INTRODUCTION

At the end of 2019, the Chinese government reported cases of pneumonia with an unknown etiological agent in the population, with the first cases being recorded in the city of Wuhan, China. The agent was later identified as a new coronavirus of Severe Acute Respiratory Syndrome (SARS), and was named SARS-CoV-2. The new coronavirus is responsible for the development of Coronavirus Disease 2019 (COVID-19), characterized as an infectious disease that affects individuals with mild or severe symptoms.^1-2

Due to the high transmissibility and mortality, the World Health Organization (WHO) declared a Public Health Emergency of International Concern³. In Brazil, the first cases were registered in February 2020, and a Public Health Emergency of National Concern was declared. In this context, the main preventive measure adopted was social distancing, culminating in the so-called lockdown in the main Brazilian cities. The measure included the blocking of borders, the cancellation of cultural events and the temporary closure of businesses and educational institutions.^4-5

Epidemiologically, by the end of March 2023, confirmed cases of COVID-19 had already surpassed 600 million worldwide, with more than 6 million deaths recorded. In the world ranking, Brazil was one of the countries with the most registered cases, accumulating more than 30 million cases and 650 thousand deaths. Regarding the distribution of morbidity and mortality from the disease in the national territory, the Southeast and Northeast regions stand out in absolute numbers, with the state of Pernambuco recording a total of more than 1 million cases and 20 thousand deaths.^6-7

In Brazil, the distribution of infectious diseases across the territory has always been marked by immense social inequality.⁸ Regarding COVID-19, the spread of the disease occurred with different magnitudes and effects in different population groups, in which socioeconomic determinants were found to be possible risk factors. The presence of inequalities reinforces an unfair scenario for the most vulnerable population, implying barriers to access to healthcare services and indicating that, in addition to the COVID-19 pandemic, the population faces an epidemic of social inequities.^9-10

Therefore, the dimension of health inequalities permeates the issue of mortality from COVID-19, and the use of different analytical strategies can contribute to a more concrete statement about the association between risk factors and the determination of mortality. The use of multilevel models is an alternative to traditional multivariate models, considering the hierarchical nature present in the data and analyzing the autocorrelation between the risk factors of a given condition at the aggregation levels.^11-12

From this perspective, the present study aimed to analyze the individual and contextual factors associated with mortality from COVID-19 in the state of Pernambuco.

METHODS

This is a case-control study with secondary data and a multilevel approach. The study used severe cases of SARS due to COVID-19 registered at the Centro de Informações Estratégicas da Vigilância em Saúde de Pernambuco (CIEVS-PE, Strategic Information Center for Health Surveillance of Pernambuco), between March 2020 and March 2022, being residents of the state of Pernambuco.

CIEVS/PE is one of the units of the Brazilian National Network for Monitoring and Response to Public Health Emergencies, and is responsible for the process of detecting, monitoring and coordinating the response to public health emergencies. In this context, it is made up of mandatory notifications, outbreaks and epidemics as well as events with health impact.¹³

For the study, confirmed deaths were considered cases, while controls were made up of recovered individuals. To compose the study sample, the inclusion criteria were to have the records made available by CIEVS/PE and to have the fields of independent variables filled in. Regarding the exclusion criteria, records without the final classification and the Cadastro Nacional de Estabelecimentos de Saúde (CNES, Brazilian National Registry of Health Establishments) number of the establishments were not considered.

Death from COVID-19 was considered as a dependent variable, and independent variables were allocated at the individual and contextual levels (subdivided into two levels, healthcare and municipalities).

Individual-level variables consisted of data made available in CIEVS-PE for each individual, such as sex, age group, presence of symptoms, O2 saturation <95, respiratory discomfort, dyspnea, presence of comorbidities, obesity, diabetes, and heart or vascular diseases.

The first contextual level (healthcare) consisted of data related to the hospital where a patient was admitted, made available in CNES, such as the establishment’s legal nature, number of Adult Intensive Care Unit (ICU) beds and number of respirators/ventilators. To define the categories of Adult ICU beds and respirators/ventilators, the average of the period analyzed was considered, and the K-means method was used, which is used to construct automatic data groupings based on the degree of similarity (clusters).¹⁴

Therefore, for both variables, the units were grouped into three clusters (low, medium and high). Regarding Adult ICU beds, the low (0-3 beds), medium (4-44 beds) and high (> 45 beds) clusters were considered. While for ventilators/respirators, the low (0-4 ventilators/respirators), medium (5-50 ventilators/respirators) and high (> 51 ventilators/respirators) clusters were considered.

