Risk factors predicting dengue hospitalization and disease severity in children and adolescents living in an endemic area of arbovirus transmission in Northeast Brazil (2017–2020)

Abstract

Background

Dengue remains a major public health threat worldwide, particularly in the pediatric population. Dengue illness can develop from infection with any of the four dengue virus serotypes (DENV1-4), ranging from mild fever to potentially life-threatening severe dengue. In children, most dengue infections present as an undifferentiated fever, making it difficult to distinguish dengue from several other common viral infections that frequently occur at early ages. Thus, the rapid identification of dengue cases followed by appropriate early interventions and management is crucial to improve clinical prognosis. In this study, we evaluated socio-demographic, clinical, and laboratory factors predictive of disease severity in children and adolescents in Northeast Brazil.

Methods

Patients aged < 1–19 years old (n = 288) hospitalized with dengue in a reference hospital between 2017 and 2020 were included in the study. Dengue patients were classified as dengue with warning signs (DwWS) and severe dengue (SD). Individual and sociodemographic information was retrieved from the hospital’s epidemiological investigation forms. Clinical data and laboratory parameters collected at the time of admission to the hospital were extracted from the medical records.

Results

Out of 288 patients, 72.2% (208/288) were classified as DwWS. We observed a steady decline in the number of hospitalizations by SD in the pediatric population during the study period, including in the epidemic year of 2019. SD cases were more frequently observed in children aged 7–12 years old (OR = 3.72, 95%CI = 1.34–13.21) and with self-identified mixed (OR = 2.28, 95%CI = 1.22–4.52) and black (OR = 5.92, 95%CI = 1.73–21.1) race. Multivariate analysis showed irritability (OR = 3.69, 95%CI = 1.36–11.19), fluid accumulation (OR = 5.50, 95%CI = 1.51–23.63), and lethargy/restlessness (OR = 19.95, 95%CI = 6.47–73.23) as significant parameters predictive of SD at hospitalization. Binary logistic regression indicated that increased levels of aspartate aminotransferase (AST; p-value = 0.003), monocytes (p-value < 0.0001), basophils (p-value = 0.0003), and red cell distribution width (RDW) percentage (p-value = 0.005) at admission are among the laboratory parameters associated with SD in the study population.

Conclusions

The socio-demographic (age and race), clinical signs and symptoms (irritability, fluid accumulation, and lethargy/restlessness), and laboratory parameters (AST, monocytes, basophils, and RDW) identified in our study can help clinicians in the prompt diagnosis of the disease and in identifying patients who require immediate intervention and intensive care.

Introduction

Dengue remains one of the most important arthropod-borne viral diseases in humans, representing a significant economic burden with a major social impact on public health in many countries [1, 2]. Dengue is currently endemic in more than 120 countries, with an estimated 4 billion people living in places at risk of dengue transmission [2, 3]. Globally, around 96 million symptomatic dengue cases are estimated to occur per year [2, 4]. The disease is caused by four distinct viral serotypes (DENV1-4), and the majority of the cases result in self-limited illness. However, a small proportion of DENV-infected patients can progress to severe forms of the disease, which can be life-threatening [5].

In the last decade, more than 22.1 million dengue cases were reported in the Americas. Brazil accounted for over half of these cases and registered more than 6,000 deaths throughout this period [6]. Over the last 30 years, Brazil has experienced extensive dengue outbreaks and an increased trend in the number of severe and fatal cases [7, 8]. Variations in dengue morbidity and mortality rates have been observed between Brazilian geographic regions, possibly related to the emergence and re-emergence of different viral serotypes, and differences in population density, climate, and environmental factors [7–15]. The incidence rates of dengue in Brazil have been typically higher among adults. However, a transient shift in the age distribution of dengue cases has been observed in recent years. For instance, in the 2007/2008 epidemic, a significant proportion of the severe dengue cases reported affected primarily children and adolescents [16]. In addition, the proportion of children under 15 years of age who required hospitalization due to severe dengue increased from 9.5% in 1998 to 46.2% in 2007 [16]. In 2022 alone, around 1.4 million dengue cases were reported in Brazil, and 24% of those were in individuals under 19 years of age [17, 18].

