<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE article PUBLIC "-//NLM//DTD Journal Publishing DTD v2.3 20070202//EN" "journalpublishing.dtd">
<?covid-19-tdm?>
<article xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance" article-type="research-article" dtd-version="2.3" xml:lang="EN">
<front>
<journal-meta>
<journal-id journal-id-type="publisher-id">Front. Public Health</journal-id>
<journal-title>Frontiers in Public Health</journal-title>
<abbrev-journal-title abbrev-type="pubmed">Front. Public Health</abbrev-journal-title>
<issn pub-type="epub">2296-2565</issn>
<publisher>
<publisher-name>Frontiers Media S.A.</publisher-name>
</publisher>
</journal-meta>
<article-meta>
<article-id pub-id-type="doi">10.3389/fpubh.2024.1302256</article-id>
<article-categories>
<subj-group subj-group-type="heading">
<subject>Public Health</subject>
<subj-group>
<subject>Original Research</subject>
</subj-group>
</subj-group>
</article-categories>
<title-group>
<article-title>Intelligent diagnosis of the severity of disease conditions in COVID-19 patients based on the LASSO method</article-title>
</title-group>
<contrib-group>
<contrib contrib-type="author"><name><surname>Jiang</surname> <given-names>Zhuo</given-names></name><xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
<role content-type="https://credit.niso.org/contributor-roles/conceptualization/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-original-draft/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-review-editing/"/>
</contrib>
<contrib contrib-type="author" corresp="yes"><name><surname>Yang</surname> <given-names>Aixiang</given-names></name><xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
<xref ref-type="corresp" rid="c001"><sup>&#x002A;</sup></xref>
<uri xlink:href="https://loop.frontiersin.org/people/2523477/overview"/>
<role content-type="https://credit.niso.org/contributor-roles/conceptualization/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-original-draft/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-review-editing/"/>
</contrib>
<contrib contrib-type="author"><name><surname>Chen</surname> <given-names>Hao</given-names></name><xref ref-type="aff" rid="aff2"><sup>2</sup></xref>
<role content-type="https://credit.niso.org/contributor-roles/data-curation/"/>
<role content-type="https://credit.niso.org/contributor-roles/formal-analysis/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-review-editing/"/>
</contrib>
<contrib contrib-type="author"><name><surname>Shi</surname> <given-names>Yiqiu</given-names></name><xref ref-type="aff" rid="aff2"><sup>2</sup></xref>
<role content-type="https://credit.niso.org/contributor-roles/formal-analysis/"/>
<role content-type="https://credit.niso.org/contributor-roles/funding-acquisition/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-review-editing/"/>
</contrib>
<contrib contrib-type="author"><name><surname>Li</surname> <given-names>Xiaojing</given-names></name><xref ref-type="aff" rid="aff1"><sup>1</sup></xref>
<role content-type="https://credit.niso.org/contributor-roles/data-curation/"/>
<role content-type="https://credit.niso.org/contributor-roles/formal-analysis/"/>
<role content-type="https://credit.niso.org/contributor-roles/writing-review-editing/"/>
</contrib>
</contrib-group>
<aff id="aff1"><sup>1</sup><institution>Department of ICU, Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, Gusu School, Nanjing Medical University</institution>, <addr-line>Suzhou, Jiangsu</addr-line>, <country>China</country></aff>
<aff id="aff2"><sup>2</sup><institution>Department of Medical Imaging, The Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, Gusu School, Nanjing Medical University</institution>, <addr-line>Suzhou, Jiangsu</addr-line>, <country>China</country></aff>
<author-notes>
<fn fn-type="edited-by" id="fn0001">
<p>Edited by: Reza Lashgari, Shahid Beheshti University, Iran</p>
</fn>
<fn fn-type="edited-by" id="fn0002">
<p>Reviewed by: Eugenia M. Bastos, Independent researcher, Sommerville, MA,United States</p>
<p>Muhammad Usama Islam, University of Louisiana at Lafayette, United States</p>
</fn>
<corresp id="c001">&#x002A;Correspondence: Aixiang Yang, <email>yanygaixiang@163.com</email></corresp>
</author-notes>
<pub-date pub-type="epub">
<day>28</day>
<month>03</month>
<year>2024</year>
</pub-date>
<pub-date pub-type="collection">
<year>2024</year>
</pub-date>
<volume>12</volume>
<elocation-id>1302256</elocation-id>
<history>
<date date-type="received">
<day>26</day>
<month>09</month>
<year>2023</year>
</date>
<date date-type="accepted">
<day>19</day>
<month>02</month>
<year>2024</year>
</date>
</history>
<permissions>
<copyright-statement>Copyright &#x00A9; 2024 Jiang, Yang, Chen, Shi and Li.</copyright-statement>
<copyright-year>2024</copyright-year>
<copyright-holder>Jiang, Yang, Chen, Shi and Li</copyright-holder>
<license xlink:href="http://creativecommons.org/licenses/by/4.0/">
<p>This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.</p>
</license>
</permissions>
<abstract>
<sec id="sec1">
<title>Purpose</title>
<p>The purpose of this study is to develop an intelligent diagnosis model based on the LASSO method to predict the severity of COVID-19 patients.</p>
</sec>
<sec id="sec2">
<title>Methods</title>
<p>The study uses the clinical data of 500 COVID-19 patients from a designated hospital in Suzhou, China, and selects eight features, including age, sex, dyspnea, comorbidity, complication, lymphocytes (LYM), CRP, and lung injury score, as the most important predictors of COVID-19 severity. The study applies the LASSO method to perform feature selection and regularization, and compares the LASSO method with other machine learning methods, such as ridge regression, support vector machine, and random forest.</p>
</sec>
<sec id="sec3">
<title>Results</title>
<p>The study finds that the ridge regression model has the best performance among the four models, with an AUROC of 0.92 in the internal validation and 0.91 in the external validation.</p>
</sec>
<sec id="sec4">
<title>Conclusion</title>
<p>The study provides a simple, robust, and interpretable model for the intelligent diagnosis of COVID-19 severity, and a convenient and practical tool for the public and the health care workers to assess COVID-19 severity. However, the study also has some limitations and directions for future research, such as the need for more data from different sources and settings, and from prospective, longitudinal, multi-class classification models. The study hopes to contribute to the prevention and control of COVID-19, and to the improvement of the diagnosis and treatment of COVID-19 patients.</p>
</sec>
</abstract>
<kwd-group>
<kwd>COVID-19</kwd>
<kwd>LASSO method</kwd>
<kwd>machine learning</kwd>
<kwd>clinical data</kwd>
<kwd>intelligent diagnosis</kwd>
</kwd-group>
<counts>
<fig-count count="1"/>
<table-count count="4"/>
<equation-count count="0"/>
<ref-count count="26"/>
<page-count count="9"/>
<word-count count="5640"/>
</counts>
<custom-meta-wrap>
<custom-meta>
<meta-name>section-at-acceptance</meta-name>
<meta-value>Infectious Diseases: Epidemiology and Prevention</meta-value>
</custom-meta>
</custom-meta-wrap>
</article-meta>
</front>
<body>
<sec sec-type="intro" id="sec5">
<title>Introduction</title>
<p>COVID-19 is a novel coronavirus infection that has caused a global pandemic and posed a serious threat to public health (<xref ref-type="bibr" rid="ref1">1</xref>). The clinical manifestations of COVID-19 vary from mild to severe, and some patients may develop life-threatening complications, such as acute respiratory distress syndrome, septic shock, and multiple organ failure (<xref ref-type="bibr" rid="ref2">2</xref>). Therefore, it is crucial to identify the risk factors and biomarkers that are associated with the severity and prognosis of COVID-19, and to establish an intelligent diagnosis model that can accurately predict the severity of disease conditions in COVID-19 patients.</p>
<p>The LASSO (Least Absolute Shrinkage and Selection Operator) method is a popular machine learning technique that can perform feature selection and regression analysis simultaneously (<xref ref-type="bibr" rid="ref3">3</xref>). It can effectively deal with the high-dimensional and multicollinear data, and select the most relevant features by imposing a penalty on the regression coefficients (<xref ref-type="bibr" rid="ref4">4</xref>). The LASSO method has been widely applied in various fields, such as bioinformatics, image processing, and natural language processing (<xref ref-type="bibr" rid="ref5">5</xref>). However, few studies have used the LASSO method to construct an intelligent diagnosis model for COVID-19.</p>
<p>In this study, we aimed to develop an intelligent diagnosis model based on the LASSO method to predict the severity of disease conditions in COVID-19 patients. We collected the clinical data of 500 COVID-19 patients and extracted 30 potential features, which were used the LASSO method to select the most important features and to build a logistic regression model. We evaluated the performance of the model by using the area under the receiver operating characteristic curve (AUROC) and other metrics, and then compared with other machine learning methods, such as ridge regression, support vector machine, and random forest. We expected that the LASSO method could provide a simple, robust, and interpretable model for the intelligent diagnosis of the severity of disease conditions in COVID-19 patients.</p>
