Clinical features and independent predictors for recurrence of positive SARS-CoV-2 RNA: A propensity score-matched analysis.

Patients with COVID-19 may be recurrence positive for severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) RNA after being cured and discharged from the hospital. The aim of this study was to explore independent influencing factors as markers for predicting positive SARS-CoV-2 RNA recurrence. The study included 601 COVID-19 patients who were cured and discharged from the Public and Health Clinic Centre of Chengdu from January 2020 to March 2021, and the recurrence positive of patients within 6 weeks after SARS-CoV-2 RNA turned negative was followed up. We used propensity score matching to eliminate the influence of confounding factors, and multivariate Logistic regression analysis was used to determine the independent influencing factors for positive SARS-CoV-2 RNA recurrence. Multivariate Logistic regression showed that the elevated serum potassium (odds ratio [OR] = 6.537, 95% confidence interval [CI]: 1.864-22.931, p = 0.003), elevated blood chlorine (OR = 1.169, 95% CI: 1.032-1.324, p = 0.014) and elevated CD3+ CD4+ count (OR = 1.003, 95% CI: 1.001-1.004, p < 0.001) were identified as independent risk factors for positive SARS-CoV-2 RNA recurrence (p < 0.05). The difference in virus shedding duration (OR = 1.049, 95% CI: 1.000-1.100, p = 0.05) was borderline statistically significant. For sensitivity analysis, we included virus shedding duration as a categorical variable in the model again and found that the OR value related to recurrence positively increased with delayed virus shedding duration, and the trend test showed a statistical difference (P trend = 0.03). Meanwhile, shortening of activated partial prothrombinase time (OR = 0.908, 95% CI: 0.824-1.000, p = 0.049) was identified as an independent protection factor for SARS-CoV-2 RNA recurrence positive. We have identified independent factors that affect the recurrence of SARS-CoV-2 RNA positive. It is recommended that doctors pay attention to these indicators when first admitted to the hospital.

INTRODUCTION

The novel coronavirus pneumonia (COVID‐19) caused by severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2) has become an international public health emergency since late 2019. According to COVID‐19 data published by the World Health Organization (https://www.who.int/data), as of May 2021, there have been over 170 million cumulative confirmed worldwide and over 3.8 million cumulative deaths. With the international epidemic prevention and control, most of the patients have been discharged after treatment. However, in some regions and countries, some discharged COVID‐19 patients have positive retest results for SARS‐CoV‐2 RNA, which has attracted the attention of researchers from all circles. , , , At present, the mechanism of recurrent positive SARS‑CoV‑2 RNA is not clear. A systematic review and meta‑analysis of 14 studies showed that the incidence of recurrent SARS‑CoV‑2 RNA positivity is estimated to be 14.8%, and patients with younger age and longer onset time have a greater risk of recurrent SARS‑CoV‑2 RNA positivity. The cause of recurrent SARS‑CoV‑2 RNA positivity may be related to false‐negative test results at the time of discharge, host status, virological factors, reinfection or immunosuppression, etc. ,  However, these possible reasons have not been confirmed by a large number of clinical studies. Studies have shown that SARS‐CoV‐2 RNA recurrence‐positive patients may have more severe immune suppression when they are admitted to the hospital. Patients with low levels of Immunoglobulin G and Immunoglobulin M (IgM) are more likely to face the risk of recurring SARS‐CoV‐2 RNA test results being positive. At the same time, another study confirmed that an increase in IgM antibodies against SARS‐CoV‐2 will be more conducive to clearing the virus. Moreover, the study by Wong et al. showed that patients with recurrent SARS‐CoV‐2 RNA positivity did not show COVID‐19 symptoms or signs during the entire quarantine period after discharge from the hospital. Meanwhile, in another study involving 123 recurrent SARS‐CoV‐2 RNA positivity patients, it was found that there were generally fewer related clinical symptoms during the recurrence period, which may be related to the higher proportion of asymptomatic or mild symptoms in patients at the first episode. Thus, the clinical manifestations of re‐positive patients have some degree of occult, and repeated reverse transcription‐polymerase chain reaction (RT‐PCR) detection during the follow‐up period is the most reliable diagnostic method. To date, there are still few clinical studies on the influencing factors of re‐positive patients. Existing studies mainly focus on the description of epidemiology and clinical characteristics, and lack of systematic studies on the influencing factors of relapse. Compared with similar studies, we used a retrospective clinical study to collect clinical data from as many patients as possible to identify the risk factors for SARS‐COV‐2 RNA recurrence‐positive patients. New findings on laboratory results and virus shedding duration provide insights into the underlying mechanism of SARS‐COV‐2 RNA repositivity.

