ABSTRACT

Objective:

To analyze the influence of individual and contextual factors of the hospital and the municipality of care on the survival of patients with Severe Acute Respiratory Syndrome due to COVID-19.

Methods:

Hospital cohort study with data from 159,948 adults and elderly with Severe Acute Respiratory Syndrome due to COVID-19 hospitalized from January 1 to December 31, 2022 and reported in the Influenza Epidemiological Surveillance Information System. The contextual variables were related to the structure, professionals and equipment of the hospital establishments and socioeconomic and health indicators of the municipalities. The outcome was hospital survival up to 90 days. Survival tree and Kaplan-Meier curves were used for survival analysis.

Results:

Hospital lethality was 30.4%. Elderly patients who underwent invasive mechanical ventilation and were hospitalized in cities with low tax collection rates had lower survival rates compared to other groups identified in the survival tree (p<0.001).

Conclusion:

The study indicated the interaction of contextual factors with the individual ones, and it shows that hospital and municipal characteristics increase the risk of death, highlighting the attention to the organization, operation, and performance of the hospital network.

Keywords:
COVID-19; Survival; Hospital care; Social environment

INTRODUCTION

Severe Acute Respiratory Syndrome (SARS) can be caused by various infectious agents, including the SARS-CoV-2 virus. It is a serious condition characterized by dyspnea, respiratory rate above 30 rpm, and oxygen saturation below 93%, requiring hospital admission¹1. Brasil. Ministério da Saúde. Secretaria de Vigilância em Saúde. Guia de vigilância epidemiológica: emergência de saúde pública de importância nacional pela doença pelo coronavírus 2019 – covid-19. Brasília: Ministério da Saúde; 2022..

Individual factors such as advanced age, male gender, presence of comorbidities and need for invasive mechanical ventilation are associated with death²2. Gupta S, Hayek SS, Wang W, Chan L, Mathews KS, Melamed ML, et al. Factors associated with death in critically ill patients with coronavirus disease 2019 in the US. JAMA Intern Med 2020; 180(11): 1436-47. https://doi.org/10.1001/jamainternmed.2020.3596
https://doi.org/10.1001/jamainternmed.20...

3. Ñamendys-Silva SA, Gutiérrez-Villaseñor A, Romero-González JP. Hospital mortality in mechanically ventilated COVID-19 patients in Mexico. Intensive Care Med 2020; 46(11): 2086-8. https://doi.org/10.1007/s00134-020-06256-3
https://doi.org/10.1007/s00134-020-06256... -⁴4. Ferreira JC, Ho Y-L, Besen BAMP, Malbouisson LMS, Taniguchi LU, Mendes PV, et al. Protective ventilation and outcomes of critically ill patients with COVID-19: a cohort study. Ann Intensive Care 2021; 11(1): 92. https://doi.org/10.1186/s13613-021-00882-w
https://doi.org/10.1186/s13613-021-00882... . Furthermore, contextual factors related to the health service or the city where patients were admitted can also influence outcomes. However, only a limited number of studies have explored this association³3. Ñamendys-Silva SA, Gutiérrez-Villaseñor A, Romero-González JP. Hospital mortality in mechanically ventilated COVID-19 patients in Mexico. Intensive Care Med 2020; 46(11): 2086-8. https://doi.org/10.1007/s00134-020-06256-3
https://doi.org/10.1007/s00134-020-06256... ,⁵5. Meng Y. COVID-19 death rates and county subdivision level contextual characteristics: a connecticut case study. Cybergeo: European Journal of Geography 2021. https://doi.org/10.4000/cybergeo.36057
https://doi.org/10.4000/cybergeo.36057...

6. Demenech LM, Dumith SC, Vieira MECD, Neiva-Silva L. Desigualdade econômica e risco de infecção e morte por COVID-19 no Brasil. Rev Bras Epidemiol 2020; 23: e200095. https://doi.org/10.1590/1980-549720200095.
https://doi.org/10.1590/1980-54972020009... -⁷7. Santana JM, Lana CNA, Souza GB, Souza LMS. Determinantes sociais da saúde e óbitos por COVID-19 nos estados da região Nordeste do Brasil. RBRASF 2020; 11(1): 18-29. https://doi.org/10.25194/rebrasf.v8i2.1305
https://doi.org/10.25194/rebrasf.v8i2.13... .

