Discussion
To control the worldwide pandemic of SARS-CoV-2 infection and efficiently allocate medical resources, the screening and management of asymptomatic or presymptomatic COVID-19 patients in the population are urgent challenges. This retrospective study systematically characterized the asymptomatic, presymptomatic, and symptomatic COVID-19 patients who were admitted to Wuhan Huoshenshan Hospital and developed a risk-stratification calculator that will help clinicians worldwide to make appropriate clinical decisions at the time of admission.
Asymptomatic and presymptomatic COVID-19 patients have similar CT findings, which indicates the urgent need for a risk-stratification model to differentiate these patients. Despite the absence of validation cohorts to confirm the performance accuracy of the differential diagnosis model, the performance of this risk-stratification model is satisfactory and accurate (based on AUCs >0.82 in both the development and validation cohorts). The package-based prediction calculator can be utilized by clinicians to perform risk stratification of asymptomatic, non-severe presymptomatic, and severe presymptomatic COVID-19 patients based on 5–10 commonly evaluated indicators. This makes the package-based calculator easy to utilize, facilitating clinical resource reallocation, such as for ICU beds and ventilators.
Furthermore, the risk-stratification model can be utilized for symptomatic and presymptomatic COVID-19 patients, which implies that the indicators identified in the analysis possibly have intrinsic associations with the development of COVID-19. A high BNP level is related to acute respiratory distress syndrome, sepsis, and congestive heart failure and contributes to higher mortality in patients with pneumonia (21, 22). PCT, as an indicator for bacterial coinfection, is approved by the US Food and Drug Administration to guide antibiotic treatment for suspected lower respiratory tract infections (23). Previously published literature reviews and meta-analyses have indicated that the results of laboratory tests for PCT could indicate progression of COVID-19 (8, 24, 25), which greatly supports our results. There is limited information about the frequency and bacterial spectrum of pulmonary coinfections and superinfections in COVID-19 patients (26); however, our analysis strengthens the vital role of BNP and PCT evaluation in the clinic to verify the possibility of coinfections associated with SARS-CoV-2 infection.
Functional CD8+ T-cell exhaustion contributes to the progression of SARS-CoV-2 infection (27, 28). As indicated in a previous report, loss of function occurs hierarchically in exhausted CD8+ T-cells (29). In the initial stage, CD8+ T-cells are unable to produce IL-2. During the intermediate stage, CD8+ T-cells may lose the ability to produce tumor necrosis factor (TNF). In the severe stage, CD8+ T-cells completely lose the ability to produce large amounts of IFNγ or there may be a physical deletion of virus-specific CD8+ T-cells. Based on the single-cell sequencing data, our re-analysis indicates that SARS-CoV-2 can induce intermediate-stage CD8+ T-cell exhaustion involving a lack of IL-2 and TNFα production but with limited IFNγ production (Fig. 5F), which may be mediated by LAG-3. Because LAG-3 cross-linking can inhibit IL-2, TNFα, and IFNγ production and T-cell proliferation (30). Therefore, LAG3+IFNγ+CD8+ T-cells may be the predominant phenotype showing exhaustion and may need to be further investigated to elucidate the pathogenic clinical severity of COVID-19.
As the retrieved and analyzed clinical data were based on all data from the clinical practice and were not derived from random selection, the data are deemed to be descriptive to some extent. For a comparable data analysis, baseline normalization by the propensity score matching method was performed, and this strategy helped reveal more information from these data. We were unable to verify the accuracy of the model due to lack of data from another research center, resulting in only partial validation of our risk-stratification model.