METHODS

Study design and participants

In this study, a retrospective clinical study was conducted to analyze the cases of 692 patients with COVID‐19 who were cured and discharged from the Public and Health Clinic Centre of Chengdu from January 2020 to March 2021. A total of 601 patients were included in the analysis after excluding severe clinical data absence, death, and minor (age <18 years) patients (n = 91). Case data were collected, including demographic characteristics, clinical characteristics, laboratory results at admission, treatment plan, and clinical outcomes. All data were collected and checked independently by two physicians. According to the results of SARS‐CoV‐2 RNA recurrence in the follow‐up period, we divided the patients into the recurrent group (n = 99) and nonrecurrent group (n = 502). To eliminate the influence of confounding factors to the greatest extent, a 1:1 propensity score matching (PSM) was conducted for the two groups of patients according to the baseline data. After matching, 94 cases were in the recurrent group and 94 cases were in the nonrecurrent group. We analyzed the potential influencing factors associated with the occurrence of recurrent SARS‐CoV‐2 RNA positivity in this result. This study was approved by the Institutional Ethics Committee of the Public and Health Clinic Centre of Chengdu. As this study is a retrospective clinical study and only involves the extraction of medical records, the need for informed consent of patients is waived.

Statistical analysis

Continuous variables were presented as mean ± SD or median (interquartile range). Categorical variables were presented as numbers (proportion). We used Student's t‐test, Mann–Whitney U test, χ 2 test, or Fisher's exact test to compare the differences between the two groups before PS matching. Student's t‐test is used to test the significance of continuous variables that conform to the normal distribution between the two groups such as age, body mass index (BMI), etc., while Mann–Whitney U test is used to test the significance of continuous variables that do not conform to the normal distribution between the two groups. The χ 2 test or Fisher's exact test is used to test the significance of dichotomous variables between two groups, such as gender, clinical severity of disease. To eliminate the influence of confounders as much as possible, the variables with statistically significant differences in baseline data were used as independent predictors to calculate the propensity score. The caliper value was set to 0.01, and the 1:1 nearest neighbor matching method was used to match patients in the two groups separately. After PS matching, the differences between the two groups were compared by paired t‐tests, Wilcoxon rank sum test or the McNemar test. The paired t‐test is used to test the significance of continuous variables that conform to the normal distribution between the two groups such as age, BMI, etc., while the Wilcoxon rank sum test is used to test the significance of the continuous variables that do not conform to the normal distribution between the two groups. McNemar test is used to test the significance of dichotomous variables between two groups, such as gender, clinical severity of disease. The predictors of recurrent SARS‐CoV‐2 RNA positivity were determined by paired univariate and multivariate logistic regression models. We used a paired univariate logistic regression model to explore the variables related to recurrent SARS‐CoV‐2 RNA positivity, and calculated odds ratio (OR). Variance inflation factor (VIF) and tolerance were used to judge the multicollinearity among independent variables. To prevent over‐fitting, five variables with statistical differences (p < 0.05) in the univariate model were selected to be included in paired multivariate logistic regression analysis to determine independent predictors related to recurrence of positive SARS‐CoV‐2 RNA. Similarly, Elnaz Gholipour et al.  used this statistical analysis method in their study of COVID‐19 victims. Meanwhile, we used the Kaplan–Meier curve to evaluate the cumulative probability of positive recurrence in COVID‐19 patients. The Benjamini–Hochberg procedure was used to adjust for the false discovery rate for multiple testing. All statistical tests were two‐sided and p < 0.05 was considered statistically significant. All analyses were performed using SPSS software version 22.0.

Study definitions

The diagnostic basis, clinical classification and discharge criteria of COVID‐19 were determined according to the Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia issued by the National Health Commission of the People's Republic of China. , , All patients were confirmed as COVID‐19 by real‐time reverse transcription‐polymerase chain reaction (RT‐PCR) testing of respiratory specimens and chest computed tomography. The time for the first negative SARS‐CoV‐2 RNA was defined as the time for the first negative result when two consecutive RT‐PCR tests were negative for respiratory specimens (sampled at least 24 h apart). The recurrence of SARS‐CoV‐2 RNA positive was defined as a positive RT‐PCR result after two consecutive SARS‐CoV‐2 RNA negative respiratory specimens (sampled at least 24 h apart) from the patient. We defined the time interval from the first SARS‐CoV‐2 RNA positive to the first SARS‐CoV‐2 RNA negative as virus shedding duration. The clinical severity of disease was divided into severe and nonsevere. Severe cases include patients who are clinically classified as severe or critical, and nonsevere cases include mild or normal. Duration of antiviral therapy was defined as the time interval between initiation of antiviral therapy and cessation of antiviral therapy during the patient's hospitalization. Cardiovascular diseases included hypertension, coronary heart disease, atherosclerosis, congenital heart disease, and so on. Chronic lung diseases included tuberculosis, chronic obstructive pulmonary disease, pulmonary fibrosis, and so on. Chronic liver diseases included chronic hepatitis B, nonalcoholic fatty liver, alcoholic fatty liver, cirrhosis, and so on.

