A Prediction Model of Mortality in COVID-19 Pneumonia Based on CT Score and Lymphocyte Count

Research Article

Austin J Pathol Lab Med. 2021; 8(1): 1030.

A Prediction Model of Mortality in COVID-19 Pneumonia Based on CT Score and Lymphocyte Count

Xie Y¹#, Dong H²#, Liao Y²#, Zhang J¹, Lv H¹, Teng X¹, Wang T¹, Zhang X¹, Xu Y¹, Wu S¹, Yang L²*, He Z¹* and Wang R¹*

¹Shanghai Jiao Tong University School of Medicine, Shanghai General Hospital, China

²Wuhan University, Wuhan Third Hospital, China #Contributed Equally

*Corresponding author: Ruilan Wang, Department of Critical Care Medicine, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, 650 New Songjiang Road, Songjiang, Shanghai 201600, P.R. China

Zhiyan He, Department of Critical Care Medicine, Shanghai General Hospital, Shanghai Jiao Tong University School of Medicine, 650 New Songjiang Road, Songjiang, Shanghai 201600, P.R. China

Luyu Yang, Intensive Care Unit, Wuhan Third Hospital, Wuhan University, China

Received: February 04, 2021; Accepted: February 26, 2021; Published: March 05, 2021

Abstract

Background: COVID-19 nucleic acid swab tests have a high false positive rate; therefore, diagnosing COVID-19 pneumonia and predicting prognosis by CT scan are very important.

Methods: In this retrospective single-centre study, we included consecutive suspected critical COVID-19 pneumonia cases in the intensive care unit of Wuhan Third Hospital from January 31, 2020, to March 16, 2020. 204 cases were confirmed by real-time RT-PCR, and all patients were evaluated with CT, cut-off values were obtained according to the Youden index and were divided into a high CT score group and a low CT score group. Epidemiological, demographic, clinical, and laboratory data were collected. Finally, Through multi-factor logistic regression model, a prediction model based on multiple prediction indicators was formed, and new joint predictive factors were calculated. The prediction model of mortality in COVID-19 pneumonia based on CT score and lymphocyte count was constructed through data processing analysis.

Results: The major imaging feature of COVID-19 pneumonia is Ground Glass Opacities (GGOs). Multivariate regression analysis found that the CT score and absolute lymphocyte count were independent risk factors for death and that the CT score predicted mortality (AUC-ROC =0.7, cut-off=1.45). When the absolute lymphocyte count was lower, the patient’s CT score was also lower. Based on this, a prediction model was established. The prediction model was: In [P/(1-P)]=0.667*gender+0.057*age-0.086CT score-0.831 lymphocyte count-3.91, the goodness of fit test of the model was P=0.041, and the area under the curve of the ROC curve of the model was 0.779.

Conclusion: CT score and absolute lymphocyte count are independent risk factors for mortality, and patients with a high CT score may have a worse prognosis. A lower absolute lymphocyte count may indicate that the patient’s CT score is also reduced. The model established by combining CT scores and lymphocyte count showed a good degree of calibration and differentiation.

Keywords: Corona virus disease; CT score; Lymphocyte count

Introduction

In December 2019, pneumonia associated with the 2019 novel coronavirus (COVID-19) was reported in Wuhan, China. Now COVID-19 has become a global problem. Between 4% and 11% of patients with COVID-19 pneumonia rapidly develop Acute Respiratory Distress Syndrome (ARDS), acute respiratory failure, and other serious complications within a short period of time and eventually deteriorate and die from multiple organ failure [1, 2].

Computed Tomography (CT) examinations, which are reproducible and objective, are used clinically to determine the severity of pneumonia and constitute an effective tool for accurately defining the management strategy. Several studies have reported a relationship between high-resolution CT and the prognosis of pulmonary fibrosis and have found that radiographic fibrosis scores based on HRCT scan reticulation and honeycomb degree predict mortality [3-5]. However, the relationship between CT and the prognosis of COVID-19 pneumonia remains unknown.

While COVID-19 nucleic acid swab tests have a false positive rate as high as 41%; lung CT abnormalities have been detected in 74% of COVID-19 cases [6]. In this study, we retrospectively analysed patients with suspected COVID-19 pneumonia to assess whether CT scores were useful predictors of mortality. We established a prediction model of mortality in COVID-19 pneumonia based on CT score and lymphocyte count.