The second contextual level (municipality) was composed of the individuals’ health macro-region of residence (I - Recife, II - Caruaru, III - Serra Talhada, IV - Petrolina), obtained from CIEVS-PE and the Centro de Integração de Dados e Conhecimentos para Saúde (CIDACS, Center for Data and Knowledge Integration for Health) Social Inequalities Index (COVID-19-SII), which is characterized as a marker of inequalities associated with COVID-19 formed by socioeconomic, sociodemographic information and difficulties in accessing healthcare services, being divided into very low, low, medium, high and very high.¹⁵

Initially, univariate analysis was performed to identify the association between the outcome (death from COVID-19) and each independent variable, calculating the Unadjusted Odds Ratio (Unadjusted OR) and its respective 95% Confidence Intervals (95% CI), in addition to statistical significance (p-value), based on the chi-square test (χ²). For multivariate analysis, the regression included the independent variables that presented statistical significance below 20% (p < 0.20).

The multilevel model used in this study was adjusted using Generalized Linear Latent and Mixed Models (GLLAMM), considering information related to individuals as the first level, the hospital of admission as the second level, and adjustment for the municipality of residence (third level). This type of three-level model allowed estimating the measures of association between exposures and outcomes, considering that observations (individuals) taken within the same group (contexts) are not independent.

At this stage, the model was adjusted using the stepwise method. Thus, the model began with all significant variables in the bivariate model, and then variables whose statistical significance was greater than 10% (p > 0.10) were removed (a process called backward). Finally, to assess the quality of regression models, the study used the Akaike criterion (AIC) and the Bayesian Information Criterion (BIC) for all stages of multivariate analysis. The best models were those that presented results with the lowest AIC and BIC values.

In descriptive data analysis and processing, the Microsoft Excel 2013 spreadsheet editor was used to store the database, and the Data Analysis and Statistical Software (STATA) version 12.0 was used to develop the statistical analyses.

The study was approved by the Research Ethics Committee, under Certificado de Apresentação para Apreciação Ética (CAAE, Certificate of Presentation for Ethical Consideration) 60944122.8.0000.5190 and Opinion 5.652.161.

RESULTS

Table 1 presents the number and proportion of cases and controls at the first level (individual) according to variables, where the absolute number was 18,198 cases and 27,647 controls. Regarding individuals’ sex, there is a predominance of men: 53.25% in cases and 54.33% in controls.

Concerning age group, it was observed that both groups had a greater predominance of individuals under 60 years of age, however, with high proportional differences. In cases, the proportion was 30.45%, and in controls, it was 61.71%. Still regarding age group, it was observed that the most advanced ages (80 years or older) were those that presented the lowest proportion of controls, i.e., individuals who were diagnosed with SARS due to COVID-19 and recovered (Table 1).

In relation to symptoms, the results indicate that they were present in almost all individuals. In cases and controls, the proportion was higher than 98%. When analyzing the presence of symptoms (O2 saturation <95%, respiratory discomfort and dyspnea), it is noted that cases presented a higher proportion than controls (Table 1).

Regarding comorbidities, there was a difference between the groups. In the cases, the predominant presence was the presence of some comorbidity, being 66.15%, while, in the cases, the absence was more predominant, with a proportion of 57.90%. When analyzing the presence of comorbidities (heart or vascular diseases, diabetes and obesity) separately, it is noted that, in both groups, the absence was more predominant. However, in general, the proportions of the presence of these comorbidities were higher in cases when compared to controls (Table 1).

Table 2 presents data on the variables of the level of healthcare, containing the number and proportion of cases and controls. As for the legal nature, it was found that the majority of establishments had a link with the Sistema Único de Saúde (SUS, Brazilian Health System), being private or public. Approximately 72.66% of the cases occurred in public SUS establishments, while controls were 66.60%.

Concerning the number of adult ICU beds available, in both groups, it is worth noting that establishments with a number considered average (4 to 44 beds) presented the highest number of cases (53.12%) and controls (45.08%) (Table 2).

Regarding the number of ventilators/respirators, similar results were observed. The value considered average (5 to 50 ventilators/respirators) was the category that presented the highest proportion of cases and controls, equivalent to 52.62% and 51.12%, respectively (Table 2).

