ABSTRACT

OBJECTIVE

To evaluate the accuracy of yellow fever (YF) suspected case definitions from the Brazilian Ministry of Health (BMH) and World Health Organization (WHO), as well as propose and evaluate new definitions of suspected cases, considering confirmed and discarded cases.

METHODS

The retrospective study was conducted at the Instituto de Infectologia Emílio Ribas (IIER), using the Epidemiologic Surveillance Form of YF cases. From the confirmed and discarded cases of YF, a logistic regression model was developed. The independent variables were used in a proposed definition of a suspected case of YF and its accuracy was evaluated.

RESULTS

In total, 113 YF suspect cases were reported, with 78 confirmed (69.0%). The definitions by BMH and WHO presented low sensitivity, 59% and 53.8%, and reduced accuracy, 53.1% and 47.8%, respectively. Predictive factors for YF were thrombocytopenia, leukopenia, and elevation of transaminases greater than twice normal. The definition including individual with acute onset of fever, followed by elevation of ALT or AST greater than twice the reference value AND leukopenia OR thrombocytopenia presented high sensitivity (88.3%), specificity (62.9%), and the best accuracy (80.4%), as proposed in the model.

CONCLUSION

The YF suspected case definitions of the BMH and the WHO have low sensitivity. The inclusion of nonspecific laboratory tests increases the accuracy of YF definition.

Yellow Fever; Accuracy; Case Definition; Epidemiologic Surveillance

INTRODUCTION

Yellow fever (YF) is an acute disease, endemic in some tropical areas of the Americas and Africa, and it can affect humans and non-human primates (NHP). It is caused by a virus of the genus Flavivirus, family Flaviviridae, and transmitted by mosquitoes. There are two cycles of transmission: sylvatic and urban. YF is a burden for public health due to its clinical severity and high potential for dissemination in urban areas¹1. Monath TP, Vasconcelos PF. Yellow fever. J Clin Virol. 2015 Mar;64:160-73. https://doi.org/10.1016/j.jcv.2014.08.030
https://doi.org/10.1016/j.jcv.2014.08.03... . Vaccination is the most important measure to prevent YF, especially its international spread. The World Health Organization (WHO) developed, by a coalition of partners (The Vaccine Alliance-GAVI and UNICEF) the comprehensive global strategy to eliminate yellow fever epidemics (EYE) (2017–2026), to face the changing epidemiology of yellow fever, resurgence of mosquitoes and increased risk of urban outbreaks and international spread²2. Banluta C. Yellow fever. Geneva: World Health Organization; (date unkown [cited 2022 May 2]). Available from: https://www.who.int/health-topics/yellow-fever#tab=tab_1/
https://www.who.int/health-topics/yellow... . The last case of urban yellow fever in Brazil occurred in 1942, despite the sporadic record of cases of sylvatic YF. The disease was endemic until 1999, especially in the North and Central-West regions. Between 2000 and 2008, an expansion of a viral circulation was observed towards the East and South regions³3. Secretaria de Estado da Saúde (SP). Coordenadoria de Controle de Doenças. Centro de Vigilância Epidemiológica. Nota Técnica sobre Febre Amarela no estado de São Paulo 2017. [cited 2020 Aug 15]. Available from: http://docs.bvsalud.org/biblioref/2020/05/1095889/famarela17_nota_tecnica.pdf
http://docs.bvsalud.org/biblioref/2020/0... .