Contextual factors encompass the conditions and environment of access to health care, including the structure of the system, financial aspects, and characteristics of the community. As such, their assessment is carried out in an aggregated manner rather than individually⁸8. Andersen RM, Davidson PL. Improving access to care in America: individual and contextual indicators. In: Andersen RM, Rice TH, Kominski GF, eds. Changing the U.S. health care system: key issues in health services, policy, and management. San Francisco: Jossey-Bass, 2001. p. 3-30.. At the hospital level, the availability of financial, human, and equipment resources can impact access to healthcare services⁹9. Travassos C, Castro MSM. Determinantes e desigualdades sociais no acesso e na utilização de serviços de saúde. In: Giovanella L, Escorel S, Lobato LVC, Noronha JC, Carvalho AI, eds. Políticas e sistema de saúde no Brasil. Rio de Janeiro: Editora FIOCRUZ; 2012. p. 183-208.. On a municipal level, socioeconomic indicators such as health, education, economic growth, social inequality, investment, and tax collection can all be taken into consideration⁸8. Andersen RM, Davidson PL. Improving access to care in America: individual and contextual indicators. In: Andersen RM, Rice TH, Kominski GF, eds. Changing the U.S. health care system: key issues in health services, policy, and management. San Francisco: Jossey-Bass, 2001. p. 3-30..

Studies that analyzed the relationship between contextual factors and COVID-19 mortality revealed notable findings. In Mexico, in 2020, care received in public services, as opposed to private services, was associated with increased mortality³3. Ñamendys-Silva SA, Gutiérrez-Villaseñor A, Romero-González JP. Hospital mortality in mechanically ventilated COVID-19 patients in Mexico. Intensive Care Med 2020; 46(11): 2086-8. https://doi.org/10.1007/s00134-020-06256-3
https://doi.org/10.1007/s00134-020-06256... , while in the United States, in 2021, the association occurred with highest municipal median income⁵5. Meng Y. COVID-19 death rates and county subdivision level contextual characteristics: a connecticut case study. Cybergeo: European Journal of Geography 2021. https://doi.org/10.4000/cybergeo.36057
https://doi.org/10.4000/cybergeo.36057... . In Brazil, in 2020, higher income inequality, as measured by the Gini coefficient, and fewer municipal beds were associated with elevated mortality⁶6. Demenech LM, Dumith SC, Vieira MECD, Neiva-Silva L. Desigualdade econômica e risco de infecção e morte por COVID-19 no Brasil. Rev Bras Epidemiol 2020; 23: e200095. https://doi.org/10.1590/1980-549720200095.
https://doi.org/10.1590/1980-54972020009... ,⁷7. Santana JM, Lana CNA, Souza GB, Souza LMS. Determinantes sociais da saúde e óbitos por COVID-19 nos estados da região Nordeste do Brasil. RBRASF 2020; 11(1): 18-29. https://doi.org/10.25194/rebrasf.v8i2.1305
https://doi.org/10.25194/rebrasf.v8i2.13... . However, despite the limited number of studies, previous research did not analyze the survival outcomes of these cases; rather, they focused solely on the correlation between these variables and death⁷7. Santana JM, Lana CNA, Souza GB, Souza LMS. Determinantes sociais da saúde e óbitos por COVID-19 nos estados da região Nordeste do Brasil. RBRASF 2020; 11(1): 18-29. https://doi.org/10.25194/rebrasf.v8i2.1305
https://doi.org/10.25194/rebrasf.v8i2.13... .

Analyzing survival, rather than just focusing on death, offers the advantage of examining the time between exposure and the event, while also handling censored data. Survival tree analysis, in particular, employs a tree-like structure with precise decision rules that are parsimonious, statistically robust, and visually interpretable¹⁰10. Wang P, Li Y, Reddy CK. Machine learning for survival analysis: a survey. ACM Comput Surv 2019; 51: 1-36. https://doi.org/10.48550/arXiv.1708.04649
https://doi.org/10.48550/arXiv.1708.0464... ,¹¹11. Linden A, Yarnold PR. Modeling time-to-event (survival) data using classification tree analysis. J Eval Clin Pract 2017; 23(6): 1299-308. https://doi.org/10.1111/jep.12779
https://doi.org/10.1111/jep.12779... . Therefore, leveraging survival analysis, especially tree-structured models, can provide insights that facilitate a more thorough examination of the factors that contribute to the increased risk of death among patients with SARS due to COVID-19.

Given that prior studies have predominantly focused on individual risk factors for death, with only a limited examination of how contextual factors influence the dynamics of COVID-19, this study aimed to investigate the association between individual and contextual factors related to both the hospital and municipality of care concerning the survival outcomes of patients hospitalized with SARS due to COVID-19.

