There were 431 people been selected in this study in total. 228 adult normal people in these areas were invited to do a telephone interview, and 82.5% (188/228) of them answer the doctor's telephone questionnaire on lifestyle. A total of 203 infected adults were admitted to the study cohort. 80.7% (164/203) of them completely answered the questionnaire via telephone follow up. There were 23 asymptomatic infections and 141 symptomatic infections. In group of asymptomatic infections, 56.52% (13/23) of patients are male and the largest age group is 20-39 which accounted for 52.17%. While in group of symptomatic infections, 48.94% (69/141) of patients are male and the largest age group is 40-59 which accounted for 45.39%.
Characteristics of COVID-19 infects
According to the report of Diamond Princess and the retrospect of the transmission of asymptomatic infections in Anyang, China, asymptomatic infections has the ability to spread and the proportion of them was relatively high 12,22. In this cohort, 14.02% (23/164) people were asymptomatic, and 47.8% (11/23) had no signs of lung infection during hospitalization (Normal lung CT). It is not easy to identify asymptomatic infects that may lead to the presence of an invisible transmission chain. From Supplementary Table 1, there were no significant differences in gender, age, BMI classification, underlying disease between asymptomatic and symptomatic infections (P>0.05). Smoke history showed significant difference between groups and proportion of people with smoke history in asymptomatic group was higher than that in symptomatic group (P=0.008). Regarding the clinical symptoms, the asymptomatic group had less dyspnea on admission (P<0.001), higher counts of white blood cells (P=0.012) and a lower C-reactive protein level (P<0.001) that reflected inflammation. In terms of lifestyle, there was no significant difference in exercise habits, regular exercise, total MET*min, walking MET*min , MET intensity classification, sedentary behavior, and the average sleep time per day (P>0.05). The difference between the two groups were found in vigorous-level physical activity MET*min (P=0.035) and moderate-level physical activity activity*min(P=0.001).
Lifestyle may affect the probability of getting COVID-19
We investigated the lifestyles of the patient group and the non-patient group in the area where the patient group is located. And we tried to find whether these lifestyle-related factors affected the probability of disease. Gender, age, BMI, smoke history, underlying disease, exercise habit, sedentary status, physical activity and sleep status were incorporated into a univariate regression model to explore the influencing factors of getting infection of COVID-19. The details are showed in Supplementary Table 2. The variables whose univariate analyses yielded p-values <0.10 were included in the multivariable logistic regression model in Table 1. Considering the possibility of collinearity, sleep stime is only included in the multiple regression model instead of sleep status. History of smoking (P=0.001), having underlying disease (P<0.001), irregular exercise (P=0.004), sedentary population (P=0.010) and too much physical activity (P<0.001) are independent risk factors for getting the disease. The lifestyle of irregular exercise can increase the risk of illness by 2.929 times (95%CI: 1.425-6.022). And the lifestyle of sedentary population can increase the risk of illness by 19.168 times (95%CI: 2.044-179.792). Among the non-sick people, there were few sedentary people. Longer sleep time significantly protected people from disease (P<0.001). For physical activity, the moderate-intensity physical activity was a protective factor against COVID-19 compared to high-intensity physical activity (P<0.001). Relative low-level physical activity was even slightly better than high-intensity physical activity (P<0.001, OR 0.069 (95%CI: 0.026-0.186). Too much physical activity may reduce immunity and be susceptible to viruses.
Table 1. Results of multiple logistic regression on illness
Items
|
Categories
|
DF
|
Estimate
|
SE
|
Χ2
|
P
|
OR (95% CI)
|
Intercept
|
|
1
|
-1.816
|
0.683
|
7.070
|
0.008
|
|
Smoke History (No as reference)
|
Yes
|
1
|
1.159
|
0.351
|
10.868
|
0.001
|
3.185(1.600-6.343)
|
Underlying Disease (No as reference)
|
Yes
|
1
|
2.022
|
0.399
|
25.729
|
<0.001
|
7.552(3.458-16.495)
|
Regular Exercise (Yes as reference)
|
No
|
1
|
1.075
|
0.368
|
8.547
|
0.004
|
2.929(1.425-6.022)
|
Sedentary population (No as reference)
|
Yes
|
1
|
2.953
|
1.142
|
6.686
|
0.010
|
19.168(2.044-179.792)
|
MET Intensity Classification (High as reference)
|
Moderate
|
1
|
-1.546
|
0.298
|
26.985
|
<0.001
|
0.213(0.119-0.382)
|
|
Low
|
1
|
-2.668
|
0.504
|
28.019
|
<0.001
|
0.069(0.026-0.186)
|
- Significant level is 0.05 and significant P values are shown in bold.
- SE: Standard Error; OR: Odds Ratio; P: P value of multiple logistic regression
Physical activity intensity and sleep status can significantly affect the hospital stay length of all COVID-19 infects.
