Epidemiology and Mortality Predictors for Severe Childhood Community-acquired Pneumonia in ICUs: A Case-control Study

Background: To identify the epidemiology and mortality predictors for severe childhood community-acquired pneumonia (CAP) and evaluate the inuence of medications on clinical outcome in the real world. Methods: We performed a retrospective case-control study among children with severe CAP aged ≤ 5 years of age, separately comparing the detailed information between the in-hospital death cases and the survival cases in two different age groups. Multivariate regression model was used to gure out mortality predictors. Results: 945 children were recruited, including 604 infants and 341 young children. Overall 88 deaths occurred (9.3%). There was low adherence to guidelines in antimicrobials and carbapenems were widely served as initial empiric regimens, but the ecacy was not superior to the guidelines recommended. In multivariate analyses, very severe pneumonia (OR: 3.55; 95% CI: 1.39-9.09), lower birth weight (OR: 3.92; 95% CI: 1.50-10.23), severe underweight (OR: 4.72; 95% CI: 1.92-11.62), mechanical ventilation (OR: 5.06; 95% CI: 1.97-12.95 (cid:0) OR (cid:0) 14.43; 95% CI 3.31-62.96) (cid:0) comorbidity including anemia (OR: 5.61; 95% CI: 2.36-13.35), neonatal asphyxia (OR: 6.03; 95% CI: 1.57-23.12), gastrointestinal hemorrhage (OR: 3.73; 95% CI: 1.21-11.48) and sedative-hypnotics ( OR: 4.32; 95% CI: 1.76-10.61; OR: 4.13; 95% CI (cid:0) 1.50-11.38) were independent risk factors for death, whereas a lower mortality was present in infants with probiotics (OR: 0.24; 95% CI: 0.10-0.54). Conclusions: Severe pneumonia remains a primary cause of death in children under 5 years of age. Clinical characteristics,


Background
Pneumonia is the most common illness and cause of hospitalization in children, consistently evaluated as the leading cause of death among children under 5 years of age [1][2][3] . Recent studies showed that pneumonia caused 0.9 million young children to die in 2013 worldwide, accounting for approximately 15% of under-5 deaths. In addition, a systematic review demonstrated the contribution of pneumonia to all-cause mortality in children younger than 5 years old descended from 1996 to 2015 in China, but still as high as 12.2% in 2015 4,5 , which is much more than the fourth Millennium Development Goal target rate of 4.4% by 2035. Therefore, it is urgent to decrease the mortality of childhood pneumonia, which can be achieved through early identi cation and management of mortality predictors.
To the best of our knowledge, several investigations performed in Africa and South America had showed more attention to clinical characteristics and comorbidities, identifying sex, days with symptoms, lack of breast-feed, and concurrent underweight for age, very severe pneumonia, severe malnutrition, or HIV infection as risk factors for death from severe pneumonia in children 2, 3, 6, 7 . Sidney F et al identi ed that inappropriate initial antibiotic therapy, including use of antibiotics except penicillin could increase the risk of death in children with community-acquired pneumonia (CAP) [8][9][10][11][12][13] . Nevertheless, there remains scarce information about antimicrobial practices for severe childhood CAP 14 . Similarly, the in uence of other concomitant medications on all-cause mortality under the real-world clinical conditions received less notice. In addition, current limited randomized controlled trials (RCT) have revealed that several medications utilized in childhood pneumonia are signi cantly associated with clinical outcome. For instance, ambroxol, probiotics and bronchodilators are considered as protective factors for mortality in critically ill children [15][16][17][18][19] . Furthermore, Smit et al sedative-hypnotics could signi cantly increase the mortality of children with pneumonia 20 . A previous study on severe mycoplasma pneumoniae pneumonia in children demonstrated that corticosteroid therapy could improve the clinical symptoms 21 . Dhruti et al, on the other hand, reported the bene cial effect of corticosteroid was not con rmed in children with pneumonia, and that resulted in systemic adverse reactions 22 . Therefore, it is controversial for children with pneumonia to utilize corticosteroid.
In our research, we collected detailed information, attempting to describe the etiology and epidemiological characteristics of severe childhood CAP de ned by World Health Organization (WHO) in pediatric intensive care unit (PICU) or neonatal intensive care unit (NICU) and to evaluate the risk factors associated with death, especially the effect of concomitant drugs and comorbidities on the basis of reallife conditions in China from 2012 to 2019.

