medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
1 Title Page
2
3 Clinical characteristics of COVID-19 and the model for predicting the
4 occurrence of critically ill patients: a retrospective cohort study.
5
6 Jing Ouyang#1, Xuefeng Shan#2, Xin Wang#3, Xue Zhang1, Yaling Chen1,
7 Miaomiao Qi1, Chao Xia1, Dongqing Gu3, Yaokai Chen*1, Ben Zhang*3
8
9 1. Chongqing Public Health Medical Center, Chongqing, China
10 2. Department of Pharmacy, The First Affiliated Hospital of Chongqing Medical
11 University, Chongqing, China
12 3. Department of Epidemiology and Biostatistics, First Affiliated Hospital, Army
13 Medical University, 30 Gaotanyan Street Shapingba District, Chongqing 400038,
14 China.
15
16 Corresponding Authors:
17 Yaokai Chen, MD, Ph.D., ([email protected]), Chongqing Public Health
18 Medical Center, 109 Baoyu road, Gele mountain, Chongqing 400038, China.
19
20 Ben Zhang, Ph.D., ([email protected]), Department of Epidemiology and
21 Biostatistics, First Affiliated Hospital, Army Medical University, 30 Gaotanyan Street
22 NOTE:Shapingba This preprint reports District, new research Chongqing that has 400038, not been certified China. by peer review and should not be used to guide clinical practice.
1 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
23 Abstract
24 Background: The present study aim to comprehensively report the epidemiological
25 and clinical characteristics of the COVID-19 patients and to develop a
26 multi-feature fusion model for predicting the critical ill probability.
27 Methods: It was a retrospective cohort study that incorporating the laboratory-
28 confirmed COVID-19 patients in the Chongqing Public Health Medical Center. The
29 prediction model was constructed with least absolute shrinkage and selection operator
30 (LASSO) logistic regression method and the model was further tested in the
31 validation cohort. The performance was evaluated by the receiver operating curve
32 (ROC), calibration curve and decision curve analysis (DCA).
33 Results: A total of 217 patients were included in the study. During the treatment, 34
34 patients were admitted to intensive care unit (ICU) and no developed death. A model
35 incorporating the demographic and clinical characteristics, imaging features and
36 laboratory findings were constructed to predict the critical ill probability and it was
37 proved to have good calibration, discrimination ability and clinic use.
38 Conclusions: The prevalence of critical ill was relatively high and the model may help
39 the clinicians to identify the patients with high risk for developing the critical ill, thus
40 to conduct timely and targeted treatment to reduce the mortality rate.
41
42 Keywords: COVID-19; Critical ill; Prevalence; Prediction model; Validation
43
44
2 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
45 1. Introduction
46 In December 2019, the Health Commission of Hubei province firstly reported a
47 group of unknown cause pneumonia patients relating to the South China seafood
48 market in Wuhan city [1]. The novel coronavirus, named severe acute respiratory
49 syndrome 2 (SARS-CoV-2) was quickly isolated and the virus-infected pneumonia
50 was later designated as coronavirus disease 2019 (COVID-19) by The World Health
51 Organization (WHO) [2]. Since then, the number of patients has increased rapidly,
52 with confirmed cases successively appearing in many Chinese provinces
53 (municipalities, autonomous regions and special administrative regions) and other 147
54 countries, areas or territories [3, 4]. As of August 1st, 2020, the globally reported
55 laboratory-confirmed cases were nearly 1,900 thousands and WHO has declared
56 COVID-19 as an international pandemic public health emergency[3, 5].
57 Up to now, a series of studies try to explain the clinical characteristics of
58 COVID-19 and the clinical spectrum was reported to range from asymptomatic
59 infection to severe pneumonia that should be admitted to the intensive care unit (ICU)
60 and even death [6-9]. With the emergence of the second or third-generation infection
61 of COVID-19, more clinical characteristics in other areas besides Wuhan need to be
62 reported. Chongqing has 40 districts under its jurisdiction, being the largest
63 municipality in China that adjacent to Wuhan. Until now, a total of 583 patients were
64 diagnosed in Chongqing, but the clinical spectrums were not reported before[4].
65 Most of the COVID-19 patients were reported to have mild symptoms and a
66 relatively better prognosis, while the critical ill patients hold poor prognosis and pose
3 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
67 high risk for death [9, 10]. There is no specialized medicine to cure COVID-19 and
68 the supportive care became the mainstay treatment regimen [11]. Therefore, to early
69 identify the patients with high risk for developing critical ill and provide targeted care
70 and treatment may prevent the patients from progression and reduce mortality risk,
71 however, the valid tools are lacking.
