Persistent inequalities in health care services utilisation in Brazil (1998–2019)

Efforts to achieve timely and effective utilisation of health care services present an ongoing challenge to the goal of providing universal health care coverage to those living in low- and middle-income countries [1]. Both financial and non-financial barriers can impede the appropriate use of these services and may lead to poorer overall health outcomes [2‐4]. Moreover, there are substantial inequalities regarding access to and utilisation of health care services even in countries that provide universal health care coverage [3]. In many cases, the most economically-advantaged members of the population utilise the vast majority of health care services in a given region. Inequalities are particularly evident with respect to the use of preventive health care services [5] and can be observed in both high-income countries that provide universal health care coverage as well as in middle-income countries [5, 6]. Inequalities in health care services utilisation that are skewed towards those of higher socioeconomic status frequently lead to substantial differences in overall health outcomes. To address these concerns, the evolution of health care services utilisation needs to be monitored. This is an essential first step towards the design of targeted policies that focus on reducing health care and outcomes disparities [2, 3].

Brazil presents a unique setting in which inequalities of health care services utilisation might be analysed. Brazil currently has one of the highest levels of economic inequality worldwide (Gini index of 48.9 in 2020) [7]. Likewise, although Brazil supports a single-payer, publicly funded national health service known as the Unified Health System/Sistema Único de Saúde (SUS), over half of the total expenditure on health care comes from private sources [8].

The SUS is a tax-based, non-contributory public health care system that provides services at all levels that are free of charge at the point of delivery. The system has achieved significant gains over the past 30 years; improvements in both coverage and access to health care services have resulted in better health outcomes overall. [9] A central feature of the national health service in Brazil is primary health care, mainly the Family Health Strategy (FHS) which covers > 60% of the population according to the coverage parameters defined by the Ministry of Health. One Family Health Team which includes a nurse and community health care workers is expected to provide care for approximately 3450 individuals [9]. The availability of primary health care services has clearly improved the health of the Brazilian population [10, 11] and has contributed to a reduction in health inequalities [12]. However, access to specialised care remains a major challenge in the public health care sector and long waiting times persist [8, 13]. Furthermore, while SUS provides universal health coverage, 28.5% of the population (primarily individuals with higher incomes who are formally employed and residing in urban centres) are also covered by one or more private health insurance schemes [14].

The highly fragmented nature of the current health care system and persistent regional disparities remain major barriers to more equitable access to health care services throughout Brazil [15]. Brazil has large geographical variations with unequal distribution of infrastructure, human resources, access to medications, and coverage by national health care programmes [16‐19]. Variations in the supply of health care resources coupled with the two-tiered (public versus private) financing system has led to disparities in both access to and quality of health care delivered as well as protection from undue financial risk [15, 20]. Results from previous studies suggest that the limited availability of health care providers, infrastructure, and medications in underserved areas remains a major challenge [11, 15‐19]. Likewise, the financial burdens associated with seeking care (e.g., transportation costs) represent an additional barrier that prevents the appropriate utilisation of health care services [21, 22].

Existing evidence suggests that the inequalities in health care services utilisation largely reflect the overall socioeconomic inequalities in Brazil [20, 23‐27]. Although results from previous studies revealed a gradual improvement in reducing the inequalities in health care services utilisation between 1998 and 2008, the existence of private health insurance schemes has been identified as one of the largest contributors to the persistence of this problem. Most of these larger published studies cover only this earlier decade [20, 23‐25]. By contrast, studies that include the more recent data (2013–2019), have focused on outcomes associated with specific health care services (e.g., preventive care) [26] or specific segments of the population (e.g., the elderly) [27].

The goal of this study is to evaluate socioeconomic inequalities and their role in the differential utilisation of health care services in Brazil from 1998 through 2019. Our study includes both past as well as more recent information focused on the utilisation of both preventive and treatment services. We also evaluated the role of private health insurance as well as the known socioeconomic and geographical disparities on health care services utilisation in Brazil.