In Table 3, it is possible to observe the results of bivariate analysis related to the municipality level. Regarding COVID-19-SII, it is noteworthy that the predominance of cases (34.58%) and controls (34.85%) was in the category listed as a high index.

As for the location of residence of individuals, it is noted that the majority of the proportion of cases and controls was concentrated in health macroregion I (Recife), with a total of 69.91% and 67.00%, respectively (Table 3).

Table 4 presents the consolidation of the final results of the multilevel model for the association between mortality from COVID-19 and individual and contextual determinants in the state of Pernambuco. Initially, the model consisted of 15 variables, of which only two did not make up the final model: one from the level of healthcare (legal nature) and one from the municipal level (COVID-19-SII).

The individual level was composed of ten variables, showing a statistical association with the final outcome and all with p = 0.000. The level of healthcare was composed of two variables: number of Adult ICUs and number of ventilators/respirators. Meanwhile, the municipal level was composed only of the variable referring to the health macroregion (Table 4).

In Pernambuco, being male (OR = 1.17) and aged 80 years or older (5.65) represented a risk factor for mortality from COVID-19. In the symptom variables, it is noteworthy that the presence of O2 saturation <95% was characterized as the greatest risk factor, being 1.56 times higher. However, the presence of respiratory distress and dyspnea was also found to be a risk factor, with an adjusted OR of 1.45 and 1.31, respectively. The presence of comorbidities (adjusted OR = 1.71) was also found to be a risk factor, with the presence of obesity (1.36) also standing out (Table 4).

Concerning healthcare, individuals admitted to establishments belonging to the medium category of the number of Adult ICU beds and ventilators/respirators had a higher chance of death from COVID-19.

The health macroregion variable was the only one at the municipal level that remained in the final model, which presented an adjusted OR lower than 1.

DISCUSSION

The results obtained in the present study showed that the individual level presented greater determining factors with mortality from COVID-19 in the state of Pernambuco, highlighting men, age group of 80 years or older, presence of symptoms, comorbidities, O2 saturation <95%, respiratory discomfort, dyspnea, heart or vascular diseases, diabetes and obesity.

In Pernambuco, there was a predominance of cases and controls in the male population, where men had a 17% higher chance of death. In a study conducted in the state of Rio Grande do Norte, the results indicate that being male was found to be a risk factor for mortality from the disease, with men having a 45% higher chance of death than women; therefore, they had a lower chance of survival. ¹⁶ A meta-analysis study found that the number of cases and fatality rates were more prevalent in the male population.¹⁷

From this perspective, it is worth noting that the chances of men dying more than women from COVID-19 can be characterized as a warning sign, and may present multivariate factors. However, the identification and recognition of this mortality factor are essential to support public policies and actions aimed at combating the disease, with ensuring equal access to healthcare services being a crucial device for men’s healthcare.

Older adults aged 80 years or older presented a higher risk of death from COVID-19, as was observed for the fatality rates from the disease.¹⁸ The pandemic attenuated the social inequalities that exist in the various scenarios in the country, being deepened in the most vulnerable older adults. The permanence in economic activities and in remote mode was found to be a barrier for this population. From this perspective, in addition to health, the pandemic brought about a series of social consequences.¹⁹

The presence of symptoms was predominant in both groups (cases and controls), being a determining factor for death, a similar result in a study that indicated cough, myalgia and other respiratory signs as the most frequent symptoms in the population.²⁰ However, it is noteworthy that the present study presents a prevalence of symptoms, as it is a study with a database of serious and hospitalized cases. This fact may have influenced the lack of differentiation between case and control groups, since almost all cases that generate hospitalization are symptomatic individuals.

Still regarding the presence of symptoms, individuals who presented O2 saturation <95%, respiratory distress and dyspnea had a higher risk of death in the state of Pernambuco. In Recife, the state capital, results indicate similar findings, where severe respiratory symptoms were also associated with death in the city.²¹ Respiratory symptoms have also been identified in patients worldwide, such as the presence of dyspnea.¹⁷ From this perspective, continuous monitoring of symptoms, through different strategies, is a strategy for preventing worsening and death from the disease.²²

The presence of comorbidities as a risk factor was more prevalent in the case group than in the control group. Among the comorbidities analyzed in the study, heart or vascular diseases and diabetes were the ones that presented the highest proportions in the case group. In a cohort study carried out to identify the clinical characteristics of patients diagnosed with COVID-19, it was found that the presence of some comorbidity was more prevalent in deaths (48%) than in those who recovered (14%), with cardiovascular diseases standing out.²³

