Prediction for the Negative Conversion Probability of Nucleic Acid Testing in Patients with Nonsevere COVID-19 Pneumonia: A Model Based on Retrospective Cohort Study

Background: We aimed to screen clinical independent predictive factors for negative conversion of nucleic acid testing and established a predictive nomogram, so as to relieve patients' anxiety and reduce unnecessary repeated nucleic acid testing. Methods: All 70 consecutive patients with nonsevere COVID-19 pneumonia were admitted to the Fangcang shelter hospital in Wuhan from February 12 th to March 8 th , 2020. We used univariate Kaplan-Meier analysis and univariate and multivariate Cox regression to identify independent predictive factors and re�t the predictive model. Area under ROC (AUR), Brier scores and calibration plots were used to assess the performance. Results: diabetes mellitus, gender and lymphocyte were deemed independent predictive factors and were incorporated into a Cox proportional hazards model. The AUR and Brier scores of the predictive model at 14 days were 0.694 [0.472; 0.890] and 0.163 [0.109; 0.219] in the internal validation set, respectively. Similarly, the AUR and Brier scores at 21 days were 0.779 [0.505; 0.957] and 0.105 [0.042; 0.175] in the internal validation set. Conclusions: By using the predictive nomogram, the clinicians could inform patients with nonsevere COVID-19 regarding a certain time to possible negative conversion, which would relieve the patients’ anxiety and reduce repeated testing.


Introduction
In December 2019, an unexplained pneumonia emerged in Wuhan and a new βcoronavirus was identi ed by nucleic acid tests, which was named SARS-CoV-2.Currently, COVID-19 is spreading rapidly worldwide and is de ned as a pandemic by World Health Organization (WHO) (1).There have been 2.24 million con rmed cases worldwide and more than 150,000 deaths as of April 19 th 2020 (2).In response to the novel coronavirus outbreak, China rst proposed the concept of "Fangcang shelter hospitals" in early February and immediately put it into use to solve the major public health problem, namely, the shortage of isolated hospital wards.The stadium, exhibition center and other public places were transformed into Fangcang shelter hospitals, large temporary hospitals, to isolate nonsevere infectious patients and provide necessary medical services (3).
A wealth of clinical data is urgently needed for a better understanding of the disease.By studying 174 consecutive patients with COVID-19, Guo (4) found that diabetes is an independent risk factor for disease progression or deterioration.Qin (5) compared 286 severe COVID-19 patients and 166 nonsevere COVID-19 patients, and found that an increased neutrophil-lymphocyte-ratio and low lymphocyte count were the risk factors that were signi cant in monitoring the progression of disease.Gender and age were also believed to be independent risk factors for deterioration in some articles (6-8).A CALL model (9) was tted to predict the risk of patients with stable COVID-19 becoming severe.
Approximately 80% of patients with COVID-19 have mild symptoms, such as cough, weakness, nasal obstruction, rhinorrhea and pharyngalgia, and do not need special treatment (10).Additionally patients with nonsevere COVID-19 quarantined from the public often feel anxious and desperate (11).When nucleic acid testings convert to negative outcomes truly matters for patients, because the negative conversion is a signi cant part of discharge criteria or recovery.Therefore an objective and practical model is necessary to help provide nonsevere patients with a rough estimate and it could relieve patients' anxiety, improve doctor-patient communication, and reduce unnecessary repeated nucleic acid testings.
As we known, the model we established is the rst to predict the negative conversion of nucleic acid testings.

Procedure
All suspected COVID-19 patients were quarantined and underwent nucleic acid testings in local primary care organization.Once RT-PCR con rmed the presence of SARS-CoV-2 in nasopharyngeal swab samples, patients who met the above inclusion and exclusion criteria were transferred to the Fangcang shelter hospital and received antivirus therapy, traditional Chinese medicine treatment (Figure 1).