RESULTS

Demographic and clinical characteristics before and after PSM

A total of 692 patients with COVID‐19 admitted between February 2020 and March 2021 were included in this study. A total 601 patients were included in the analysis after severe clinical data absence, death, and minor (age <18 years) patients (n = 91) were excluded. A total of 99 patients (16.5%) showed recurrence positive after SARS‐CoV‐2 RNA turned negative, and 502 patients (83.5%) did not show positive RT‐PCR results again by the end of the follow‐up period. The baseline characteristics of 601 patients with COVID‐19 before PSM were listed in Table 1. The male ratio (69/99, 69.7%) and BMI (23.14 ± 3.24) in the recurrent group were lower than those in the nonrecurrent group (male: 408/502, 81.3%, BMI: 24.19 ± 3.90), the differences were statistically significant (p < 0.05). The mean age of patients with SARS‐CoV‐2 RNA recurrence positive was 42.23 ± 15.19 years old compared with 38.85 ± 12.09 years old in patients without recurrence, and the difference was statistically significant (p < 0.05). In addition, patients in the recurrent group were more likely to have COVID‐19‐related clinical symptoms at the time of a visit than patients in the nonrecurrent group (48/99, 48.5% vs. 163/502, 32.5%, p = 0.002). There were also more patients with SARS‐CoV‐2 RNA relapse‐positive who were diagnosed as severe (11/99, 11.1% vs. 20/502,4%, p = 0.003). The two groups of patients were statistically different in gender, age, BMI, clinical symptoms and severity of disease (p < 0.05), and there was no statistical difference in the other variable groups (p > 0.05). According to gender, age, BMI, clinical symptoms and severity of disease, we performed PSM between SARS‐CoV‐2 RNA recurrence‐positive patients and nonrecurrence‐positive patients according to the 1:1 nearest neighbor matching method, and 94 cases were successfully matched (Table 1). After PS matching, the covariates between the two groups were not statistically significant (p > 0.05). The 94 patients in the recurrent group all recurred positively within 42 days (6 weeks) after discharge after the first SARS‐CoV‐2 RNA turned negative. We used the Kaplan–Meier curve to assess the cumulative probability of positive recurrence in COVID‐19 patients (Figure 1). Five patients (7.45%) had recurrence‐positive results within 14 days after the first SARS‐CoV‐2 RNA negative, and 74 patients (39.36%) had recurrence‐positive results within 28 days after the first SARS‐CoV‐2 RNA negative.

Table 1

The demographic and clinical characteristics of patients with COVID‐19

Variables	Before PS matching				After PS matchinga
	Total (n = 601)	Recurrence (n = 99)	Nonrecurrence (n = 502)	p valueb	Total (n = 188)	Recurrence (n = 94)	Nonrecurrence (n = 94)	p valuec
Gender, n (%)				0.009d				0.349
Male	477 (79.4)	69 (69.7)	408 (81.3)		131 (69.7)	69 (73.4)	62 (66)
Female	124 (20.6)	30 (30.3)	94 (18.7)		57 (30.3)	25 (26.6)	32 (34)
Age (years)	39.41 ± 12.70	42.23 ± 15.19	38.85 ± 12.09	0.039d	39.81 ± 14.30	41.14 ± 14.44	38.49 ± 14.10	0.145
BMI (Kg/m2)	24.02 ± 3.82	23.14 ± 3.24	24.19 ± 3.90	0.013d	23.60 ± 3.56	23.19 ± 3.22	24.00 ± 3.85	0.077
Clinical symptoms, n (%)				0.002d				0.121
Yes	211 (35.1)	48 (48.5)	163 (32.5)		96 (51.1)	43 (45.7)	53 (56.4)
No	390 (64.9)	51 (51.5)	339 (67.5)		92 (48.9)	51 (54.3)	41 (43.6)
Clinical severity of disease, n (%)				0.003d				0.057
Severe	31 (5.2)	11 (11.1)	20 (4)		22 (11.7)	7 (7.4)	15 (16)
Nonsevere	570 (94.8)	88 (88.9)	482 (96)		166 (88.3)	87 (92.6)	79 (84)
Comorbidities, n (%)
None	258 (42.9)	47 (47.5)	211 (42.03)	0.317	91 (48.4)	45 (47.9)	46 (48.9)	1.000
Diabetes	42 (7)	9 (9.1)	33 (6.6)	0.369	14 (7.4)	7 (7.4)	7 (7.4)	1.000
Cardiovascular diseases	83 (13.8)	15 (15.2)	68 (13.5)	0.672	28 (14.9)	12 (12.8)	16 (17)	0.503
Chronic pulmonary disease	16 (2.7)	4 (4)	12 (2.4)	0.555	4 (2.1)	2 (2.1)	2 (2.1)	1.000
Chronic liver disease	179 (29.8)	27 (27.3)	152 (30.3)	0.550	47 (25)	26 (27.7)	21 (22.3)	0.500

Matched controls based on gender, age, BMI, symptoms, clinical severity of disease.

Calculated using the χ 2 test, Fisher's exact test, t‐test, or Mann–Whitney U test.

Calculated using paired t‐tests, Wilcoxon rank sum test, or the McNemar test.

Significant at p < 0.05.

Abbreviation: BMI, body mass index.

