Survival and Mortality in Hospitalized Children with COVID-19: A Referral Center Experience in Yazd, Iran

Introduction COVID-19 prognostic risk factors, therapeutic protocols, and clinical outcomes in pediatric cases are still under investigation. Materials and Methods This historical cohort study evaluated the survival time of hospitalized children (1 month–18 years old) with COVID-19 admitted from March 2020 to August 2021 to an educational hospital in Yazd, Iran. The follow-up of patients was performed at least one month after discharge. Results From 183 hospitalized cases, 24 children were deceased. The median age of patients was 5.41, and 54.2% were male. The survival rate after one-month follow-up was 0.88, and the most significant predictors associated with survival time were the male sex, positive history of hospitalization, lymphopenia, hypoxia, and length of stay more than two weeks using Bayesian Cox regression analysis. Conclusion Accurate estimation of the impact of predictors on poor outcomes may help healthcare providers use therapeutic protocols based on risk factors and healthcare requirements of each patient to improve their survival.


Introduction
Coronavirus disease (COVID-19) caused by severe acute respiratory syndrome coronavirus 2 (SARS-Cov-2) started in Wuhan, China, in December 2019 and was stated as a pandemic and public health emergency by the World Health Organization. Iran reported the frst death due to COVID-19 among Middle East countries, and a mortality rate of 2% (7, 588, 477 patients and 145, 269 deaths of COVID -19) has been announced in Iran until March 2023 [1]. Although the impact of epidemiological characteristics, laboratory fndings, clinical symptoms, and other risk factors in adults was studied through various types of research [2][3][4][5][6], chief complaint, treatment protocols, and prognosis risk factors are not clearly defned in children infected with COVID-19. Moreover, limited studies have explored pediatric cases' survival and long-term clinical symptoms after discharge in this age group. Some studies revealed that children with COVID-19 are mild or asymptomatic [7,8]; however, there is a lack of information about pediatric severe pulmonary manifestations in new virus variants [9,10]. In addition, a better understanding of signifcant chronic complications after COVID-19 infection [11,12] and the impact of multisystem infammatory syndrome (MIS-C) on the outcome of infection in children may prevent the increasing mortality rate and the growing treatment costs in the future. Based on available evidence, a few studies estimated survival probabilities regarding risk factors, laboratory fndings, and treatments of children with COVID-19 that would be highlighted in this study using diferent statistical approaches.

Study Design and Data Collection.
Te present paper is written based on the STROBE statement for reporting historical cohort studies [13]. In this research, children aged 1 month-18 years old were admitted consecutively to Shahid Sadoughi COVID-19 referral hospital, including those directly referred to the emergency ward or referred from other hospitals in Yazd province. Tese patients were diagnosed based on laboratory-confrmed SARS-CoV-2-positive swabs (RT-PCR), clinically confrmed by the clinicians according to defned symptoms or confrmed through CT-scan radiographic fndings. Te hospitalized patients were enrolled from March 1, 2020, to August 1, 2021, and demographics, underlying diseases, laboratory fndings, and treatment protocols were gathered through the hospital information system. Outcomes of patients were recorded through follow-ups for at least one month. Te study was approved by the Research Ethics Committee of the Shahid Sadoughi University of Medical Sciences and Health Services, Yazd (IR.S-SU.REC.1401.049). Data anonymization was performed before analysis, and the technique of data masking was used for data anonymization in this study. At frst step, all the information of each patient was recorded in an excel sheet, and at the second step, the identifying number (defned specifc codes for this study) was given to each patient. At the last, step data swapping was performed; patients' privacy and data security were prioritized at all levels, and the study was accomplished under the Helsinki Declaration (2013) guidelines.