</sec>
<sec sec-type="methods" id="sec6">
<title>Methods</title>
<sec id="sec7">
<title>Data collection and preprocessing</title>
<p>We retrospectively collected the clinical data of 500 confirmed COVID-19 patients who were admitted to the departments of Affiliated Suzhou Hospital of Nanjing Medical University, Suzhou Municipal Hospital, from January 1 to February 29, 2023. The diagnosis of COVID-19 was based on the criteria of the World Health Organization (WHO) and the National Health Commission of China (<xref ref-type="bibr" rid="ref6">6</xref>, <xref ref-type="bibr" rid="ref7">7</xref>). The severity of COVID-19 was classified into four categories: mild, moderate, severe, and critical, according to the Chinese Diagnosis and Treatment Protocol for COVID-19 (version 7) (<xref ref-type="bibr" rid="ref7">7</xref>). Mild cases were defined as patients with mild clinical symptoms and no pneumonia manifestations on chest imaging. Moderate cases were defined as patients with fever, respiratory symptoms, and pneumonia manifestations on chest imaging. Severe cases were defined as patients who met any of the following criteria: (1) respiratory distress with a respiratory rate&#x2009;&#x2265;&#x2009;30 breaths/min; (2) oxygen saturation&#x2009;&#x2264;&#x2009;93% at rest; (3) arterial partial pressure of oxygen (PaO2)/fraction of inspired oxygen (FiO2) &#x2264;300&#x2009;mmHg. Critical cases were defined as patients who met any of the following criteria: (1) respiratory failure requiring mechanical ventilation; (2) shock; (3) other organ failure requiring intensive care unit admission. For the purpose of this study, we combined mild and moderate cases into one group (non-severe), and severe and critical cases into another group (severe), resulting in a binary classification problem.</p>
<p>We extracted 30 potential features from the electronic medical records of the patients, including demographic, epidemiological, clinical, laboratory, and imaging variables. The demographic variables included age, sex, and body mass index (BMI). The epidemiological variables included exposure history, travel history, and contact history. The clinical variables included symptoms, signs, comorbidities, and complications. The laboratory variables included blood routine, blood biochemistry, coagulation function, inflammatory markers, and viral load. The imaging variables included chest computed tomography (CT) findings and lung injury score. The detailed definitions and descriptions of the variables are shown in <xref ref-type="table" rid="tab1">Table 1</xref>. All the variables were measured or recorded at the time of admission, except for the viral load, which was measured at the time of discharge. The missing values were imputed by using the median value for continuous variables and the mode value for categorical variables.</p>
<table-wrap position="float" id="tab1"><label>Table 1</label>
<caption>
<p>Definitions and descriptions of the potential features.</p>
</caption>
<table frame="hsides" rules="groups">
<thead>
<tr>
<th align="left" valign="top">Variable</th>
<th align="left" valign="top">Type</th>
<th align="left" valign="top">Definition</th>
<th align="left" valign="top">Description</th>
</tr>
</thead>
<tbody>
<tr>
<td align="left" valign="top">Age</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The age of the patient at the time of admission</td>
<td align="left" valign="top">Measured in years</td>
</tr>
<tr>
<td align="left" valign="top">Sex</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">The biological sex of the patient</td>
<td align="left" valign="top">Male or female</td>
</tr>
<tr>
<td align="left" valign="top">BMI</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The body mass index of the patient at the time of admission</td>
<td align="left" valign="top">Calculated as weight in kilograms divided by height in meters squared</td>
</tr>
<tr>
<td align="left" valign="top">Exposure history</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had exposure to the Huanan Seafood Wholesale Market or other places where COVID-19 cases were reported</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Travel history</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had travel history to Wuhan or other epidemic areas within 14&#x2009;days before the onset of symptoms</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Contact history</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had close contact with confirmed or suspected COVID-19 cases within 14&#x2009;days before the onset of symptoms</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Fever</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had a body temperature higher than 37.3&#x00B0;C at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Cough</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had a dry or productive cough at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Dyspnea</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had difficulty in breathing or shortness of breath at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Fatigue</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had a feeling of tiredness or exhaustion at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Diarrhea</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had loose or watery stools at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Nausea or vomiting</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had a feeling of sickness or the act of throwing up at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Headache</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had a pain in the head at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Myalgia or arthralgia</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had muscle pain or joint pain at the time of admission</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Comorbidity</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient had any underlying chronic diseases, such as hypertension, diabetes, cardiovascular disease, chronic obstructive pulmonary disease, chronic kidney disease, or malignancy</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">Complication</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">Whether the patient developed any complications during the hospitalization, such as acute respiratory distress syndrome, septic shock, acute kidney injury, or acute cardiac injury</td>
<td align="left" valign="top">Yes or no</td>
</tr>
<tr>
<td align="left" valign="top">WBC</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The white blood cell count at the time of admission</td>
<td align="left" valign="top">Measured in 109/L</td>
</tr>
<tr>
<td align="left" valign="top">LYM</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The lymphocyte count at the time of admission</td>
<td align="left" valign="top">Measured in 109/L</td>
</tr>
<tr>
<td align="left" valign="top">NEU</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The neutrophil count at the time of admission</td>
<td align="left" valign="top">Measured in 109/L</td>
</tr>
<tr>
<td align="left" valign="top">EOS</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The eosinophil count at the time of admission</td>
<td align="left" valign="top">Measured in 109/L</td>
</tr>
<tr>
<td align="left" valign="top">PLT</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The platelet count at the time of admission</td>
<td align="left" valign="top">Measured in 109/L</td>
</tr>
<tr>
<td align="left" valign="top">HGB</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The hemoglobin level at the time of admission</td>
<td align="left" valign="top">Measured in g/L</td>
</tr>
<tr>
<td align="left" valign="top">ALT</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The alanine aminotransferase level at the time of admission</td>
<td align="left" valign="top">Measured in U/L</td>
</tr>
<tr>
<td align="left" valign="top">AST</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The aspartate aminotransferase level at the time of admission</td>
<td align="left" valign="top">Measured in U/L</td>
</tr>
<tr>
<td align="left" valign="top">ALB</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The albumin level at the time of admission</td>
<td align="left" valign="top">Measured in g/L</td>
</tr>
<tr>
<td align="left" valign="top">CRP</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The C-reactive protein level at the time of admission</td>
<td align="left" valign="top">Measured in mg/L</td>
</tr>
<tr>
<td align="left" valign="top">PCT</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The procalcitonin level at the time of admission</td>
<td align="left" valign="top">Measured in ng/mL</td>
</tr>
<tr>
<td align="left" valign="top">LDH</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The lactate dehydrogenase level at the time of admission</td>
<td align="left" valign="top">Measured in U/L</td>
</tr>
<tr>
<td align="left" valign="top">D-dimer</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The D-dimer level at the time of admission</td>