The results shown in Table 1 summarise the five variables of interest, including (1) any visits to a physician during the past year; (2) any hospitalisations during the past year, excluding those for labour and delivery; (3) any surgical procedures during the past year; (4) use of Pap smears during the past three years by women between 25 and 59 years of age; and (5) use of mammograms during the past two years by women between 50 and 69 years of age. In 2019, the health care services used most frequently were physician visits (80.8%; 95% [Confidence Interval] CI, 80.4%–81.3%). By contrast, only 6.5% (95% CI, 6.2%–6.9%) of the adult population (18 years of age and older) reported being hospitalised for reasons other than labour and delivery. Interestingly, when hospital admissions for labour and delivery were included in this calculation, 7.6% of the population reported at least one hospitalisation during the past year (95% CI, 7.2%–7.9%; see Additional file 1 Appendix Table A1). Likewise, only 3.0% of the population (95% CI, 2.7%–3.3%) underwent a surgical procedure during the past year.

With respect to preventive screening services, 58.3% (95% CI, 56.9%–59.6%) of those eligible underwent a mammogram while 82.3% (95% CI, 81.6%–83%) had a Pap smear. Over time, the utilisation of health care services (defined as the share of respondents that have used a specific service) has increased (Table 1). The largest increase in health care services utilisation between 1998 and 2019 was in the frequency of physician visits at 23.4 percentage points. By contrast, the utilisation of inpatient services remained virtually unchanged during this period. The use of preventive services also increased between 2003 and 2019 by 8.7 and 13.5 percentage points for Pap smears and mammograms, respectively. The remaining variables are summarised and discussed in Additional file 1 Appendix Table A2 and Table A1.

Disaggregation of these results by socioeconomic status provides some explanation for our findings (Additional file 1 Appendix Figure A1). First, the number of physician visits, surgical procedures, and use of preventive services per capita was highest among the respondents in the highest income quintiles; this pattern has remained stable over time. By contrast, the prevalence of hospitalisations across income quintiles has changed over time. While those in the lowest income quintile reported fewer hospitalisations in 2019 compared to 1998, the prevalence of hospitalisation among those in the most wealthy segments of the population increased slightly during the same period.

Erreygers-corrected CInds and the concentration curves for each of the five variables of interest are shown in Table 2 and Additional file 1 Appendix Figure A2, respectively (as well as Table A3 for the hospitalization variables). As noted above, a CInd with a negative coefficient represents a situation in which inequalities are more prevalent among those of lower socioeconomic status (i.e., pro-poor). By contrast, a positive coefficient represents an unequal distribution of the outcome variable in a direction that favours those of higher socioeconomic status (i.e., pro-rich). Our analysis of the data collected in 2019 revealed that all health care service utilisation variables were skewed significantly in the pro-rich direction, with Erreygers-corrected CInds of 0.114 (95% CI, 0.103–0.124) for physician visits, 0.010 (95% CI, 0.004–0.016) for hospitalisations, 0.016 (95% CI, 0.012–0.021) for surgical procedures, 0.137 (95% CI, 0.121–0.152) for Pap smears, and 0.265 (95% CI, 0.235–0.294) for mammograms.

The change of CInd values over time reveals somewhat different trends. CInds for preventive screening services decreased over time, indicating a reduction in the utilisation gap between rich and poor. By contrast, although the CInd values increased over time for surgical procedures, no statistically significant differences were identified for physician visits. Furthermore, inequalities with respect to hospital care, as previously depicted in the descriptive charts showing service utilisation by income, indicated a significant pro-poor distribution in 1998 (Erreygers-corrected CInds of -0.004, 95% CI, -0.007–0.001) followed by a reversal to a significant pro-rich inequality in 2019 (Erreygers-corrected CInds of 0.010, 95% CI, 0.004–0.016).

Figure 1 presents the results of the decomposition analysis of the CInds determined from the results of the 2019 survey that document the relative contribution of each critical variable to the overall CInd. The full set of results including all potential variables is included in Additional file 1 Appendix Table A4. The results of the decomposition analysis revealed that the use of private health insurance and socioeconomic status are the two main factors associated with the inequality of health care services utilisation in 2019. Our findings revealed that the use of private health insurance coverage contributed ~ 30%–140%, holding socioeconomic status constant, while differences in socioeconomic status explain ~ 30%–70% of the observed inequalities across all outcomes. Region of residence and educational attainment had positive but relatively smaller contributions to the overall inequality in health care services utilisation. The contributions of all other variables (notably, those identified as predisposing variables) were comparatively small compared to those discussed above.