Patient's Characteristics, Treatments, and Outcomes.
Demographic data (age and sex), underlying diseases, preadmission, on-admission clinical symptoms, length of hospital stay (day), PICU admission, vital signs on admission (respiratory rate, systolic and diastolic blood pressure, and oxygen saturation in an air room (O 2 sat)), patients who sufered from multisystem infammatory syndrome of children (MIS-C) which was diagnosed based on the WHO criteria, laboratory fndings (the frst valid value of laboratory parameters was considered), and chest CT-scan fndings (pattern and severity score) of 183 patients were collected from medical records. Te severity of the disease is defned based on WHO recommendations defnition in three categories: nonsevere (absence of signs of severe or critical diseases), severe (O 2 sat lower than 94% on room air, tachypnea, and signs of severe respiratory distress), and critical (requires life-sustaining treatment, acute respiratory distress syndrome, sepsis, and septic shock) [14,15]. Based on the WHO defnition and observed symptoms, the severity of each patient's condition was recorded in the electronic profle. Te primary outcome was determined as the death from COVID-19 versus survival after at least one-month follow-up, and the patients who were alive at the last followup time were defned as censored.

Statistical Analysis.
Continuous and categorical variables were described using the mean ± SD and frequency (percentage). In order to compare the diference of explanatory variables in survived and deceased groups, the Pearson chi-square and Fisher exact tests were used. Cramer's V and Eta were considered as the measure of association between independent variables and outcome. Te Kaplan-Meier method and the log-rank test were performed to estimate and compare the survival functions in diferent categories of each explanatory variable. Using a stepwise process, the Cox proportional hazards model was ftted, including signifcant independent variables in univariate analysis. In the multivariate approach, Bayesian survival analysis was conducted to estimate signifcant independent variables to overcome a few outcomes. Te survival analysis was conducted using RStudio (RStudio Team (2020), RStudio: integrated development for R. RStudio, PBC, Boston, MA URL https://www.rstudio. com/) and packages survival, KMsurv, spBayesSurv, and coda. Te signifcance level of 0.05 was regarded as statistically signifcant.

Characteristics, Clinical Findings, and Outcome of Children with COVID-19.
A total of 183 children aged 1mo-18 y were included in this study; 54.2% and 45.8% were males and females, respectively, with a median age of 5.41 (mean-� 6.95 and standard deviation � 5.59). Tirteen percent (n � 24) of children with COVID-19 died that most aged 1-5; however, the association between the age groups and death was insignifcant. Te previous hospitalization, PICU admission, abnormal long CT fndings, O 2 sat, patient intubation, ventilator support, and length of hospital stay were signifcant among all variables. Te strength of the relationship between death and variables, including ventilatory support (Cramer's V � 0.52), PICU admission (Cramer's V � 0.49), and patient intubation (Cramer's V � 0.48), was the highest ( Table 1). As shown in Table 1, the main preadmission symptoms were fever, cough, dyspnea, and vomiting. On-admission time, fever, and cough were observed among sixty-nine and thirty-four patients, and cough was signifcantly higher in deceased children. Neurologic diseases were the frequent underlying disease, and MIS-C was observed in 24.6 percent of children. Moreover, 12.6 percent of patients were critical, and as expected, most deceased children were in the severe and critical groups. Table 2 compares the range of laboratory fndings between deceased and surviving patients. According to the age of patients, increased AST, CPK, and lymphopenia showed marginal signifcance.

Drugs Used for Patients.
In this study, 28 children received methylprednisolone (21 in survived and 7 in deceased groups), and the diferences between the two groups were signifcantly meaningful (P � 0.04). Fifty-three children received remdesivir, and the death rate was 33%, showing no signifcant diference between the two groups. In addition, only 10.5 percent of children received favipiravir, and all survived. No signifcant diference in the mortality rate was observed among other drugs used for patients (Table 3). . Te survival rate of hospitalized children and their risk factors were evaluated by Kaplan-Meier estimates (Figure 1(a)). Te overall         (Figure 1(b)). Based on the log-rank test, the variables signifcantly increased the death rate were length of hospital stay, the severity of disease, positive history of previous hospitalization, low O 2 sat on admission, PICU admission, patient intubation, and ventilatory support, respectively ( Figure 2). According to the age of the patients, lymphopenia had a slight impact on the survival rate. Te hazard ratio of each signifcant variable was      than one week could increase the risk of death by more than threefold. As shown in Table 4, based on the backward method of Cox multivariate analysis, none of the variables showed a considerable impact on the risk of death. Te multivariate Cox regression failed to estimate the actual efect of covariates because of a few samples in some categories of predictive variables.