The diagnosis of dengue in children and adolescents is challenging. Similar to adults, most of the cases in the pediatric population are asymptomatic or present with undifferentiated fever, which makes the early detection of cases difficult [19–23]. When symptomatic, signs and symptoms of the disease in the pediatric population include fever, vomiting, headache, and abdominal pain, which occur with a higher frequency when compared to adults. Severe dengue is more frequently observed among children under 15 years of age, and the progression to the severe forms of disease usually occurs abruptly in children and adolescents. In addition, severe symptoms like plasma leakage and hemoconcentration are usually more frequently observed in children when compared to adults, while cavitary effusions are more frequent in younger children than in adolescents [21, 22, 24–27]. The greater permeability of the microcirculation in children increases the risk of shock, and this risk is even greater in infants, which leads to a greater number of deaths in this pediatric group.

Despite the differences in signs and symptoms predictive of severe dengue between children and adults, few studies have sought to identify the epidemiological and clinical features predictive of disease progression in the pediatric population [19, 22, 26–29]. Here, we analyze the association between sociodemographic, clinical, and laboratory factors and dengue hospitalization in children and adolescents living in an area of intense arthropod-borne viral exposure. The information generated here can provide support for the implementation of early clinical interventions, which can contribute to better management of pediatric dengue cases. This can lead to a reduction in hospitalizations for dengue in this age group and a consequent reduction in the costs associated with these health interventions, especially in epidemic periods.

Methods

Study design and population

This retrospective, cross-sectional study included children and adolescents (newborn to 19 years old) [30] with confirmed dengue who were admitted to the hospital from 2017 to 2020 at the Hospital Barão de Lucena (HBL). HBL is a publicly funded reference hospital for clinical and surgical pediatric specialties located in the city of Recife, Pernambuco state, Northeast Brazil. With an estimated population of 9.5 million in 2019 and a population density of 89.62 inhabitants per km², Pernambuco ranks as Brazil’s seventh most populous state [31]. Recife, Pernambuco’s capital and largest urban center, is located on the Atlantic coast and has an estimated population of around 1.6 million inhabitants [31]. Previous epidemiological studies have confirmed a high endemicity of arboviral transmission in this setting [32]. Between 2017 and 2022, around 25,910 cases of dengue were reported in the population by the local arboviral surveillance system [17, 18]. Of those, 42% of the cases were reported in children and adolescents under 19 years of age. In the same period, twenty-four deaths in this age group were reported.

Data collection and case definitions

We included all cases with laboratory and/or clinical-epidemiological diagnosis of dengue in children and adolescents admitted to hospitalization at HBL during the study period. Dengue cases included patients with acute febrile illness and two or more of the following symptoms: nausea/vomiting, rash, aches and pains, tourniquet test positive, leukopenia, or any warning sign, as defined by the 2009 World Health Organization (WHO) dengue guideline [33] and the 2024 dengue clinical management guideline from the Brazilian Ministry of Health [34]. Laboratory confirmation of dengue included the detection of viral genomes by Reverse transcription polymerase chain reaction (RT-PCR) and/or the detection of IgM antibodies by enzyme-linked immunosorbent assay (ELISA) [34]. Dengue cases were classified as dengue with warning signs (DwWS) and severe dengue (SD) following the criteria adopted by the 2009 WHO guideline [33]. All dengue cases that were reported to the Brazilian Information System for Notifiable Diseases (SINAN) by the HBL surveillance system were included in the study. Sociodemographic information (age, biological sex, self-reported race, and schooling) was retrieved from the hospital epidemiologic investigation forms using a standardized questionnaire. Clinical data and results of laboratory tests performed during hospitalization were extracted from the hospital medical records using a standardized form. We excluded dengue cases with incomplete or partially complete clinical records and/or epidemiologic investigation forms.