72 Thus, in the present study, we aim to describe the epidemiological and clinical
73 characteristics of COVID-19 patients who were diagnosed in the Chongqing city and
74 then to construct and validate a nomogram for predicting the risk for developing
75 critical ill patients, to help the clinicians have an early identification of the high risk
76 patients and tailor targeted treatment regimens and reduce the mortality risk.
77
78 2. Materials and Methods
79 2.1. Study design and Participants
80 It was a retrospective cohort study. All of the COVID-19 patients who were
81 diagnosed in the Chongqing Public Health Medical Center (Chongqing, China)
82 between 24th January and 16th February 2020 were included and the data cutoff for the
83 follow-up was 9th March, 2020. The patients who were laboratory diagnosed
84 according to WHO interim guidance in urban and surrounding areas of Chongqing
85 were admitted to this hospital without selectivity[12].
86 For the development of the model for predicting the critical ill patients, all of the
87 included patients were randomly splitting into two cohorts, namely the construction
88 cohort (70%) and validation cohort (30%). The study was approved by the
4 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
89 institutional review board of the Chongqing Public Health Medical Center
90 (2020-017-01-KY).
91
92 2.2. Data sources
93 The demographic and epidemiological characteristics such as age, gender, past
94 medical history, exposure history to Wuhan or the infected patients were collected by
95 one professional personnel using a structured face-to-face questionnaire. The clinical
96 characteristics such as the symptoms, imaging features, laboratory findings, treatment
97 regimens, comorbidities and outcomes (under treatment, discharged or dead) for all of
98 the patients were retrospectively reviewed and extracted from the electronic medical
99 record using a unified data collection form.
100
101 2.3. Factors and Outcome definition
102 The primary end point of the present study was the critical ill pneumonia, which
103 was defined as the COVID-19 patients who were admitted to the intensive care unit
104 (ICU) required mechanical ventilation or have ≥ 60% inspired oxygen (FiO2) [11].
105 Comorbidities of acute kidney injury was defined based on the highest serum
106 creatinine level and urine output[13]. The incubation period was defined as the
107 interval between the earliest date of exposure to the Wuhan or the contact of the
108 infected patients and the earliest date of symptom onset. For the patients who were
109 usually worked in Wuhan, the latest date of exposure was considered as the time of
110 infection[8]. Acute respiratory distress syndrome (ARDS) were defined according to
5 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
111 the WHO interim guidance[12]. Secondary infection was made basing on the
112 occurrence of symptoms or sign of hospital-acquired new pathogen infection after
113 admission [9], and blood sepsis was defined as life-threatening organ dysfunction
114 caused by infection according to the Third International Consensus Definition for
115 Sepsis and Septic Shock (Sepsis-3)[14]. Recurrent COVID-19 was defined as the
116 patients hospitalized again due to a positive test after discharge.
117
118 2.4. Statistical analysis
119 For the construction of the model for predicting the occurrence of the critical ill,
120 the features with p < 0.1 in the construction cohort were further analysed by using the
121 least absolute shrinkage and selection operator (LASSO) method for dimensionality
122 reduction and feature selection[15]. A nomogram was constructed basing on the
123 logistic regression analysis to provide a much more understandable measure. The
124 calibration ability of the nomogram was evaluated by the calibration curve. The
125 discrimination ability of the nomogram was assessed by the area under the receiver
126 operating curve (AUC). The difference in the AUC estimates was compared by
127 Delong non-parametric test[16]. The Decision curve analysis (DCA) was additionally
128 applied to calculate the net benefit by using the nomogram at different threshold
129 probabilities[17].
130 Statistical analyses were conducted using Statistical Package for the Social
131 Sciences (SPSS) version 23.0 software package for Windows (SPSS Inc), and R
132 version 3.4.1 (R Foundation for Statistical Computing, Vienna, Austria;
133 www.r-project.org). Statistically significant levels were two-tailed and set at P<0.05.