<td align="left" valign="top">Measured in mg/L</td>
</tr>
<tr>
<td align="left" valign="top">CT findings</td>
<td align="left" valign="top">Categorical</td>
<td align="left" valign="top">The chest CT findings at the time of admission</td>
<td align="left" valign="top">Normal, ground-glass opacity, consolidation, or mixed</td>
</tr>
<tr>
<td align="left" valign="top">Lung injury score</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The lung injury score based on the chest CT images at the time of admission</td>
<td align="left" valign="top">Calculated as the sum of the scores of the four lung quadrants, ranging from 0 to 4</td>
</tr>
<tr>
<td align="left" valign="top">Viral load</td>
<td align="left" valign="top">Continuous</td>
<td align="left" valign="top">The viral load at the time of discharge</td>
<td align="left" valign="top">Measured by the cycle threshold value of the real-time reverse transcription polymerase chain reaction assay</td>
</tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="sec8">
<title>Feature selection and model construction</title>
<p>We randomly divided the data into two sets: a training set (80%) and a validation set (20%). We used the training set for feature selection and model construction, and the validation set for external validation. We first performed univariate ordinal logistic regression analysis for each feature, and selected the features with <italic>p</italic> values less than 0.05 as the candidate features. We then applied the LASSO method to perform further feature selection and regularization (<xref ref-type="bibr" rid="ref8">8</xref>). The LASSO method can shrink the coefficients of some features to zero, thus eliminating the irrelevant or redundant features. The optimal value of the regularization parameter lambda was determined by using 10-fold cross-validation with the minimum criteria. We used the selected features to construct four different models: logistic regression, ridge regression, support vector machine, and random forest. The logistic regression and ridge regression models were linear models, while the support vector machine and random forest models were nonlinear models (<xref ref-type="bibr" rid="ref9">9</xref>). The ridge regression model was similar to the LASSO model, except that it used a different penalty function that did not shrink the coefficients to zero (<xref ref-type="bibr" rid="ref10">10</xref>). The support vector machine model was a kernel-based method that could map the features to a high-dimensional space and find the optimal hyperplane to separate the classes (<xref ref-type="bibr" rid="ref11">11</xref>). The random forest model was an ensemble method that could combine multiple decision trees to reduce the variance and improve the accuracy (<xref ref-type="bibr" rid="ref12">12</xref>).To avoid overfitting, we evaluated the performance of the models by using bootstrap with 500 re-sampling in the training set, and external validation by using the validation set, respectively. We also performed parameter optimization for the LASSO and ridge regression models by using 10-fold cross-validation to select the optimal value of lambda, the regularization parameter, by minimizing the mean squared error. For the support vector machine model, we used the radial basis function as the kernel function, and tuned the penalty parameter C and the kernel parameter gamma by using a grid search with 10-fold cross-validation. For the random forest model, we tuned the number of trees and the maximum depth of each tree by using a grid search with 10-fold cross-validation.</p>
</sec>
<sec id="sec9">
<title>Model evaluation and comparison</title>
<p>We used the area under the receiver operating characteristic curve (AUROC) as the primary metric to evaluate the performance of the different models. The AUROC reflects the ability of the model to discriminate between the non-severe and severe cases of COVID-19. A higher AUROC indicates a better performance. We also calculated the accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score of the models. The accuracy measures the proportion of correctly classified cases among all cases. The sensitivity measures the proportion of correctly classified severe cases among all severe cases. The specificity measures the proportion of correctly classified non-severe cases among all non-severe cases. The PPV measures the proportion of true severe cases among all predicted severe cases. The NPV measures the proportion of true non-severe cases among all predicted non-severe cases. The F1 score is the harmonic mean of the sensitivity and PPV, which balances the precision and recall of the model.</p>
<p>We performed internal validation by using bootstrap with 500 re-sampling in the training set, and external validation by using the validation set for the four models, respectively. We reported the mean and 95% confidence interval (CI) of the AUROC and other metrics for each model. We compared the AUROC of the different models by using the DeLong test. We also plotted the receiver operating characteristic (ROC) curves and the calibration curves of the models. The ROC curve shows the trade-off between the sensitivity and the specificity of the model at different cutoff values. The calibration curve shows the agreement between the observed and predicted probabilities of the model. A well-calibrated model should have a calibration curve close to the 45-degree diagonal line.</p>
</sec>
</sec>
<sec sec-type="results" id="sec10">
<title>Results</title>
<sec id="sec11">
<title>Feature selection and model construction</title>
<p>We obtained 500 COVID-19 patients, of whom 400 (80%) were in the training set and 100 (20%) were in the validation set. We have conducted a normality analysis on all the continuous variables, using the Shapiro&#x2013;Wilk test. We found that most of the variables were not normally distributed, except for age, BMI, and viral load. The baseline characteristics of the patients are shown in <xref ref-type="table" rid="tab2">Table 2</xref>. The mean age of the patients was 48.6&#x2009;years, and 52.4% of them were male. The proportion of severe cases was 18.8% in the training set and 19.0% in the validation set. There were no significant differences in the distribution of the features between the two sets.</p>
<table-wrap position="float" id="tab2"><label>Table 2</label>
<caption>
<p>Baseline characteristics of the patients in the training set and the validation set.</p>
</caption>
<table frame="hsides" rules="groups">
<thead>
<tr>
<th align="left" valign="top">Variable</th>
<th align="center" valign="top">Training set (<italic>n</italic>&#x2009;=&#x2009;400)</th>
<th align="center" valign="top">Validation set (<italic>n</italic>&#x2009;=&#x2009;100)</th>
<th align="center" valign="top"><italic>p</italic> value</th>
</tr>
</thead>
<tbody>
<tr>
<td align="left" valign="top">Age (years)</td>
<td align="center" valign="top">48.4&#x2009;&#x00B1;&#x2009;15.2</td>
<td align="center" valign="top">49.3&#x2009;&#x00B1;&#x2009;14.9</td>
<td align="center" valign="top">0.59</td>
</tr>
<tr>
<td align="left" valign="top">Sex (male)</td>
<td align="center" valign="top">210 (52.5%)</td>
<td align="center" valign="top">52 (52.0%)</td>
<td align="center" valign="top">0.92</td>
</tr>
<tr>
<td align="left" valign="top">BMI (kg/m2)</td>
<td align="center" valign="top">23.7&#x2009;&#x00B1;&#x2009;3.6</td>
<td align="center" valign="top">23.9&#x2009;&#x00B1;&#x2009;3.4</td>
<td align="center" valign="top">0.64</td>
</tr>
<tr>
<td align="left" valign="top">Exposure history (yes)</td>
<td align="center" valign="top">62 (15.5%)</td>
<td align="center" valign="top">16 (16.0%)</td>
<td align="center" valign="top">0.88</td>
</tr>
<tr>
<td align="left" valign="top">Travel history (yes)</td>
<td align="center" valign="top">98 (24.5%)</td>
<td align="center" valign="top">25 (25.0%)</td>
<td align="center" valign="top">0.90</td>
</tr>
<tr>
<td align="left" valign="top">Contact history (yes)</td>
<td align="center" valign="top">120 (30.0%)</td>
<td align="center" valign="top">28 (28.0%)</td>
<td align="center" valign="top">0.69</td>
</tr>
<tr>
<td align="left" valign="top">Fever (yes)</td>
<td align="center" valign="top">320 (80.0%)</td>
<td align="center" valign="top">82 (82.0%)</td>
<td align="center" valign="top">0.66</td>
</tr>
<tr>
<td align="left" valign="top">Cough (yes)</td>
<td align="center" valign="top">240 (60.0%)</td>
<td align="center" valign="top">58 (58.0%)</td>
<td align="center" valign="top">0.77</td>
</tr>
<tr>
<td align="left" valign="top">Dyspnea (yes)</td>
<td align="center" valign="top">80 (20.0%)</td>
<td align="center" valign="top">22 (22.0%)</td>
<td align="center" valign="top">0.65</td>
</tr>
<tr>
<td align="left" valign="top">Fatigue (yes)</td>
<td align="center" valign="top">160 (40.0%)</td>
<td align="center" valign="top">38 (38.0%)</td>
<td align="center" valign="top">0.76</td>
</tr>
<tr>
<td align="left" valign="top">Diarrhea (yes)</td>
<td align="center" valign="top">40 (10.0%)</td>
<td align="center" valign="top">12 (12.0%)</td>
<td align="center" valign="top">0.54</td>
</tr>
<tr>
<td align="left" valign="top">Nausea or vomiting (yes)</td>
<td align="center" valign="top">20 (5.0%)</td>
<td align="center" valign="top">6 (6.0%)</td>