From December 2016 to May 2017, 792 cases of sylvatic YF were reported and 274 deaths were confirmed (lethality 34.5%), predominating in the states of Minas Gerais (61.4%) and Espírito Santo (32.8%), with also 642 confirmations in NHP⁴4. Ministério da Saúde (BR). Secretaria de Vigilância em Saúde. Centro de Operações de Emergências em Saúde Pública sobre Febre Amarela. Informe – Nº 43/207. Monitoramento dos casos e óbitos de febre amarela no Brasil. Brasília, DF: Ministério da Saúde; 2017 [cited 2020 May 2]. Available from: https://sbim.org.br/images/files/ms-febreamarela-170531.pdf
https://sbim.org.br/images/files/ms-febr... . From July 2017 to April 2018, 1,127 human cases of sylvatic YF and 331 deaths were confirmed (lethality 29.4%), predominating in the states of Minas Gerais (42.6%), São Paulo (40.2%) and Rio de Janeiro (16.6%)⁵5. Ministério da Saúde (BR). Secretaria de Vigilância em Saúde. Informe - Nº 21. Monitoramento do Período Sazonal de Fabre Amarela - Brasil - 2017/2018. Brasília, DF: Ministério da Saúde; 2018 [cited 2020 May 2]. Available from: https://sbim.org.br/images/files/informe-fa-21-11abr18-c.pdf
https://sbim.org.br/images/files/informe... .

YF is a notifiable disease in Brazil. Its definition of a suspect case is an individual with fever (up to 7 days) of sudden occurrence followed by jaundice and/or hemorrhagic manifestations, residing in (or coming from) a YF risk area, or in locations with confirmed epizootic in NHP or of isolation of virus in vector mosquitoes in the 15 days prior, not vaccinated against YF, or with ignored vaccine status⁶6. Ministério da Saúde (BR). Secretaria de Vigilância em Saúde. Coordenacao-Geral de Desenvolvimento da Epidemiologia em Servicos. Guia de vigilancia em saude: volume único. 3a ed. Brasilia, DF: Ministerio da Saude; 2019.. In 2009, a study analyzed 28 confirmed cases of sylvatic YF in the State of São Paulo. However, only 50% of the cases matched the definition of suspect case⁷7. Mascheretti M, Tengan CH, Sato HK, Suzuki A, Souza RP, Maeda M, et al. Yellow fever: reemerging in the state of Sao Paulo, Brazil, 2009. Rev Saude Publica. 2013 Oct;47(5):881-9. https://doi.org/10.1590/S0034-8910.2013047004341
https://doi.org/10.1590/S0034-8910.20130... .

This study aimed to describe the suspect cases attended at the Instituto de Infectologia Emílio Ribas (IIER), to evaluate sensitivity and specificity of YF case definitions from the Brazilian Ministry of Health and WHO during the 2018 outbreak, and to propose highly accurate case definitions.

METHODS

Study Design: Retrospective Observational Study

Data sources: data from patients with suspected YF treated at the IIER were collected using the Epidemiological Surveillance Forms, selecting sociodemographic, clinical, and laboratory variables. Most patients were from the State of São Paulo (96.2%), mostly from the municipality of Mairiporã (41.0%).

Inclusion requirement: suspect cases had a clinical sample collected for laboratorial etiological confirmation provided by Instituto Adolfo Lutz, Public Health Laboratory. Suspected cases of YF were confirmed or discarded by YF virus detection by real-time polymerase chain reaction (RT-PCR) technique (91% and 91.4%) or positive immunoglobulin M (IgM) serology (8% and 5.7%), respectively.

Dependent variable: case classification (confirmed or discarded)

Independent variables: sociodemographic, clinical (signs and symptoms), and laboratory tests.

Analysis Procedure

Characteristics and the confirmed and discarded cases for YF were evaluated by Pearson’s chi-squared test and Fisher’s exact test. Univariate analysis of signs, symptoms, and laboratorial alterations of suspect cases were performed in order to identify possible variables predictive of YF, considering odds ratio (OR) values. The variables with p-value ≤ 0.20 in non-adjusted analysis were considered for a logistic regression model to identify independent factors associated to YF. The final models were evaluated by Hosmer-Lemeshow test.

Receiver operating characteristic (ROC) curves were constructed for each of the explanatory models, and the predictive factors were considered to create proposed definitions (1 and 2). Those proposed definitions were evaluated by sensitivity, specificity, predictive value positive (PVP), predictive value negative (PVN), and accuracy, which were also used to evaluate the following official definitions of suspect case.