6 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
134 3. Results
135 3.1. Demographic and epidemiologic characteristics
136 A total of 217 patients were finally included in the current study, most of them
137 were lived Chongqing city (N=208), seven from Hubei province and two from
138 Sichuan province. (Figure 1) For these patients, 30.9% have exposure history to
139 Wuhan, and 48.8% have contact history to infected patients (Table 1, Figure 2A).
140 Median period for the incubation and time interval form illness onset to hospital were
141 7.00±7.3 days and 4.5±5.0 days, respectively. (Figure 2B) About half of the patients
142 have infected family members (N=102, 47.0%), and the median infected family
143 members was 1.0±2.0. (Figure 2C)
144 3.2. Clinical characteristics
145 As shown in Table 1, the cardiovascular disease was the most common
146 comorbidities for these patients (31.1%), followed by diabetes mellitus (10.6%), and
147 digestive disease (4.1%). At admission, the most common symptoms were cough
148 (64.5%) and fever (49.3%) Almost all of the patients have bilateral infiltration
149 pulmonary infection (95.4%). For the laboratory findings, the leukopenia and
150 lymphocytopenia occurred in 28.6% and 30.4% of the patients, respectively. Most of
151 the patients received interferon therapy (99.5%) and antiviral therapy (98.6%). During
152 the treatment, 32.7% presented liver injury, and 9.2% have acute respiratory distress
153 syndrome (ARDS). The median time interval between admission and discharge time
154 was 19.0±14.0 days. At the end of follow-up, six patients (4.3%) experienced
155 recurrent COVID-19 after discharge.
7 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
156 3.3. Prevalence of critical ill
157 A total of 34 critical ill patients (15.7%) were admitted to ICU. The prevalence
158 of critical ill was significantly increased with age and peaked at 61-70 years (33.3%;
159 95%CI: 19.5%-50.8%) then decreased. (Figure 2D) The prevalence in male and
160 females presented different trends with age, however no difference was found in the
161 total critical ill prevalence between them (18.8% for male, 12.9% for female, ꭓ
162 2=1.41; p=0.24). (Figure 2D)
163
164 3.4. Differences between the critical ill and other patients in construction cohort
165 A total of 159 participants with 25 critical ill patients were incorporated into the
166 construction cohort. Results suggested the critical ill patients presented high
167 proportion of advanced age, married or divorced status, diabetes mellitus, more
168 symptoms of fever, expectoration, short of breath and fatigue, imaging presentation of
169 ground-glass opacity, pleural thickening, pleural effusion and more patients with high
170 blood glucose, and lymphocytopenia. (Table 2)
171 3.5. The model for predicting critical ill probability
172 A total of 13 factors (two for demographic factors, six for clinical factors, three
173 for imaging features and two for laboratory findings) were finally selected. (Figure
174 3A-B) And then a nomogram was constructed for predicting the critical ill occurrence
175 probability. (Figure 3C)
176 The AUC for the predicting nomogram was 92.2% (95%CI: 87.8%-96.6%).
177 (Figure 3D) When comparing the performance among different prediction models, the
8 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
178 clinical factors model presented comparable performance with the full model (p=0.06)
179 and have significantly higher performance than demographic (p=0.01), imaging
180 features (p=0.03) and laboratory findings model (p=0.02). The calibration curve
181 suggested good performance of the nomogram. (Figure 4A) Decision curve analysis
182 also suggested with the threshold probability is over 20%, the application of the
183 nomogram for predict critical ill and conducted targeted treatment adds more benefit
184 than treating all or none of the patients. (Figure 4B)
185 When applying the nomogram in the validation cohort, the results showed
186 relatively good agreement between predicted and observed critical ill probability with
187 AUC of 87.3% (95%CI: 77.1%-97.5%), and no difference was found between the
188 construction and validation cohort (92.2% versus 87.3%, D=0.87; p=0.39). (Figure
189 4C-D)
190
191 4. Discussion
192 The present study comprehensively described the epidemiological and clinical
193 characteristics of COVID-19 in Chongqing. One third of the 217 patients have
194 exposure history to Wuhan and another half reported to have contact with infected
195 patients, which further confirmed the human to human transmission of the COVID-19
196 and indicated most of the included patient maybe the second or more generations of
197 infection. Moreover, the family cluster outbreak make up the main source of patients
198 in this study. About 47% of the patients were reported to have infected family
199 members and the median number of infected family member was 1 (maximum 7
9 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
200 members), which confirm the previous family cluster reports [18] and further
201 indicated the importance of family protection. Mean age of them were 46.5 years,
202 nearly half of them were male and teenagers were seldom infected, which were
203 similar to the findings from a national multicenter investigation[8]. However, the
204 incubation period of the present study was relatively longer than before [8, 19, 20].
