chi-square test to assess the bivariate associations between the study variables. Our outcome variable OOP health expenditure was typically non-parametric and positively skewed with influential outliers. Traditional ordinary least square regressions with log-transformation and retransformations are too inconsistent to handle skewness and provide inferences in natural units of mean expenditures [26]. Generalized linear regression models (GLM) are flexible to model skewed expenditure data and avoid the issue of outcome transformation [27]. Survey GLM with gamma distribution and log link function [28, 29], was employed to assess various determinants of OOP health expenditures and account for the complex survey design.

To analyse the determinants of catastrophic health expenditure, we used non-food expenditures of the household as a close measure for the household’s capacity to pay. We estimated the share of OOP health expenditure in the household’s non-food expenditure (E _j ) to define catastrophic health expenditure:

Where, HE _j is the household’s average monthly OOP health expenditure; NFE _j denotes the household’s average monthly non-food expenditure. We dichotomized and coded the outcome as ‘1’ , when E _j exceeded the threshold of 40% [7], ‘0’ if otherwise. We employed survey logit model to assess the determinants of catastrophic health expenditure in households. We performed all analyzes using statistical software STATA 12.1.

A total of 7279 adults and 1018 older participants were included in the present study. Majority of the study participants were women. Most of the adult participants had primary education (86%), while only 14% had more than primary education. Among the older participants, the majority were women, lacked formal education and had past occupation as manual laborers. Only 35% of households had safe drinking water access, while 92% households had access to toilet facilities. The average family size of the households was 5.1 (SD 2.8).

The self-reported monthly income differed significantly between the adults and older participants (Mann- Whitney U = 3442425; p < 0.001). Adult women (U = 5559127.0; p < 0.001) and those who worked as manual laborers (U = 5328429.5; p < 0.001) had a significantly lower income. Older women had a significant lower income compared to older men (U = 115054.0; p < 0.001). Adult women (χ ² = 6.3; p < 0.001) and adult participants (χ ² = 5.4; p < 0.001) who lacked formal education reported higher disability. However, older women did not report higher disabilities (χ ² = 2.8; p = 0.09).

About 1030 (14.2%) of the adult participants had visited medical health services and 117 participants (1.6%) had visited traditional healers within the past month. Older participants, who visited medical health services, were 190 (18.7%). The adult participants utilized traditional healers or faith-based services significantly more than the older participants (χ ² = 5.7; p = 0.01). However, private healthcare utilization did not significantly differ among the adults and older participants (χ ² = 3.6; p = 0.06). The mean OOP health expenditures among the adult participants, over the past month was 2.2 (SD 9.5) US$, while the older participants’ OOP health expenditure was 0.7 (SD 6.3) US$. Total OOP expenditures were spent mostly on health professional fees 0.8 (SD 7.5) US$, followed by over -the-counter medications 0.2 (SD 1.1) US$ and hospital admissions 0.1 (SD 3.6) US$.

We present the bivariate and multivariate analyses for the correlates of adult OOP health expenditures in Table 2. Factors such as female gender, occupation as unskilled manual laborer, hearing defects and functional disability were significantly associated with higher adult OOP health expenditures. Traditional healer visits increased the OOP health expenditures among the adult participants. The higher OOP health expenditure among women could be attributed to expenditure on their reproductive health, because women aged above 45 years, had significantly less OOP health expenditures (β = −0.37; 95% CI −0.68, −0.04; p = 0.03), than men.

We present the bivariate and multivariate analyses for the factors associated with OOP health expenditures among the older participants in Table 3. Presence of visual impairment, Functional disability and traditional healthcare visits significantly increased the OOP health expenditure, after adjusting for potential confounders. Older participants with occupations as unskilled manual laborer spent significantly less out-of their pockets, despite a higher prevalence of disability. Among the older participants aged above 65 years (n = 744), unskilled manual laborers (β = −0.71; 95% CI −1.30, −0.12; p = 0.01) and those who lacked formal education (β = −0.62; 95% CI −1.21, −0.02; p = 0.04) had lower OOP health expenditures. While, factors such as functional disability (β = +0.89; 95% CI +0.32, +1.46; p = 0.02), and visits to traditional healers (β = +1.40; 95% CI +0.41, +2.39; p = 0.006), increased their OOP health expenditures.