Brazil’s Ministry of Health’s – Brazil’s definition (2017)

An individual with acute onset of fever (up to 7 days), followed by jaundice and/or hemorrhagic manifestations, residing in (or coming from) a risk area for YF, or from locations with occurrence of epizootic confirmed in NHP, or isolation of virus in vector mosquitoes, in the 15 days prior, not vaccinated against YF, or with ignored vaccine status. In cases of outbreak, it is recommended to adjust the definition of suspect case making it more sensitive for detection of the highest number of cases possible, taking in consideration the wide clinical spectrum of the disease⁶6. Ministério da Saúde (BR). Secretaria de Vigilância em Saúde. Coordenacao-Geral de Desenvolvimento da Epidemiologia em Servicos. Guia de vigilancia em saude: volume único. 3a ed. Brasilia, DF: Ministerio da Saude; 2019..

World Health Organization’s – WHO definition (2015)

Any individual with acute onset of fever and jaundice occurrence within 14 days of onset of the first symptoms⁸8. World Health Organization. Yellow fever: case definitions for public health surveillance, 2015. Geneva: World Health Organization; (date unkown [cited 2022 May 2]). Available from: http://www.who.int/csr/disease/yellowfev/case-definition/en/
http://www.who.int/csr/disease/yellowfev... .

For the analysis of both definitions (Brazil and WHO), a vaccinated individual was considered as one with immunization received in, at least, 10 days from the beginning of the symptoms⁹9. Barrett AD, Teuwen DE. Yellow fever vaccine: how does it work and why do rare cases of serious adverse events take place? Curr Opin Immunol. 2009 Jun;21(3):308-13. https://doi.org/10.1016/j.coi.2009.05.018
https://doi.org/10.1016/j.coi.2009.05.01... . Moreover, the cases with non-reported traveling to or residence in YF risk areas were included as possible exposition to sylvatic environments during the YF outbreak. Laboratorial value patterns considered to describe observed alterations were the following: thrombocytopenia with platelet count equal or under 150,000/mm3; leukopenia with leukocyte count equal or under 4,500/mm3; leukocytosis with leukocyte count equal or over 11,000/mm3; renal function alterations with serum urea values over 40mg/dL and/or creatinine over 1.3 mg/dL; hyperbilirubinemia with total bilirubin serum dosage over 2 mg/dL; and increase of twice the reference value for transaminases considering the maximum value of 40 U/L for alanine aminotransferase (ALT) and aspartate aminotransferase (AST).The significance level adopted was 5% for all hypothesis tests. Analyses were performed using SPSS for Windows v.25 and Stata/MP 14.0 for Windows software. The study used data from hospital epidemiological surveillance. The data was approved by the Ethics Committee of IIER (Protocol 024329/2018).

RESULTS

In total, 113 suspect cases of YF were reported at IIER from January to November 2018, with 78 confirmed cases (69.0%). Of these, 23 individuals died (lethality of 29.5%). Males predominated among confirmed (80.8%) and discarded (74.3%) cases, without statistical difference between both groups (p=0.436). Analysis of age group among suspected population showed a predominance of adult cases (from 18 to 59 years old) (77.9%) and similar distribution between confirmed and discarded (p = 0.061). Whites were prevalent among suspect cases (69.9%), without statistical difference among confirmed and discarded (p = 0.183). Educational level was not different between those groups (p = 0.094), even though a larger proportion of individuals with up to 8 years of education or more is observed in the discarded group (91.4%) compared to the confirmed ones (78.4%). A higher proportion of individuals whose disease was discarded were previously vaccinated against YF (14.3%) compared to those who had the disease (3.8%), with no statistical difference (p = 0.105). Hospitalization occurred in 89.7% of confirmed cases, with significant difference (p < 0.001) if compared to the discarded cases (54.3%), as well as evolution to death, which occurred with 8.3% of those discarded and 29.5% of those confirmed, p = 0.016).