Figure 1

The cumulative probability of positive recurrence in COVID‐19 patients

Laboratory findings

The laboratory results of 188 COVID‐19 patients after PS matching are listed in Table 2. The lymphocyte count (1.87 ± 0.81 × 109/L), monocyte count (0.50 ± 0.19 × 109/L) and eosinophilic cell count (0.08 × 109/L [interquartile range {IQR}: 0.03–0.18]) in the recurrent group were all higher than those in the nonrecurrent group (lymphocyte count: 1.47 ± 0.69 × 109/L, monocyte count: 0.44 ± 0.18 × 109/L, eosinophilic cell count: 0.04 × 109/L [IQR: 0.01–0.10]), with statistical significance (p < 0.05). Compared with SARS‐CoV‐2 RNA without recurrence‐positive patients (neutral lymphatic ratio: 2.49 × 109/L [IQR: 1.69–3.86], C‐reactive protein: 4.41 mg/L [IQR: 0.80–10.70]), the neutral lymphatic ratio (1.68 × 109/L [IQR: 1.28–2.85]) and C‐reactive protein (1.13 mg/L [IQR: 0.50–5.27]) were reduced in recurrence‐positive patients. The difference was statistically significant (p < 0.05). Liver and renal function tests showed that the globulin of in the recurrent group was 28.10 g/L (IQR: 25.88–31.43) higher than that of in the nonrecurrent group, which was 26.20 g/L (IQR: 23.20–28.93). In the recurrent group, the albumin‐globulin ratio of 1.56 ± 0.36 and blood glucose of 5.31 mmol/L (IQR: 4.86–5.82) were both lower than those in the non‐recurrent group (albumin‐globulin ratio: 1.71 ± 0.34, blood glucose: 5.57 mmol/L [IQR: 4.99–6.77]), and the difference was statistically significant (p < 0.05). The serum potassium (4.04 ± 0.46 mmol/L) and the blood chlorine (106.82 ± 4.03 mmol/L) of SARS‐CoV‐2 RNA recurrence‐positive patients were higher than those of SARS‐CoV‐2 RNA nonrecurrence‐positive patients (serum potassium: 3.71 ± 0.48 mmol/L, blood chlorine: 104.73 ± 4.01 mmol/L), with the statistical difference (p < 0.05). Compared with the nonrecurrent group, patients in the recurrent group had lower lactate dehydrogenase (172.00 U/L [IQR: 149.00–202.25] vs. 186.00 U/L [IQR: 163.00–224.75], p = 0.026), hydroxybutyrate dehydrogenase (129.00 U/L [IQR: 114.50–153.25] vs. 140.50 U/L [IQR: 122.75–168.00], p = 0.021), activated partial thromboplastin time (27.95 s [IQR: 25.20–30.43] vs. 29.30 s [IQR: 27.40–31.33], p = 0.001) and fibrinogen (3.00 ± 1.04 g/L vs. 3.34 ± 1.16 g/L, p = 0.033). In addition, the thrombin time of patients in the recurrent group was 16.30 s (IQR: 15.00–17.33) longer than that of patients in the nonrecurrent group (15.25 s [IQR: 14.40–16.43]), with statistical significance (p < 0.05). The results of lymphocyte subsets showed that the CD3+ count (1323.5 ± 651.3 cells/μl), the CD3+CD4+ count (773.50 ± 381.68 cells/μl), the CD3+CD8+ count (464.29 ± 268.59 cells/μl), lymphocyte count (1824.61 ± 883.26 cells/μl) and lymphocyte percentage (22.07% [IQR: 14.49–28.06]) in patients with SARS‐CoV‐2 RNA relapse‐positive were all higher than those of patients with SARS‐CoV‐2 RNA nonrelapse‐positive (CD3+ count: 942.63 ± 507.08 cells/μl, CD3+CD4+ count: 535.18 ± 322.96 cells/μl, CD3+CD8+ count: 342.74 ± 209.96cells/μl, lymphocyte count: 1307.95 ± 677.35 cells/μl, Lymphocyte percentage: 15.88% [IQR: 10.62–23.69]), the differences were statistically significant (p < 0.05).