The inclusion of regional fixed-effects in our analysis of the 2019 wave was intended to capture the contribution of all non-observed differences across the location of residence (e.g., availability of medical equipment, inequality, and/or general standard of living). We performed a series of robustness checks that included some of the non-observed factors that vary at the aggregate level as covariates. The results of this analysis are shown in Additional file 1 Appendix 3 – Robustness tests. In the first test, we substituted regional fixed-effects for state-level covariates that indicate poverty, income inequality, and availability of mammographers, as well as, whether the household was registered at a primary health care facility. The results revealed that state-level poverty, income inequality, and availability of mammographers were positively associated with the pro-rich distribution of utilisation of health care services, whereas being registered in a public health care facility was negatively associated, albeit at a much smaller magnitude. Crucially, the magnitude of the contributions of all other contributing factors does not change significantly. In the second test, we used the Wagstaff correction instead of the Erreygers correction of CInd [36]. The results of this secondary analysis followed the same pattern as our primary findings. Finally, in the third test, we analysed additional preventive care outcomes to determine whether our results represented a more general trend. Among our findings, the decreasing trends we observed with respect to the use of mammograms were also identified in other services that targeted a specific segment of the adult population (i.e., blood sugar tests), but not for other services (i.e., blood pressure and cholesterol testing). These results suggest that the decreasing trends we observed with respect to the use of mammograms, which is a preventive service provided to women only, will also be relevant to services that target the adult population (i.e., blood sugar tests) and also may reflect improvements in primary care services over time [39, 40]. However, the fact that inequalities persist in other preventive services suggests that many of these improvements might not be general in nature.

The main objective of this study was to evaluate health care services utilisation in Brazil and identify the main contributors to inequalities over the past 20 years (1998–2019). The data used in this study were collected from two nationally-representative surveys, PNAD conducted in 1998, 2003, and 2008, and PNS conducted in 2013 and 2019, respectively. The results of our analysis revealed a persistent pro-rich inequality with respect to most health care services utilisation outcomes that has diminished somewhat over time. The only exception to this trend is hospitalisations, which moved from an earlier pro-poor inequality to one that is currently pro-rich. Our findings revealed that while the wealthiest individuals were more likely to use preventive, outpatient, and inpatient care services compared to members of the lower socioeconomic groups, there was a consistent downward trend in pro-rich inequalities specifically with respect to the use of preventive health care services. In fact, the use of mammograms and Pap smears, which were the services with the largest initial level of inequality in utilisation were those that showed the largest decline, with both CInds falling over 30% during this period. The results from the decomposition analysis revealed that enabling indicators (e. g., socioeconomic status and private health insurance coverage) are the factors most significantly associated with the observed inequalities. Furthermore, our results are consistent with one another at both national and regional levels. Specifically, we found that the magnitude of inequality with respect to health care services utilisation was particularly large in the North and North-East regions of the country. Socioeconomic factors and private health insurance coverage were also identified as the largest contributors to inequality in these regions. Overall, while our analysis did not permit us to identify the factors driving this downwards trend along with persistent socioeconomic inequalities in health care utilisation, one hypothesis is that they may be related to the large expansion of public policies that granted access to primary health care to a large portion of the population specifically within a context of persistently high income inequality. For example, in 1998 just 5.6% of the population was covered by the FHS, which is Brazil’s main public primary health care program; by 2019, this percentage had grown to 62.7% [41]. Whereas income inequality has decreased between the two time points, Brazil has consistently ranked among the ten countries with the highest level of income inequality worldwide, as measured by the Gini Index [7].