205 In line with the previous reports, the present study also suggested fever and
206 cough were the most common symptoms at admission [6, 9, 10, 20]. Additionally we
207 found COVID-19 may present chilly and poor appetite, which was seldom reported
208 before. CT imaging feature also plays important roles on the identification of the
209 COVID-19 patients as it is reported that the CT imaging findings predate the RT-PCR
210 positive results [21, 22]. In this study, most of the patients present bilateral infiltration
211 with ground-glass opacity on the CT scan, which was in line with the previous studies
212 [6, 9, 10, 20]. Moreover, we also found pleural thickening in half of the patients and a
213 small fraction of them presented pleural effusion, which was seldom reported at this
214 time. In keeping with the finding in other studies, we also found part of the patients
215 presented with leukopenia, lymphocytopenia, anemia and thrombocytopenia; however,
216 due to the sample size and distribution of the clinical spectrum of the patients, the
217 proportions were not consistent [7-9, 10, 20]. For the lack of the specific treatment, all
218 of the patients underwent supportive care. A lot of marketed and unmarketed drugs
219 and Chinese herbs such as Remdesivir and Lianhua qingwen capsule are in clinical
220 trials; the curative effect will be revealed later and applied to the clinic as soon as
221 possible.
10 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
222 The results also showed the prevalence of critical ill was relatively smaller than
223 the previous studies reported by Huang (31.7%) [9], Zhou (26%) [10] and Wang
224 colleagues (26.1%)[6]. Moreover, all of the included patients in this study presented
225 relatively smaller proportion of comorbidities and death than them. The phenomenon
226 may be partly explained by the relatively smaller time interval from illness to
227 admission of the present study [6, 9, 10]. Thus early diagnosis and have timely
228 treatment of the infected patients might may benefit of the patients from high risk of
229 critical ill progress, comorbidities occurrence and even death. In addition, although
230 with low frequency, we also found 2.8% of the patients have recurrent COVID-19
231 after hospital discharge. It should be served as a reminder that patients who have been
232 discharged from the hospital need to continue to be isolated and protected to some
233 extent avoiding the recurrent and lead to transmission again.
234 Critical ill predispose patients high risk to death [10, 11]. Comparing with the
235 controls, the present study showed the critical ill patient presented advanced age, high
236 proportion of diabetes mellitus, short of breath, and lymphopenia, which was
237 consistent with Wang et al[6]. However, Huang and colleagues did not suggest this
238 difference, which may due to the relatively smaller sample size [9]. Additionally, we
239 also found the critical ill patients present higher proportion of imaging findings of
240 ground-glass opacity, pleural thickening and pleural effusion, which were seldom
241 reported. It could be partly explained by the report that the changes in the chest CT
242 imaging features were associated with the clinical manifestation from diagnosis to
243 recovery and may predict the disease progression[21].
11 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
244 Basing on the aforementioned factors, the present study comprehensively
245 developed a multi-feature fusion model to help the clinicians have early identification
246 of the critical ill patients. After evaluation, the model was proved to have good
247 internal and external performance and may provide important value for the medical
248 decision making. Among these included factors, the clinical features showed a
249 relatively better performance than the others, thus the symptoms and past disease
250 history were the primary information for evaluating the severity of the patients. The
251 demographic factors, imaging features and laboratory findings with comparable
252 predicting value, were good supplement for the clinical characteristics. However, due
253 to the limited sample size of patients, we cannot carry out clinical trials for validating
254 its clinical application in this study.
255 Our study has some limitations. Firstly, the sample size incorporated into the
256 research was relatively small, which may partly affect the statistic power of the results.
257 Secondly, not all of the laboratory tests were done in all of the included patients such
258 as D-dimer and proinflammatory cytokines, which were proved to play important
259 roles in the critical ill occurrence [9]. But with all of the included features, the
260 prediction model was proved to be have good performance and clinical use, thus they
261 may not significantly affect the results.
262 To sum up, this study comprehensively describes the clinical characteristics of
263 the COVID-19 patients and then to construct and validate a model for predicting the
264 occurrence of critical ill probability. The results may help the clinicians to have an
265 unrivalled understanding of the characteristics of COVID-19 patients and then to
12 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
266 stratified the patients into different risk subgroups for developing critical ill and tailor
267 targeted treatment regimens for the high risk patients and reduce the mortality rate.
268
269 Acknowledgments: This work was supported by National Natural Science
270 Foundation of China (81673255, 81874283, 81903398); Chongqing Special Research
271 Project for Prevention and Control of Novel Coronavirus Pneumonia
272 (No.cstc2020jscx-fyzx0074); The Recruitment Program for Young Professionals of