The proportion of households that experienced catastrophic health expenditure was 18%. We present the bivariate and multivariate analyses for the correlates of the households’ catastrophic health expenditure in Table 4. During the multivariate analysis, factors such as household members with chronic illnesses, household head’s occupation as a manual laborer, healthcare visits to a traditional healer, a household size of more than five and domestic violence against women significantly increased the likelihood of experiencing catastrophic health expenditure, after adjusting for all bivariate significant variables. Wealthy households were spared catastrophic health expenditure.

Women of reproductive age had a significantly higher OOP expenditure than the older women in our study. This is in line with a Tanzanian study, which reported the prevalence of high informal payments and increased need for health service utilization among younger women and their children [30]. With the meagre resources and a higher prevalence of disability, these women are at great risk of being confronted with the financial burden of OOP health expenditures. Many people rely on the services of traditional healers and face increased out-of-pocket health expenditure in Africa [22, 31]. Our findings show that OOP costs for traditional healers are high among the adult participants. Culturally ingrained beliefs for various illnesses and inconsistencies in health service access urge them to seek the traditional healers with high OOP expenses [32]. Akin to earlier studies, obesity [33] and disability [34, 35] were also associated with higher OOP health expenditures in our study population. Disabilities during youth may be attributable to poor health status, related poor nutrition, living conditions and a high incidence of infectious diseases. Disabilities due to physical, functional, and psychiatric morbidities can affect the nature of health service utilization and demand high OOP health expenses among older people. A higher prevalence of disabilities among elderly are known to be associated with increased OOP health expenditure in LMICs [36]. Economic disadvantages related to the older peoples’ disabilities can exacerbate their untreated medical illness, increase dependency and restrict their access to health services. Due to the absence of social security systems, the older people either suffer financial burden through OOP payments or remain disabled without seeking health care.

The socioeconomic status of both adult and older participants was inversely associated with OOP health expenditures in our study. African studies [37, 38] have consistently reported the differences in health care utilization and expenditure in terms of socioeconomic status. A study from Tanzania [39] also found that, occurrence of higher health care seeking and subsequent expenditure among the people with a higher socioeconomic status. OOP health expenditures remain an obstacle for health service utilization, and consistently increase the burden of diseases among people with poor literacy and low income laborers compared to the more affluent. Economic deprivation worsens the health status of these individuals and leads to decreased labor productivity [19]. The vicious cycle of further economic loss and increased vulnerability to illnesses predisposes to catastrophic health expenditures [40].

OOP expenditure depends on the resources and healthcare needs of an individual. These resources and needs have a complex bidirectional interaction. The rich have better living conditions and better health, thereby reducing their needs for OOP health expenditure. However, they are endowed with better economic resources, education, enhanced awareness towards the need for health care and the thresholds at which they access that service are correspondingly lower. On the contrary, the poor and those with low literacy are less likely to use health services [16], especially preventive services, leading to worse health outcomes and subsequently increased need for OOP expenditures [40]. The need for OOP health expenditure that arises out of poor health determine the link between the economic inequality and inequities in the delivery of health care to people in LMICs. Gender inequality and unbalanced access to education additionally contribute and sustain healthcare inequities. Affordability remains a stronger determinant for OOP health expenditure than the valid need for essential health services.

Tanzanian households had a higher prevalence (18%) of catastrophic health expenditure at the 40% threshold, compared to the observed prevalence in Burkina Faso (10.8%) [7] and a lower prevalence than in Nigeria (27%) [41]. Our results show that a low socioeconomic status of the household increased the probability for catastrophic expenses. The association between domestic violence against women and high health expenditures are well documented [42]. Women’s welfare is vital to the household and injustice against them will affect their income contribution, health and well-being. Our finding chronic disease morbidity as an important determinant of catastrophic health expenditure has also been reported elsewhere [7]. The increased prevalence of chronic diseases in LMICs demands cost-effective control programs similar to infectious disease control programs to reduce the risk of catastrophic household health expenditure.

Our study highlighted the various determinants, potentially responsible for high OOP health expenditures among the adult and older population in Tanzania. To ensure equitable health care delivery for all, OOP health expenditure should be minimized. Hence, to reduce the financial barrier and improve the accessibility to health services the following are suggested,