Data Collection
The retrospective cohort study started on February 12 th and ended on March 8 th in 2020.After admission to our Fangcang shelter hospital, we collected 70 consecutive patients' general conditions, symptoms, underlying diseases and basic laboratory tests.General conditions included age, gender, temperature and so forth.Symptoms consisted of cough, weakness, nasal obstruction, rhinorrhea, pharyngalgia, chest distress, diarrhea and dyspnea.Underlying diseases included hypertension, stable coronary heart disease (CHD) and diabetes mellitus (DM) for at least 6 months.Basic laboratory tests recorded only the lymphocyte count and WBC count.All factors mentioned above were regarded as predictive variables and were recorded by two clinicians independently.The outcomes variables of our study were the negative conversion time of nucleic acid testing and the negative outcomes of the tests.The negative conversion time of nucleic acid testing refers to the period from the rst positive outcomes of nucleic acid testing to negative outcomes of two consecutive nucleic acid tests and the negative outcomes were con rmed by RT-PCR twice.All patients in our cohort were followed up until March 13 th , 2020.

Statistical Analysis
Statistical analyses were performed using R software, version 3.6.2) (Figure 3).

variables associated with negative conversion of nucleic acids testing in multivariate Cox regression
model were screened by lasso again to avoid over tting (Supplementary Figure 2, 3).The remaining 3 variables (Gender, lymphocyte count and DM) were deemed the nal predictive factors and the Cox proportional hazard model was re tted (Table 3).

Predictive Model for the Probability of Negative Conversion
We made the model diagnosis and found no in uence cases (Supplementary Figure 4) or multicollinearity (VIF<2).All 3 variables met the proportional hazards assumption (Supplementary Figure 5).
The bootstrap strategy was used to validate the model internally.The predictive model showed a certain level of accuracy in estimating the probability of negative conversion of nucleic acid testing, with a Cindex of 0.664 (95% CI 0.60-0.74).
ROC curves were used to assess the discrimination of the predictive model, while Brier scores and calibration plots were used for calibration of the model.4).
Finally, the model was presented with a nomogram to facilitate clinical practice (Figure 4).Each variable in the nomogram is assigned to a certain score on a point scale from 0 to 100 according to its weight in the model.Patients with nonsevere COVID-19 pneumonia could add the points of each variable and calculate the total scores.Then the total scores were projected to the risk lines of 14-day or 21-day positive probability and the negative conversion probability equaled to one minus positive probability.

Discussion
The Fangcang shelter hospital, a new concept applied by China to cope with serious public health problems, is used to accommodate nonsevere COVID-19 pneumonia patients during the SARS-CoV-2 outbreak(3).The exponentially increasing number of patients with COVID-19 aggravated the burden of global health care resources.Age, gender, temperature, comorbidities and some laboratory tests were deemed independent risk factors for disease progression (4-8), and there have been 10 prognostic models for predicting the severe progression risk or mortality risk before March 24 th , 2020 (8).Moreover, the majority of patients has mild symptoms and does not need special medical intervention (10).Hence a model that could predict the probability of negative conversion of nucleic acid testing for mild patients becomes important, because it is a signi cant part of recovery or discharge standards and could relieve the patient's anxiety, reduce repeated valueless testings, and decrease waste of public health resources.
We collected 70 patients' general clinical characteristics and routine blood examinations results in the Fangcang shelter hospital.All factors were screened by univariate analysis and the factors that were statistically signi cant (P<0.05) or veri ed as predictors for progression by other articles were incorporated into a multivariate Cox regression.Gender, DM and lymphocyte count were deemed independent predictors and used to re t the nal model.In our predictive model, diabetes and low lymphocyte count were independent risk factors for reducing the probability of negative conversion of testings, which were in concordance with factors related to disease progression(4-6).It is speculated that dysregulation of the immune response plays an important role in the pathogenesis of COVID 19 and the lymphocyte count decrease in blood may be caused by the transfer of lymphocytes in peripheral blood to the lungs (15).
Dysregulation of the immune system in diabetic patients may be susceptible to infection(16) and be related to the prolonged nucleic acid detection.Regarding the mechanism, the inhibition of neutrophil chemotaxis, altered cytokine, phagocytic cell dysfunction and hyperglycemia may contributed the immune dysfunction (17).Compared with those without diabetes, diabetic patients were more likely to progress to severe illness and had a higher fatality rate(4, 18) and the majority of retrospective cohort studies veri ed that diabetes was a risk factor for disease progression (4,19).
Interestingly, Ji (9) and Wang (6) believed that being female was a protective factor against disease progression, which in our models was deemed a disadvantage factor for negative conversion.Regarding the reasons, We think it is based on the different composition ratio of patients.The patients in our Fangcang hospital were not only con rmed by RT-PCR, but also met the inclusion and exclusion critiria.In