Data analysis

Data processing and statistical analyses were carried out using the R software. Data were summarized, and frequency distributions were displayed in tables and graphs. Descriptive statistics for categorical variables were presented as percentages, and continuous variables as mean ± standard deviation. The temporal distribution of the dengue cases per year in the study population was described and compared with the number of reported cases in the general population of the study site, according to the national surveillance system [17, 18]. The normality and homogeneity of the variables were tested using the Shapiro-Wilk test. For paired analysis, the paired Student’s T test or the Wilcoxon test was used. The frequency distribution of the sociodemographic characteristics, clinical manifestations, and laboratory markers and their association with dengue outcome (DwWS and SD) were analyzed. Univariate and multivariate analyses were performed using Logistic regression. The magnitude of the associations between the independent variables and dengue disease severity was estimated with the Odds Ratio (OR) and the 95% confidence intervals (CI). For the multivariate logistic regression, three groups of variables were used: sociodemographic characteristics, clinical symptoms, and laboratory results. The laboratory variables were categorized according to the reference values obtained from the hospital laboratory, according to age and sex. All variables with a p-value of < 0.25 in the univariate analysis were initially included in the multivariate model. The use of a less stringent threshold for univariable prescreening is meant to avoid the drop of important adjustment variables from the multivariable model due to stochastic variability [35, 36]. Then, variables were ranked from lowest to highest p-value and inserted into the multivariate model. The final model included only the variables that showed a significance of p < 0.05.

Ethical consideration

This work used a secondary database, and thus informed consent from the participants was waived. The protocol for the study was approved by the Research Ethics Committee of the State University of Pernambuco (CAEE: 61283822.9.0000.5207).

Results

During the study period (2017–2020), a total of 645 dengue cases were admitted to the hospital at the HBL. Of those, 46.8% (302/645) of the cases were reported in children and adolescents, and a total of 288 cases were included in this study. Based on the clinical symptoms and laboratory records, 27.8% (80/288) of the cases were classified as SD and 72.2% (208/288) as DwWS. Fourteen cases with incomplete/partially complete clinical and epidemiological records we excluded from the study.

The temporal distribution (2017–2020) of the dengue cases reported in the study population and the number of total dengue cases in children and adolescents reported by the arboviral surveillance system of Pernambuco state are depicted in Fig. 1. Overall, we found a steady increase in the number of hospitalizations by DwWS in children and adolescents throughout 2017–2020 in our study population. At the population level, there was a significant increase in the number of dengue cases in children and adolescents reported by the surveillance system in the study setting in 2019. A small increase in the number of hospitalizations for DwWS was also noted in the same year in our study population. On the contrary, a steady decline in the number of hospitalizations by SD in children and adolescents was observed during the study period, including in the year of 2019 (Fig. 1).

Fig. 1.

Temporal distribution of the dengue with warning signs (DwWS; black solid line) and severe dengue (SD, dashed line) cases among children and adolescents hospitalized with dengue in a reference hospital and the number of reported dengue cases in children and adolescents reported by the official surveillance system (gray bars) from Recife, Pernambuco, during the study period (2017–2019)

Table 1 shows the crude and adjusted odds ratios (OR) of the association between dengue clinical outcome and sociodemographic characteristics in the study population. Interestingly, the frequency of dengue cases was higher in the older age groups, regardless of dengue disease severity. However, SD cases were more frequently observed in the age group of 7–12 years old when compared to DwWS cases (58.8% versus 38.0%, OR = 3.72, 95%CI = 1.34–13.21; Table 1). We found no significant differences in dengue disease severity between biological sex (p = 0.334) and area of residence (p = 0.826). There was a significant increase in the frequency of SD in children and adolescents with self-identified mixed (73.7% versus 62.9%, OR = 2.28, 95%CI = 1.22–4.52) and black (8.8% versus 2.9%, OR = 5.92, 95%CI = 1.73–21.1) race when compared to those self-identified as white. The associations between age group and self-identified race with dengue clinical outcomes remained after adjusting for the effects of these variables in the multivariate analysis.

Table 1.