273 China (No number); Funds from Army Medical University and First Affiliated
274 Hospital of Army Medical University (2018XLC1004, SWH2018BJKJ-12).
275 Chongqing Natural Science Foundation Program (cstc2019jcyj-msxmX0466).
276
277 Author Contributions: Drs Jing Ouyang, Xuefeng Shan, and Xin Wang had full
278 access to all of the data in the study and take responsibility for the integrity of the data
279 and the accuracy of the data analysis. Drs Ben Zhang, Yaokai Chen contributed
280 equally as senior authors. Concept and design: Ben Zhang, Yaokai Chen; Acquisition,
281 analysis, or interpretation of data: Xuefeng Shan, Xin Wang, Jing Ouyang, Xue Zhang,
282 Miaomiao Qi, Chao Xia, Yaling Chen; Drafting of the manuscript: Xuefeng Shan, Xin
283 Wang, Jing Ouyang; Critical revision of the manuscript for important intellectual
284 content: Ben Zhang, Yaokai Chen; Statistical analysis: Xin Wang, Xue Zhang, Yaling
285 Chen; Supervision: Ben Zhang, Yaokai Chen.
286
287 Declaration of Competing Interest: The authors declare no conflict of interest.
13 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
288 References
289 1. Zhu N, Zhang D, Wang W, et al. A Novel Coronavirus from Patients with
290 Pneumonia in China, 2019. New England Journal of Medicine. 2020; 382(8):
291 727-733. Doi: 10.1056/NEJMoa2001017.
292 2. World Health Organization. WHO Director-General's remarks at the media briefing
293 on 2019-nCoV on 11 February 2020. (accessed 11 Feb 2020).
294 https://www.who.int/dg/speeches/detail/who-director-general-s-remarks-at-the-med
295 ia-briefing-on-2019-ncov-on-11-february-2020.
296 3. World Health Organization. WHO. Novel coronavirus (COVID-19) Situation.
297 https://www.who.int/emergencies/diseases/novel-coronavirus-2019. 2019 (accessed
298 25 May 2020)
299 4. Chinese Center for Disease Control and Prevention. Novel coronavirus (COVID-19)
300 Situation. http://2019ncov.chinacdc.cn/2019-nCoV/, 2019 (accessed 25 May 2020).
301 5. Center for Systems Science and Engineering. COVID-19 Dashboard by the CCSE
302 at Johns Hopkins University.
303 https://gisanddata.maps.arcgis.com/apps/opsdashboard/
304 index.html#/bda7594740fd40299423467b48e9ecf6. 2019 (accessed 25 May 2020)
305 6. Wang D, Hu B, Hu C, et al. Clinical Characteristics of 138 Hospitalized Patients
306 With 2019 Novel Coronavirus-Infected Pneumonia in Wuhan, China. JAMA. 2020;
307 323(11): 1061-1069. Doi: 10.1001/jama.2020.1585.
308 7. Chen N, Zhou M, Dong X, et al. Epidemiological and clinical characteristics of 99
309 cases of 2019 novel coronavirus pneumonia in Wuhan, China: a descriptive study.
310 Lancet. 2020; 395(10223): 507-513. Doi: 10.1016/S0140-6736(20)30211-7.
311 8. Guan WJ, Ni ZY, Hu Y, et al. Clinical Characteristics of Coronavirus Disease 2019
312 in China. New England Journal of Medicine. 2020; 382(18): 1708-1720. Doi:
313 10.1056/NEJMoa2002032.
314 9. Huang C, Wang Y, Li X, et al. Clinical features of patients infected with 2019 novel
315 coronavirus in Wuhan, China. Lancet. 2020; 395(10223): 497-506. Doi:
316 10.1016/S0140-6736(20)30183-5.
14 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
317 10. Zhou F, Yu T, Du R, et al. Clinical course and risk factors for mortality of adult
318 inpatients with COVID-19 in Wuhan, China: a retrospective cohort study. Lancet.
319 2020; 395(10229): 1054-1062. Doi: 10.1016/S0140-6736(20)30566-3.
320 11. Yang X, Yu Y, Xu J, et al. Clinical course and outcomes of critically ill patients
321 with SARS-CoV-2 pneumonia in Wuhan, China: a single-centered, retrospective,
322 observational study. Lancet Respir Med. 2020; 8(5): 475-481. Doi:
323 10.1016/S2213-2600(20)30079-5.
324 12. World Health Organization. Clinical management of severe acute respiratory
325 infection when novel coronavirus (nCoV) infection is suspected. 2020. (accessed
326 11 Jan 2020).
327 13. Kidney disease: improving global outcomes (KDIGO) acute kidney injury work
328 group. KDIGO clinical practice guideline for acute kidney injury. 2012.
329 https://kdigo.org/wp-content/uploads/2016/10/KDIGO-2012-AKI-Guideline-Engli
330 sh.pdf.
331 14. Singer M, Deutschman CS, Seymour CW, et al. The Third International
332 Consensus Definitions for Sepsis and Septic Shock (Sepsis-3). JAMA. 2016;
333 315(8): 801-810. Doi: 10.1001/jama.2016.0287.
334 15. Tibshirani R. Regression shrinkage and selection via the lasso: a retrospective.