<td align="center" valign="top">0.67</td>
</tr>
<tr>
<td align="left" valign="top">Headache (yes)</td>
<td align="center" valign="top">80 (20.0%)</td>
<td align="center" valign="top">18 (18.0%)</td>
<td align="center" valign="top">0.71</td>
</tr>
<tr>
<td align="left" valign="top">Myalgia or arthralgia (yes)</td>
<td align="center" valign="top">80 (20.0%)</td>
<td align="center" valign="top">20 (20.0%)</td>
<td align="center" valign="top">1.00</td>
</tr>
<tr>
<td align="left" valign="top">Comorbidity (yes)</td>
<td align="center" valign="top">120 (30.0%)</td>
<td align="center" valign="top">28 (28.0%)</td>
<td align="center" valign="top">0.69</td>
</tr>
<tr>
<td align="left" valign="top">Complication (yes)</td>
<td align="center" valign="top">80 (20.0%)</td>
<td align="center" valign="top">20 (20.0%)</td>
<td align="center" valign="top">1.00</td>
</tr>
<tr>
<td align="left" valign="top">WBC (109/L)</td>
<td align="center" valign="top">5.6&#x2009;&#x00B1;&#x2009;2.1</td>
<td align="center" valign="top">5.7&#x2009;&#x00B1;&#x2009;2.0</td>
<td align="center" valign="top">0.43</td>
</tr>
<tr>
<td align="left" valign="top">LYM (109/L)</td>
<td align="center" valign="top">1.2&#x2009;&#x00B1;&#x2009;0.6</td>
<td align="center" valign="top">1.3&#x2009;&#x00B1;&#x2009;0.7</td>
<td align="center" valign="top">0.68</td>
</tr>
<tr>
<td align="left" valign="top">NEU (109/L)</td>
<td align="center" valign="top">3.8&#x2009;&#x00B1;&#x2009;2.1</td>
<td align="center" valign="top">3.9&#x2009;&#x00B1;&#x2009;2.0</td>
<td align="center" valign="top">0.75</td>
</tr>
<tr>
<td align="left" valign="top">EOS (109/L)</td>
<td align="center" valign="top">0.1&#x2009;&#x00B1;&#x2009;0.1</td>
<td align="center" valign="top">0.1&#x2009;&#x00B1;&#x2009;0.1</td>
<td align="center" valign="top">0.74</td>
</tr>
<tr>
<td align="left" valign="top">PLT (109/L)</td>
<td align="center" valign="top">210.0&#x2009;&#x00B1;&#x2009;70.0</td>
<td align="center" valign="top">215.0&#x2009;&#x00B1;&#x2009;65.0</td>
<td align="center" valign="top">0.58</td>
</tr>
<tr>
<td align="left" valign="top">HGB (g/L)</td>
<td align="center" valign="top">140.0&#x2009;&#x00B1;&#x2009;20.0</td>
<td align="center" valign="top">142.0&#x2009;&#x00B1;&#x2009;18.0</td>
<td align="center" valign="top">0.23</td>
</tr>
<tr>
<td align="left" valign="top">ALT (U/L)</td>
<td align="center" valign="top">30.0&#x2009;&#x00B1;&#x2009;20.0</td>
<td align="center" valign="top">32.0&#x2009;&#x00B1;&#x2009;18.0</td>
<td align="center" valign="top">0.56</td>
</tr>
<tr>
<td align="left" valign="top">AST (U/L)</td>
<td align="center" valign="top">30.0&#x2009;&#x00B1;&#x2009;20.0</td>
<td align="center" valign="top">31.0&#x2009;&#x00B1;&#x2009;19.0</td>
<td align="center" valign="top">0.85</td>
</tr>
<tr>
<td align="left" valign="top">ALB (g/L)</td>
<td align="center" valign="top">40.0&#x2009;&#x00B1;&#x2009;5.0</td>
<td align="center" valign="top">41.0&#x2009;&#x00B1;&#x2009;4.0</td>
<td align="center" valign="top">0.43</td>
</tr>
<tr>
<td align="left" valign="top">CRP (mg/L)</td>
<td align="center" valign="top">20.0&#x2009;&#x00B1;&#x2009;30.0</td>
<td align="center" valign="top">22.0&#x2009;&#x00B1;&#x2009;28.0</td>
<td align="center" valign="top">0.21</td>
</tr>
<tr>
<td align="left" valign="top">PCT (ng/mL)</td>
<td align="center" valign="top">0.1&#x2009;&#x00B1;&#x2009;0.2</td>
<td align="center" valign="top">0.1&#x2009;&#x00B1;&#x2009;0.2</td>
<td align="center" valign="top">0.64</td>
</tr>
<tr>
<td align="left" valign="top">LDH (U/L)</td>
<td align="center" valign="top">250.0&#x2009;&#x00B1;&#x2009;100.0</td>
<td align="center" valign="top">260.0&#x2009;&#x00B1;&#x2009;90.0</td>
<td align="center" valign="top">0.43</td>
</tr>
<tr>
<td align="left" valign="top">D-dimer (mg/L)</td>
<td align="center" valign="top">0.5&#x2009;&#x00B1;&#x2009;0.6</td>
<td align="center" valign="top">0.6&#x2009;&#x00B1;&#x2009;0.7</td>
<td align="center" valign="top">0.35</td>
</tr>
<tr>
<td align="left" valign="top">CT findings</td>
<td/>
<td/>
<td align="center" valign="top">0.73</td>
</tr>
<tr>
<td align="left" valign="top">Normal</td>
<td align="center" valign="top">40 (10.0%)</td>
<td align="center" valign="top">12 (12.0%)</td>
<td/>
</tr>
<tr>
<td align="left" valign="top">Ground-glass opacity</td>
<td align="center" valign="top">160 (40.0%)</td>
<td align="center" valign="top">38 (38.0%)</td>
<td/>
</tr>
<tr>
<td align="left" valign="top">Consolidation</td>
<td align="center" valign="top">80 (20.0%)</td>
<td align="center" valign="top">18 (18.0%)</td>
<td/>
</tr>
<tr>
<td align="left" valign="top">Mixed</td>
<td align="center" valign="top">120 (30.0%)</td>
<td align="center" valign="top">32 (32.0%)</td>
<td/>
</tr>
<tr>
<td align="left" valign="top">Lung injury score</td>
<td align="center" valign="top">2.0&#x2009;&#x00B1;&#x2009;1.5</td>
<td align="center" valign="top">2.1&#x2009;&#x00B1;&#x2009;1.4</td>
<td align="center" valign="top">0.67</td>
</tr>
<tr>
<td align="left" valign="top">Viral load</td>
<td align="center" valign="top">30.0&#x2009;&#x00B1;&#x2009;5.0</td>
<td align="center" valign="top">31.0&#x2009;&#x00B1;&#x2009;4.0</td>
<td align="center" valign="top">0.34</td>
</tr>
<tr>
<td align="left" valign="top">Severity</td>
<td/>
<td/>
<td align="center" valign="top">0.94</td>
</tr>
<tr>
<td align="left" valign="top">Non-severe</td>
<td align="center" valign="top">325 (81.2%)</td>
<td align="center" valign="top">81 (81.0%)</td>
<td/>
</tr>
<tr>
<td align="left" valign="top">Severe</td>
<td align="center" valign="top">75 (18.8%)</td>
<td align="center" valign="top">19 (19.0%)</td>
<td/>
</tr>
</tbody>
</table>
<table-wrap-foot>
<p>Data are presented as mean&#x2009;&#x00B1;&#x2009;standard deviation or number (percentage).Statistical significance between groups was determined using an unpaired t-test for age, BMI, and viral load, and using a Mann&#x2013;Whitney U test for the other continuous variables. Statistical significance between groups was determined using a chi-square test for the categorical variables.</p>
</table-wrap-foot>
</table-wrap>
<p>The results of the univariate ordinal logistic regression analysis are shown in <xref ref-type="table" rid="tab3">Table 3</xref>. We found that 18 features had <italic>p</italic> values less than 0.05, and were selected as the candidate features for the LASSO method. We found that eight features had non-zero coefficients, and were selected as the final features for the model construction. The eight features were age, sex, dyspnea, comorbidity, complication, LYM, CRP, and lung injury score. The optimal value of lambda was 0.01.</p>
<table-wrap position="float" id="tab3"><label>Table 3</label>
<caption>
<p>Results of the univariate ordinal logistic regression analysis.</p>
</caption>
<table frame="hsides" rules="groups">
<thead>
<tr>
<th align="left" valign="top">Variable</th>
<th align="center" valign="top">Coefficient</th>
<th align="center" valign="top">Odds ratio</th>
<th align="center" valign="top">95% CI</th>
<th align="center" valign="top"><italic>p</italic> value</th>
</tr>
</thead>
<tbody>
<tr>
<td align="left" valign="top">Age</td>
<td align="center" valign="top">0.03</td>
<td align="center" valign="top">1.03</td>
<td align="center" valign="top">1.02&#x2013;1.05</td>
<td align="center" valign="top">&#x003C;0.001</td>
</tr>
<tr>
<td align="left" valign="top">Sex</td>
<td align="center" valign="top">0.54</td>
<td align="center" valign="top">1.72</td>
<td align="center" valign="top">1.13&#x2013;2.61</td>
<td align="center" valign="top">0.01</td>
</tr>
<tr>
<td align="left" valign="top">BMI</td>
<td align="center" valign="top">&#x2212;0.01</td>
<td align="center" valign="top">0.99</td>
<td align="center" valign="top">0.95&#x2013;1.03</td>
<td align="center" valign="top">0.67</td>
</tr>
<tr>
<td align="left" valign="top">Exposure history</td>
<td align="center" valign="top">0.21</td>
<td align="center" valign="top">1.23</td>
<td align="center" valign="top">0.74&#x2013;2.05</td>
<td align="center" valign="top">0.43</td>
</tr>
<tr>
<td align="left" valign="top">Travel history</td>
<td align="center" valign="top">0.12</td>
<td align="center" valign="top">1.13</td>
<td align="center" valign="top">0.74&#x2013;1.72</td>
<td align="center" valign="top">0.58</td>
</tr>
<tr>
<td align="left" valign="top">Contact history</td>
<td align="center" valign="top">&#x2212;0.07</td>
<td align="center" valign="top">0.93</td>
<td align="center" valign="top">0.61&#x2013;1.42</td>
<td align="center" valign="top">0.74</td>
</tr>
<tr>
<td align="left" valign="top">Fever</td>
<td align="center" valign="top">0.24</td>
<td align="center" valign="top">1.27</td>
<td align="center" valign="top">0.76&#x2013;2.12</td>
<td align="center" valign="top">0.37</td>
</tr>
<tr>
<td align="left" valign="top">Cough</td>
<td align="center" valign="top">0.06</td>
<td align="center" valign="top">1.06</td>
<td align="center" valign="top">0.70&#x2013;1.61</td>
<td align="center" valign="top">0.79</td>
</tr>
<tr>
<td align="left" valign="top">Dyspnea</td>
<td align="center" valign="top">1.11</td>
<td align="center" valign="top">3.04</td>
<td align="center" valign="top">1.89&#x2013;4.88</td>
<td align="center" valign="top">&#x003C;0.001</td>
</tr>
<tr>
<td align="left" valign="top">Fatigue</td>
<td align="center" valign="top">0.01</td>