Crude and adjusted odds ratio (OR) of the association between dengue clinical outcome and sociodemographic characteristics in children and adolescents hospitalized with dengue in a reference hospital, Northeast Brazil, 2017–2020

Characteristics	Dengue Outcome, n (%)		OR (95% CI)	P value	Adjusted OR (95% CI)	P value
	DwWS	SD
Age group, years
≤ 2 years	25 (12.0)	4 (5.0)	1.00
3–6 years	30 (14.4)	5 (6.2)	1.04 (0.25–4.60)	0.955	0.80 (0.18–3.65)	0.765
7–12 years	79 (38.0)	47 (58.8)	3.72 (1.34–13.21)	0.021	3.38 (1.23–12.58)	0.003
13–19 years	74 (35.6)	24 (30.0)	2.03 (0.70–7.39)	0.229	2.11 (0.71–7.83)	0.211
Biological sex
Female	96 (46.2)	42 (52.5)	1.00
Male	112 (53.8)	38 (47.5)	0.78 (0.46–1.30)	0.334
Self-identified race
White	71 (34.2)	14 (17.5)	1.00
Mixed	131 (62.9)	59 (73.7)	2.28 (1.22–4.52)	0.012	2.24 (1.18–4.47)	0.017
Black	6 (2.9)	7 (8.8)	5.92 (1.73–21.1)	0.004	7.35 (1.98–28.99)	0.003
Area of residence
Rural	9 (4.3)	3 (3.7)	1.00
Urban	199 (95.7)	77 (96.3)	1.16 (0.34–5.33)	0.826

Bold numerical values in the table represent statistical significance

DwWS Dengue with warning signs, SD Severe dengue, OR Odds ratio, CI Confidence interval

Following, we investigated the association between clinical signs and symptoms predictive of dengue disease severity in the study population. Clinical signs and symptoms of dengue, such as edema, irritability, fluid accumulation, persistent vomiting, a positive tourniquet test result, presence of rash, retro-orbital pain, abdominal pain, and lethargy, were risk factors associated with SD in children and adolescents in the crude analysis (Table 2). However, when adjusting the effect of these variables in the multivariate analysis, the clinical symptoms that remained associated with disease severity were irritability, fluid accumulation, and lethargy. The chance of developing SD among children and adolescents who reported fluid accumulation and irritability was five (OR = 5.50, 95%CI = 1.51–23.63, p-value = 0.013) and four (OR = 3.69, 95%CI = 1.36–11.19, p-value = 0.014) times greater than those who did not report these clinical manifestations, respectively. In addition, patients who reported lethargy/restlessness were also at higher risk of presenting with SD when compared to patients with DwWS (OR = 19.95, 95%CI = 6.47–73.23, p-value = < 0.0001). However, the large confidence interval indicates the need for increasing the number of the population studied (Table 2).

Table 2.

Crude and adjusted odds ratio (OR) of the association between dengue clinical outcome and clinical symptoms in children and adolescents hospitalized with dengue in a reference hospital, Northeast Brazil, 2017–2020

Clinical signs and symptoms of the disease	Dengue Outcome, n (%)		OR (95% CI)	P value	Adjusted OR (95% CI)	P value
	DwWS	SD
Thrombocytopenia	180 (90.9)	69 (92.0)	1.15 (0.46–3.28)	0.776	2.01 (0.44–11.96)	0.397
Edema	38 (27.7)	43 (68.2)	5.60 (2.97–10.9)	< 0.0001
Irritability	11 (5.45)	19.0 (26.7)	6.34 (2.89–14.59)	< 0.0001	5.50 (1.51–23.63)	0.013
Fluid accumulation	44 (40.0)	42 (70.0)	3.50 (1.81–6.97)	0.0003	3.69 (1.36–11.19)	0.014
Headache	114 (96.3)	42 (94.0)	0.74 (0.14–5.45)	0.729
Vomiting	159 (83.7)	69 (86.2)	1.22 (0.60–2.67)	0.595
Persistent vomiting	20 (11.3)	17 (25.0)	2.62 (1.26–5.38)	0.008
Positive tourniquet test	59 (57.8)	52 (82.5)	3.45 (1.66–7.66)	0.0014
Rash	28 (45.16)	27 (71.05)	2.98 (1.28–7.27)	0.0129
Nausea	47 (50.0)	9 (56.2)	1.29 (0.44–3.87)	0.644
Petechiae	26 (83.8)	65 (81.25)	0.83 (0.25–2.40)	0.747
Retro-orbital pain	18 (10.5)	7 (77.8)	29.75 (6.62–210.3)	0.0001
Abdominal pain	18 (10.5)	65 (89.0)	69.06 (30.1-178.2)	< 0.0001
Lethargy/restlessness	11 (6.9)	33 (47.1)	11.92 (5.67–26.81)	< 0.0001	19.65 (6.47–73.23)	< 0.0001
Mucosal bleeding	85 (53.8)	39 (58.2)	1.2 (0.67–2.14)	0.5431
Liver enlargement	22 (28.6)	24 (30.0)	0.93 (0.47–1.86)	0.8441
Prior admission to care*	125 (60.1)	57 (71.2)	1.65 (0.95–2.92)	0.0802
Comorbidities**	21 (10.1)	12 (15.0)	1.56 (0.71–3.30)	0.2505