335 Journal of the Royal Statistical Society. 2011; 73(3): 273-282.
336 16. DeLong ER, DeLong DM, Clarke-Pearson DL. Comparing the Areas under Two
337 or More Correlated Receiver Operating Characteristic Curves: A Nonparametric
338 Approach. Biometrics. 1988; 44(3): 837-845. Doi: 10.2307/2531595.
339 17. Steyerberg EW, Vickers AJ. Decision curve analysis: a discussion. Med Decis
340 Making. 2008; 28(1): 146-149. Doi: 10.1177/0272989X07312725.
341 18. Chan JF, Yuan S, Kok-KH, et al. A Familial Cluster of Pneumonia Associated
342 With the 2019 Novel Coronavirus Indicating Person-To-Person Transmission: A
343 Study of a Family Cluster. Lancet. 2020; 395(10223): 514-523. Doi:
344 10.1016/S0140-6736(20) 30154-9.
345 19. Li Q, Guan X, Wu P, et al. Early Transmission Dynamics in Wuhan, China, of
346 Novel Coronavirus-Infected Pneumonia. New England Journal of Medicine. 2020;
15 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
347 382(13): 1199-1207. Doi: 10.1056/NEJMoa2001316.
348 20. Xu XW, Wu XX, Jiang XG, et al. Clinical findings in a group of patients infected
349 with the 2019 novel coronavirus (SARS-Cov-2) outside of Wuhan, China:
350 retrospective case series. BMJ. 2020; 368: m606. Doi: 10.1136/bmj.m606.
351 21. Pan F, Ye T, Sun P, et al. Time Course of Lung Changes on Chest CT During
352 Recovery From 2019 Novel Coronavirus (COVID-19) Pneumonia. Radiology.
353 2020; 295(3): 715-721. Doi: 10.1148/radiol.2020200370.
354 22. Huang P, Liu T, Huang L, et al. Use of Chest CT in Combination with Negative
355 RT-PCR Assay for the 2019 Novel Coronavirus but High Clinical Suspicion.
356 Radiology. 2020; 295(1): 22-23. Doi: 10.1148/radiol.2020200330.
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
16 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
377 Tables.
378 Table1. Baseline demographic and clinical characteristics and outcomes in admitted to 379 Chongqing Public Health Hospital. Total Construction Validation Character cohort cohort cohort (N=217, 100%) (N=159, 70%) (N=58, 30%) Demographic characteristics Age (mean±SD, years) 46.5±16.0 45.6±16.6 49.1±14.0 Male 101(46.5) 73(45.9) 28(48.3) Married or divorce 189(87.1) 133(83.6) 56(96.6) Pregnant 0(0.0) 0(0.0) 0(0.0) Smoke 36(16.6) 28(17.6) 8(13.8) Drink 18(8.3) 12(7.5) 6(10.3) Past history Cardiovascular disease 68(31.3) 49(30.8) 19(32.8) Digestive disease 9(4.1) 5(3.1) 4(6.9) Nervous system disease 4(1.8) 3(1.9) 1(1.7) Cancer 1(0.5) 1(0.6) 0(0.0) Diabetes disease 23(10.6) 18(11.3) 5(8.6) Asthma 1(0.5) 0(0.0) 1(1.7) COPD 0(0.0) 0(0.0) 0(0.0) Clinical symptoms Fever (≥ 37.3℃) 107(49.3) 75(47.2) 32(55.2) Cough 140(64.5) 107(67.3) 33(56.9) Expectoration 75(34.6) 56(35.2) 19(32.8) Shortness of breath 29(13.4) 23(14.5) 6(10.3) Muscular soreness 26(12.0) 