METHODS

A cohort study was carried out with reported cases of SARS due to COVID-19, drawing data from the Influenza Epidemiological Surveillance Information System (Sistema de Informação de Vigilância Epidemiológica da Influenza – SIVEP-Gripe). SIVEP-Gripe is an information system created by the Ministry of Health to document cases and deaths from SARS and COVID-19 in Brazil. COVID-19 notifications are compulsory and encompass information from both public and private hospitals. The dataset was obtained from the OpenDataSUS website (https://opendatasus.saude.gov.br/), considering the database updated as of March 30^th, 2023.

This study included cases reported in SIVEP-Gripe and hospitalized between January 1^st, 2022 and December 31^st, 2022, focusing only cases reported in 2022, as vaccination efforts throughout the country had already started. This circumstance prevented the saturation of the health system during that year, allowing for the examination of contextual variables’ influence on the survival of patients with COVID-19 in situations of greater control of the epidemic.

Information on the establishment where the cases were admitted was acquired from the National Registry of Health Establishments (Cadastro Nacional de Estabelecimentos de Saúde – CNES), using the Microdatasus¹²12. Saldanha RF, Bastos RR, Barcellos C. Microdatasus: pacote para download e pré-processamento de microdados do Departamento de Informática do SUS (DATASUS). Cad Saúde Pública 2019; 35(9): e00032419. https://doi.org/10.1590/0102-311x00032419
https://doi.org/10.1590/0102-311x0003241... package of the R software. Subsequently, a linkage was established between CNES data and the data from SIVEP-Gripe using the same software. This linkage process involved key fields such as: establishment identification (CNES and SIVEP-Gripe), competence (month and year of establishment update, available solely in CNES), and month and year of hospitalization of the individual (SIVEP-Gripe). Following the linkage process, details regarding the establishment where each case was admitted were incorporated for every notification recorded in SIVEP-Gripe.

Municipal variables were also gathered, including indicators comprising the 2020 Sustainable Cities Development Index – Brazil, designed to assist cities in monitoring performance in accordance with the 17 Sustainable Development Goals (SDGs) of the United Nations¹³13. Índice de Desenvolvimento Sustentável das Cidades. Brasil 2023. [acessado em 28 abr. 2022]. Disponível em: https://www.cidadessustentaveis.org.br/paginas/idsc-br
https://www.cidadessustentaveis.org.br/p... , and the Social Inequalities Index for COVID-19 2022 (IDS-COVID-19)¹⁴14. Brasil. Ministério da Saúde. Fundação Oswaldo Cruz. Índice de desigualdades sociais para covid-19. [cited on Apr 28, 2022]. Available at: https://cidacs.bahia.fiocruz.br/idscovid19/ids-covid-19/
https://cidacs.bahia.fiocruz.br/idscovid... . The linkage of these data was carried out using the municipal code. The selected indicators were those likely to be associated with health conditions and inequality within municipalities, particularly in relation to the dynamics of COVID-19 in these regions.

The study included only cases aged 20 years old or older (adults and aged people), with a final classification of SARS due to COVID-19 who were admitted to a hospital. Postpartum women, pregnant women, and those who presented missing information or typing errors on the date of hospital admission, date of discharge, information on the evolution of the case (death or discharge) were excluded. Additionally, establishments with fewer than five registered beds and those lacking information about the inpatient establishment after linkage were excluded from the analysis.

Study variables

Individual variables represent characteristics of cases hospitalized with SARS due to COVID-19 and are categorized into: sociodemographic — gender (male, female), age (20 to 39, 40 to 59, 60 to 79, ≥80 years), race/color (White, Black, Yellow, Brown, Indigenous); clinical — Admission to the Intensive Care Unit (ICU) (Yes, No), Mechanical Ventilation (Invasive, Non-Invasive, None), Multimorbidity (Yes, No), COVID-19 vaccination schedule (Not Immunized – not vaccinated or with incomplete vaccination schedule, two doses, booster dose).

The variable multimorbidity refers to the number of comorbidities reported by patients at the time of hospitalization, which included the following risk factors: Chronic Cardiovascular Disease, Chronic Hematological Disease, Chronic Liver Disease, Asthma, Diabetes Mellitus, Chronic Neurological Disease, Chronic Pneumopathy, Immunosuppression, Chronic Kidney Disease, and Obesity.