This population requires differentiated healthcare, in which mapping these individuals can be a fundamental tool for healthcare. In this context, the pandemic reaffirmed the importance of coordination between health surveillance services and Primary Care, with multidisciplinary and interdisciplinary action constituting one of the main healthcare strategies within SUS.²⁴

At the contextual level, in the context of healthcare, the number of ICU beds was found to be a determining factor in mortality, with individuals admitted to facilities with the average number of beds having the highest risk. Sometimes, hospitals with the highest number of beds received patients with the most critical conditions, which can increase mortality rates in these facilities. A national survey found that the health regions with the highest mortality rates were marked by a shortage of beds and equipment.²⁵ However, it is worth noting that the present study did not necessarily seek to analyze the availability of beds and ventilators/respirators, but rather the dimension of healthcare.

As for the second contextual level, initially, due to the history of social disparities in the distribution of infectious and parasitic diseases, the presence of higher risk variables was expected, but this result was not observed. This result may indicate that proximal (individual) factors are stronger than distal (contextual) factors. However, this finding does not imply that there is no association between COVID-19 and the context of individuals. The pandemic has led to the development of several studies, with different methodological approaches, with the source of information being one of the main differences between the studies.

Therefore, this result points to the need for future studies that consider other contextual variables and methodologies. Even with the advancement of coverage and quality of health information systems, limitations regarding the completion and completeness of variable records are constant. However, despite the limitations exposed, the construction of research based on secondary databases allows the development of longitudinal studies with a low operational cost.

Given the scenario identified in the results of this study, it is essential to humanize the data. The deaths that occurred in the state of Pernambuco are not just epidemiological numbers, but information about individuals with names, families and stories. Part of the Brazilian population died because the right to life was denied, mainly due to living conditions and access to healthcare services.

The fight against COVID-19 and the search for improvements in the quality of life of the population of the state of Pernambuco necessarily involve strengthening SUS as the main state policy for guaranteeing citizenship and the right to health, reducing social inequalities and inequities. Finally, the pandemic has exposed the Brazilian health system weaknesses, but it has also made clear the greatness of SUS.

REFERENCES

1. World Health Organization. WHO. Coronavirus disease (COVID-19). 2023. https://www.who.int/health-topics/coronavirus.

2. Guo Y, Cao Q, Hong Z, et al. The origin, transmission and clinical therapies on coronavirus disease 2019 (COVID-19) outbreak - an update on the status. Military Med Res. 2020; 7: 11. https://dx.doi.org/10.1186/s40779-020-00240-0.

3. World Health Organization. - WHO. WHO declares Public Health Emergency on novel coronavirus. 2020. https://www.paho.org/en/news/30-1-2020-who-declares-public-health-emergency-novel-coronavirus.

4. Ministério da Saúde. (BR). Portaria n° 188, de 3 de fevereiro de 2020. Declara Emergência em Saúde Pública de importância Nacional (ESPIN) em decorrência da Infecção Humana pelo novo Coronavírus (2019-nCoV). Diário Oficial da República Federativa do Brasil, Brasília (DF), 2020 fev 3; Seção 1:1. https://bvsms.saude.gov.br/bvs/saudelegis/gm/2020/prt0188_04_02_2020.html.

5. Silva LLSD, Lima AFR, Polli DA, et al. Medidas de distanciamento social para o enfrentamento da COVID-19 no Brasil: caracterização e análise epidemiológica por estado. Cad. Saúde Pública. 2020; 36 (9): 1-15. https://dx.doi.org/10.1590/0102-311X00185020.

6. World Health Organization. - WHO. Coronavirus (COVID-19) Dashboard. 2023. https://covid19.who.int.

7. Ministério da Saúde. (BR). Painel Coronavírus. 2023. https://covid.saude.gov.br.

8. Albuquerque MV, Ribeiro LHL. Desigualdade, situação geográfica e sentidos da ação na pandemia da COVID-19 no Brasil. Cadernos de Saúde Pública. 2020; 36 (12): 1-14. https://dx.doi.org/10.1590/0102-311X00208720.

9. Reigada CLC, Smiderle CASL. Atenção à saúde da mulher durante a pandemia COVID-19: orientações para o trabalho na APS. Revista Brasileira de Medicina de Família e Comunidade. 2021; 16 (43): 2535. https://dx.doi.org/10.5712/rbmfc16(43)2535.

