Prediction Model for Severe Mycoplasma Pneumoniae Pneumonia in Pediatric Patients by Admission Laboratory Indicators

OBJECTIVE
The purpose of this study was to develop a model for predicting severe Mycoplasma pneumoniae pneumonia (SMPP) in pediatric patients with Mycoplasma pneumoniae pneumonia (MPP) on admission by laboratory indicators.


METHODS
Pediatric patients with MPP from January 2019 to December 2020 in our hospital were enrolled in this study. SMPP was diagnosed according to guideline for diagnosis and treatment of community-acquired pneumonia in children (2019 version). Prediction model was developed according to the admission laboratory indicators. Receiver operating characteristic curve and Goodness-of-fit test were analyzed for the predictive value.


RESULTS
A total of 233 MPP patients were included in the study, with 121 males and 112 females, aged 4.541 (1-14) years. Among them, 84 (36.1%, 95% CI 29.9-42.6%) pediatric patients were diagnosed as SMPP. Some admission laboratory indicators (immunoglobulins M (IgM), eosinophil proportion, eosinophil count, hemoglobin, erythrocyte sedimentation rate (ESR), total protein, albumin and prealbumin) were found statistically different (p < 0.05) between non-SMPP group and SMPP group. Logistic regress analysis showed IgM, eosinophil proportion, eosinophil count, ESR and prealbumin were independent risk factors for SMPP. According to these five admission laboratory indicators, the prediction model for SMPP in pediatric patients was developed. The area under curve of the prediction model was 0.777, and the goodness-of-fit test showed that the predicted SMPP incidence by the model was consistent with the actual incidence (χ2 = 244.51, p = 0.203).


CONCLUSION
We developed a model for predicting SMPP in pediatric patients by admission laboratory indicators. This model has good discrimination and calibration, which provides a basis for the early identification SMPP on admission. However, this model should be validated by multicenter studies with large sample.


Introduction
Mycoplasma pneumoniae pneumonia (MMP) is a common respiratory infection in children. In recent years, MMP accounts for 10%-40% of community acquired pneumonia (CAP) in children, especially in Asia [1]. If the host cannot effectively clear the pathogen of mycoplasma pneumoniae, immune response will continuously damage of the respiratory epithelium and cilia [2,3]. In addition, drug resistant MP strains are more and more common. In China, more than 85% of MP strains among pediatric patients have been reported as macrolide-resistant [4]. In Taiwan, the rate of macrolide-resistant strains in the pediatric population was 23.6% [5]. Therefore, severe mycoplasma pneumoniae pneumonia (SMMP) has become a signi cant issue all around the world. SMMP cases will give rise to numerous complications, such as organ dysfunction, pleural effusion, capillary leak syndrome, and plastic bronchitis [6,7]. Furthermore, cardiovascular dysfunction, liver injury, or multiple organ dysfunction syndrome were associated with longer mechanical ventilation duration, delayed PICU discharge, and high hospital mortality [6]. Therefore, early identi cation and prevention of the occurrence of SMMP is very important.
Risk assessment is the rst step for early identi cation and prevention of SMMP. Some studies have assessed the risk factors of the SMMP. Shin  Some studies also have found laboratory indicators were also related with SMMP in pediatric patients. Huang X et al reported C-reactive protein (CRP), lactate dehydrogenase (LDH) and D-dimer (D-D) as independent risk factors for SMMP, and D-D had the highest predictive power (P < 0.01) [11]. Wang M et al found that the serum concentrations of tumor necrosis factor alpha (median 114.5 pg/ml, range 49.1-897.9 pg/ml) and interferon gamma (median 376.9 pg/ml, range 221.4-1997.6 pg/ml) were signi cantly higher in children with SMMP [12]. Laboratory indicators are easy to obtain for each hospitalized pediatric patients. In this study, we aim to develop a model for predicting SMMP in pediatric patients by admission laboratory indicators.

Patients
Pediatric patients with MPP from January 2019 to December 2020 in our hospital were enrolled in this study.
Selection criteria: pediatric patients with MPP should meet the following three conditions: (1) clinical manifestations of cough with or without fever; (2) imaging ndings of lobar in ltration, lobular patchy in ltration or interstitial changes in the lung; (3) laboratory evidence of the titer of single MP antibody was ≥1:160. Informed consent was obtained from their parents or legal guardians. SMPPdiagnostic criteria SMMP was diagnosed according to the guideline for diagnosis and treatment of community acquired pneumonia in children (2019 version) [13]. Any of the following manifestations is considered as SMPP: (1) disturbance of consciousness; (2) hypoxemia, shown as: rapid breathing, RR ≥ 70 / min (infant), RR ≥ 50 / min (over 1 year old); assisted breathing (groan, nasal fan, three concave sign); intermittent apnea; oxygen saturation < 92%; (3) persistent high fever for more than 5 days; (4) dehydration; (5) chest X-ray showed that multiple lobes were involved or complicated with pleural effusion; (6) with serious extrapulmonary complications.