The contextual variables at the hospital level obtained from the CNES include: hospital management and structure – Teaching Activity (Yes, No), Type of Management (Mixed, State, Municipal), Link with SUS (Yes, No), Size of the Hospital (Small – 5 to 49 beds, Medium – 50 to 149 beds, Large – 150 or more beds) and hospital indicators: Ratio of Doctors/bed, Nurses/bed, Physiotherapists/bed, Nursing Technicians/bed, Infusion Pump/bed, Electrocardiogram Monitor/bed, Mechanical Ventilator/bed, and Defibrillator/bed. These indicators were calculated according to the study by Botega et al.¹⁵15. Botega LA, Andrade MV, Guedes GR. Profile of general hospitals in the Unified Health System. Rev Saúde Pública 2020; 54: 81. https://doi.org/10.11606/s1518-8787.2020054001982
https://doi.org/10.11606/s1518-8787.2020... .

The contextual variables at the municipal level used were: Families registered in the Single Registry for social programs (%), Life expectancy at birth (years), Municipal health budget (in reais, per capita), Population served by municipal family health teams (%), GDP per capita (R$ per capita), Gini coefficient, Access to basic health care equipment (%), Public investment (R$ per capita), Total revenue collected (%), IDS- COVID-19. Further details regarding the analyzed indicators can be seen in Chart 1.

The primary outcome of interest was survival time (in days) until in-hospital death from COVID-19. For cases that resulted in death, the survival time was calculated as the duration from hospital admission to the date of death. For those who survived, the survival time was determined from the date of hospitalization until hospital discharge. Survival time was observed up to 90 days after hospitalization; Cases with survival times exceeding 90 days or discharged before 90 days were regarded as censored. Censorship was applied at 90 days, as beyond this timeframe, cases had similar probabilities of survival.

Data analysis

Data were analyzed using R 4.2.3 software (http://www.r-project.org/). For the variables race/color (16.5%), ICU admission (8.1%), Mechanical Ventilation (12.1%), Health Professionals (9.5%), Equipment (4.8%), and Total Revenues Collected (2.7%) that presented missing values, single imputation was performed with the Fully Conditional Specification (FCS) method implemented in the R MICE¹⁶16. van Buuren S, Groothuis-Oudshoorn K. mice : Multivariate Imputation by Chained Equations in R. J Stat Softw 2011; 45(3): 1-67. https://doi.org/10.18637/jss.v045.i03
https://doi.org/10.18637/jss.v045.i03... package. After the imputation procedure, a descriptive analysis was carried out, which included proportions, means, standard deviations, medians, interquartile ranges, as well as minimum and maximum values for the variables under examination.

Survival analysis was performed to evaluate factors associated with mortality from COVID-19 within 90 days of hospitalization. For this, survival trees were constructed, a non-parametric technique that incorporates tree-structured regression models to analyze survival time¹⁷17. Bou-Hamad I, Larocque D, Ben-Ameur H. A review of survival trees. Statist Surv 2011; 5: 44-71. https://doi.org/10.1214/09-SS047
https://doi.org/10.1214/09-SS047... . This technique offers flexibility by not requiring the specification of variable distributions and automatically identifying how interactions among two or more explanatory variables influence the outcome of interest¹⁷17. Bou-Hamad I, Larocque D, Ben-Ameur H. A review of survival trees. Statist Surv 2011; 5: 44-71. https://doi.org/10.1214/09-SS047
https://doi.org/10.1214/09-SS047... . Interaction is the impact of an explanatory variable on other explanatory variables and is represented by the subdivisions of the tree nodes. Furthermore, unlike linear regression models, no assumptions need to be made about the independence of explanatory variables (collinearity). If two explanatory variables are correlated, the decision tree selects the variable that provides the best split for a given node, in this case, based on a measure of node deviation between a saturated log-likelihood model and a maximized log-likelihood¹⁸18. LeBlanc M, Crowley J. Relative risk trees for censored survival data. Biometrics 1992; 48(2): 411-25. PMID: 1637970.. The terminal nodes, which represent risk groups identified by the tree, present survival curves estimated using the Kaplan-Meier method. The trees were implemented via Survival, LTRCtrees, and Party.kit packages¹⁹19. Hothorn T, Zeileis A. partykit: A Modular Toolkit for Recursive Partytioning in R. J Mach Learn Res 2015; 16(118): 3905-9.

20. Fu W, Simonoff J, Jing W. LTRCtrees: survival trees to fit left-truncated and right-censored and interval-censored survival data. R package version 1.1.1; 2021. [cited on Jun 19, 2023]. Available at: https://CRAN.R-project.org/package=LTRCtrees
https://CRAN.R-project.org/package=LTRCt... -²¹21. Therneau TM. A package for survival analysis in R. R package version 3.5-5; 2023. [cited on Jun 19, 2023]. Available at: https://CRAN.R-project.org/package=survival
https://CRAN.R-project.org/package=survi... .