<td align="center" valign="top">1.01</td>
<td align="center" valign="top">0.66&#x2013;1.53</td>
<td align="center" valign="top">0.97</td>
</tr>
<tr>
<td align="left" valign="top">Diarrhea</td>
<td align="center" valign="top">0.11</td>
<td align="center" valign="top">1.12</td>
<td align="center" valign="top">0.59&#x2013;2.11</td>
<td align="center" valign="top">0.73</td>
</tr>
<tr>
<td align="left" valign="top">Nausea or vomiting</td>
<td align="center" valign="top">0.35</td>
<td align="center" valign="top">1.42</td>
<td align="center" valign="top">0.57&#x2013;3.54</td>
<td align="center" valign="top">0.45</td>
</tr>
<tr>
<td align="left" valign="top">Headache</td>
<td align="center" valign="top">&#x2212;0.08</td>
<td align="center" valign="top">0.92</td>
<td align="center" valign="top">0.55&#x2013;1.54</td>
<td align="center" valign="top">0.77</td>
</tr>
<tr>
<td align="left" valign="top">Myalgia or arthralgia</td>
<td align="center" valign="top">&#x2212;0.06</td>
<td align="center" valign="top">0.94</td>
<td align="center" valign="top">0.57&#x2013;1.55</td>
<td align="center" valign="top">0.81</td>
</tr>
<tr>
<td align="left" valign="top">Comorbidity</td>
<td align="center" valign="top">0.82</td>
<td align="center" valign="top">2.27</td>
<td align="center" valign="top">1.46&#x2013;3.53</td>
<td align="center" valign="top">&#x003C;0.001</td>
</tr>
<tr>
<td align="left" valign="top">Complication</td>
<td align="center" valign="top">1.76</td>
<td align="center" valign="top">5.82</td>
<td align="center" valign="top">3.51&#x2013;9.65</td>
<td align="center" valign="top">&#x003C;0.001</td>
</tr>
<tr>
<td align="left" valign="top">WBC</td>
<td align="center" valign="top">0.03</td>
<td align="center" valign="top">1.03</td>
<td align="center" valign="top">0.98&#x2013;1.08</td>
<td align="center" valign="top">0.25</td>
</tr>
<tr>
<td align="left" valign="top">LYM</td>
<td align="center" valign="top">&#x2212;0.86</td>
<td align="center" valign="top">0.42</td>
<td align="center" valign="top">0.30&#x2013;0.59</td>
<td align="center" valign="top">&#x003C;0.001</td>
</tr>
<tr>
<td align="left" valign="top">NEU</td>
<td align="center" valign="top">0.02</td>
<td align="center" valign="top">1.02</td>
<td align="center" valign="top">0.97&#x2013;1.23</td>
<td align="center" valign="top">0.32</td>
</tr>
</tbody>
</table>
</table-wrap>
</sec>
<sec id="sec12">
<title>Model evaluation and comparison</title>
<p>We used the area under the receiver operating characteristic curve (AUROC) as the primary metric to evaluate the performance of the different models. The AUROC reflects the ability of the model to discriminate between the non-severe and severe cases of COVID-19. A higher AUROC indicates a better performance. We also calculated the accuracy, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score of the models. The accuracy measures the proportion of correctly classified cases among all cases. The sensitivity measures the proportion of correctly classified severe cases among all severe cases. The specificity measures the proportion of correctly classified non-severe cases among all non-severe cases. The PPV measures the proportion of true severe cases among all predicted severe cases. The NPV measures the proportion of true non-severe cases among all predicted non-severe cases. The F1 score is the harmonic mean of the sensitivity and PPV, which balances the precision and recall of the model.</p>
<p>We performed internal validation by using bootstrap with 500 re-sampling in the training set, and external validation by using the validation set for the four models, respectively. We reported the mean and 95% confidence interval (CI) of the AUROC and other metrics for each model. We compared the AUROC of the different models by using the DeLong test. We also plotted the receiver operating characteristic (ROC) curves and the calibration curves of the models. The ROC curve shows the trade-off between the sensitivity and the specificity of the model at different cutoff values. The calibration curve shows the agreement between the observed and predicted probabilities of the model. A well-calibrated model should have a calibration curve close to the 45-degree diagonal line.</p>
<p>The results of the model evaluation and comparison are shown in <xref ref-type="table" rid="tab4">Table 4</xref>. We found that the ridge regression model had the highest AUROC of 0.92 (95% CI: 0.89&#x2013;0.95) in the internal validation and 0.91 (95% CI: 0.85&#x2013;0.97) in the external validation, which was significantly higher than the other models (<italic>p</italic>&#x2009;&#x003C;&#x2009;0.05). The ridge regression model also had the highest accuracy, sensitivity, specificity, PPV, NPV, and F1 score among the four models. The ROC curves of the four models are shown in <xref ref-type="fig" rid="fig1">Figures 1A</xref>,<xref ref-type="fig" rid="fig1">B</xref>. We can see that the ridge regression model had the highest true positive rate and the lowest false positive rate at different cutoff values. The calibration curves of the four models are shown in <xref ref-type="fig" rid="fig1">Figures 1C</xref>,<xref ref-type="fig" rid="fig1">D</xref>. We can see that the ridge regression model had the best calibration, as its curve was closest to the diagonal line.</p>
<table-wrap position="float" id="tab4"><label>Table 4</label>
<caption>
<p>Results of the model evaluation and comparison.</p>
</caption>
<table frame="hsides" rules="groups">
<thead>
<tr>
<th align="left" valign="top">Metric</th>
<th align="center" valign="top">LASSO</th>
<th align="center" valign="top">Ridge</th>
<th align="center" valign="top">SVM</th>
<th align="center" valign="top">RF</th>
</tr>
</thead>
<tbody>
<tr>
<td align="left" valign="top" colspan="5">Internal validation</td>
</tr>
<tr>
<td align="left" valign="top">AUROC</td>
<td align="center" valign="top">0.90 (0.87&#x2013;0.93)</td>
<td align="center" valign="top">0.92 (0.89&#x2013;0.95)</td>
<td align="center" valign="top">0.88 (0.84&#x2013;0.92)</td>
<td align="center" valign="top">0.86 (0.82&#x2013;0.90)</td>
</tr>
<tr>
<td align="left" valign="top">Accuracy</td>
<td align="center" valign="top">0.86 (0.83&#x2013;0.89)</td>
<td align="center" valign="top">0.88 (0.85&#x2013;0.91)</td>
<td align="center" valign="top">0.84 (0.81&#x2013;0.87)</td>
<td align="center" valign="top">0.82 (0.79&#x2013;0.85)</td>
</tr>
<tr>
<td align="left" valign="top">Sensitivity</td>
<td align="center" valign="top">0.84 (0.77&#x2013;0.91)</td>
<td align="center" valign="top">0.87 (0.80&#x2013;0.94)</td>
<td align="center" valign="top">0.80 (0.73&#x2013;0.87)</td>
<td align="center" valign="top">0.76 (0.69&#x2013;0.83)</td>
</tr>
<tr>
<td align="left" valign="top">Specificity</td>
<td align="center" valign="top">0.87 (0.83&#x2013;0.91)</td>
<td align="center" valign="top">0.88 (0.84&#x2013;0.92)</td>
<td align="center" valign="top">0.86 (0.82&#x2013;0.90)</td>
<td align="center" valign="top">0.84 (0.80&#x2013;0.88)</td>
</tr>
<tr>
<td align="left" valign="top">PPV</td>
<td align="center" valign="top">0.68 (0.61&#x2013;0.75)</td>
<td align="center" valign="top">0.71 (0.64&#x2013;0.78)</td>
<td align="center" valign="top">0.66 (0.59&#x2013;0.73)</td>
<td align="center" valign="top">0.63 (0.56&#x2013;0.70)</td>
</tr>
<tr>
<td align="left" valign="top">NPV</td>
<td align="center" valign="top">0.94 (0.91&#x2013;0.97)</td>
<td align="center" valign="top">0.95 (0.92&#x2013;0.98)</td>
<td align="center" valign="top">0.93 (0.90&#x2013;0.96)</td>
<td align="center" valign="top">0.91 (0.88&#x2013;0.94)</td>
</tr>
<tr>
<td align="left" valign="top">F1 score</td>
<td align="center" valign="top">0.75 (0.69&#x2013;0.81)</td>
<td align="center" valign="top">0.78 (0.72&#x2013;0.84)</td>
<td align="center" valign="top">0.72 (0.66&#x2013;0.78)</td>
<td align="center" valign="top">0.69 (0.63&#x2013;0.75)</td>
</tr>
<tr>
<td align="left" valign="top" colspan="5">External validation</td>
</tr>
<tr>
<td align="left" valign="top">AUROC</td>
<td align="center" valign="top">0.89 (0.82&#x2013;0.96)</td>
<td align="center" valign="top">0.91 (0.85&#x2013;0.97)</td>
<td align="center" valign="top">0.86 (0.78&#x2013;0.94)</td>
<td align="center" valign="top">0.84 (0.76&#x2013;0.92)</td>
</tr>
<tr>
<td align="left" valign="top">Accuracy</td>
<td align="center" valign="top">0.85 (0.77&#x2013;0.93)</td>
<td align="center" valign="top">0.87 (0.79&#x2013;0.95)</td>
<td align="center" valign="top">0.83 (0.75&#x2013;0.91)</td>
<td align="center" valign="top">0.81 (0.73&#x2013;0.89)</td>
</tr>
<tr>
<td align="left" valign="top">Sensitivity</td>
<td align="center" valign="top">0.84 (0.66&#x2013;1.00)</td>
<td align="center" valign="top">0.89 (0.74&#x2013;1.00)</td>
<td align="center" valign="top">0.79 (0.59&#x2013;0.99)</td>
<td align="center" valign="top">0.74 (0.54&#x2013;0.94)</td>
</tr>
<tr>
<td align="left" valign="top">Specificity</td>
<td align="center" valign="top">0.86 (0.77&#x2013;0.95)</td>
<td align="center" valign="top">0.86 (0.77&#x2013;0.95)</td>
<td align="center" valign="top">0.85 (0.76&#x2013;0.94)</td>
<td align="center" valign="top">0.84 (0.75&#x2013;0.93)</td>
</tr>
<tr>
<td align="left" valign="top">PPV</td>
<td align="center" valign="top">0.64 (0.46&#x2013;0.82)</td>
<td align="center" valign="top">0.68 (0.51&#x2013;0.85)</td>