Table 2

Laboratory data of patients with COVID‐19 on admission to hospital

Variables	Normal range	Total (n = 188)	Recurrence (n = 94)	Nonrecurrence (n = 94)	p value
White blood cell count (×109/L)	3.5–9.5	5.95 (4.62–7.21)	6.23 (4.63–7.27)	5.82 (4.60–7.12)	0.538
Neutrophil count (×109/L)	2–7	3.56 (2.74–4.75)	3.42 (2.59–4.60)	3.67 (2.83–4.84)	0.212
Lymphocyte count (×109/L)	0.8–4	1.67 ± 0.77	1.87 ± 0.81	1.47 ± 0.69	<0.001a
NLR (×109/L)	–	2.09 (1.47–3.42)	1.68 (1.28–2.85)	2.49 (1.69–3.86)	0.002a
Monocyte count (×109/L)	0.12–1.2	0.47 ± 0.18	0.50 ± 0.19	0.44 ± 0.18	0.027a
Eosinophil count (×109/L)	0.02–0.5	0.07 (0.02–0.14)	0.08 (0.03–0.18)	0.04 (0.01–0.10)	0.001a
Hemoglobin (g/L)	110–160	141.4 ± 18.72	142.34 ± 18.34	140.47 ± 19.15	0.464
Platelet count (×109/L)	100–300	213.85 ± 68.04	219.94 ± 64.24	207.76 ± 71.45	0.155
MPV (fL)	7–11	10.78 ± 1.28	10.88 ± 1.26	10.69 ± 1.30	0.330
C‐reactive protein (mg/L)	0–5	1.98 (0.50–8.77)	1.13 (0.50–5.27)	4.41 (0.80–10.70)	0.001a
ALT (U/L)	0–37	23.00 (14.00–40.75)	22.00 (15.00–37.25)	25.00 (14.00–46.25)	0.186
AST (U/L)	0–37	22.00 (18.00–29.00)	21.00 (17.00–28.00)	24.00 (18.75–31.00)	0.063
ALP (U/L)	40–150	63.00 (55.00–79.75)	65.00 (55.00–83.25)	62.00 (54.00–76.25)	0.052
GGT (U/L)	0–50	20.00 (13.00–35.00)	20.00 (13.00–35.00)	21.50 (13.00–34.50)	0.864
Albumin (g/L)	35–55	43.66 ± 4.54	43.40 ± 4.53	43.91 ± 4.56	0.420
Globulin (g/L)	20–35	27.15 (24.55–30.00)	28.10 (25.88–31.43)	26.20 (23.20–28.93)	<0.001a
A/G	–	1.64 ± 0.36	1.56 ± 0.36	1.71 ± 0.34	0.003a
Total bilirubin (μmol/L)	0–20.5	8.45 (6.40–12.38)	8.80 (6.50–12.13)	8.35 (6.20–12.90)	0.772
blood urea nitrogen (mmol/L)	2–6.9	4.10 ± 2.31	4.14 ± 2.14	4.06 ± 2.48	0.638
Creatinine (μmol/L)	40–133	67.00 (56.00–74.78)	67.00 (56.75–76.00)	66.00 (55.00–72.00)	0.447
Blood glucose (mmol/L)	3.9–6.1	5.40 (4.93–6.38)	5.31 (4.86–5.82)	5.57 (4.99–6.77)	0.017a
Ca (mmol/L)	2.2–2.55	2.31 ± 0.15	2.32 ± 0.15	2.29 ± 0.15	0.211
Na (mmol/L)	135–155	141.30 (140.02–143.18)	141.45 (140.48–143.13)	141.30 (139.88–143.20)	0.409
K (mmol/L)	3.5–5.3	3.87 ± 0.49	4.04 ± 0.46	3.71 ± 0.48	<0.001a
Cl (mmol/L)	90–110	105.78 ± 4.14	106.82 ± 4.03	104.73 ± 4.01	<0.001a
Lactate dehydrogenase (U/L)	109–245	179.50 (154.00–217.50)	172.00 (149.00–202.25)	186.00 (163.00–224.75)	0.026a
HBDH (U/L)	72–182	135.50 (118.00–159.75)	129.00 (114.50–153.25)	140.50 (122.75–168.00)	0.021a
Creatine kinase (U/L)	25–196	84.50 (63.00–119.00)	84.50 (62.75–119.00)	84.50 (62.75–125.00)	0.948
Creatine kinase‐MB (U/L)	0–24	11.00 (9.00–14.00)	11.00 (9.00–15.00)	11.00 (9.00–13.00)	0.137
Prothrombin time (s)	10–14	13.25 ± 0.91	13.18 ± 0.87	13.31 ± 0.95	0.278
APTT (s)	22–38	28.70 (26.33–30.78)	27.95 (25.20–30.43)	29.30 (27.40–31.33)	0.001a
Fibrinogen (g/L)	2–4	3.17 ± 1.11	3.00 ± 1.04	3.34 ± 1.16	0.033a
thrombin time (s)	14–21	15.90 (14.70–17.00)	16.30 (15.00–17.33)	15.25 (14.40–16.43)	<0.001a
D‐dimer (μg/ml)	0–1	0.68 (0.56–0.86)	0.72 (0.58–0.92)	0.62 (0.53–0.83)	0.176
CD3+ count (cells/μl)	770–2041	1133.06 ± 612.62	1323.5 ± 651.3	942.63 ± 507.08	<0.001a
CD3+CD4+ count (cells/μl)	414–1123	654.34 ± 372.28	773.50 ± 381.68	535.18 ± 322.96	<0.001a
CD3+CD8+ count (cells/μl)	238–874	403.52 ± 248.02	464.29 ± 268.59	342.74 ± 209.96	<0.001a
CD3+ percentage	66–82	70.96 ± 10.35	69.81 ± 12.00	72.12 ± 8.29	0.113
CD3+CD4+ percentage	40–58	40.93 ± 7.99	41.46 ± 7.30	40.41 ± 8.63	0.386
CD3+CD8+ percentage	15–32	25.74 ± 7.06	24.93 ± 7.16	26.57 ± 6.90	0.125
Lymphocyte count (cells/μl)	–	1566.28 ± 826.59	1824.61 ± 883.26	1307.95 ± 677.35	<0.001a
Lymphocyte percentage	–	18.54 (12.50–26.32)	22.07 (14.49–28.06)	15.88 (10.62–23.69)	0.004a
CD4+/CD8+ ratio	–	1.75 ± 0.70	1.83 ± 0.69	1.68 ± 0.70	0.156