We analyzed signs, symptoms, and laboratorial alterations to verify possible predictive characteristics of YF among suspect cases attended. Initially, univariate analysis was performed, comparing confirmed and discarded cases of YF (Table 1). All variables with p < 0.20 were included in the logistic regression model. Two models show the significant and adjustment variables (Table 2).

The quality of adjustment of the multivariate models were evaluated by Hosmer-Lemeshow test, whose p-value was not statistically significant (Model 1: p = 0.801; Model 2: p = 0.904), indicating good fit. In Model 1, significant predictive factors identified were leukopenia (OR= 4.26; 95%CI: 1.29–14.05) and the AST elevation superior to twice the reference value (OR= 14.27; 95%CI: 2.62–77.79) and thrombocytopenia (OR= 2.86; 95%CI: 0.96–8.54), p < 0.06. In Model 2, significant predictive factors were thrombocytopenia (OR= 3.39; 95%CI: 1.11–0.32), leukopenia (OR= 5.88; 95%CI: 1.51–22.95), and the ALT elevation superior to twice the reference value (OR= 53.67; 95%CI: 5.26–547.36). Figures 1 and 2 show the ROC curves for the Model 1 and Model 2:

Based on such factors, three different case definitions were proposed.

1. Proposed definition 1: individual with acute onset of fever (reported or measured), followed by elevation of AST superior to twice the reference value AND leukopenia AND thrombocytopenia.

2. Proposed definition 2: individual with acute onset of fever (reported or measured), followed by elevation of ALT superior to twice the reference value AND leukopenia AND thrombocytopenia.

3. Proposed definition 3: individual with acute onset of fever (reported or measured), followed by elevation of ALT or AST superior to twice the reference value AND leukopenia OR thrombocytopenia.

Accuracy analysis was also compared with official definitions from Brazilian Ministry of Health (2017) and from World Health Organization (2015) (Table 3).

Evaluation of sensitivity of suspect cases definitions indicates that only 59.0% met Brazil’s suspect case definition and 53.8% met WHO’s definition. Specificity was higher for Brazil’s definition (40.0%), while for WHO’s this indicator presented a lower value of 34.3%. PVP presented similar values, in both definitions, with higher index for Brazil’s definition (68.7%). PVN presented values of 30.4% and 25.0% in Brazil’s and WHO’s definitions, respectively. The definition with the most accuracy in the analyzed population was Brazil’s (53.1%), while WHO’s presented 47.8% accuracy in the analyzed population.

Proposed definition 1 presented low sensitivity (54.5%), high specificity (88.6%), high PVP (91.3%), low PVN (47.0%), and moderate accuracy (65.2%). Proposed definition 2 presented low sensitivity (55.8%), high specificity (91.4%), high PVP (93.5%), low PVN (48.5%), and moderate accuracy (67.0%). Proposed definition 3 presented high sensitivity (88.3%), moderate specificity (62.9%), high PVP (84.0%), moderate PVN (71.0%), and high accuracy (80.4%). Considering the criteria of all definitions described, proposed definition 3 was the most accurate.

DISCUSSION

Case definition is an important tool for epidemiological surveillance, allowing detection of cases, estimation of incidence, and identification of epidemics. Standardization of case definition also contributes to evaluation of effectiveness of control measures and incidence comparison in different regions and periods. Case definition must be simple and useful, combining clinical, laboratorial, and epidemiological characteristics, depending on the monitoring objectives for each disease or health condition. A definition with high sensitivity and specificity is desirable, but these attributes must be balanced, considering the increase of sensitivity of great importance when extension of an epidemic needs to be assessed¹⁰10. Fitzner J, Qasmieh S, Mounts AW, Alexander B, Besselaar T, Briand S, et al. Revision of clinical case definitions: influenza-like illness and severe acute respiratory infection. Bull World Health Organ. 2018 Feb;96(2):122-8. https://doi.org/10.2471/BLT.17.194514
https://doi.org/10.2471/BLT.17.194514... . Balance must be reached between the need of high sensitivity at the expense of screening false-positive cases¹¹11. Teutsch SM, Churchill RE, editors. Principles and practice of public health surveillance. New York: Oxford University Press; 2000..