Bold numerical values in the table represent statistical significance

DwWS Dengue with warning signs, SD Severe dengue, OR Odds ratio, CI Confidence interval

^*Prior admission to care includes patients who seek medical assistance in other health units before being transferred to HBL for hospitalization

^**Co-morbidities included diabetes, obesity, bronchial asthma, rhinitis, hematological disease, malnutrition, diabetes, heart disease, and hypertension

We next analyzed the association between laboratory findings and dengue disease severity in the study population. Binary logistic regression was used to assess the association between continuous independent laboratory parameters and dichotomous outcomes (DwWS and SD). Differences in the levels of liver enzymes (ALT: alanine aminotransferase, and AST: aspartate aminotransferase), hyponatremia (low sodium levels), low neutrophil count, high monocyte and basophil count, high red blood cell (RBC) and hemoglobin levels, and high red cell distribution width (RDW) percentage were among the laboratory parameters associated with SD in the crude analysis. In the multivariate analysis, an increase in the levels of AST (OR = 1.25, 95%CI = 1.10–1.50, p-value = 0.003), high monocyte (OR = 1.00, 95%CI = 1.00-1.01, p-value < 0.0001) and basophil (OR = 1.19, 95%CI = 1.10–1.33, p-value = 0.0003) counts, and an increase in the RDW percentage (OR = 3.66, 95%CI = 1.60-10.38, p-value = 0.005) at admission remained as statistically significant laboratory parameters associated with SD in the study population. In addition, children and adolescents had a 26% lower chance of having SD with every increase (million/mm³) in RBC counts (OR = 0.84, 95%CI = 0.69–0.95, p-value = 0.025; Table 3).

Table 3.

Crude and adjusted odds ratio (OR) of the association between dengue clinical outcome and hematological and biochemical parameters in children and adolescents hospitalized with dengue in a reference hospital, Northeast Brazil, 2017–2020