Abbreviations: A/G, albumin‐globulin ratio; ALP, alkaline phosphatase; APTT, activated partial thromboplastin time; AST, aspartate aminotransferase; Cl: blood chlorine; GGT, γ‐glutamyl transferase; HBDH, hydroxybutyrate dehydrogenase; K, serum potassium; LR, neutrophil count/lymphocyte count ratio; MPV, mean platelet volume; NALT, alanine aminotransferase.

Significant at p < 0.05.

The treatment regimen for patients with COVID‐19

The treatment of 188 COVID‐19 patients after PS matching were listed in Table 3. In both groups, 183 patients (97.3%) were treated with at least one antiviral drug during treatment. 120 patients (63.8%) received oxygen therapy. Of these, 10 patients (5.3%) received mechanical ventilation. The mean length of hospital stay of 188 patients was 18.97 ± 11.53 days, the mean virus shedding duration was 15.74 ± 11.30 days, and the mean duration of antiviral therapy was 14.49 ± 8.46 days. 44 patients (46.8%) in the recurrent group had been treated with oxygen, significantly less than 76 patients (80.9%) in the nonrecurrent group, and the difference between the two groups was statistically significant (p < 0.05). In addition, the average virus shedding duration (18.85 ± 12.51 days) and the average length of hospital stay (21.76 ± 13.29 days) in the recurrent group were higher than those in the nonrecurrent group (virus shedding duration 12.64 ± 9.00 days, length of hospital stay 16.19 ± 8.67), and the difference was statistically significant (Table 3).

Table 3

The treatment of COVID‐19

Variables	Total (n = 188)	Recurrence (n = 94)	Nonrecurrence (n = 94)	p value
Antiviral therapy, n (%)				0.375
Yes	183 (97.3)	90 (95.7)	93 (98.9)
No	5 (2.7)	4 (4.3)	1 (1.1)
Oxygen therapy, n (%)				<0.001a
Yes	120 (63.8)	44 (46.8)	76 (80.9)
No	68 (36.2)	50 (53.2)	18 (19.1)
Mechanical ventilation, n (%)	10 (5.3)	4 (4.3)	6 (6.4)	0.727
Length of hospital stay (day)	18.97 ± 11.53	21.76 ± 13.29	16.19 ± 8.67	0.001a
Virus shedding duration (day)	15.74 ± 11.30	18.85 ± 12.51	12.64 ± 9.00	<0.001a
Duration of antiviral therapy (day)	14.49 ± 8.46	14.38 ± 8.57	14.60 ± 8.39	0.852

Significant at p < 0.05.

Univariate and multivariate analysis of SARS‐CoV‐2 RNA recurrence in COVID‐19 patients

The results of univariate analysis are listed in Table 4. In univariate analysis, paired logistic regression model was used to analyze the relationship between laboratory tests, treatment regimen and positive events of SARS‐CoV‐2 RNA recurrence. Among them, there were no significant differences in neutral lymphocyte ratio, lactate dehydrogenase, hydroxybutyrate dehydrogenase, thrombin time and lymphocyte percentage between the two groups (p > 0.05). We used VIF and tolerance to judge the multicollinearity problem among independent variables. To simplify the model and prevent overfitting, after excluding the variables with multicollinearity, we decided to include the variables in the multivariate analysis by referring to the literature and discussing the model. Finally, the serum potassium, blood chlorine, activated partial thromboplastin time, CD3+CD4+ count and virus shedding duration were included in the multivariate model for adjustment. Paired multivariate logistic analysis showed that the serum potassium (OR = 6.537, 95% confidence interval [CI]: 1.864–22.931, p = 0.003), blood chlorine (OR = 1.169, 95% CI: 1.032–1.324, p = 0.014) and CD3+CD4+ count (OR = 1.003, 95% CI: 1.001–1.004, p < 0.001) was an independent risk factor for positive SARS‐CoV‐2 RNA recurrence (p < 0.05). The virus shedding duration (OR = 1.049, 95% CI: 1.000–1.100, p = 0.05) had borderline statistical significance with positive SARS‐CoV‐2 RNA recurrence (Figure 2). For sensitivity analysis, we included virus shedding duration as a categorical variable again in the analysis and conducted a trend test. The results (Table 5) showed that the OR value related to recurrence positive increased with delayed virus shedding duration, and this trend was statistically significant (P trend = 0.003). Meanwhile, in this model, shortening of activated partial prothromboplasminogen time (OR = 0.908, 95% CI: 0.824–1.000, p = 0.049) was considered to be an independent influencing factor associated with positive SARS‐CoV‐2 RNA recurrence, and the difference was statistically significant (p < 0.05).