Overall, our results are consistent with those reported in studies from other Latin American countries that have quantified income-related inequalities in health care utilisation [42‐48]. Overall, these reports have identified a consistent pro-rich inequality with respect to the utilisation of physician visits and preventive care services, specifically Pap smears and mammograms, throughout Latin America. By contrast, inequalities in the use of hospital care are mixed. Similar to our findings, utilisation of hospital-based health care skews pro-rich in both Mexico [42] and Peru [45]. By contrast, the utilisation of these services is skewed toward the poor in Chile [44]. There was no evidence of inequality in Colombia in a study that specifically targeted older members of the population [46]. In addition, although our findings are consistent with previous results reported for Brazil [20, 23‐27], we have observed several new trends during the past decade. Specifically, our results revealed that the earlier pro-poor inequalities in hospitalisations converted to pro-rich over time when we excluded deliveries. Moreover, we found that the pro-rich inequality in utilisation of physician visits and surgical procedures remained unchanged from 2013 through 2019, based on data that were not evaluated in previous publications. Finally, our results that focused on the utilisation of preventive care services, specifically Pap smear and mammogram screening, revealed a continuous trend of declining inequalities, although these remained at higher levels compared to the other health care services. Collectively, these results suggest that there are still significant challenges to be met to provide equitable access to health care in low- and middle-income countries and that these problems persist even after the establishment of a large publicly-funded national health system that provides universal health care coverage to all.

Our finding of pro-rich inequalities in hospitalisation services (excluding those associated with labour and delivery) in 2019 complements the results reported in the existing literature. By covering the entire adult population (instead of just the elderly population [27]) and excluding deliveries from total hospitalisations, our results provide a clearer understanding of the direction of inequality with respect to the need-based utilisation of inpatient care services. These findings indicate that the rate of utilisation for those 18 years of age or older in the lowest income quintile decreased significantly across most areas of Brazil, particularly in the comparatively impoverished North and Northeast regions. Furthermore, decomposition analysis of the CInds for hospitalisation services suggested that the observed pro-rich inequality may be largely driven by the availability of private insurance to individuals in the higher income groups.

Another important finding relates to the downward trend in the pro-rich inequality observed in the utilisation of preventative health care services (i.e., mammograms and Pap smear) between 2003 and 2019. This result is particularly relevant because it coincides with a period during which both the coverage and quality of primary health care services increased in Brazil [9, 10]. Moreover, when we include a covariate in the model indicating coverage of the household by the FHS, the results of the decomposition of CInd for the preventive care services, specifically Pap smears, revealed that being covered by this service was negatively associated with the pro-rich inequality in utilisation of the health care services. This result indicates that the program has effectively targeted the poorest groups (see Additional file 1 Appendix 3 – Robustness tests). This result also complements findings from previous studies that have included these recent findings (up to 2019) [25‐27]. Although these findings suggest that the gap between the rich and the poor with respect to the use of preventive health care services has been closing, there remains significant variation across the regions in the country. Persistent geographic disparities in the use of mammograms could be explained at least in part by the unequal distribution of imaging equipment. Amaral et al. [49] reported that while overall mammography capacity is sufficient to meet existing needs in Brazil, specific resources are not distributed proportionally based on the population of each region. Notably, only the South region has mammography equipment that is sufficient to serve the entire population.

As noted above, our findings documenting health care services utilisation at the regional level are consistent with those that have emerged at the national level. Both analyses point to significant inequalities, specifically in the North-East, when compared to the rest of the country. This is most likely a result of both economic and healthcare-related factors. When ranked by region, the North-East has one of the lowest levels of socioeconomic development. The North-East is currently ranked last of the five designated regions in Brazil with respect to Gross Regional Product per capita [50]. The high concentration of inequalities in the North-East may also reflect the limited availability of health care infrastructure. Of note, our findings suggested a large improvement in utilisation of preventive care services among poorer women across all regions of Brazil, with the largest improvements reported for the Midwest region and revealing a clear trend of improvement over time across all regions [51, 52].