Te Bayesian Hazard Estimation of Variables.
Bayesian survival analysis was regarded as a practical statistical approach for predicting survival rates in such data where the occurrence of outcome was not high. As mentioned in the result section, only 13% of hospitalized children with COVID-19 died, and the patients' overall survival rate on the 10 th day of follow-up was higher than 90%. In addition, some diagnostic procedures, such as CT scans or some treatments, including intubation and ventilatory support, were not used for many children. As a result, the actual estimate of covariates in the survival rate prediction could not be calculated. In the last step of the analysis, the Bayesian Cox regression was conducted based on normal priors for the coefcient's parameters. Trace diagrams were considered the measures of model assessment. All trace diagrams were converged, and the coefcients of variables are presented in Table 5. As shown, the risk of death was 50% higher for boys; however, the survival rate of age groups was not signifcantly diferent. Te risk of death for children with a positive history of hospitalization diseases and underlying diseases would increase to 2.77 and 1.75, respectively. In children with O 2 sat lower than 94%, death occurred 2.3 times more than in patients with normal O 2 sat, and lymphopenia patients had a higher risk of death (HR � 1.27%, 95%CI: 1.18-1.36).
In addition, the hospitalization duration of more than two weeks would increase the hazard of death to 89% for children.

Discussion
Since the global research on mortality and survival in children with COVID-19 was insufcient or the data collected were limited, this study was undertaken to record and report a large subset of relevant variables. Moreover, in this research, each patient was followed at least one month after discharge, which was not investigated in other studies. According to the data, 183 patients aged between 1 month and 18 years old were admitted to Shahid Sadoughi Hospital in Yazd between March 2020 and August 2021. Te fndings of this study included detailed demographics, clinical characteristics, paraclinical data, medications, and their association with survival probabilities. Te mortality rate of hospitalized children with COVID-19 was reported as 5.3% in Iran and tripled in children with underlying diseases [16]. In this study, 13% of patients died, two times higher than the overall mortality rate in similar studies in Iran. Te mortality rate of 18 years old or younger children with COVID-19 in Europe up to April 2020 was reported as 0.69% and 2.2% in two other studies [8,17,18].
Similarly, Oliveira et al. found that the mortality rate in children younger than 20 was 7.8% (15). Te death rate could vary in diferent studies based on substantial diferences among countries in epidemiological features, healthcare system, access to standard treatments, and length of followup. From studies in China and a meta-analysis in Taiwan, the majority of children with COVID-19 (more than 90%) were reported in mild to moderate status, signifcantly higher than the observed rate of this research [7,19]. Shahid Sadoughi Hospital was a referral center for severe patients of COVID-19, and half of the admitted children with COVID-19 were in severe or critical status, which caused a high death rate in this group of patients. Moreover, at the beginning of the COVID-19 epidemic period, physicians' knowledge about the standard treatments for this infection was limited, and only 30% and 15% of patients were treated by Rem and Meth, respectively.  [19][20][21][22]. Compared with other age groups, children aged between 5 and 13 years old were at a lower risk of death, consistent with the Armin et al. study [23]. No association was found between gender and the death rate in this study. However, slightly more deaths occurred in boys, consistent with other studies [8,24]. In this research, male children were at a higher risk of severe disease, as in the review of Gallo Marin et al. [25], but the male gender did not seem to be an independent risk factor for severe COVID-19. 27% of children had a positive history of hospitalization, which caused a considerable increase in the death rate. In Iran, the frequency of having at least one underlying disease in patients younger than 18 years old was reported from 8% to 50%, with a death rate of 15% [16,23]. Te data showed a death rate of 45.5% for children with a underlying disease, which was in line with the results of the United States CDC report (42.3%) but signifcantly higher than the fndings of a Brazilian study (28%) and a multicenter cohort study of European pediatric patients (25%) [8,18,26]. In some studies, respiratory and neurological diseases were reported as the most common comorbidities in deceased cases, but this study showed malignancies of neurological diseases as the common underlying diseases among children who died due to COVID-19 [27,28].
Furthermore, previous studies represented a signifcantly higher mortality rate in hospitalized children with comorbidities, which was not found in this study. In this study, 24% of the patients were complicated by MIS-C, which was higher than in similar studies [29]. As mentioned, half of the admitted patients were severe, which could cause critical outcomes.