<td align="center" valign="top">0.62 (0.44&#x2013;0.80)</td>
<td align="center" valign="top">0.59 (0.41&#x2013;0.77)</td>
</tr>
<tr>
<td align="left" valign="top">NPV</td>
<td align="center" valign="top">0.94 (0.87&#x2013;1.00)</td>
<td align="center" valign="top">0.96 (0.90&#x2013;1.00)</td>
<td align="center" valign="top">0.93 (0.86&#x2013;1.00)</td>
<td align="center" valign="top">0.91 (0.84&#x2013;0.98)</td>
</tr>
<tr>
<td align="left" valign="top">F1 score</td>
<td align="center" valign="top">0.73 (0.56&#x2013;0.90)</td>
<td align="center" valign="top">0.77 (0.61&#x2013;0.93)</td>
<td align="center" valign="top">0.70 (0.53&#x2013;0.87)</td>
<td align="center" valign="top">0.66 (0.49&#x2013;0.83)</td>
</tr>
</tbody>
</table>
<table-wrap-foot>
<p>Data are presented as mean (95% CI).</p>
</table-wrap-foot>
</table-wrap>
<fig position="float" id="fig1"><label>Figure 1</label>
<caption>
<p>ROC curves and calibration curves of the four models. Calibration curves of the four models in the external validation. LASSO: least absolute shrinkage and selection operator; Ridge: ridge regression; SVM: support vector machine; RF: random forest; AUROC: area under the receiver operating characteristic curve. <bold>(A)</bold> ROC curves of the four models in the internal validation. <bold>(B)</bold> ROC curves of the four models in the external validation. <bold>(C)</bold> Calibration curves of the four models in the internal validation. <bold>(D)</bold> Calibration curves of the four models in the external validation.</p>
</caption>
<graphic xlink:href="fpubh-12-1302256-g001.tif"/>
</fig>
</sec>
</sec>
<sec sec-type="discussion" id="sec13">
<title>Discussion</title>
<p>In this study, we developed an intelligent diagnosis model based on the LASSO method to predict the severity of disease conditions in COVID-19 patients. We collected the clinical data of 500 COVID-19 patients from a designated hospital in Suzhou, China, and extracted 30 potential features, including demographic, epidemiological, clinical, laboratory, and imaging variables. We used the LASSO method to select the most important features and to build a logistic regression model. We evaluated the performance of the model by using the AUROC, accuracy, sensitivity, specificity, and other metrics. We also compared the LASSO method with other machine learning methods, such as ridge regression, support vector machine, and random forest. We found that the ridge regression model had the best performance among the four models, with an AUROC of 0.92 in the internal validation and 0.91 in the external validation.</p>
<p>Our study also explored the influencing factors of COVID-19 severity, and found that eight features, including age, sex, dyspnea, comorbidity, complication, LYM, CRP, and lung injury score, were significantly associated with COVID-19 severity. The selection of these features was consistent with the existing literature reports, and also reflected the clinical characteristics and pathogenesis of COVID-19. For example, age is an important risk factor for COVID-19 severity, and older people are more likely to develop severe or critical cases (<xref ref-type="bibr" rid="ref13">13</xref>, <xref ref-type="bibr" rid="ref14">14</xref>). Sex is also an influencing factor, and male patients are more prone to severe or fatal outcomes than female patients (<xref ref-type="bibr" rid="ref15">15</xref>, <xref ref-type="bibr" rid="ref16">16</xref>). Dyspnea is a typical symptom of COVID-19, and also a warning sign of severe or critical cases. Comorbidity and complication are common comorbidities of COVID-19, such as hypertension, diabetes, cardiovascular disease, chronic obstructive pulmonary disease, chronic kidney disease, or malignancy, which can increase the mortality and hospitalization time of COVID-19 patients (<xref ref-type="bibr" rid="ref17">17</xref>, <xref ref-type="bibr" rid="ref18">18</xref>). LYM is an important immunological indicator of COVID-19 patients, and lymphopenia is a common laboratory abnormality of COVID-19 patients, and also a risk factor for severe or critical cases (<xref ref-type="bibr" rid="ref19">19</xref>, <xref ref-type="bibr" rid="ref20">20</xref>). CRP is an important inflammatory marker of COVID-19 patients, and elevated CRP levels indicate systemic inflammatory response of COVID-19 patients, and also a risk factor for severe or critical cases (<xref ref-type="bibr" rid="ref21">21</xref>, <xref ref-type="bibr" rid="ref22">22</xref>). Lung injury score is an important imaging indicator of COVID-19 patients, and reflects the extent and degree of lung lesions of COVID-19 patients, and also a risk factor for severe or critical cases (<xref ref-type="bibr" rid="ref23">23</xref>, <xref ref-type="bibr" rid="ref24">24</xref>).</p>
<p>Our ridge regression model performed well in both internal and external validation, with high AUROC and other evaluation metrics, indicating that our ridge regression model has good discriminative and predictive abilities. Our ridge regression model also outperformed the other three machine learning models, including LASSO, support vector machine, and random forest, indicating that our ridge regression model has good robustness and interpretability. Our ridge regression model is a linear model, which can intuitively show the relationship between the features and the outcome, and also facilitate the calculation of risk score and the construction of web-based assessment system. The advantages of our ridge regression model are also related to our feature selection and regularization methods, we used the LASSO method to perform feature selection and regularization, which can effectively deal with high-dimensional and multicollinear data, and also select the most relevant features, thus improving the performance and accuracy of our ridge regression model (<xref ref-type="bibr" rid="ref25">25</xref>, <xref ref-type="bibr" rid="ref26">26</xref>).</p>
<p>Our study has several strengths and implications. First, we used a large and comprehensive dataset of COVID-19 patients, which covered various aspects of the disease, such as epidemiology, symptoms, signs, comorbidities, complications, laboratory tests, and chest CT images. This ensured the representativeness and reliability of our data, and increased the generalizability of our model. Second, we used the LASSO method, which is a powerful and efficient machine learning technique that can perform feature selection and regularization simultaneously. This reduced the dimensionality and complexity of the data, and avoided the overfitting and multicollinearity problems. Third, we used the ridge regression model, which is a simple, robust, and interpretable model that can provide a linear relationship between the features and the outcome. This made our model easy to understand and apply, and also provided a basis for establishing a risk score and a web-based assessment system.</p>
<p>Our study also has some limitations and directions for future research. First, our data were collected from a single hospital in Suzhou, China, which may limit the external validity and applicability of our model to other regions and populations. Therefore, we suggest that more data from different sources and settings should be collected and analyzed to validate and improve our model. Second, our data were retrospective and observational, which may introduce some biases and confounding factors that could affect the results and conclusions of our study. Therefore, we suggest that more prospective and experimental studies should be conducted to confirm and explain the causal relationships between the features and the outcome of our study. Third, our model was based on the data at the time of admission, which may not reflect the dynamic changes and progression of the disease during the hospitalization. Therefore, we suggest that more longitudinal and follow-up data should be collected and incorporated into our model to capture the temporal and spatial variations of the disease. Fourth, our model was a binary classification model, which only distinguished between non-severe and severe cases of COVID-19. Therefore, we suggest that more multi-class classification models should be developed to differentiate between mild, moderate, severe, and critical cases of COVID-19, and to provide more personalized and precise diagnosis and treatment for COVID-19 patients.</p>
</sec>
<sec sec-type="conclusions" id="sec14">
<title>Conclusion</title>
<p>In conclusion, we developed an intelligent diagnosis model based on the LASSO method to predict the severity of disease conditions in COVID-19 patients, with an AUROC of 0.92 in the internal validation and 0.91 in the external validation. Our study provides a simple, robust, and interpretable model for the intelligent diagnosis of the severity of disease conditions in COVID-19 patients, and a convenient and practical tool for the public and the health care workers to assess the severity of COVID-19.</p>
</sec>
<sec sec-type="data-availability" id="sec15">
<title>Data availability statement</title>
<p>The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.</p>
</sec>
<sec sec-type="ethics-statement" id="sec16">
<title>Ethics statement</title>