Study design and eligibility criteria
This was a retrospective case-control study, which enrolled infants and young children under 5 years who suffered severe or very severe CAP in ICUs between January 2012 and January 2019 at two large teaching hospitals over the northwest China. Disease diagnosis and treatment decisions were made by the physicians. The exclusion criteria for cases comprised: (1) patients who had previously been included other clinical trials during the same research period; or (2) treatment without antimicrobials in the rst 2 days after ICU admission 11 ; or (3) duration of antimicrobial therapy < 48h 11 ; or (4) children with hospital-acquired pneumonia; or (5) cases of which data records were imperfect.
Because the classi cation of severe CAP, common comorbidities, and therapeutic regimens are different between young infants de ned as children under 2 months and children 2 to 59 months of age, data were collected and analyzed separately in the two age groups.

Variables and de nitions
We collected detailed information from medical records containing demographics, medical history (previous surgical history, antimicrobial therapy prior to admission), clinical features of pneumonia at ICU admission (fever, severity of pneumonia, days with symptom), biochemical examinations [white blood cell (WBC) count and lactate dehydrogenase (LDH) were determined with blood samples obtained in the rst 2 days of ICU hospitalization 6,23,24 ], microbiological data, comorbidities (according to clinical diagnosis and suspicion), initial antimicrobial therapy, other treatments (including supportive treatments and concomitant medications) in the period of hospitalization, and outcomes at discharge. WHO classi cation of severity of pneumonia 25 : (1) severe pneumonia: cough and/or di culty in breathing, with lower chest indrawing for young children aged 2 to 59 months, and with lower chest indrawing or tachypnea for young infants. (2) very severe pneumonia: severe pneumonia plus at least one of the following items: unconscious, lethargy or convulsions; severe dyspnea; inability to drink and breastfeed; oxygen saturation < 90% or central cyanosis; serious complications, including heart failure, respiratory failure, shock, empyema, sepsis, and MODS.
Weight-for-age was classi ed on the basis of WHO standards in 2006 5 . Moderate underweight and undernutrition were considered as weight-for-age ranging − 3~-2 SD from the median, severe undernutrition and underweight were de ned as weight-for-age < -2 SD from the median. Fever was de ned as temperature above 38℃ 3, 10 .
Pathogens were isolated and detected directly from the samples of sputum, blood, bronchoalveolar wash/aspirate and pleural effusion gathered during hospitalization 8 .
Initial antimicrobial therapy was de ned as the antimicrobial regimens used within the rst 2 days of ICU admission 10 . Children who received at least 2 days of antimicrobial treatment could be recruited.

Study procedures and outcomes
All participants were distributed in two standardized charts named "Severe CAP Inpatient Recording Chart for young children aged 2-59 months" and "Severe CAP Inpatient Recording Chart for infants aged < 2 months" according to age, which were deserved to collect accurately requisite information recorded in the medical history during hospitalization in ICUs.
The primary outcome evaluated was in-hospital mortality in the period of our investigation 26,27 .

Statistical analysis
Characteristics of participants were strati ed into survivor group and dead group according to outcome at discharge. All categorical variables between two groups were compared with Pearson χ 2 and Fisher's exact tests. The Cochran-Armitage trend test (Z) was utilized to estimate the trend in the rate of death from pneumonia with age.
Univariate logistic regression was used to preliminarily evaluate all variables previously con rmed and suspected to be associated with mortality in different age groups according to P value, odds ratios (ORs) and 95% con dence intervals (95% CIs). Results were represented as frequency and percentages of the group. Among all the statistical tests, a value of P < 0.05 was considered signi cant. In order to determine independent risk factors for death and to explore the in uence of comorbidities and medications on mortality in children with severe pneumonia, multivariate logistic regression was applied to investigate all data besides therapeutic outcome (considered as the result of severe pneumonia, not a risk factor) by a multiple stepwise regression model. Variables with P values < 0.1 were entered into the multivariable model, then variables with resulting P values > 0.05 were removed. The adjusted ORs and 95% CIs were used to estimate the strength of correlation. All statistical analyses were conducted with SAS version 9.4 (SAS Institute, Cary, NC).