Symptoms.
Most studies reported fever and cough as the frequent symptoms among hospitalized children with COVID-19. Dyspnea and chest pain were declared as the third and fourth most frequent symptoms in children with severe disease who were admitted to Children's Hospital in New York City, but no signifcant diference was observed in the frequency of these symptoms in comparison with patients without a severe disease [30]. Similar to a study in Vietnam, dyspnea on-admission time had a statistically  signifcant association with the death rate [31]. In this study, vomiting and diarrhea were frequent among hospitalized children on preadmission or on admission to the hospital. However, any of these symptoms could not be considered a risk factor for death in patients aged <18 years old [32,33]. O 2 sat lower than 95% was the presenting symptom in 41% of hospitalized Brazilian children, but its association with the mortality rate was not investigated [18]. Based on the fndings of this research and other studies, O 2 sat lower than 94% was common among infected children, and there was a negative association between death frequency and oxygen saturation percent [8,18,34].

Laboratory Findings and Drugs Used.
Te result showed that the mortality rate in patients with high AST (U/L) and CPK (U/L) and children with lymphopenia would increase considerably. According to one systematic review published in 2020, hypoalbuminemia, lymphopenia, leukocytosis, elevated interleukin level 6, and prolonged PT time were efective laboratory features associated with the mortality rate in children with COVID-19 [35]. Like this study, a cohort study in Korea presented that lymphopenia and its severity could be considered prognostic variables for clinical outcomes, such as mortality, need for intensive care, and need for intubation and ventilatory support [36].

Survival and Mortality.
In this research, the signifcant risk factors of death related to COVID-19 were a positive history of hospitalization, hypoxia, lymphopenia, elevated serum levels of AST and CPK, need for intubation and ventilatory support, and length of hospital stay of more than one week. Compared with studies of Oliveira et al., Gӧtzinger et al., and Swann et al., there was a weak relationship between underlying diseases and mortality risk [8,18,22]. Te actual efect of each risk factor was not evaluated, considering the survival of hospitalized pediatric patients. Te reason for not reporting the Cox regression analysis result could be a few samples in diferent categories of predictive variables, which were adjusted using the Bayesian Cox approach in this investigation. Considering the Bayesian Cox regression, patients aged between 1 and 13 years old compared to patients older than 13 years and children without any underlying diseases had a signifcantly lower risk of death, consistent with previous reports [8,18,22]. In addition, lymphopenia showed a signifcant efect on the survival of pediatric patients. Gallo Marin et al. revealed that O 2 sat <90.5% seemed to be a reliable predictor for patient survival which was not found in this investigation [25].
Te most feature of this study was the estimation of the survival rate of COVID-19 inpatient children and its association with predictive factors through two diferent statistical approaches. Also, this study had two other strengths. First, all of the important variables related to the mortality of COVID-19 patients were recorded, and the survival rate based on the diferent categories of these variables was evaluated. Second, the children were followed precisely at least one month after discharge. However, some important limitations of the current study were the small population of patients in a single center. In addition, the data were collected from the electronic records of the patients, which had some missing information that afected the outcomes of the study.

Conclusions
Since SARS-CoV-2 is a novel virus and the number of infected children is not limited, this study provided a proper statistical method to evaluate the association of demographical and clinical profles and laboratory fndings with the survival of hospitalized children. Te fndings suggested that specifc subgroups of children, including boys and those with a positive history of hospitalization and underlying diseases, are at higher mortality risk. Te presence of lymphopenia and hypoxia also signifcantly increases the mortality risk in children with COVID-19. In addition, hospitalization for more than two weeks increases the risk of developing complications from COVID-19 and even death. Tese results underscore the need for improved therapeutic protocols to control hospital mortality rates. It is recommended that physician pay more attention to clinical and paraclinical risk factors which could increase the severity and mortality in pediatric patients.

Data Availability
Te registered data used to provide the fndings of this study are restricted by the Ethics Committee of Shahid Sadoughi University of Medical Sciences and Health Services in order to protect patient privacy. Data are available from Mehran Karimi at drmehrankarimi@yahoo.com for researchers who meet the criteria for access to confdential data.