First, a tree was generated solely based on individual variables (gender, age, race/color, ICU admission, mechanical ventilation, and multimorbidities) to estimate the proportional risk for each patient. Next, three groups were established according to the tertiles of proportional risk estimated by the tree: low, moderate, and high.

After creating the tree with the individual variables, a subsequent tree was generated, incorporating the risk identified from the individual variables along with the variables related to establishments and municipalities. The objective was to discern the influence of hospital and municipal structures on the survival time of individuals.

The survival curves of cases within each terminal node were compared using the Kaplan-Meier method, along with the logrank test to ascertain differences between groups, with a significance level of 5%.

The study was approved by the Research Ethics Committee of the University Hospital of Universidade Federal do Maranhão and by the National Research Ethics Committee of the National Health Council (Conselho Nacional de Saúde – CNS), CAAE No. 32206620.0.0000.5086, on June 19th, 2020, as per resolutions 466/12 and 510/16 of CNS²²22. Brasil. Ministério da Saúde. Conselho Nacional de Saúde. Resolução no 510, de 7 de abril de 2016. O Plenário do Conselho Nacional de Saúde em sua Quinquagésima Nona Reunião Extraordinária, realizada nos dias 06 e 07 de abril de 2016, no uso de suas competências regimentais e atribuições conferidas pela Lei n o 8.080, de 19 de setembro de 1990, pela Lei n o 8.142, de 28 de dezembro de 1990, pelo Decreto n o 5.839, de 11 de julho de 2006, e. Brasília: Diário Oficial República Federativa do Brasil de 24 maio de 2016 [cited on Jun 17, 2023]. Available at: https://bvsms.saude.gov.br/bvs/saudelegis/cns/2016/res0510_07_04_2016.html
https://bvsms.saude.gov.br/bvs/saudelegi... ,²³23. Brasil. Ministério da Saúde. Conselho Nacional de Saúde. Resolução no 466, de 12 de dezembro de 2012. diretrizes e normas regulamentadoras de pesquisa envolvendo seres humanos. Brasília: Diário Oficial República Federativa do Brasil de 12 de dezembro de 2012 [cited on Jun 17, 2023]. Available at: https://bvsms.saude.gov.br/bvs/saudelegis/cns/2013/res0466_12_12_2012.html
https://bvsms.saude.gov.br/bvs/saudelegi... .

RESULTS

Out of the 200,626 reported SARS cases that met the inclusion criteria, 40,678 (20.3%) were excluded, resulting in a final sample of 159,948 cases (Figure 1). Among these, 30.4% (n = 48,688) resulted in death. The median hospital stay was 6 days among censored cases and 8 days among those who died (Table 1).

A higher 90-day lethality rate was observed among men (32.1%), individuals aged 80 years old or older (39%), those of black color/race (35.6%), from the Northeast region (35.6%), with multimorbidities (36.4%), who were admitted to the ICU (50.1%), and individuals who required invasive mechanical ventilation (77.9%) (Table 1).

Death was more prevalent among cases admitted to small hospitals (31.2%), those linked to SUS (32.7%), establishments without teaching activities (34.2%), and municipalities with IDS COVID of 2 to 5 (33.2%). The average ratio of defibrillators/beds in hospitals and the Gini coefficient in municipalities of individuals who died were 0.09 (±0.0) and 0.54 (±0.06), respectively, while the average proportion of collection was 24.83% (±12.44%) (Table 2).

The initial survival tree created from individual variables generated nine terminal nodes and employed mechanical ventilation, ICU admission, and age as decision variables (Figure 2). The stratification of groups into a categorical variable called “individual risk” occurred as follows: in the first tertile (low risk), cases belonging to nodes 5, 6, 8, and 11 and those not subjected to mechanical ventilation or adults who received non-invasive ventilation were included; the second tertile (moderate risk) comprised cases belonging to nodes 9 and 12, which included aged individuals undergoing non-invasive mechanical ventilation; Finally, the third tertile (high risk) encompassed cases undergoing invasive ventilation: nodes 15, 16, and 17. The higher the risk, the lower the survival rate of these patients.

The survival tree generated with the identified risk based on individual characteristics, hospital, and municipal characteristics included the following variables: link with SUS, defibrillator/bed ratio, Gini coefficient, revenue collected, IDS COVID, and individual risk (Figure 3). The root node, which conducts the initial division, utilized individual risk as a decision variable, ultimately identifying 8 (eight) terminal nodes.