Brazilian Ministry of Health and World Health Organization’s case definitions presented lower sensitivity, specificity, PVN, and accuracy, with moderate PVP. YF in Brazil presents low incidence, with significant increase of cases in 2017 and 2018, especially in the Southeastern region. A more sensitive case definition is necessary for diseases with low incidence, thus reducing the number of false negatives, considering the severity of disease and the possibility of prevention¹²12. Centers for Disease Control and Prevention. Case definitions for infectious conditions under public health surveillance. MMWR Recomm Rep. 1997 May;46 (RR-10):1-55..

Predictive factors of YF in the analyzed population were transaminases greater than twice the reference value, leukopenia, and thrombocytopenia. A study¹³13. Ribeiro AF, Cavalin RF, Suleiman JMAH, Costa JA, Vasconcelos MJ, Málaque CMS, et al. Yellow fever: factors associated with death in a hospital of reference in infectious diseases, São Paulo, Brazil, 2018. Am J Trop Med Hyg. 2019 Jul;101(1):180-8. https://doi.org/10.4269/ajtmh.18-0882
https://doi.org/10.4269/ajtmh.18-0882... analyzed risk factors for death in 72 confirmed cases of YF and it identified, among recovered cases, the presence of leukopenia and increase of transaminases and thrombocytopenia. The analysis of regression showed independent factors related to death the increase of transaminases, age, and creatinine increase¹³13. Ribeiro AF, Cavalin RF, Suleiman JMAH, Costa JA, Vasconcelos MJ, Málaque CMS, et al. Yellow fever: factors associated with death in a hospital of reference in infectious diseases, São Paulo, Brazil, 2018. Am J Trop Med Hyg. 2019 Jul;101(1):180-8. https://doi.org/10.4269/ajtmh.18-0882
https://doi.org/10.4269/ajtmh.18-0882... . Another study, which analyzed 76 cases of YF, showed leukopenia, thrombocytopenia, and increase of transaminases among the cases. When risk factors for death were analyzed, there was statistical significance for the increase of transaminases, neutrophilia, and viral load¹⁴14. Kallas EG, D’Elia Zanella LG, Moreira CH, Buccheri R, Diniz GB, Castiñeiras AC, et al. Predictors of mortality in patients with yellow fever: an observational cohort study. Lancet Infect Dis. 2019 Jul;19(7):750-8. https://doi.org/10.1016/S1473-3099(19)30125-2
https://doi.org/10.1016/S1473-3099(19)30... . Analysis of 52 YF confirmed cases showed increase of transaminases and thrombocytopenia. However, logistic regression showed that alteration of renal excretion and increase of transaminases were significantly associated to death¹⁵15. Escosteguy CC, Pereira AG, Marques MR, Lima TR, Galliez RM, Medronho RA. Yellow fever: profile of cases and factors associated with death in a hospital in the State of Rio de Janeiro, 2017-2018. Rev Saúde Pública. 2019 Oct;53:89. https://doi.org/10.11606/s1518-8787.2019053001434
https://doi.org/10.11606/s1518-8787.2019... . A study analyzing confirmed cases of YF attended in ICU identified factors associated to death, serum lipase and, as protection, factor V. Nonetheless, in the univariate analysis there was increase of transaminases, bilirubin, lipase, INR and lactate in patients who died, and no alteration of viral load when compared to cases which evolved to cure¹⁶16. Casadio LV, Salles AP, Malta FM, Leite GF, Ho YL, Gomes-Gouvêa MS, et al. Lipase and factor V (but not viral load) are prognostic factors for the evolution of severe yellow fever cases. Mem Inst Oswaldo Cruz. 2019;114:e190033. https://doi.org/10.1590/0074-02760190033
https://doi.org/10.1590/0074-02760190033... .