Participants characteristics
From 2012 to 2019, there were a total of 958 cases under 5 years of age who suffered from severe CAP in PICU and NICU, from which 12 were not eligible for duration of antimicrobial therapy <48h and 1 was excluded because of imperfect data records. Finally, 945 children were recruited for analysis, including 341 young children aged 2-59 months and 604 infants younger than 2 months. 88 deaths occurred during the investigation, with the overall all-cause mortality being 9.3%. Cochran-Armitage trend test showed that there was no signi cant difference between age and mortality (Z=0.089, P .929). The baseline features of participants were demonstrated in Table 1. In infants younger than 2 months, 63% were male. Almost half of the infants were born with low birth weight (41.1%) and most of the dead had underweight (73.5%) or very severe pneumonia (75.5%). 102 infants (16.9%) used mechanical ventilation during hospitalization, including 22 non-survivors (21.6%). In children aged 2-59 months, 54% were male and 62.4% were with very severe pneumonia, answering for more deaths. Young children with underweight accounted for 66.7% of the fatalities and there were only 11 children (3.2%) who used mechanical ventilation. The majority of severe CAP cases aged 0-59 months were hospitalized for 7-14 days.

Comorbidity
In children less than 2 months, most of the patients had >3 diseases simultaneously (74.8%), and it revealed that more comorbidity could signi cantly increase the risk of death (P <.001). Congenital heart disease (CHD) was the most common comorbidity, but no association was identi ed between CHD and severe CAP death (31.9% vs 40.8%, P =.296). The secondary was acute respiratory distress syndrome (ARDS), which was statistically related to a high mortality (3.4% vs 20.4%, P <.001). Besides, death was more likely to occur in the cases combined with anemia, pulmonary hypertension, encephalopathy, sepsis, neonatal asphyxia, or gastrointestinal hemorrhage.
In young children aged 2-59 months, mortality was also higher in the patients with >3 combined diseases than that in the patients with less comorbidity (P <.001). Being alike to the young infants, CHD was the most prevalent among the children with severe CAP. Compared with survivors, a higher risk of death could result from severe CAP complicated with pulmonary hypertension.

Antimicrobial treatment and concomitant medications
As shown in Table 1, medications used during hospitalization may also be responsible for the mortality. The vast majority of initial antimicrobial treatment was empirical owing to the di culty in identifying pathogens, of which monotherapy was the most frequent choice for young infants aged <2 months (93.5%) with 7.6% for mortality, including second-generation cephalosporin (54.3%), carbapenem (29.7%), third-generation cephalosporin (12.7%), other β-lactams (2.5%), and else (0.8%). However, over half of the children aged 2-59 months used ≥2 antimicrobials (72.4%), mainly antiviral drugs plus third-generation cephalosporin (68.2%) or carbapenem (24.5%). Of the 121 cases using the single-drug therapy in this group, third-generation cephalosporin was the most universal choice (50.4%), followed by carbapenem (44.6%), and the mortality was as high as 14.9%, much more than the mortality of children who used ≥2 antimicrobials (9.5%). The multivariate analyses described there was no correlation between antimicrobial regimens and mortality.
In the real world, other medications including antiasthmatic, expectorant, immunotherapy, dietary supplement, and diuretic etc. were also widely employed in children with severe CAP in ICU. Tabulated data pointed out that there was a higher mortality in infants and young children with sedative-hypnotics