Cases classified as having mild individual risk and admitted to hospitals not linked to SUS (node 3) had a lower risk of death with a median survival time of 90 days. Cases with high individual risk who lived in cities with revenue collected less than 19.5% (node 17) had a higher risk of death and a median survival time of 10 days. The 90-day hospital survival curve showed a statistical difference between the cases of terminal nodes generated by the tree (p<0.001) (Figure 4).

DISCUSSION

The findings suggest that individual factors, as well as factors related to the hospital structure and municipalities of care, significantly influenced the survival of patients hospitalized with SARS due to COVID-19. Their interaction defined unequal risks of death, with heightened risk observed among those who required mechanical ventilation, the aged, those admitted to hospitals linked to SUS, with limited availability of defibrillators, residing in municipalities characterized by lower Gini coefficients and percentages of revenue collected, and higher IDS-COVID.

Based on individual variables, the use of invasive mechanical ventilation and advanced age were identified as factors that reduce survival in the study group, as documented in the literature²2. Gupta S, Hayek SS, Wang W, Chan L, Mathews KS, Melamed ML, et al. Factors associated with death in critically ill patients with coronavirus disease 2019 in the US. JAMA Intern Med 2020; 180(11): 1436-47. https://doi.org/10.1001/jamainternmed.2020.3596
https://doi.org/10.1001/jamainternmed.20...

3. Ñamendys-Silva SA, Gutiérrez-Villaseñor A, Romero-González JP. Hospital mortality in mechanically ventilated COVID-19 patients in Mexico. Intensive Care Med 2020; 46(11): 2086-8. https://doi.org/10.1007/s00134-020-06256-3
https://doi.org/10.1007/s00134-020-06256... -⁴4. Ferreira JC, Ho Y-L, Besen BAMP, Malbouisson LMS, Taniguchi LU, Mendes PV, et al. Protective ventilation and outcomes of critically ill patients with COVID-19: a cohort study. Ann Intensive Care 2021; 11(1): 92. https://doi.org/10.1186/s13613-021-00882-w
https://doi.org/10.1186/s13613-021-00882... .

Those admitted to SUS hospitals had lower survival rates, possibly due to lower availability of equipment. Before the pandemic, 72% of regions already had less than 10 ICU beds per 100,000 inhabitants, which represents limited bed availability for 61% of the Brazilian population without health insurance²⁴24. Rache B, Rocha R, Nunes L, Spinola P, Malik AM, Massuda A. Necessidades de infraestrutura do SUS em preparo ao COVID-19: leitos de UTI, respiradores e ocupação hospitalar. Instituto de Estudos para Políticas de Saúde 2020; 1-5. [cited on Apr 29, 2023]. Available at: https://observatoriohospitalar.fiocruz.br/sites/default/files/biblioteca/ESTUDO%20ANA%20MALIK%20NT3-vFinal.pdf_0.pdf
https://observatoriohospitalar.fiocruz.b... . Despite the SUS receiving the largest number of people with conditions that require hospital admission, the system only has 48% of ICU beds in Brazil²⁵25. Noronha KVMS, Guedes GR, Turra CM, Andrade MV, Botega L, Nogueira D, et al. Pandemia por COVID-19 no Brasil: análise da demanda e da oferta de leitos hospitalares e equipamentos de ventilação assistida segundo diferentes cenários. Cad Saúde Pública 2020; 36(6): e00115320. https://doi.org/10.1590/0102-311X00115320
https://doi.org/10.1590/0102-311X0011532... .

In 2020, only 0.2% of locations lacked defibrillators, yet many regions possess up to 5 pieces of equipment per 10,000 inhabitants²⁶26. Portela MC, Pereira CCA, Andrade CLT, Lima SML, Braga Neto FC, Soares FRG, et al. As regiões de saúde e a capacidade instalada de leitos de UTI e alguns equipamentos para o enfrentamento dos casos graves de Covid-19. Rio de Janeiro: Fiocruz/ENSP, 2020., which may compromise patient care. Furthermore, those admitted to hospitals with a defibrillator ratio lower than 0.123 are in the North region of the country, historically characterized by limited equipment availability²⁴24. Rache B, Rocha R, Nunes L, Spinola P, Malik AM, Massuda A. Necessidades de infraestrutura do SUS em preparo ao COVID-19: leitos de UTI, respiradores e ocupação hospitalar. Instituto de Estudos para Políticas de Saúde 2020; 1-5. [cited on Apr 29, 2023]. Available at: https://observatoriohospitalar.fiocruz.br/sites/default/files/biblioteca/ESTUDO%20ANA%20MALIK%20NT3-vFinal.pdf_0.pdf
https://observatoriohospitalar.fiocruz.b... .