Proposed case definition 2 presented moderate (67,6%), sensitivity of 56,6% specificity of 91.4%, PVP of 93.5%. This case definition includes fever, increase of transaminases (AST superior to twice), and thrombocytopenia or leukopenia. The exams are simple, inexpensive, and improve sensitivity for detection of cases. Proposed definition 1 presents accuracy of 65.8%, sensitivity of 55.3%, and specificity of 88.6%, PVP 91.3%. In this proposed definition, the presence of fever was followed by laboratorial alterations, increase of transaminases (ALT superior to twice), leukopenia, thrombocytopenia. This proposed definition has lower sensitivity and higher specificity. Proposed definition 3 presented high sensitivity (88.3%), moderate specificity (62.9%), high PVP (84.0%), moderate PVN (71.0%), and high accuracy (80.4%). In this proposed definition, the presence of fever was followed by elevation of ALT or AST greater than twice the reference value and leukopenia or thrombocytopenia. Considering the epidemiological situation with high or low case incidence, we must adapt the case definition regarding sensitivity and specificity. In a scenario of low incidence of disease, a higher sensitivity is desirable, even if it increases the number of false positives, to minimize the number of non-detected cases. However, a surveillance system with high PVP, low false positive reports, would not lead to wasted resources on cases that do not actually exist¹¹11. Teutsch SM, Churchill RE, editors. Principles and practice of public health surveillance. New York: Oxford University Press; 2000.. Therefore, definitions proposed by Brazilian Ministry of Health and WHO presented low sensitivity, thus requiring more sensitive definitions to detect most of YF cases. Other studies developed definitions with the use of similar methodology¹⁷17. Braga JU, Bressan C, Dalvi AP, Calvet GA, Daumas RP, Rodrigues N, et al. Accuracy of Zika virus disease case definition during simultaneous Dengue and Chikungunya epidemics. PLoS One. 2017 Jun;12(6):e0179725. https://doi.org/10.1371/journal.pone.0179725
https://doi.org/10.1371/journal.pone.017... . Incorporation of laboratorial results made the definition more sensitive, since transaminases alterations in serum are earlier identified than clinical manifestation of jaundice, a signal included in both case definitions (Brazil and WHO) that may be absent in some cases¹⁸18. Xavier AR, Freitas GS, Santos CF, Januzzi WA, Lacerda GS, Carvalho ER, et al. Yellow fever: laboratorial diagnosis and clinical manifestations. J Bras Patol Med Lab. 2018;54(5):296–305. https://doi.org/10.5935/1676-2444.20180050
https://doi.org/10.5935/1676-2444.201800... .

Limitation of study may be attributed to the use of data from one hospital only, consequently with a small sample size that can lead to inaccuracies in the association measures. However, Instituto de Infectologia Emílio Ribas is a reference public hospital for infectious disease in the state of São Paulo. Furthermore, cases were identified during a YF epidemic in some regions of the state. Thus, increasing the chance of virus detection in investigated cases. Most cases confirmed the YF virus detection by real-time polymerase chain reaction (RT-PCR) technique (91%). Only seven cases were confirmed by serology, an exam that can cross-react with other flaviviruses. However, four of these cases had jaundice, one associated with hemorrhage that progressed to death. These cases were from regions with yellow fever transmission in the same period. The use of laboratorial exams to case definition can be difficult in regions with lack of access to laboratory diagnosis. Nevertheless, the inclusion of laboratorial criteria, associated to symptomatology, increased sensitivity of case definition, making it useful in situations of outbreak and in sentinel investigation of feverish syndromes for differential diagnosis, especially considering coexistence of many arboviruses in Brazil with similar symptomatology. Furthermore, detection of YF virus in densely populated peri-urban regions with a high rate of Aedes aegypti infestation increases the risk of urban YF epidemic. Therefore, case definitions with high accuracy allow implementation of more timely preventive measures, such as vaccination of population under risk.

Acknowledgements

To the professionals of the Epidemiology Service and Emergency Room of the Instituto de Infectologia Emílio Ribas and the Instituto Adolfo Lutz, Public Health laboratory, for their assistance in investigating the cases, considering the epidemiological, clinical, and laboratory data.

REFERENCES

Publication Dates

History