12(7.5) 14(24.1) Unconsciousness 0(0.0) 0(0.0) 0(0.0) Headache 23(10.6) 15(9.4) 8(13.8) Dizziness 23(10.6) 15(9.4) 8(13.8) Poor appetite 16(7.4) 11(6.9) 5(8.6) Pharyngeal pain and discomfort 29(13.4) 24(15.1) 5(8.6) Chest pain 3(1.4) 3(1.9) 0(0.0) Nausea and vomiting 7(3.2) 5(3.1) 2(3.4) Running nose 13(6.0) 9(5.7) 4(6.9) Diarrhea 19(8.8) 13(8.2) 6(10.3) Fatigue 48(22.1) 30(18.9) 18(31.0) Chilly 25(11.5) 16(10.1) 9(15.5) Nasal obstruction 8(3.7) 5(3.1) 3(5.2) Imaging feature Pulmonary infection Unilateral infiltration 8(3.7) 8(5.0) 0(0.0) Bilateral infiltration 207(95.4) 149(93.7) 58(100.0)
17 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
Ground-glass opacity 180(82.9) 128(80.5) 52(89.7) Pleural thickening 111(51.2) 78(49.1) 33(56.9) Pleural effusion 6(2.8) 4(2.5) 2(3.4) Laboratory findings Blood glucose (mmol/L) <6.1 117(53.9) 85(53.5) 32(55.2) 6.1-7.0 65(30.0) 47(29.6) 18(31.0) >7.0 35(16.1) 27(17.0) 8(13.8) White blood cell (Per109/L) <4 62(28.6) 41(25.8) 21(36.2) 4-19 148(68.2) 114(71.7) 34(58.6) >10 7(3.2) 4(2.5) 3(5.2) Lymphocyte count (Per109/L) <1.1 66(30.4) 46(28.9) 20(34.5) 1.1-3.2 149(68.7) 111(69.8) 38(65.5) >3.2 2(0.9) 2(1.3) 0(0.0) Platelets (Per109/L) <125 34(15.7) 24(15.1) 10(17.2) 125-350 173(79.7) 127(79.9) 46(79.3) >350 10(4.6) 8(5.0) 2(3.4) Hemoglobin (g/L) <130 82(37.8) 59(37.1) 23(39.7) 130-175 135(62.2) 100(62.9) 35(60.3) D-dimer (ug/L) ≤ 0.5 177(81.6) 128(80.5) 49(84.5) 0.5-1.0 30(13.8) 24(15.1) 6(10.3) >1.0 10(4.6) 7(4.4) 3(5.2) C-reactive protein (mg/L) 0.0-5.0 89(41.0) 67(42.1) 22(37.9) >5.0 128(59.0) 92(57.9) 36(62.1) Treatment Oxygen therapy 148(68.2) 110(69.2) 38(65.5) Antiviral therapy 214(98.6) 156(98.1) 58(100.0) Interferon therapy 216(99.5) 159(100.0) 57(98.3) Antibiotic therapy 50(23.0) 36(22.6) 14(24.1) Antifungal therapy 2(0.9) 1(0.6) 1(1.7) Glucocorticoid therapy 19(8.8) 10(6.3) 9(15.5) Immunoglobulin therapy 8(3.7) 6(3.8) 2(3.4) Chinese herbs treatment 114(52.5) 84(52.8) 30(51.7) Protect liver treatment 58(26.7) 46(28.9) 12(20.7) Nutrition therapy 15(6.9) 11(6.9) 4(6.9) Comorbidities and Outcomes Acute kidney injury 2(0.9) 0(0.0) 29(3.4) Liver injury 71(32.7) 52(32.7) 19(32.8)
18 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
ARDS 20(9.2) 14(8.8) 6(10.3) Secondary infection 0(0.0) 0(0.0) 0(0.0) Blood sepsis 5(2.3) 2(1.3) 3(5.2) Admitted to ICU 34(15.7) 25(15.7) 9(15.5) Admission to discharge time 19.0±14.0 19.0±14.0 19.5±14.0 (median± quartiles, days) Treatment outcome Under treatment 78(35.9) 61(38.4) 17(29.3) Hospital discharge 139(64.1) 98(61.6) 41(70.7) Death 0(0.0) 0(0.0) 0(0.0) Recurrent SARS-Cov-2 5 (2.3) 4(2.5) 1(1.7) 380 Abbreviations: COPD: chronic obstructive pulmonary disease; ARDS: Acute respiratory distress 381 syndrome; ICU: Intensive Care Unit.
