This project investigates the relationship between healthcare spending and the distribution of health outcomes in Brazil and seeks to answer the question: Does increased healthcare spending correlates with improved and equitably distributed health outcomes?
It adopts a spatial and temporal comparison at both the state and municipal levels, combined with a case study to ensure better comparability. The project begins by analyzing healthcare expenditure growth across states and its distribution across spending categories. It then conducts a correlation analysis at the municipal level. To ensure better comparability, the project includes a case study on adequacy of prenatal care, paired with a corresponding expenditure category. This approach enables a granular analysis of targeted spending outcomes while highlighting disparities across racial groups.
This project relies on publicly available data from the Brazilian Ministry of Health. Health expenditure per capita (in Brazilian reais) is sourced from the Public Health Budget Information System (SIOPS), while spending by category is obtained from the Ambulatory Information System (SIA). Data on various health outcomes are accessible via DataSUS Tabnet, the Ministry's health data department’s portal. Detailed information on prenatal care quality and racial distribution is provided by the Live Birth Information System (SINASC), available through the Live Birth Monitoring Dashboard.
All datasets can be exported in CSV format after selecting the desired series, granularity, and time period. Additionally, data—except for SIOPS—can be fetched automatically using the R package microdatasus by specifying the start year, end year, and desired database. The data is published exclusively in Portuguese, and all labels were translated into English for this project.
This graph illustrates the distribution of healthcare spending across Brazilian states from 2010 to 2023. While all states experienced significant increases, some saw much steeper growth than others. In 2023, the state with the highest average per capita spending, Mato Grosso do Sul, allocated BRL 1,890 per person, while Amapá, the lowest spender, allocated less than half that amount at BRL 777. This spatial disparity aligns with the broader economic divide between the wealthier southern and poorer northern regions of the country.
It’s also relevant to consider how these budgets are distributed. The second graph, which allows filtering by state, reveals that surgical procedures consistently represent the largest spending category. Interestingly, northern states allocate a higher proportion of their budgets to childbirth and obstetric care. However, as later sections will demonstrate, these states also tend to report lower prenatal care quality scores.
The graph below allows for an interactive exploration of the correlation between healthcare spending and various health outcomes by selecting specific indicators. To maximize the number of observations, the dataset focuses on the 3,000 municipalities with the highest healthcare expenditure. Surprisingly, the analysis reveals a negative correlation between healthcare spending and the number of hospital beds, as well as an unexpected positive correlation between higher spending and both gross mortality and hospitalization rates (per 100,000 inhabitants).Similarly, there appears to be a neutral relationship with prenatal care adequacy and avoidable cause mortality.
These findings suggest that increased healthcare spending does not automatically translate into improved health outcomes, pointing instead to potential inefficiencies in resource allocation and systemic management.
*Zoom in on the graph to better align with the selected indicators, click on the region legend to highlight specific areas and reveal outcome patterns, and hover over a dot to view detailed information about the chosen variable.
Many factors influence the relationship between healthcare spending and health outcomes. To ensure better comparability, this section focuses on one key indicator over time: adequacy of prenatal care.
The first graph shows that spending on doctor appointments increased from BRL 40 million to BRL 65 million, while the percentage of women receiving adequate prenatal care rose from 64% to 70%. While this upward trend is positive, it’s equally important to examine who is benefiting from these investments.
The bar graph below disaggregates prenatal care adequacy by race over the years. In 2023, nearly 83% of white women received adequate prenatal care, compared to 72% of Black women and only 50% of Indigenous women. Although these figures represent a clear improvement from previous years, they highlight a persistent gap. This analysis underscores the importance of looking beyond overall health outcome averages. Without targeted measures to ensure investments reach historically underserved groups, increased spending alone may not effectively address existing inequalities.
One of the main challenges I faced was selecting the most effective visualizations—those capable of presenting complex information in a clear and digestible way. On one hand, using data at the state level (26 states) provided too small of a sample for meaningful correlation analysis, while on the other hand, analyzing all 5,500+ Brazilian municipalities introduced visual cluttering. After trying out different visualisation formats for the third visualisation, I decided to filter the top 3,000 municipalities by spending and allow users to interactively select specific regions, striking a balance between granularity and visual clarity. Another challenge was narrowing the project's scope. With countless factors influencing the relationship between healthcare spending and outcomes, there were many possibilities of data series to explore. However, to maintain focus and coherence, I concentrated on health outcomes and a targeted case study. Finally, working directly with Vega-Lite to code visualizations revealed limitations in achieving specific features, especially the scatter plots and dual-axis time series graphs. To overcome these constraints, I integrated the Altair library in Python, which offered me greater control.
This project demonstrates that increased healthcare spending in Brazil does not consistently lead to improved or equitably distributed health outcomes. Even in regions with high levels of expenditure, inefficiencies persist, as evidenced by neutral or even negative correlations with key indicators such as the number of hospital beds, gross mortality, and avoidable cause mortality. While targeted spending on medical consultations has shown a positive correlation with improvements in prenatal care adequacy, this trend does not consistently extend to broader health outcomes. Even where spending correlates with positive impacts, these benefits are not evenly distributed across the population, as racial and regional disparities continue to shape health outcomes.
It is important to emphasize that these findings showcase correlation, not causation. Evaluating the true impact of specific policies requires rigorous causal analysis tailored to local contexts. Future research is needed to identify which spending categories yield the highest returns on health outcomes and understand how non-financial factors—such as infrastructure quality, workforce capacity, and socio-economic conditions—shape health outcomes. Ultimately, in cases where increased healthcare spending is proven effective, it must be paired with inclusive and targeted strategies to ensure equitable and sustainable health improvements across all population groups.