Methods

Study design and participants

For this retrospective single-centre study, we collected consecutive patients with suspected COVID-19 pneumonia in the intensive care unit of Wuhan Third Hospital, China, from January 31, 2020, to March 16, 2020. The study was approved by the ethics committee and exempted from written informed consent.

Inclusion criteria: All patients were suspected of infection with SARS-CoV-2.

Exclusion criteria: Patients with missing data were excluded.

Collected data

We obtained epidemiological, demographic, clinical, laboratory, management and outcome data from the patient records. Clinical results were followed up until March 16, 2020.

Swabs of the upper respiratory tract from all patients at admission were kept in viral transport media. Real-time RT-PCR was used to detect for SARS-CoV-2. All patients underwent chest CT examinations.

All of the patients’ radiographs included Ground Glass Opacities (GGOs), consolidation, pleural effusion, seepage, and number of involved layers. According to the acute exacerbation of idiopathic pulmonary fibrosis score (AE - IPF), each patient underwent chest CT [3]. Two physicians analysed the chest CT images of each patient and calculated the CT score for each layer as follows: CT score = normal lung tissue (%) ×1+GGO (%) ×2+ consolidation (%) ×3. Finally, the scores of all levels were averaged to obtain the final score, and the higher score from the two physicians was used. The physicians were blinded to the clinical profile and outcome of the patient during the scoring process.

Our treatment plan was as follows: the antiviral drug administered was Arbidol, the initial antibiotic administered was moxifloxacin, and the antibiotic was adjusted according to the culture results. Anticoagulant therapy was employed according to the condition of the patient. When the absolute lymphocyte count was <0.5*109/L, intravenous immunoglobulin was given for 5 days, and supplemental albumin was provided. If after 5 days the absolute lymphocyte number was still low, thymosin was added to enhance the immune function. Small doses of hormone (1-2 mg/kg) were administered for 3-5 days depending on the patient’s level of exudation. The rest of the treatment was based on the guidelines of the WHO [7].

Outcome

The primary outcome was the 28-day mortality. We also described the demographics, physical signs on admission, laboratory results, chest CT scores, and clinical outcomes.

Statistical analysis

If data were normally distributed, we represent the continuous measurements as the mean (SD); otherwise, they are represented as the median (IQR). The classification variables are represented as counts (%). With regard to the laboratory results, we also evaluated whether the measurements were outside the normal ranges. Logistic regression analysis was used to evaluate the outcomes based on the risk factors selected through univariate analysis. The diagnostic value of the CT score for predicting mortality was evaluated by calculating the Area Under the Receiver Operating Characteristic curve (AUC ROC). AUC ROC analysis was performed by comparing survivors with nonsurvivors. The optimal cut-off value was determined by the highest value of the Youden index, maximizing the specificity and sensitivity, as shown in the AUC ROC analysis. We used SPSS (version 26.0) for analyses.

Through multi-factor logistic regression model, a prediction model based on multiple prediction indicators was formed, and new joint predictive factors were calculated. With 28-day mortality or not as the outcome, the area under the receiver operating characteristic curve Area Under the ROC curve (AUROC) of the combined predictors and each original index was compared to determine the optimal critical value, calculate the operating performance parameters such as sensitivity, specificity and accuracy of prediction, and finally carry out individual prediction by substituting individual values. Stata 10.0 software was used for statistical analysis and mapping, and Stata 10.0 command statement, operation process and result output.

Results

The study included 39 patients with suspected COVID-19. All patients tested positive for SARS-CoV-2 nucleic acid. Of these, 24 (61.5%) patients were male, with an average age of 60 years (51-66) (Table 1). At admission, 14 (35.9%) of the patients had tachypnoea. The platelet counts were lower than normal in 2 patients (5.13%) and higher than normal in 1 patient (2.56%). Three patients (7.69%) had abnormal liver function. Four (10.25%) patients had abnormal renal function. Three patients (7.69%) had an abnormal myocardial enzyme spectrum. In most patients, D-dimer levels were greater than 35 mg/L during the course of the disease.