Result of univariate logistic regression (Supplementary Table 3) showed that smoke history, MET intensity classification, and sleep status were statistically significant (P<0.10). These variables and those with clinically significant variables in the study were then employed into the ordinal logit model. Result of the ordinal logit regression (Table 2) showed that MET intensity classification and sleep status had significant effects on the hospital stay. Taking the high MET intensity level as a reference, hospital stay would increase by 1.812 times (95% CI: 0.887-3.701) with no significance when the level is moderate (P>0.05) and significantly increase by 6.674 times (95% CI: 1.613-27.613) when the level is low (P=0.009). As for the sleep status, compared with the recommended, the risk of prolong hospital stay would increase by 2.287 times (95% CI: 0.951-5.502) with no statistically significant (P>0.05) if the sleep status was may be appropriate. While lack of sleep can significantly increase the risk of 5.525 times (95% CI: 1.284-23.770, P=0.022).
Table 2. Results of multiple logistic regression model on inpatient days
Variables
|
Categories
|
Estimate
|
SE
|
Χ2
|
P
|
OR (95% CI)
|
Intercept
|
Inpatient days ≥20
|
-1.802
|
0.291
|
38.316
|
<0.001
|
|
Intercept
|
Inpatient days between 10 and 19
|
1.778
|
0.295
|
36.382
|
<0.001
|
|
MET Intensity Classification
|
Moderate
|
0.594
|
0.365
|
2.656
|
0.103
|
1.812(0.887-3.701)
|
|
Low
|
1.898
|
0.725
|
6.864
|
0.009
|
6.674(1.613-27.613)
|
Sleep Status
|
Maybe Appropriate
|
0.827
|
0.448
|
3.415
|
0.065
|
2.287(0.951-5.502)
|
|
Lack of sleep
|
1.709
|
0.745
|
5.271
|
0.022
|
5.525(1.284-23.770)
|
- SE: Standard Error; OR: odds ratio; P: P value of X2 test.
- Score Test for the Proportional Odds Assumption: Χ2=0.3325, DF=2, P=0.8468, which suggested that the model satisfies the assumption.
- The reference for MET Intensity Classification and Sleep status are High and recommended respectively.
Moderate physical activity plays an important role in reducing the hospital stay length of all COVID-19 infects.
The classification of MET intensity is mainly based on the weekly frequency in addition to each MET * min, which is a comprehensive evaluation result 19. To look more closely at the impact of each physical activity category, the comparisons of vigorous activity MET*min, moderate activity MET*min, walking MET*min, and the sum of vigorous with moderate activity MET*min in groups of hospital stays were conducted using Kruskal-Wallis H test. The results in Table 3 showed moderate activity MET*min (P=0.015) and the sum of vigorous activity MET*min with moderate activity MET*min (P=0.025) had a significant influence on the length of hospital stay, which suggested that those who have a moderate physical activity level before being infected may recover faster from COVID-19 than others.
Table 3. Results of Kruskal-Wallis H test in groups of hospital stay
Variables
|
Hospital stay days
|
N
|
Mean Score
|
DF
|
H
|
P
|
Vigorous Activity MET*min
|
0-9
|
16
|
78.41
|
2
|
0.249
|
0.883
|
|
10-19
|
108
|
82.60
|
|
|
|
|
≥20
|
39
|
81.82
|
|
|
|
Moderate Activity MET*min
|
0-9
|
16
|
94.75
|
2
|
8.441
|
0.015
|
|
10-19
|
108
|
86.75
|
|
|
|
|
≥20
|
39
|
63.63
|
|
|
|
Walking MET*min
|
0-9
|
16
|
82.22
|
2
|
0.658
|
0.720
|
|
10-19
|
108
|
83.27
|
|
|
|
|
≥20
|
38
|
76.16
|
|
|
|
Vigorous & Moderate Activity MET*min
|
0-9
|
16
|
97.41
|
2
|
7.406
|
0.025
|
|
10-19
|
108
|
86.44
|
|
|
|
|
≥20
|
40
|
65.89
|
|
|
|
Significant level is 0.05 and significant P values are shown in bold. DF is degrees of freedom.
Physical activity intensity and sleep status can significantly affect the hospital stay length of symptomatic patients.
In the previous study of the entire cohort, we found that good sleep and moderate physical activity can affect the length of hospital stay. Since some studies have shown that the clinical manifestations and epidemiological characteristics of asymptomatic infects are elusive 23,24, and our cohort contains asymptomatic patients, we re-select symptomatic patients to conduct the analysis. The results of univariate logistic regression of each independent variable for symptomatic patients are in Supplementary Table 4. Based on the results of the ordinary logit model for symptomatic patients in Supplementary Table 5, the MET intensity classification and sleep status were also statistically significant (P<0.05). The inpatient days increased with the reduced physical activity intensity rating (P<0.05), which could increase by 2.289 times (95%CI: 1.051-4.983) and 11.370 times (95%CI: 1.969-65.644) for the moderate and low level of the MET intensity, respectively. Lack of sleep remained a significant risk factor (P<0.05), in which the OR was 4.816 (95% CI: 1.108-20.937) compared with the recommended sleep status. Supplementary Table 6 shows the results of Kruskal-Wallis H test of each physical activity intensity in groups of hospital stay in symptomatic patients. The same as the whole cohort, moderate activity MET*min (P=0.002) and the sum of vigorous with moderate activity MET*min were significant (P=0.009).