Table 4

Univariate regression analysis for relevant factors of patients with recurrence of SARS‐CoV‐2 RNA positivity

Variables	Univariate OR	95% CI	p value
Lymphocyte count	3.333	1.817–6.114	<0.001a
NLR	0.915	0.836–1.001	0.054
Monocyte count	6.476	1.168–35.913	0.033a
Eosinophil count	184.014	5.493–6164.678	0.004a
C‐reactive protein	0.961	0.931–0.992	0.015a
Globulin	1.127	1.045–1.216	0.002a
A/G	0.262	0.100–0.685	0.006a
Blood glucose	0.760	0.609–0.949	0.015a
K	7.672	2.906–20.254	<0.001a
Cl	1.170	1.067–1.283	0.001a
Lactate dehydrogenase	0.997	0.992–1.001	0.115
HBDH	0.995	0.989–1.001	0.111
APTT	0.886	0.808–0.972	0.010a
Fibrinogen	0.742	0.559–0.984	0.039a
thrombin time	1.175	0.996–1.429	0.107
CD3+ count	1.002	1.001–1.003	<0.001a
CD3+CD4+ count	1.003	1.002–1.004	<0.001a
CD3+CD8+ count	1.003	1.001–1.005	0.001a
Lymphocyte count	1.001	1.001–1.002	<0.001a
Lymphocyte percentage	0.999	0.995–1.002	0.511
Oxygen therapy	0.179	0.080–0.401	<0.001a
Length of hospital stay	1.048	1.017–1.081	0.003a
virus shedding duration	1.058	1.024–1.093	0.001a

Abbreviations: A/G, albumin‐globulin ratio; APTT, activated partial thromboplastin time; CI, confidence interval; Cl, blood chlorine; HBDH, hydroxybutyrate dehydrogenase; K, serum potassium; NLR, neutrophil count/lymphocyte count ratio; OR, odd ratio; SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2.

Significant at p < 0.05.

Figure 2

Multivariate regression analysis for relevant factors of patients with recurrence of SARS‐CoV‐2 RNA positivity. aSignificant at p < 0.05. APTT, activated partial thromboplastin time; CI, confidence interval; CI, blood chlorine; K, serum potassium; OR, odd ratio

Table 5

Multivariate regression analysis for relevant factors of patients with recurrence of SARS‐CoV‐2 RNA positivity

Variables	Multivariate OR	95% CI	p value
K	7.010	1.955–25.137	0.003a
CL	1.165	1.027–1.321	0.017a
APTT	0.908	0.824–1.000	0.049a
CD3+CD4+ count	1.003	1.001–1.004	＜0.001a
virus shedding duration
≤14	Reference	–	0.265
15–28	1.458	0.448–4.744	0.531
>28	3.137	0.793–12.404	0.103
P trend	0.003a

Abbreviations: APTT, activated partial thromboplastin time; CI, confidence interval; Cl, blood chlorine; K, serum potassium; OR, odd ratio; SARS‐CoV‐2, severe acute respiratory syndrome coronavirus 2.

Significant at p < 0.05.

DISCUSSION

This study investigated the demographic and clinical characteristics, laboratory tests at admission, treatment options, and the recurrence of SARS‐CoV‐2 RNA positive of COVID‐19 patients who have been cured and discharged from the hospital. We used the method of PSM analysis to eliminate the influence of some confounding factors. At the same time, we also used logistic regression analysis to discuss the potential influencing factors related to the positive recurrence of RT‐PCR results. The potential independent influencing factors for positive SARS‐CoV‐2 RNA relapse identified in this study included increased CD3+CD4+ count, increased serum potassium and blood chlorine, shortened activated partial prothromboplastin time and delayed virus shedding duration. Among them, logistic regression analysis showed that elevated CD3+CD4+ counts, elevated blood potassium, elevated blood chlorine, and delayed nucleic acid shedding time were potential risk factors for the recurrence of SARS‐CoV‐2 RNA positive, and shortened activated partial prothrombin time was the potential protective factor.