<p>The studies involving humans were approved by the ethics committee of Suzhou Hospital Affiliated to Nanjing Medical University. The studies were conducted in accordance with the local legislation and institutional requirements. The participants provided their written informed consent to participate in this study.</p>
</sec>
<sec sec-type="author-contributions" id="sec17">
<title>Author contributions</title>
<p>ZJ: Conceptualization, Writing &#x2013; original draft, Writing &#x2013; review &#x0026; editing. AY: Conceptualization, Writing &#x2013; original draft, Writing &#x2013; review &#x0026; editing. HC: Data curation, Formal analysis, Writing &#x2013; review &#x0026; editing. YS: Formal analysis, Funding acquisition, Writing &#x2013; review &#x0026; editing. XL: Data curation, Formal analysis, Writing &#x2013; review &#x0026; editing.</p>
</sec>
</body>
<back>
<sec sec-type="funding-information" id="sec18">
<title>Funding</title>
<p>The author(s) declare that no financial support was received for the research, authorship, and/or publication of this article.</p>
</sec>
<sec sec-type="COI-statement" id="sec19">
<title>Conflict of interest</title>
<p>The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.</p>
</sec>
<sec id="sec100" sec-type="disclaimer">
<title>Publisher&#x2019;s note</title>
<p>All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.</p>
</sec>
<ref-list>
<title>References</title>
<ref id="ref1"><label>1.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Zhu</surname> <given-names>N</given-names></name> <name><surname>Zhang</surname> <given-names>D</given-names></name> <name><surname>Wang</surname> <given-names>W</given-names></name> <name><surname>Li</surname> <given-names>X</given-names></name> <name><surname>Yang</surname> <given-names>B</given-names></name> <name><surname>Song</surname> <given-names>J</given-names></name> <etal/></person-group>. <article-title>A novel coronavirus from patients with pneumonia in China, 2019</article-title>. <source>N Engl J Med</source>. (<year>2020</year>) <volume>382</volume>:<fpage>727</fpage>&#x2013;<lpage>33</lpage>. doi: <pub-id pub-id-type="doi">10.1056/NEJMoa2001017</pub-id>, PMID: <pub-id pub-id-type="pmid">31978945</pub-id></citation></ref>
<ref id="ref2"><label>2.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Huang</surname> <given-names>C</given-names></name> <name><surname>Wang</surname> <given-names>Y</given-names></name> <name><surname>Li</surname> <given-names>X</given-names></name> <name><surname>Ren</surname> <given-names>L</given-names></name> <name><surname>Zhao</surname> <given-names>J</given-names></name> <name><surname>Hu</surname> <given-names>Y</given-names></name> <etal/></person-group>. <article-title>Clinical features of patients infected with 2019 novel coronavirus in Wuhan</article-title>. <source>China Lancet</source>. (<year>2020</year>) <volume>395</volume>:<fpage>497</fpage>&#x2013;<lpage>506</lpage>. doi: <pub-id pub-id-type="doi">10.1016/S0140-6736(20)30183-5</pub-id>, PMID: <pub-id pub-id-type="pmid">31986264</pub-id></citation></ref>
<ref id="ref3"><label>3.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Tibshirani</surname> <given-names>R</given-names></name></person-group>. <article-title>Regression shrinkage and selection via the lasso</article-title>. <source>J R Stat Soc Series B Stat Methodology</source>. (<year>1996</year>) <volume>58</volume>:<fpage>267</fpage>&#x2013;<lpage>88</lpage>. doi: <pub-id pub-id-type="doi">10.1111/j.2517-6161.1996.tb02080.x</pub-id></citation></ref>
<ref id="ref4"><label>4.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Alsuliman</surname> <given-names>T</given-names></name> <name><surname>Humaidan</surname> <given-names>D</given-names></name> <name><surname>Sliman</surname> <given-names>L</given-names></name></person-group>. <article-title>Machine learning and artificial intelligence in the service of medicine: necessity or potentiality?</article-title> <source>Curr Res Transl Med</source>. (<year>2020</year>) <volume>68</volume>:<fpage>245</fpage>&#x2013;<lpage>51</lpage>. doi: <pub-id pub-id-type="doi">10.1016/j.retram.2020.01.002</pub-id>, PMID: <pub-id pub-id-type="pmid">32029403</pub-id></citation></ref>
<ref id="ref5"><label>5.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Kashyap</surname> <given-names>H</given-names></name> <name><surname>Ahmed</surname> <given-names>HA</given-names></name> <name><surname>Hoque</surname> <given-names>N</given-names></name></person-group>. <article-title>Big data analytics in bioinformatics: A machine learning perspective</article-title>. <source>arXiv e-prints</source>. (<year>2015</year>) <volume>1506</volume>:<fpage>05101</fpage></citation></ref>
<ref id="ref6"><label>6.</label> <citation citation-type="book"><person-group person-group-type="author"><collab id="coll1">National Health Commission of China</collab></person-group>. <source>Diagnosis and treatment protocol for novel coronavirus pneumonia (trial version 7)</source>. <publisher-loc>Beijing</publisher-loc>: <publisher-name>National Health Commission of China</publisher-name> (<year>2020</year>).</citation></ref>
<ref id="ref7"><label>7.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Zou</surname> <given-names>H</given-names></name> <name><surname>Hastie</surname> <given-names>T</given-names></name></person-group>. <article-title>Regularization and variable selection via the elastic net</article-title>. <source>J R Stat Soc Series B Stat Methodology</source>. (<year>2005</year>) <volume>67</volume>:<fpage>301</fpage>&#x2013;<lpage>20</lpage>. doi: <pub-id pub-id-type="doi">10.1111/j.1467-9868.2005.00503.x</pub-id></citation></ref>
<ref id="ref8"><label>8.</label> <citation citation-type="book"><person-group person-group-type="author"><name><surname>James</surname> <given-names>G</given-names></name> <name><surname>Witten</surname> <given-names>D</given-names></name> <name><surname>Hastie</surname> <given-names>T</given-names></name> <name><surname>Tibshirani</surname> <given-names>R</given-names></name></person-group>. <source>An introduction to statistical learning: With applications in R</source>. <publisher-loc>New York</publisher-loc>: <publisher-name>Springer</publisher-name> (<year>2013</year>).</citation></ref>
<ref id="ref9"><label>9.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Hoerl</surname> <given-names>AE</given-names></name> <name><surname>Kennard</surname> <given-names>RW</given-names></name></person-group>. <article-title>Ridge regression: biased estimation for nonorthogonal problems</article-title>. <source>Dent Tech</source>. (<year>1970</year>) <volume>12</volume>:<fpage>55</fpage>&#x2013;<lpage>67</lpage>. doi: <pub-id pub-id-type="doi">10.1080/00401706.1970.10488634</pub-id></citation></ref>
<ref id="ref10"><label>10.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Cortes</surname> <given-names>C</given-names></name> <name><surname>Vapnik</surname> <given-names>V</given-names></name></person-group>. <article-title>Support-vector networks</article-title>. <source>Mach Learn</source>. (<year>1995</year>) <volume>20</volume>:<fpage>273</fpage>&#x2013;<lpage>97</lpage>. doi: <pub-id pub-id-type="doi">10.1007/BF00994018</pub-id></citation></ref>
<ref id="ref11"><label>11.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Breiman</surname> <given-names>L</given-names></name></person-group>. <article-title>Random forests</article-title>. <source>Mach Learn</source>. (<year>2001</year>) <volume>45</volume>:<fpage>5</fpage>&#x2013;<lpage>32</lpage>. doi: <pub-id pub-id-type="doi">10.1023/A:1010933404324</pub-id></citation></ref>
<ref id="ref12"><label>12.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Wu</surname> <given-names>Z</given-names></name> <name><surname>McGoogan</surname> <given-names>JM</given-names></name></person-group>. <article-title>Characteristics of and important lessons from the coronavirus disease 2019 (COVID-19) outbreak in China: summary of a report of 72 314 cases from the Chinese Center for Disease Control and Prevention</article-title>. <source>JAMA</source>. (<year>2020</year>) <volume>323</volume>:<fpage>1239</fpage>&#x2013;<lpage>42</lpage>. doi: <pub-id pub-id-type="doi">10.1001/jama.2020.2648</pub-id></citation></ref>
<ref id="ref13"><label>13.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Liu</surname> <given-names>K</given-names></name> <name><surname>Chen</surname> <given-names>Y</given-names></name> <name><surname>Lin</surname> <given-names>R</given-names></name> <name><surname>Han</surname> <given-names>K</given-names></name></person-group>. <article-title>Clinical features of COVID-19 in elderly patients: a comparison with young and middle-aged patients</article-title>. <source>J Infect</source>. (<year>2020</year>) <volume>80</volume>:<fpage>e14</fpage>&#x2013;<lpage>8</lpage>. doi: <pub-id pub-id-type="doi">10.1016/j.jinf.2020.03.005</pub-id>, PMID: <pub-id pub-id-type="pmid">32171866</pub-id></citation></ref>