Laboratorial Parameters	Dengue Outcome					Univariate analysis		Multivariate analysis
	DwWS		SD		P value
	N	Median (Q3-Q1)	N	Median (Q3-Q1)		Crude OR (95%CI)	P value	Adjusted OR (95%CI)	P value
ALT (U/L)	208	85.0 (158.5–44.0)	80	80.5 (104.0–44.0)	0.032	0.99 (0.99-1.00)	0.001
AST (U/L)	208	48.0 (55.0–44.0)	80	55.0 (59.2–50.0)	< 0.0001	1.10 (1.06–1.15)	< 0.0001	1.25 (1.10–1.50)	0.003
Albumin (g/dL)	89	4.2 (4.8–3.8)	57	4.1 (5.0-3.9)	0.527	1.36 (0.90–2.15)	0.161
Sodium level (mEq/L)	208	137.0 (141.0-109.7)	80	128.5 (137.0-102.0)	0.005	1.00 (0.99–1.01)	0.638
Platelets (cells/mm3)	188	111,000 (125000 − 87000)	76	112,000 (121000 − 96500)	0.660	1.00 (1.00–1.00)	0.353
Leukocytes (cells/mm3)	188	3870 (8900 − 2215)	76	3525 (8790 − 2545)	0.766	1.00 (1.00–1.00)	0.662
Neutrophil (cells/mm3)	188	4506 (4843 − 4209)	75	3120 (4641 − 2398)	< 0.0001	1.00 (1.00–1.00)	0.157
Lymphocyte (cells/mm3)	188	2226 (3120 − 1502)	75	1851 (2838 − 1511)	0.641	1.00 (1.00–1.00)	0.668
Monocyte (cells/mm3)	188	967 (1080 − 498)	74	3099 (3623 − 2902)	< 0.0001	1.00 (1.00–1.00)	< 0.0001	1.00 (1.00-1.01)	< 0.0001
Eosinophil (cells/mm3)	208	46.5 (80.0-24.7)	80	39.5 (67.2–33.0)	0.932	1.01 (1.00-1.01)	0.001
Basophil (cells/mm3)	188	0.0 (2.2-0.0)	75	26.0 (47.0–0.0)	< 0.0001	1.19 (1.14–1.25)	< 0.0001	1.19 (1.10–1.33)	0.0003
HCT (%)	208	50.0 (56.0-43.2)	80	49.1 (52.3–44.0)	0.255	0.97 (0.93-1.00)	0.069
RBC (million/mm3)	188	4.5 (4.9–4.3)	75	4.8 (5.1–4.3)	0.007	2.22 (1.41–3.56)	0.0007	0.84 (0.69–0.95)	0.025
Hemoglobin (g/dL)	188	15.4 (17.9–14.5)	75	17.6 (18.2–15.1)	0.049	1.19 (1.03–1.37)	0.022
MCV (fL)	188	83.9 (88.8–82.7)	75	88.1 (92.3–83.2)	0.064	1.06 (1.00-1.12)	0.069
MCH (pg)	188	31.2 (33.1–27.9)	75	29.7 (32.4–27.8)	0.197	0.96 (0.90-1.00)	0.183
MCHC (g/dL)	208	34.1 (35.6–32.5)	80	34.1 (35.1–32.6)	0.756	0.97 (0.89–1.05)	0.459
RDW (%)	188	12.2 (13.0-11.6)	75	13.0 (13.9–12.1)	< 0.0001	2.12 (1.59–2.89)	< 0.0001	3.66 (1.60-10.38)	0.005

Bold numerical values in the table represent statistical significance. Units: U/L (units per liter), g/dL (grams per deciliter), mEq/L (milliequivalents per liter), cells/mm3 (cells per cubic millimeter), % (percentage), fL (femtoliters), pg (picograms)

DwWS Dengue with warning signs, SD Severe dengue, OR Odds ratio, CI Confidence interval, Q3 third quartile, Q1 first quartile, ALT Alanine aminotransferase, AST Aspartate aminotransferase, HCT Hematocrit, RBC Red blood cell, MCV Mean cell volume, MCH Mean cell hemoglobin, MCHC Mean cell hemoglobin concentration, RDW Red cell distribution width

Finally, we used multivariate logistic regression to assess the overall association of sociodemographic, clinical, and laboratory parameters with SD in the study population. When all risk factors were analyzed combined, we found that self-identified mixed race, clinical signs of lethargy and irritability, and the laboratory parameters of monocyte counts and sodium levels were predictive of SD in children and adolescents at admission for hospitalization (Fig. 2). Individuals who reported having mixed race had six times higher risk of SD when compared to those who reported having self-reported white race (OR = 6.51, 95%CI = 2.05–25.05, p-value = 0.003). The chance of being admitted to hospitalization with SD was fifteen and thirty-eight times greater in children and adolescents showing clinical signs and symptoms of irritability (OR = 15.05, 95%CI = 4.28–61.86, p-value = 0.003) and lethargy (OR = 38.81, 95%CI = 11.96-159.76, p-value < 0.0001), respectively. Low sodium levels (hyponatremia) (OR = 6.94, 95%CI = 2.72–19.65, p-value = 0.0001) and increased monocyte counts (OR = 8.64, 95%CI = 2.73–33.63, p-value = 0.0007) were also associated with a greater risk of SD in the multivariate model. Children and adolescents with low platelet counts were less likely to be admitted with SD when compared to DwWS (OR = 0.14, 95%CI = 0.03–0.69, p-value = 0.0175; Fig. 2).

Fig. 2.