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
19 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
398 Table 2: Difference in the demographic and clinical characteristics between severe and 399 non-severe pneumonia patients in the construction cohort. Critical ill Non-critical ill Characteristics ꭓ 2/Z p-value (N, %) (N, %) Demographic characteristics Age (years) 2.94 <0.01 ≤ 40 4(16.0) 59(44.0) 41-60 12(48.0) 53(39.6) >60 9(36.0) 22(16.4) Sex, (Male) 13(52.0) 60(44.8) 0.44 0.51 Marriage (Married or divorce) 25(100.0) 108(80.6) 5.80 0.02 Smoke 4(16.0) 24(17.9) 0.05 0.82 Drink 3(12.0) 9(6.7) 0.84 0.36 Exposure to Wuhan 9(37.5) 38(28.4) 0.81 0.37 Past history Cardiovascular disease 8(32.0) 41(30.6) 0.02 0.89 Digestive disease 1(4.0) 4(3.0) 0.07 0.79 Nervous system disease 0(0.0) 3(2.2) 0.57 0.45 Diabetes disease 8(32.0) 10(7.5) 12.64 <0.01 Clinical symptoms Fever (≥ 37.3℃) 18(72.0) 57(42.5) 7.34 0.01 Cough 21(84.0) 86(64.2) 3.76 0.05 Expectoration 18(72.0) 38(28.4) 17.59 <0.01 Shortness of breath 9(36.0) 14(10.4) 11.12 <0.01 Muscular soreness 0(0.0) 12(9.0) 2.42 0.12 Headache 3(12.0) 12(9.0) 0.23 0.63 Dizziness 2(8.0) 13(9.7) 0.07 0.79 Poor appetite 3(12.0) 8(6.0) 1.19 0.28 Pharyngeal pain and 2(8.0) 22(16.4) 1.17 0.28 discomfort Chest pain 0(0.0) 3(2.2) 0.57 0.45 Nausea and vomiting 1(4.0) 4(3.0) 0.07 0.79 Running nose 1(4.0) 8(6.0) 0.15 0.70 Diarrhea 0(0.0) 13(9.7) 2.64 0.10 Fatigue 9(36.0) 21(15.7) 5.69 0.02 Chilly 4(16.0) 12(9.0) 1.16 0.28 Nasal obstruction 0(0.0) 5(3.7) 0.96 0.33 Imaging features Pulmonary infection 1.60 0.21 Unilateral infiltration 0(0.0) 8(6.1) Bilateral infiltration 25(100.0) 124(93.9) Ground-glass opacity 25(100.0) 103(76.9) 7.18 0.01 Pleural thickening 20(80.0) 58(43.3) 11.37 <0.01 Pleural effusion 3(12.0) 1(0.7) 10.88 <0.01
20 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
Laboratory findings Blood glucose (mmol/L) 2.78 0.01 <6.1 8(32.0) 77(57.5) 6.1-7.0 8(32.0) 39(29.1) >7.0 9(36.0) 18(13.4) White blood cell (Per109/L) 0.59 0.56 <4 7(28.0) 34(25.4) 4-19 15(60.0) 99(73.9) >10 3(12.0) 1(0.7) Lymphocyte count (Per109/L) 3.72 <0.01 <1.1 15(60.0) 31(23.1) 1.1-3.2 10(40.0) 101(75.4) >3.2 0(0.0) 2(1.5) Plates (Per109/L) 0.17 0.87 <125 5(20.0) 19(14.2) 125-350 17(68.0) 110(82.1) >350 3(12.0) 5(3.7) Hemoglobin (g/L) 2.82 0.09 <130 13(52.0) 46(34.3) 130-175 12(48.0) 88(65.7) C-reactive protein (mg/L) 14.18 <0.01 0.0-5.0 2(8.0) 65(48.5) >5.0 23(92.0) 69(51.5) D-dimer (ug/L) 1.94 0.05 ≤ 0.5 17(68.0) 111(82.8) 0.5-1.0 4(16.0) 20(14.9) >1.0 4(16.0) 3(2.2)
400
401
402
403
404
405
406
407
408
21 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
409 Figures
410
411
412
413 Figure 1: Geographical distribution of the included COVID-19 patients in the
414 present study.
415
416
417
418
419
420
22 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
421 422
423 Figure 2: Distribution of the epidemiological characteristics of the included
424 COVID-19 patients. A: Proportion of exposure history of included patients; B:
425 Incubation and time interval between illness and hospital admission distribution; C:
426 Number of infected family member and nucleic acid test before admission of included
427 patients; D: The age and sex distribution of the critical ill patients.
428
429
430
431
432
433
434
23 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
435
436
437 Figure 3: Factors selection by the least absolute shrinkage and selection operator
438 (LASSO), the model construction and the comparison of the discrimination
439 among different models. A: The LASSO coefficient profile of the thirteen selected
440 factors; B: Selection of the tuning parameter (λ) in the LASSO model using 10-fold
441 cross-validation via binomial deviance minimization criteria. C: The nomogram for
442 predicting the probabilities of critical ill occurrence; D: The comparison of the AUC
443 among different models for predicting critical ill occurrence.
444
445
446
447
448
24 medRxiv preprint doi: https://doi.org/10.1101/2020.08.13.20173799; this version posted August 14, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license .
449
450
451
452 Figure 4: Validation of the performance, clinical use and transportability of the
453 nomogram. A: The calibration curve for the nomogram in the construction cohort; B:
454 Decision curve analysis of the net benefit by using the prediction nomogram in the
455 construction cohort; C: The calibration curve of the nomogram in the validation
456 cohort; D: The difference in the discrimination ability for nomogram between the
457 construction and validation cohort.
458 459
460
461
462
463
464
25