SUS users often face worse socioeconomic conditions, which hinders access to health services and contribute to worse assessment of their health status²⁷27. Ribeiro MCSA, Barata RB, Almeida MF, Silva ZP. Perfil sociodemográfico e padrão de utilização de serviços de saúde para usuários e não-usuários do SUS-PNAD 2003. Ciên Saúde Coletiva 2006; 11(4): 1011-22. https://doi.org/10.1590/S1413-81232006000400022
https://doi.org/10.1590/S1413-8123200600... . Furthermore, individuals without private health insurance have a higher prevalence of chronic non-communicable diseases, which may increase the risk of developing a severe form of COVID-19²⁸28. Malta DC, Bernal RTI, Lima MG, Silva AG, Szwarcwald CL, Barros MBA. Socioeconomic inequalities related to noncommunicable diseases and their limitations: National Health Survey, 2019. Rev Bras Epidemiol 2021; 24(suppl 2): e210011. https://doi.org/10.1590/1980-549720210011.supl.2
https://doi.org/10.1590/1980-54972021001... .

Cases hospitalized in municipalities with a Gini lower than 0.575 had lower survival rates. This coefficient is part of SDG 10 (Reduction of Inequalities) and measures the concentration of income in each municipality²⁹29. Instituto de Pesquisa Econômica Aplicada. Objetivos do Desenvolvimento Sustentável. Redução da desigualdades. [cited on Apr 28, 2023]. Available at: https://www.ipea.gov.br/ods/ods10.html
https://www.ipea.gov.br/ods/ods10.html... . Although a previous study has indicated that income inequality correlated with a higher risk of death from COVID-19³⁰30. Elgar FJ, Stefaniak A, Wohl MJA. The trouble with trust: time-series analysis of social capital, income inequality, and COVID-19 deaths in 84 countries. Soc Sci Med 2020; 263: 113365. https://doi.org/10.1016/j.socscimed.2020.113365
https://doi.org/10.1016/j.socscimed.2020... , our findings suggest that despite these patients living in places with lower income concentration, they were admitted to hospitals with fewer health professionals and equipment, situated in municipalities with a substantial percentage of individuals with low income. Therefore, even amid lower income inequality, the lack of hospital and social resources may reduce survival. These inequalities lead to groups of people with reduced access to diagnostic tests and an elevated risk of infection, hospitalization, and death³¹31. Ahmed F, Ahmed N, Pissarides C, Stiglitz J. Why inequality could spread COVID-19. Lancet Public Health 2020; 5(5): e240. https://doi.org/10.1016/S2468-2667(20)30085-2
https://doi.org/10.1016/S2468-2667(20)30... .

Similar patterns were observed among individuals in municipalities that failed to meet the target and are below the green threshold of 19.7% of total revenue collected. This indicator is part of SDG 17 (Partnerships and Means of Implementation) and reflects the municipality’s capacity for tax collection, indicating its reliance on resources from the State or the Union³²32. Instituto de Pesquisa Econômica Aplicada. Objetivos do Desenvolvimento Sustentável. Parcerias e meios de implementação. [cited on Apr 28, 2023]. Available at: https://www.ipea.gov.br/ods/ods17.html
https://www.ipea.gov.br/ods/ods17.html... . The revenue collected by municipalities directly affects people’s health. A study delineating the evolution of municipal financing of SUS, from 2004 to 2019, shows an increase in non-own health expenses following the 2015 crisis, indicating greater fiscal dependence for healthcare funding. This means that municipalities, especially smaller ones, have become increasingly reliant on state health funds³³33. Cruz WGN, Barros RD, Souza LEPF. Financiamento da saúde e dependência fiscal dos municípios brasileiros entre 2004 e 2019. Ciên Saúde Coletiva 2022; 27(6): 2459-69. https://doi.org/10.1590/1413-81232022276.15062021
https://doi.org/10.1590/1413-81232022276... . This situation becomes more challenging due to insufficient resources to cover healthcare expenses³³33. Cruz WGN, Barros RD, Souza LEPF. Financiamento da saúde e dependência fiscal dos municípios brasileiros entre 2004 e 2019. Ciên Saúde Coletiva 2022; 27(6): 2459-69. https://doi.org/10.1590/1413-81232022276.15062021
https://doi.org/10.1590/1413-81232022276... .