Forest plot of the multivariate logistic regression with the sociodemographic, clinical, and laboratory parameters predictive of severe dengue in the study population

Discussion

The identified trends of dengue-reported cases in the study area are in line with the patterns of dengue incidence in Brazil during the study period. The introduction and circulation of Zika virus in the country (2015/16) was followed by a significant decline in the number of annually reported dengue cases in 2017 and 2018 [14, 15]. This decline in the number of dengue cases was not unique to Brazil, as other regions of the Americas and the Caribbean also experienced a similar downward trend. Dengue incidence resurged in Brazil and elsewhere in 2019, which is reflected in the number of reported cases in Pernambuco state for that year [14]. Despite these variations in the yearly reported cases registered in the study area, the number of children and adolescents hospitalized with DwWS remained relatively steady throughout the study period, while the number of SD cases dropped during the same time frame. Increased public health awareness following the co-circulation of multiple arboviruses (Zika and Chikungunya) in the area likely led to improved clinical management and a consequent reduction in the likelihood of developing severe cases may have contributed to the observed patterns.

Overall, there is a lack of clearly defined individual and sociodemographic factors predictive of SD in the pediatric population. In our study, children aged between 7 and 12 years old were more likely to be admitted to hospitalization with SD when compared to the younger age group (≤ 2 years old). Similar findings have been reported in other epidemiological studies carried out in the pediatric population in which children aged >5 years old were found to be at a significantly higher risk for developing severe forms of the disease [21, 28]. Previous reports suggest that dengue is a disease of school-aged children and that the increase in outdoor activities in this age group compared to younger groups might contribute to greater exposure to mosquito bites [21, 22, 37]. In addition, older age likely increases the probability of secondary dengue infections, a known risk factor for SD, particularly in endemic areas of DENV transmission [38]. However, other studies have also shown the predisposition of younger children, especially those aged < 1 year old, to severe forms of dengue. This likely reflects the dual roles of maternally-transferred dengue antibodies in infants: protective in the early months, but detrimental when antibodies wane to non-protective levels [39–43].

In addition to older age, self-reported mixed and black race was a sociodemographic predictor of SD in our study population. Genetic ancestry has been associated with differences in the incidence and severity of dengue in several epidemiological studies, particularly in Caribbean countries [44–46]. However, it is important to note that Brazil is a multiracial country [47] and that information regarding race in our study was self-reported. Thus, inferences regarding biological differences related to genetic ancestry and predisposition to severe dengue disease are beyond the scope of our work. The higher risk of hospitalization by SD among mixed/black race individuals observed in our study likely reflects the unfavorable socio-economic conditions and racial inequities in health that are significantly higher in those who identify their race as mixed and black compared to those who identify as white [47, 48].

Our multivariate analysis identified irritability, fluid accumulation, and lethargy/restlessness as the clinical signs and symptoms predictive of SD in our patients. These results align with previous studies in the pediatric population carried out in Brazil and other dengue-endemic countries [19, 21, 22, 26, 28, 49, 50]. Lethargy/restlessness and irritability were independent clinical signs associated with SD, confirming the observation of other studies in children analyzing clinical and laboratory factors associated with dengue severity [19, 26, 49, 50]. Fluid accumulation is a known consequence of plasma leakage that can occur in the critical phase of the disease. Age-related differences in baseline microvascular permeability between children and adults have been described [51]. The greater microvascular permeability coupled with less developed hemodynamic compensatory mechanisms in children and adolescents likely increases the risk of SD in the pediatric population [51]. Other independent clinical signs and symptoms associated with SD in our study population included edema, persistent vomiting, rash, retro-orbital and abdominal pain, and a positive tourniquet test. Those are all well-known warning signs of severe dengue and can also be considered valuable clinical signs for monitoring children and adolescents with dengue [19].

Our multivariate analysis of the routine hematological and biochemical parameters indicated that AST, monocyte and basophil counts, RBC, and RDW are among the warning factors predictive of SD in children and adolescents. The median values of AST and ALT have been observed to be higher in both children and adults with SD, suggesting a correlation between increased transaminase levels and disease severity. However, in our study population, only increased levels of AST remained associated with SD after adjusting for the effects of other variables. Different from ALT, AST is not primarily hepatic and can be derived from other sources such as erythrocytes, heart, and striated muscle. Therefore, the increased median levels of AST found in our study population might not solely reflect hepatic damage [52]. A transient monocytosis has been previously correlated with an increased risk of SD in children [27]. Monocytes represent one of the main cellular targets for DENV infection, but these phagocytic immune cells also play a critical role in restricting DENV dissemination and contributing to viral clearance [27].