Discussion
In our retrospective case-control study of 945 critically ill children who were hospitalized with WHOde ned severe CAP, we found the mortality was as high as 9.3%, practically in line with a previous research with a larger cohort of 15709 cases 7 .
Several independent risk factors associated with death identi ed in our multivariate analyses were also observed in previous studies. Very severe pneumonia, severe underweight and lower birth weight were eminent mortality predictors of severe CAP 3,8,26 . Thus, we should precisely distinguish the severity of CAP and weight categories for children at admission. Comorbidity played an important role in the death from severe CAP in our study. Death was more likely to occur in the patients with more comorbidity, which was widely reported in published studies 28,29 . Our ndings were similar to the results of a prospective cohort study carried out by Penelope et al, who identi ed the mortality of severe childhood pneumonia would increase when combined with anemia and gastrointestinal hemorrhage 7 . It was obvious that the presence of neonatal asphyxia in infants with severe pneumonia was related to a higher mortality, similarly to what was demonstrated in foregoing investigation 8,30 . Consequently, accurately diagnosing and treating actively are essential to reduce the severe CAP mortality. In our exploration, the strongest independent predictor for mortality was an appropriately 14 times relative raise related to mechanical ventilation. However, Children who use mechanical ventilation have higher mortality rate, which re ected acute conditions and severity of severe pneumonia rather than the in uence of mechanical ventilation on mortality.
All of forementioned correlations were expected, and the extraordinary features of this research was that detailed concomitant medications during hospitalization in the real world were collected and analyzed. In the analysis for sedative-hypnotics, we excluded medications for sedation in mechanically ventilated patients. In nal, there were only four sedative drugs, including phenobarbital, chloral hydrate, midazolam and diazepam used in research population. Adjusted for severity of severe CAP and mechanical ventilation, our multivariate results suggested that sedative-hypnotics were signi cantly relevant to an increased mortality, which was consistent with earlier studies. In a systematic meta-analysis of 12 controlled trials including 982 infants, phenobarbital could signi cantly increase the need for mechanical ventilation in children on account of inhibiting respiratory function, resulting in apnea and respiratory failure 20 . In their retrospective analyses, Lützen et al demonstrated that life-threatening respiratory depression would occur in patients with pneumonia who used phenobarbital, despite of the low probability of phenobarbital-induced respiratory insu ciency 31 . Therefore, sedative-hypnotics were generally not recommended for severe childhood CAP. If necessary, arterial blood gas measurement should be performed regularly to monitor End tidal CO 2 , and the children are at risk of apnea when the End tidal CO 2 is < 30 mmHg or > 50mmHg 32 . What is more, probiotics were independently associated with a reduced risk of death, which was alike to the result of a prospective multicenter RCT carried out in 9 NICUs from Colombia, in which the investigator observed a lower morbidity and mortality of nosocomial infection in the probiotic group, including pneumonia 19 . In the other RCTs, Biswal et al described that prophylactic probiotics could apparently decrease the incidence of ventilator-associated pneumonia in children in PICU and NICU 18 . This might be on account of inhibiting overgrowth of pathogens through rehabilitating non-pathogenic bacteria to compete with pathogens as well as optimizing local and systemic immunity. Moreover, colonization of probiotics in the gastrointestinal tract can reduce intestinal permeability and competitively restrain attachment of pathogens, thereby depressing the possibility of colonization and translocation. In conclusion, probiotics could decrease the incidence and all-cause mortality of nosocomial infections, including severe pneumonia 18,19,33 . On the other hand, diarrhea is very common in infants and young children, accounting for almost 10% of the mortality, so did our results. A recent research showed probiotics played a protective role in preventing the diarrhea-related fatality 34 . Critically ill children possibly bene t from probiotics. Therefore, probiotics should be of particular concern in children with severe CAP in ICU. In addition, a higher mortality was observed in severe pneumonia children with corticosteroids, although there was no signi cant association in our multivariable analysis, Identically, Dhruti et al demonstrated that the bene cial effect of corticosteroid including inhaled corticosteroids was not con rmed in childhood pneumonia, and resulted in systemic adverse reactions 22 . In contrast, several researches proved corticosteroids signi cantly improved the severity of CAP, shortened the length of hospitalization, and decreased mortality in adults 35 , and similar results were observed in children with severe mycoplasma pneumoniae pneumonia 21 . Therefore, the validity of corticosteroid is controversial in childhood pneumonia. Only those combined with acute wheezing or with inhaled corticosteroids in the low to medium dose range might bene t from corticosteroid [36][37][38] . More and larger randomized, placebo-controlled trials are needed to establish the e cacy and safety of inhaled corticosteroids.
According to analysis for pathogens detected in 122 patients, we found Gram-negative bacteria were the most frequent in severe pediatric pneumonia in ICU, particularly Enterobacteriaceae and Klebsiella spp.
being in line with the published studies 39 . Furthermore, severe pneumonia resulted from Gram-negative bacteria was highly associated with death, which should receive more attention. In terms of antimicrobial therapy in severe CAP, narrow-spectrum therapy (i.e. penicillin or ampicillin) are recommended for the children aged 2-59 months, and a combination of ampicillin/penicillin and gentamicin or the use of broader-spectrum antimicrobials alone, including the third-generation cephalosporins and chloramphenicol was recommended as the rst choice for the cases with very severe pneumonia or the young infants with severe pneumonia by WHO guidelines. [9][10][11] However, a wide range of extended spectrum antimicrobials, particularly carbapenems were widely selected as initial empiric antimicrobial regimens in critically ill children with severe CAP according to our analysis, which disclosed low adherence to guidelines, and that in the multivariate model, no signi cant difference in e cacy was observed between the guideline-recommended antimicrobials and others. Therefore, initial antimicrobial regimens for severe childhood CAP should be chosen on the basis of WHO classi cation of severity and guideline-recommended treatment consistent with the severity, which may promote rational use of antimicrobials for severe childhood CAP and consequently prevent antimicrobial resistance.
Despite detailed information from a real world and rigorous criteria are the strengths, our research has several limitations. First, our retrospective study might have contributed to bias to some extent. Variables that were not documented in the medical record but potentially associated with death cannot be analyzed, including vaccination status, duration of breastfeeding, adverse drug reaction, oxygen saturation as well as some in ammation markers level and so on. Second, a few comorbidities diagnosed by physicians lacked laboratory evidence such as sepsis or anemia, which signi cantly affected our results. Therefore, it could not be decided whether these situations were misdiagnosed. The last limitation is small sample size, especially in the group of young children aged 2-59 months, restricting the capacity to adjust for confounders.