The COVID-19 pandemic exacerbated challenges related to healthcare spending. A previous study indicated that the majority of states in the Southeast region of Brazil were not prepared for a drop in revenue, as they were already operating at the brink of their fiscal health. Indeed, in April 2020, the peak period of the pandemic, there was an impact on revenue among the states analyzed³⁴34. Borges MGB. Impactos da covid-19 nas receitas tributárias e na condição financeira dos estados do sudeste do Brasil. In: XX USP International Conference in Accounting; 2020 jul 29-31: São Paulo, Brasil. São Paulo: USP; 2020. Available at: https://congressousp.fipecafi.org/anais/20UspInternational/ArtigosDownload/3010.pdf
https://congressousp.fipecafi.org/anais/... . This resulted in a greater need to transfer resources from the Union to states and municipalities. However, by the end of June 2020, only 39.5 and 33.9% of the planned resources had been transferred to states and municipalities, respectively. It was not until July onwards, with already 100 thousand deaths resulting from COVID-19, that resources were transferred in greater volume³⁵35. Servo LMS, Santos MAB, Vieira FS, Benevides RPS. Financiamento do SUS e Covid-19: histórico, participações federativas e respostas à pandemia. Saúde Debate 2020; 44: 114-29. https://doi.org/10.1590/0103-11042020E407
https://doi.org/10.1590/0103-11042020E40... . Therefore, there existed a disparity between local needs and the Union’s transfer, and the delay in resource allocation underscores the Union’s lack of preparedness during a health system crisis.

IDS-COVID is another indicator that highlights social inequalities in health related to COVID-19¹⁴14. Brasil. Ministério da Saúde. Fundação Oswaldo Cruz. Índice de desigualdades sociais para covid-19. [cited on Apr 28, 2022]. Available at: https://cidacs.bahia.fiocruz.br/idscovid19/ids-covid-19/
https://cidacs.bahia.fiocruz.br/idscovid... and has been demonstrated to be a predictor of the risk of death from the disease. Cases originating from municipalities with an IDS-COVID greater than or equal to 2, that is, greater inequality, have a lower survival rate. Thus, these findings underscore the significance of identifying a cutoff point that allows greater attention to municipalities with this characteristic.

The limitations of this study are associated with the use of a secondary database, which, may potentially contain typing errors and incomplete information. Additionally, since the study focuses on hospitalized patients, the results cannot be extrapolated to all cases of COVID-19 but rather to those with a severe form of the disease. Nevertheless, inconsistent data exclusion criteria and missing data imputation techniques were employed for this analysis. Given that it is the largest national database containing information about COVID-19, it enables inference regarding the disease’s course in the Brazilian population.

Another limitation is due to the data obtained from CNES, in which 5% of the cases were hospitalized in hospitals with fewer than five beds or without correspondence with the SIVEP-Gripe data, which made their exclusion from the study necessary. The remaining variables with missing data went through the imputation process. Despite these limitations, this is one of the first studies that uses data referring to health establishments, as well as social indicators with the intention of verifying survival in this group.

This study underscores the development of models based on survival trees, enabling the integration of hierarchical structures. The algorithm employed for tree construction automatically discerns these structures, eliminating the necessity to specify the hierarchical levels of each variable within the model. Moreover, it facilitates a transparent visualization of the relationships among variables and the hierarchical arrangement of variables constituting the final model.

In conclusion, this study highlights the interaction between individual and contextual factors, revealing that hospital and municipal characteristics heighten the risk of death, even within a context of widespread vaccination that resulted in fewer hospitalized cases. These findings, when viewed through the lens of hospital and municipal indicators, underscore the ongoing challenges surrounding SUS financing and the subsequent availability of equipment and professionals. This challenge is exacerbated in municipalities characterized by a lower percentage of revenue collected and historical inequalities. Consequently, these combined factors may contribute to heightened vulnerability among patients. Hence, there is a pressing need for increased attention to the organization, functioning, and performance of the small hospital network, which often receives fewer resources. Additionally, municipalities with pronounced inequality in COVID-19-related indicators and limited resources warrant heightened scrutiny to address social and health-related demands effectively.

ACKNOWLEDGMENTS:

We thank the Maranhão Scientific and Technological Research and Development Support Foundation (Fundação de Amparo à Pesquisa e ao Desenvolvimento Científico e Tecnológico do Maranhão – FAPEMA) for the program to support the publication of articles and the Coordination for the Improvement of Higher Education Personnel (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – CAPES) [Funding code No.: 001].

REFERENCES

Publication Dates

History