When considering barriers to health care, the results of our study are comparable to those reported previously. Collectively, our findings suggested that the availability of private health care insurance is positively associated with the pro-rich inequalities in health care services utilisation in Brazil. The use of private health insurance is tied directly to employment status and the ability to pay for care and is thus concentrated in regions with the highest income levels. The availability of private insurance tends to exacerbate inequalities in health care services utilisation and provides users with advantages over those who rely solely on public health services [53]. This may lead to an implicit two-tier system, in which those with higher income have ‘double coverage’ from both private health insurance and the national system. While two-tier systems exist in other Latin American countries [54], Brazil’s case is unique given that an extensive private system co-exists with a national health service that provides health care coverage for all. Private medical insurance spending in Brazil is the highest in the region; this is coupled with public spending on healthcare which is well below the average for Latin American nations [7]. Focusing specifically on Brazil, where there is a large body of evidence indicating the existence of ongoing socioeconomic inequalities in many dimensions of health (i.e., inequalities in health according to socioeconomic status), our results suggest that lower rates of utilisation of preventive health care services might be one of the reasons why lower-income individuals present, on average, with inferior health outcomes.

This study has some limitations. While Andersen’s behavioral model of health services utilisation [30] provides a framework that permits us to consider many of these factors and their interrelationships, it exhibits only limited capacity to generalise these results. The Andersen model also offers guidance toward an appropriate interpretation of these results by facilitating the identification of the factors that contribute to income-associated inequalities in health care services utilisation. Likewise, as applied to our results, the Cind and decomposition analysis facilitate the acquisition of correlational data and do not provide insight into causal inference. Furthermore, socioeconomic inequalities measured by the Cind (i.e., based on income distribution) represent only a partial approach to health care inequalities. While the focus of our analysis is on income-related inequalities, other potentially relevant factors, such as racial and gender-based inequalities have not been considered. For example, results from a previous study revealed that multiple inequities play a much larger role than income-related inequity alone [55]. There are some further limitations associated with the decomposition analysis used in this paper as previously reported [56]. First, it is one-dimensional because it focuses on the degree of variation in health but ignores rank; second, it can only correctly decompose one form of rank dependent index because it assumes a constant weighting function [56]. Finally, as discussed in the text, we needed to exclude individuals residing in the rural areas of the North region to maintain sample consistency over time. This may have led to an underestimate of inequalities, as access to health care services is particularly restricted in this part of the country [57]. Nonetheless, and despite these limitations, we present a full overview and analysis of the inequalities in health care services utilisation in Brazil over two decades. Our study has revealed the main factors that contribute to these inequalities and identified private health insurance as a key driver of these observations in Brazil.

Our results support the view that efforts to achieve equality in the delivery of health care services need to be intensified and should rank high among policy concerns in Brazil. Our findings also have implications for policy development in other middle-income countries that provide universal health care coverage. While the Brazilian system has a unique public/private health care mix, our evaluation still offers valuable lessons for other middle-income countries in Latin America and beyond. Specifically, our results reveal that reductions in inequalities in access to health care can be achieved in a system in which public and private financing systems co-exist. However, caution should be exercised as some of the challenges associated with a dual (public/private) system tend to be persistent. First, our findings are consistent with results reported by several Organisation for Economic Cooperation and Development (OECD)-aligned countries and the Americas that have achieved universal or near-universal health care coverage for their populations. Specifically, OECD findings revealed that inequalities in access to health care services persist over time despite increases in health care expenditures and investments directed at expanding critical infrastructure [2, 3, 58, 59]. Second, our findings revealed that, although utilisation of health care services has increased over time, inequalities in health care services utilisation remain high and favour those of higher socioeconomic status; this is particularly the case for the utilisation of preventive care services. While the reduction in inequalities observed may have been the direct result of programmes and policies implemented over the last few years (e.g., FHS), this question goes beyond the scope of this paper. More importantly, the first decade of the 2000s coincided with significant economic growth. This may have directly or indirectly resulted in an overall reduction in the pro-rich inequality with respect to access to the selected interventions. Policy interventions might be developed to target both financial and non-financial barriers, particularly those with the greatest impact on poor and vulnerable individuals residing in the lower-resourced regions of the country. Moreover, given the apparent associations between these inequalities and the availability of private health insurance, future research might be designed to explore the possibility of a causal relationship between the availability of this type of health care coverage and the utilisation of health services.