Figure 4 The
Figure 4

Table 1 )
The model was presented with a nomogram to facilitate clinical practice.AUR, Brier scores and calibration plots were used to assess the performance of the model at 14 and 21 days., weakness, nasal obstruction, rhinorrhea, pharyngalgia, chest distress, diarrhea, and dyspnea were 38(54%), 17(24%), 3(4%), 4(6%), 6(9%), 18(26%), 11(16%), and 8(11%), respectively.There were 12(17%), 10(14%), 2(3%) patients with hypertension, diabetes mellitus and CHD, respectively (.Additionally, the number of patients with negative conversion at 14 days and 21 days was shown in 1 (R Foundation for StatisticalComputing, Vienna, Austria).Continuous predictive variables, such as temperature, WBC count, and lymphocyte count, were converted to categorical variables in light of normal values.According to the relationship between predictive variables and outcomes (Supplementary Figure1), age was converted to a ternary variable: ≤40, 40-50, ≥50 years of age.All recorded factors mentioned above were screened rst by univariate Cox analysis and univariateKaplan-Meier analysis.The factors with statistical signi cance (p<0.05) were incorporated into a multivariate analysis and then the independent variables mentioned as risk factors for COVID-19 progression in published articles were also incorporated.Additionally, factors believed to be predictive factors for recovery by specialists were included in the multivariate analysis as well.The incorporated variables in the multivariate Cox regression model were screened by lasso to avoid over tting.The remaining variables were deemed the nal predictive factors, and the nal Cox proportional hazard model was re tted.We made the model diagnosis based on the proportional hazards assumption, in uential cases and multicollinearity.The bootstrap strategy was used to validate the model internally.2020.67 of 70 (Approximately 95.7%) patients had two consecutive negative outcomes of nucleic acid testings and the average negative conversion time was 17.50 ± 3.87 days.32 (46%) patients were male and the median age of patients was 50 [40-57] years.The median WBC count of patients was 5.26 [4.06-6.44]*10^9/Land the average number of lymphocytes was 1.30 ± 0.49*10^9/L.The number of patients with a normal temperature was 14 (20%), and the others had different levels of fever.The numbers of patients with cough

Table 1 .
(5,9,14)ent Predictive Factors Associated with Negative Conversion 16 recorded factors were rst screened by univariate Cox analysis and univariate Kaplan-Meier analysis (Table1, 2).The factors with statistical signi cance (p<0.05) were age, gender and DM status (Figure2).Lymphocyte count and temperature were mentioned as risk factors for COVID-19 progression or deterioration in recently published articles(5,9,14).Cough and dyspnea were also believed to be predictive factors for recovery in light of specialists' suggestions.
Wang's study, gender in different groups was compared by only univariate analysis (c 2 test), so the in uence of other confounding factors cannot be ruled out without multivariate analysis.

Table 2
Univariate and Multivariate Cox Analysis of Negative Conversion

Table 3
The Re tted Model with 3 Independent Predictors