Study design
This study is a longitudinal analysis of publicly-funded hospitalization rates in Brazil from 2010–2022 and its correlation with data completeness. We investigated patterns in hospitalization rates from January 2012 to December 2022 and the proportion of missing data over time.
Data sources
Data from the Hospital Information System in Brazil (SIH) were obtained (https://datasus.saude.gov.br/) from January 2010 to June 2022. This data includes 145,554,343 hospital admissions records for all publicly-funded admissions.
The Hospital Information System of DataSUS (SIH/SUS) is the main source of data on hospital admissions within the Unified Health System (SUS) in Brazil. Information collected includes diagnoses, procedures, length of stay, and patients’ personal data. These data are generated as part of the administrative activity of the State, specifically in the provision of health services by SUS and the receipt of financial counterpart. All health units that serve patients through SUS must report admission data to SIH/SUS monthly, making them mandatory administrative records. These data are made publicly available in an anonymized format. Thus, SIH/SUS has also become a crucial source of data for epidemiological and public health analyses.
According to Brazilian legislation [31], the collection of the racial category and filling out the race/ethnicity field are required in healthcare services based on the user’s self-declaration criteria, as per Brazilian Institute of Geography and Statistics (IBGE) standards. In the case of newborns, deaths, or when the user is unable to self-declare, family members or guardians will declare their race or ethnicity. If there is no responsible party, healthcare professionals will complete the race/ethnicity field during the service.
The official racial/skin colour categories in Brazil include Black, Brown, White, Indigenous, and Yellow (Asian) [32].
Individuals categorize themselves into racial groups based on skin tone, hair texture, facial structure, and cultural and ethnic traits [33].
We aggregated individual admission records by year and quarter, primary cause of admission (by International Classification of Diseases (ICD-10) codes), and by race/color. We grouped hospital admissions into three groups: Blacks, Whites, and missing. Therefore, the Black group includes all Black and Mixed race individuals (Brown or Pardo)—given the historical degree of racial mixing, and as frequently done in the scientific literature [34, 35]. Asians and Indigenous peoples were not analyzed due to low numbers and also, especially related to indigenous people, given unique characteristics related to living conditions and specificities related to access to health services and other indigenous policies.
We use data from the Continuous National Household Sample Survey (PNAD-Contínua) of the IBGE for population estimates for the period of 2012–2022. This survey is carried out continuously throughout the year, collecting data on the population, employment, income, education, and other socio-economic characteristics of Brazilian households. It is through the PNAD-Contínua that, for example, the official unemployment rate in the country is calculated. It provides a comprehensive view of the country’s socio-economic characteristics, allowing for the tracking of trends over time and the comparison of different population groups with quarterly data. We obtained estimates of the population by race/color by each quarter for the entire period as the denominator for rates. SIH data is collected on a monthly basis, while PNAD-Contínua are collected quarterly. Month/quarter compatibility was used to calculate the rates.
We considered the following causes that are recognized for their significant racial inequalities in Brazil [15,16,17,18,19, 36]: motorcycle accidents (ICD-10 V20-V29); assaults (ICD-10 X92-Y09), tuberculosis (ICD-10 A15-A19); hypertensive diseases (ICD-10 I10-I15); at-risk hospitalizations during pregnancy (known as “near miss” [37], see Appendix 1).
The data from the Hospital Information System (SIH) and IBGE are publicly accessible and anonymized. Ethical principles of transparency, privacy, confidentiality, and protection of individual rights of participants are therefore respected.
Analysis
We first estimated hospitalization rates for Black and White individuals per 10,000 inhabitants (of respective race/color), for each group of causes of admission, using hospitalization data from SIH [38] and national population data from PNAD-Continua as a population denominator for calculating rates. Secondly, we calculated the share of admissions for each category (Whites, Blacks and missing values), summing up 100% for each quarter-year. Thirdly, we computed the ratio of hospitalization rates between Blacks and Whites to quantify the inequalities in hospitalization rates between Blacks and Whites over time. Therefore, we are left with an indication of the rate ratio between Black and White, in which if there is a value equal to 1, it means that there are no racial differences. A value below 1 indicates inequalities in the disadvantage experienced by the White population, while a value above 1 indicates inequalities in the disadvantage experienced by the Black population.
Time series were plotted for hospital admissions of all causes and by groups of causes using 12-month moving averages.
Add Comment