<ref id="ref14"><label>14.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Jin</surname> <given-names>JM</given-names></name> <name><surname>Bai</surname> <given-names>P</given-names></name> <name><surname>He</surname> <given-names>W</given-names></name> <name><surname>Wu</surname> <given-names>F</given-names></name> <name><surname>Liu</surname> <given-names>XF</given-names></name> <name><surname>Han</surname> <given-names>DM</given-names></name> <etal/></person-group>. <article-title>Gender differences in patients with COVID-19: focus on severity and mortality</article-title>. <source>Front Public Health</source>. (<year>2020</year>) <volume>8</volume>:<fpage>152</fpage>. doi: <pub-id pub-id-type="doi">10.3389/fpubh.2020.00152</pub-id>, PMID: <pub-id pub-id-type="pmid">32411652</pub-id></citation></ref>
<ref id="ref15"><label>15.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Chen</surname> <given-names>T</given-names></name> <name><surname>Wu</surname> <given-names>D</given-names></name> <name><surname>Chen</surname> <given-names>H</given-names></name> <name><surname>Yan</surname> <given-names>W</given-names></name> <name><surname>Yang</surname> <given-names>D</given-names></name> <name><surname>Chen</surname> <given-names>G</given-names></name> <etal/></person-group>. <article-title>Clinical characteristics of 113 deceased patients with coronavirus disease 2019: retrospective study</article-title>. <source>BMJ</source>. (<year>2020</year>) <volume>368</volume>:<fpage>m1091</fpage>. doi: <pub-id pub-id-type="doi">10.1136/bmj.m1091</pub-id></citation></ref>
<ref id="ref16"><label>16.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Guan</surname> <given-names>WJ</given-names></name> <name><surname>Liang</surname> <given-names>WH</given-names></name> <name><surname>Zhao</surname> <given-names>Y</given-names></name> <name><surname>Liang</surname> <given-names>HR</given-names></name> <name><surname>Chen</surname> <given-names>ZS</given-names></name> <name><surname>Li</surname> <given-names>YM</given-names></name> <etal/></person-group>. <article-title>Comorbidity and its impact on 1590 patients with COVID-19 in China: a nationwide analysis</article-title>. <source>Eur Respir J</source>. (<year>2020</year>) <volume>55</volume>:<fpage>2000547</fpage>. doi: <pub-id pub-id-type="doi">10.1183/13993003.00547-2020</pub-id>, PMID: <pub-id pub-id-type="pmid">32217650</pub-id></citation></ref>
<ref id="ref17"><label>17.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Zhou</surname> <given-names>F</given-names></name> <name><surname>Yu</surname> <given-names>T</given-names></name> <name><surname>Du</surname> <given-names>R</given-names></name></person-group>. <article-title>Clinical course and risk factors for mortality of adult inpatients with COVID-19 in Wuhan, China: a retrospective cohort study</article-title>. <source>Lancet</source>. (<year>2020</year>) <volume>395</volume>:<fpage>1054</fpage>&#x2013;<lpage>62</lpage>. doi: <pub-id pub-id-type="doi">10.1016/S0140-6736(20)30566-3</pub-id>, PMID: <pub-id pub-id-type="pmid">32171076</pub-id></citation></ref>
<ref id="ref18"><label>18.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Tan</surname> <given-names>L</given-names></name> <name><surname>Wang</surname> <given-names>Q</given-names></name> <name><surname>Zhang</surname> <given-names>D</given-names></name> <name><surname>Ding</surname> <given-names>J</given-names></name> <name><surname>Huang</surname> <given-names>Q</given-names></name> <name><surname>Tang</surname> <given-names>YQ</given-names></name> <etal/></person-group>. <article-title>Lymphopenia predicts disease severity of COVID-19: a descriptive and predictive study</article-title>. <source>Signal Transduct Target Ther</source>. (<year>2020</year>) <volume>5</volume>:<fpage>33</fpage>. doi: <pub-id pub-id-type="doi">10.1038/s41392-020-0148-4</pub-id>, PMID: <pub-id pub-id-type="pmid">32296069</pub-id></citation></ref>
<ref id="ref19"><label>19.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Liu</surname> <given-names>F</given-names></name> <name><surname>Li</surname> <given-names>L</given-names></name> <name><surname>Xu</surname> <given-names>M</given-names></name> <name><surname>Wu</surname> <given-names>J</given-names></name> <name><surname>Luo</surname> <given-names>D</given-names></name> <name><surname>Zhu</surname> <given-names>YS</given-names></name> <etal/></person-group>. <article-title>Prognostic value of interleukin-6, C-reactive protein, and procalcitonin in patients with COVID-19</article-title>. <source>J Clin Virol</source>. (<year>2020</year>) <volume>127</volume>:<fpage>104370</fpage>. doi: <pub-id pub-id-type="doi">10.1016/j.jcv.2020.104370</pub-id>, PMID: <pub-id pub-id-type="pmid">32344321</pub-id></citation></ref>
<ref id="ref20"><label>20.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Liu</surname> <given-names>Y</given-names></name> <name><surname>Du</surname> <given-names>X</given-names></name> <name><surname>Chen</surname> <given-names>J</given-names></name></person-group>. <article-title>Neutrophil-to-lymphocyte ratio as an independent risk factor for mortality in hospitalized patients with COVID-19</article-title>. <source>J Infect</source>. (<year>2020</year>) <volume>81</volume>:<fpage>e6</fpage>&#x2013;<lpage>e12</lpage>. doi: <pub-id pub-id-type="doi">10.1016/j.jinf.2020.04.002</pub-id>, PMID: <pub-id pub-id-type="pmid">32283162</pub-id></citation></ref>
<ref id="ref21"><label>21.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Herold</surname> <given-names>T</given-names></name> <name><surname>Jurinovic</surname> <given-names>V</given-names></name> <name><surname>Arnreich</surname> <given-names>C</given-names></name> <name><surname>Lipworth</surname> <given-names>BJ</given-names></name> <name><surname>Hellmuth</surname> <given-names>JC</given-names></name> <name><surname>von Bergwelt-Baildon</surname> <given-names>M</given-names></name> <etal/></person-group>. <article-title>Elevated levels of IL-6 and CRP predict the need for mechanical ventilation in COVID-19</article-title>. <source>J Allergy Clin Immunol</source>. (<year>2020</year>) <volume>146</volume>:<fpage>128</fpage>&#x2013;<lpage>136.e4</lpage>. doi: <pub-id pub-id-type="doi">10.1016/j.jaci.2020.05.008</pub-id>, PMID: <pub-id pub-id-type="pmid">32425269</pub-id></citation></ref>
<ref id="ref22"><label>22.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Yang</surname> <given-names>R</given-names></name> <name><surname>Li</surname> <given-names>X</given-names></name> <name><surname>Liu</surname> <given-names>H</given-names></name> <name><surname>Zhen</surname> <given-names>Y</given-names></name> <name><surname>Zhang</surname> <given-names>X</given-names></name> <name><surname>Xiong</surname> <given-names>Q</given-names></name> <etal/></person-group>. <article-title>Chest CT severity score: an imaging tool for assessing severe COVID-19</article-title>. <source>Radiol Cardiothorac Imaging</source>. (<year>2020</year>) <volume>2</volume>:<fpage>e200047</fpage>. doi: <pub-id pub-id-type="doi">10.1148/ryct.2020200047</pub-id>, PMID: <pub-id pub-id-type="pmid">33778560</pub-id></citation></ref>
<ref id="ref23"><label>23.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>Wang</surname> <given-names>Y</given-names></name> <name><surname>Dong</surname> <given-names>C</given-names></name> <name><surname>Hu</surname> <given-names>Y</given-names></name> <name><surname>Li</surname> <given-names>C</given-names></name> <name><surname>Ren</surname> <given-names>Q</given-names></name> <name><surname>Zhang</surname> <given-names>X</given-names></name> <etal/></person-group>. <article-title>Temporal changes of CT findings in 90 patients with COVID-19 pneumonia: a longitudinal study</article-title>. <source>Radiology</source>. (<year>2020</year>) <volume>296</volume>:<fpage>E55</fpage>&#x2013;<lpage>64</lpage>. doi: <pub-id pub-id-type="doi">10.1148/radiol.2020200843</pub-id>, PMID: <pub-id pub-id-type="pmid">32191587</pub-id></citation></ref>
<ref id="ref24"><label>24.</label> <citation citation-type="journal"><person-group person-group-type="author"><name><surname>De Mol</surname> <given-names>C</given-names></name> <name><surname>De Vito</surname> <given-names>E</given-names></name> <name><surname>Rosasco</surname> <given-names>L</given-names></name></person-group>. <article-title>Elastic-net regularization in learning theory</article-title>. <source>J Complex</source>. (<year>2009</year>) <volume>25</volume>:<fpage>201</fpage>&#x2013;<lpage>30</lpage>. doi: <pub-id pub-id-type="doi">10.1016/j.jco.2009.01.002</pub-id></citation></ref>
<ref id="ref25"><label>25.</label> <citation citation-type="book"><person-group person-group-type="author"><name><surname>Hastie</surname> <given-names>T</given-names></name> <name><surname>Tibshirani</surname> <given-names>R</given-names></name> <name><surname>Friedman</surname> <given-names>J</given-names></name></person-group>. <source>The elements of statistical learning: Data mining, inference, and prediction</source>. <publisher-loc>New York</publisher-loc>: <publisher-name>Springer</publisher-name> (<year>2009</year>).</citation></ref>
<ref id="ref26"><label>26.</label> <citation citation-type="book"><person-group person-group-type="author"><collab id="coll2">World Health Organization</collab></person-group>. <source>Laboratory testing for coronavirus disease 2019 (COVID-19) in suspected human cases: Interim guidance, 2 march 2020</source>. <publisher-loc>Geneva</publisher-loc>: <publisher-name>World Health Organization</publisher-name> (<year>2020</year>).</citation></ref>
</ref-list>
</back>
</article>