Conclusions
In conclusion, 9.3% of children with severe CAP in ICU died, and those with very severe pneumonia, severe underweight or lower birth weight were at a higher risk of death. Gram-negative bacteria are prevalent in severe childhood CAP, resulting in a higher mortality. In addition, there is a signi cant in uence of concomitant medications (e.g., sedative-hypnotics and probiotics) and comorbidity on clinical outcome. Severe CAP is a primary cause of death in children as before, and particular attention should be shown to identify the mortality predictors and establish prophylactic measures (e.g., adequate initial antibiotic drugs, supplement of oxygen therapy, evaluating laboratory results and clinical conditions, along with bene cial drugs) in order to reduce mortality. The related materials of this research, including the protocol, the registration form of verbal informed consent, and "Severe CAP Inpatient Recording Chart", have been submitted to and approved by the Ethics Committees of Shaanxi Provincial People's Hospital and Xijing Hospital.

Abbreviations
Because this is a retrospective study and the research hospitals did not require the written informed consent, we contacted the participants by phone and ask for their verbal consent after obtaining the approval of the ethics committees.

Consent for publication:
The participants agreed to the publication of the article.

Competing interests:
The Authors declare that there is no con ict of interest.

Fundings:
This work was supported by the Department of science and technology of Shaanxi Province [2019SF-188] and the Shaanxi Provincial People's Hospital [2016YX-11]. However, these funding source had no role in study design, data analysis, decision to publish, or preparation of the manuscript.
Authors' contributions: "JWW" (corresponding authors) conceptualized and designed the study, coordinated and supervised data collection, reviewed and critically revised the manuscript for important intellectual content. "LC" ( rst authors) collected the data, carried out the initial analyses, drafted the initial manuscript, and reviewed and revised the manuscript. "ZL" and "ZHJ" designed the data collection instruments and reviewed and revised the manuscript. "YZ" conceptualized the study, reviewed and revised the manuscript. All authors approved the nal manuscript as submitted and agree to be accountable for all aspects of the work.