To simplify the model and prevent overfitting, variables with collinearity were excluded through VIF and Tolerance. Considering the problem of paired sample size (n = 188), after reviewing the literature and discussing the model, we only included 5 variables with statistically significant differences between groups in univariate analysis for multivariate analysis. After adjustment for multivariate analysis, it was found that patients in the recurrent group had higher serum potassium and blood chlorine than those in the nonrecurrent group. Previous studies have noted that the disease course of COVID‐19 involves renal and gastrointestinal changes, and that fluid and electrolyte balance is controlled by the kidney and the gastrointestinal tract. Angiotensin‐converting enzyme 2 (ACE2) is a receptor for SARS‐CoV‐2 to enter host cells, and it is highly expressed in human tissues and organs. ACE2 was highly expressed in the brush border of proximal tubule cells and small intestinal epithelial cells, which enhanced the sensitivity of SARS‐CoV‐2 to enter kidney and gastrointestinal host cells and further caused renal and gastrointestinal damage. Another study also confirmed that electrolyte disturbance in COVID‐19 patients is a common renal complication, highlighted by hyperkalemia. From this point of view, the increase of serum potassium and blood chlorine in the recurrent group may be due to the COVID‐19‐related renal and gastrointestinal damage caused by SARS‐CoV‐2 entering the host cells, leading to the disturbance of fluid and electrolyte. However, there may be many other explanations waiting to be explored. Furthermore, lymphocyte subsets play a stabilizing role in immune function. In multivariate regression analysis, CD3+CD4+ count was identified as an independent risk factor for positive SARS‐CoV‐2 RNA recurrence. The possible explanation for this finding is that there may be an associated immune response disorder during novel coronavirus infection, and its changes are significantly correlated with the clinical characteristics and severity of COVID‐19 patients. , Different from our conclusion, another study on the risk factors of SARS‐CoV‐2 RNA relapse positive indicated that increased lymphocyte count was an independent risk factor for recurrence, but this study has not discussed the related indicators of lymphocyte subsets such as CD3+CD4+ count. As can be seen from our results, CD3+ count, CD3+CD4+ count, CD3+CD8+ count, lymphocyte count and CD4+/CD8+ ratio in patients with the recurrent group were all higher than those in the nonrecurrent group, suggesting that there may be some special abnormal changes in lymphocytes and their subsets in patients with COVID‐19 RNA re‐positivity. In our results, shortening of activated partial prothrombin time was identified as an independent protective factor for SARS‐CoV‐2 RNA recurrence. These results were similar to those of Wei Xu et al.'s dynamic study of coagulation parameters in COVID‐19 patients, which showed an association between APTT and disease severity in patients. Similarly, previous studies suggested that APTT presented abnormal waveforms in the analysis of clot waveform, and it was speculated that COVID‐19 may have special coagulation abnormalities.

Notably, the virus shedding duration (OR = 1.049, 95% CI: 1.000–1.100, p = 0.05) was found to be of borderline statistical significance in the results of multivariate analysis. To verify the results again and improve the sensitivity analysis, we included the virus shedding duration as a categorical variable in the multivariate analysis. The results showed that the OR value related to recurrence positive increased with delayed virus shedding duration, and the test of trend was statistically significant (P trend <0.05). This result indicates that with the delaying of virus shedding duration, the risk of SARS‐CoV‐2 RNA recurrence positive tends to increase. In previous studies, it was confirmed that older age, excessive 200 mg cumulative corticosteroid and treatment with Arbidol are associated with delayed COVID‐19 RNA shedding. , And our results suggest that COVID‐19 patients with delayed virus shedding duration may have a higher risk of positive recurrence. Consistent with the current research results, Hong et al.29 reported that the median virus shedding duration in SARS‐CoV‐2 re‐positive cases was 17 days at the initial hospitalization, which was significantly longer than the median time of 12 days in nonrepositive cases. The same result was also reported in the study of Ao et al.

The contribution of this study is that these potentially independent influencing factors may help clinicians identify patients at positive risk for SARS‐CoV‐2 RNA recurrence, thereby improving the prognostic management of patients. The main advantage of this study is the use of PSM to eliminate the influence of confounding factors. However, this study needs to acknowledge certain limitations. First, due to the nature of the retrospective study, the bias of the trial cannot be completely controlled. Prospective studies with a larger sample size cohort are warranted for further investigation. Second, single‐center clinical studies may have introduced selective bias into the results. Moreover, due to the relatively small sample size after PS matching, all the statistically significant variables in univariate analysis were not included in the multivariate analysis, which may result in that some potential influencing factors were not all found. Notwithstanding these limitations, this study did confirm some of the contributing factors for SARS‐CoV‐2 RNA recurrence positive, which could still provide some evidence for further research.

In summary, this study reports the independent factors influencing the recurrence of SARS‐CoV‐2 RNA positive in COVID‐19 patients. Studies have found that COVID‐19 patients with elevated CD3+CD4+ counts, elevated serum potassium, elevated blood chloride and delayed virus shedding duration on admission may have a high risk for recurrent SARS‐CoV‐2 RNA positivity. The shortened activated partial thromboplastin time may be a potential protective factor for recurrent SARS‐CoV‐2 RNA positivity in COVID‐19 patients during the follow‐up period. We recommend that these indicators be looked at during treatment.

CONFLICT OF INTERESTS

The authors declare that there are no conflict of interests.

AUTHOR CONTRIBUTIONS

Ke Liu: writing—original draft, writing—review and editing, Methodology. Xiuli Yang: writing—review and editing, writing—original draft, formal analysis. Chen Feng: Data curation. Mei Chen: formal analysis. Chuantao Zhang: supervision, resources, funding acquisition, methodology, writing—original draft. Yuelian Wang: data curation, funding acquisition, methodology.