10. Barata RB. Como e por que as desigualdades sociais fazem mal à saúde. Rio de Janeiro: Editora FIOCRUZ; 2009. https://books.scielo.org/id/48z26.

11. Santos CAST, Amorim LDAF, Oliveira NF. Métodos de análise multinível em epidemiologia. In: Almeida Filho N, Barreto ML, organizadores. Epidemiologia & saúde: fundamentos, métodos, aplicações. Rio de Janeiro: Guanabara Koogan; 2011. p. 265-72.

12. Puente-Palacios KL, Laros JA. Análise multinível: contribuições para estudos sobre efeito do contexto social no comportamento individual. Estudos de Psicologia. 2009; 26 (3): 349-361. https://dx.doi.org/10.1590/S0103-166X2009000300008.

13. Centro de Informações Estratégicas de Vigilância em Saúde. - CIEVS. Pernambuco: Centro de Informações Estratégicas de Vigilância em Saúde. 2020. https:// www.cievspe.com/home.

14. Jain AK. Data clustering: 50 years beyond Kmeans. Pattern Recognit Lett. 2010; 31: 651-66. https://dx.doi.org/10.1016/j.patrec.2009.09.011.

15. Grimberg A, Ara A, Ferreira B, et al. Guia sobre o índice de desigualdades sociais para Covid-19: IDS-COVID-19. Salvador: CIDACS; 2022. https://cidacs.bahia.fiocruz.br/idscovid19/wp-content/uploads/2022/06/Guia-do-IDS-COVID-19.pdf.

16. Galvão MHR, Roncalli AG. Fatores associados a maior risco de ocorrência de óbito por COVID-19: análise de sobrevivência com base em casos confirmados. Revista Brasileira de Epidemiologia. 2021; 23. https://dx.doi.org/10.1590/1980-549720200106.

17. Li LQ, Huang T, Wang YQ, et al. COVID‐19 patients' clinical characteristics, discharge rate, and fatality rate of meta‐analysis. Journal of medical virology. 2020; 92 (6): 577-583. https://dx.doi.org/10.1002/jmv.25757.

18. Bonanad C, García-Blas S, Tarazona-Santabalbina F, et al. The effect of age on mortality in patients with COVID-19: a meta-analysis with 611,583 subjects. Journal of the American Medical Directors Association, 2020; 21 (7); 915-918. https://dx.doi.org/10.1016/j.jamda.2020.05.045.

19. Romero DE, Muzy J, Damacena GN, et al. Idosos no contexto da pandemia da COVID-19 no Brasil: efeitos nas condições de saúde, renda e trabalho. Cadernos de Saúde Pública. 2021; 37 (3): e00216620. https://dx.doi.org/10.1590/0102-311X00216620.

20. Wong MCS, Huang J, Wong YY, et al. Epidemiology, symptomatology, and risk factors for long COVID symptoms: population-based, multicenter study. JMIR Public Health and Surveillance, 2023; 9 (1): e42315. https://dx.doi.org/10.2196/42315.

21. Souza IVSD, Holanda ERD, Barros MBSC. Fatores associados ao óbito por covid-19 em Recife, Pernambuco, 2020: estudo transversal com dados do sistema “Notifique Aqui”. Epidemiologia e Serviços de Saúde. 2023; 32: e2022701. https://dx.doi.org/10.1590/S2237-96222023000200014.

22. Jansen-Kosterink S, Hurmuz M, Ouden MD, et al. Predictors to use mobile apps for monitoring COVID-19 symptoms and contact tracing: survey among dutch citizens. JMIR Formative Research, 2021; 5 (12): e28416. https://dx.doi.org/10.2196/28416.

23. Chen T, Wu DI, Chen H, et al. Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study. BMJ. 2020; 368. https://dx.doi.org/10.1136/bmj.m1091.

24. Prado NMDBL, Biscarde DGDS, Pinto Junior EP, et al. Ações de vigilância à saúde integradas à Atenção Primária à Saúde diante da pandemia da Covid-19: contribuições para o debate. Ciência & Saúde Coletiva. 2021; 26: 2843-2857. https://dx.doi.org/10.1590/1413-81232021267.00582021.

25. Moreira RS. COVID-19: unidades de terapia intensiva, ventiladores mecânicos e perfis latentes de mortalidade associados à letalidade no Brasil. Cadernos de Saúde Pública. 2020; 36 (5): e00080020. https://dx.doi.org/10.1590/0102-311X00080020.

IR