In addition to altered liver enzymes, other hematological and biochemical parameters predictive of SD included RBC counts, RDW, and sodium levels. RBC count conferred a significant protective effect for SD in our study population. This likely leads to better awareness of this early warning sign and, consequently, more appropriate management and assistance. High RDW has been associated with bleeding in DENV-infected patients. High RDW can also be linked to anemia, which can aggravate dengue infections, particularly in children. Interestingly, hyponatremia was associated with SD in the crude analysis and in the final model, which included socio-demographic, clinical, and laboratory factors. Hyponatremia has been reported as a common electrolyte disturbance in dengue-associated complications in both children and adults [25, 26, 28, 53]. The reasons for hyponatremia in dengue patients include dysfunction of the sodium-potassium pump, loss of sodium in the urine, salt depletion, and excess water due to increased metabolism. However, severe hyponatremia (less than 120 mEq/L) in dengue patients is rarely described [54].

The use of secondary data from the hospital’s epidemiologic investigation forms and notification system has some inherent limitations. Although we included a comprehensive dataset of sociodemographic, clinical, and laboratory factors, other important variables were not included in our analysis due to the lack of or inconsistency of completion in the notification forms. This includes information on the days of illness, the type of medical care attention before hospitalization, and the duration of the hospitalization. Information on the type of infection (primary or secondary infection) was not available in the epidemiologic forms. However, given the high dengue endemicity at early ages in our study area [11, 32, 42, 55], we expect most of the infections to be secondary. Whether the previous dengue immunity profile impacted disease outcome, as in previous studies [38], would need to be confirmed, ideally in prospective cohort studies. Despite these limitations, our study was able to identify important risk factors associated with DwWS and SD in children and adolescents that can be useful in the clinical management of these patients.

Conclusion

In summary, we described several socio-demographic, clinical, and laboratory parameters predictive of SD in children and adolescents. Irritability, fluid accumulation, and lethargy were among the clinical signs and symptoms particularly helpful in identifying the cases that progressed to SD. Although these signs are considered nonspecific, lethargy/restlessness and irritability have been consistently associated with increased risk for severe dengue, especially in pediatric patients [56, 57]. Clinical awareness and assessment of these signs are crucial to monitor patients, especially in infants and younger children. These signs can be useful to guide clinical decision-making in the critical phase of the disease, particularly when more specific warning signs (e.g., abdominal bleeding, liver enlargement, persistent vomiting, and others) are absent. In addition, laboratory markers such as AST, monocyte count, RBC, and RDW were also present in patients progressing to SD. The factors identified here can be used as early recognition features of DwWS and SD at the time of admission to the hospital. Monitoring of these clinical and laboratory parameters by health professionals is essential to support appropriate clinical management and potentially reduce morbidity and mortality in the pediatric population.

Abbreviations

AST

aspartate aminotransferase

CI

confidence interval

DENV

dengue virus

DwWS

dengue with warning signs

ELISA

enzyme-linked immunosorbent assay

HBL

Hospital Barão de Lucena

IgM

Immunoglobulin M

OR

Odds ratio

RDW

red cell distribution width

RT-PCR

Reverse transcription polymerase chain reaction

SD

severe dengue

SINAN

Brazilian Information System for Notifiable Diseases

WHO

World Health Organization

Authors’ contributions

M.P.S. collected and organized data, and assisted in study design, interpretation of the results, and manuscript writing. A.K.L.B. assisted in data collection and organized the data. G.T.N.D. performed the statistical analyses and assisted in the writing of the manuscript and the interpretation of the results. P.M.S.C. designed the study, interpreted the results, and prepared the final version of the manuscript. All authors have read and approved the manuscript.

Funding

No funding was received for this study.

Data availability

All data generated or analyzed during this study are included in this published article.

Ethics approval and consent to participate

This work used a secondary database, and thus informed consent from the participants was waived. The protocol for the study was approved by the Research Ethics Committee of the State University of Pernambuco (CAEE: 61283822.9.0000.5207).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.