Admission Laboratory examination
Within 24 hours after admission, venous blood was collected and sent to the laboratory. MP-IgM and MP-IgG were detected by enzyme linked immunosorbent assay (ELISA). White blood cell (WBC) count, neutrophil proportion, neutrophil count, eosinophil proportion, eosinophil count, hemoglobin, and platelet were detected by automatic blood cell analyzer. Lactate dehydrogenase (LDH), total protein, albumin, and pre-albumin were detected by automatic biochemical analyzer. Hypersensitive C-reactive protein (HsCRP) were detected by automatic chemiluminescence instrument. Erythrocyte sedimentation rate (ESR) was detected by automatic ESR analyzer.

Statistical analysis
The count data were expressed by the number of cases and percentage, chi square test was used for comparison. The normal distribution of the admission laboratory data was expressed by the mean ± standard deviation, and the t test was used for comparison.
Prediction model was developed according the following steps: rst, all the admission laboratory indicators were analyzed by logistic stepwise regression analysis using forward likelihood ratio method (the inclusion and exclusion criteria were 0.20); then, Nomograph prediction model was constructed based on logistic regression; nally, the predictive value of the model was evaluated by discrimination and calibration (discrimination was described by ROC and AUC, calibration was described by goodness of t test).
All analyses were conducted using Stata 14.0. Statistical signi cance was de ned as P < 0.05 in the tests.

Results
Study population U/L, and there was a possibility of intra-airway mucus formation [14]. Although bronchial mucus plugs in children with MPP was not different with SMMP, their predictive model has similar characteristics to our model, and both included many laboratory indicators.
In our prediction model, IgM was recognized as an important laboratory indicator for SMMP. MP-IgM is the rstly produced antibody after the onset MP infection, followed by speci c IgG antibodies in the early stage of MPP, and many MP-IgM showed unchanged higher titers during subsequent course of the disease [15]. MP-IgM was a sensitive indicator of MP infection in children with a high consistency and correlation with the reference positive standard of PA titer ≥ 1:160, and a 4-fold increase in MP-IgG could be the supplementary diagnosis method [16]. In our study, we found MP-IgM was important indicator for SMMP, but MP-IgG was not.
In our study, we found eosinophil was related with SMMP, not only eosinophil proportion, but also eosinophil count. Kim JH et al reported in pediatric patients with MMP, serum eosinophil cationic protein (ECP) levels were signi cantly higher in atopic patients at all three time points tested, and eosinophil counts were higher in the clinical recovery and follow-up phases [17]. Bao YX believed atopy was a risk factor for the presence and SMMP due to the high pathogen load in airway, they found more children in the high-MP-load group presented with increased serum IgE and ECP (P < 0.05). Their views are the same as ours. We included eosinophil proportion and eosinophil count for the model predicting SMMP [9].
ESR and prealbumin were also enrolled in our prediction model. As an important indicator of SMMP, ESR has been con rmed by many studies. Huang X et al found ESR was signi cantly higher in the SMPP group than those in the general MPP group (P < 0.05) [11]. Lu A et al reported that age, LDH, and ESR were the signi cant factors in predicting refractory M. pneumoniae pneumonia by logistic regression [18]. Prealbumin was also an important indicator for SMMP. In Zhang J`s study, they found prealbumin levels were lower in in the mucous group of children with MMP [14]. Zhang Y et al reported the levels of prealbumin were lower than that in SMPP group than those in the general MPP group (P < 0.01) [19].

Limitation
There are some limitations in our study. First, we only included 233 MMP pediatric patients. The sample is small, and we can't do external validation in other patient cohorts. Second, the patients were come from a medical center. The prediction effect of the model needs to be further veri ed by other medical centers.

Page 7/12
We developed a model for predicting SMMP in pediatric patients by admission laboratory indicators. This model has good discrimination and calibration, which provides a basis for the early identi cation SMMP on admission.

Declarations
Ethics approval and consent to participate: Ethical approval was granted from Ethical Committee of Wuxi No.8 People's Hospital and Wuxi Occupational Disease Hospital.
Consent for publication: Not applicable.
Availability of data and materials: Data for this article can be accessed by contacting the corresponding author.
Competing interests: The authors had no con icts of interest to declare in relation to this article.
Funding: Scienti c research projects of health committee of Wuxi city in 2019.
Authors' contributions: Qing Chang, concept designer, data analysis; Hong-Lin Chen, writing support, statistical analysis; Neng-Shun Wu, Yan-Min Gao, Rong Yu, Wei-Min Zhu, data gathering, data interpretation and clinical data analysis. All authors have read and approved the manuscript.  ROC for Nomograph prediction model for SMPP in pediatric patients