Clinical characteristics and predictors of pneumonia in patients presenting fever or respiratory symptoms with normal or low leukocyte counts: a retrospective study

Background Many patients went to the hospital presenting with acute fever, or respiratory symptoms, most of whom have a normal or low leukocyte counts. The aim of this study was to investigate the clinical characteristics and predictors of pneumonia in those patients. In this retrospective study, adult patients ( ≥ 18 years old) presenting with acute fever or respiratory symptoms with normal or low leukocyte counts ( ≤ 9.5 × 10 9 /L) in Peking Union Medical College Hospital between 26 January 2020 and 10 March 2020 were included. Patients were categorized into groups with pneumonia or upper respiratory tract infection (URTI) according to chest CT scans. Logistic regression was used to explore predictors of pneumonia.


Background
Fever or cough is the most common symptom of respiratory infections. Over 10 million patients visit the emergency department because of fever or cough in the United States per year. [1] Pneumonia is in the differential diagnosis of respiratory tract symptoms that are the most common cause of urgent emergency department visits, with high hospitalization and mortality rates worldwide. [2][3][4][5][6][7] The diagnosis of pneumonia usually depends on symptoms and chest imaging. [7,8] However, it is not feasible to obtain chest radiography in all patients admitted to the hospital who present with acute fever or respiratory symptoms. Since delays in correct diagnosis increase the risk of poor outcomes [9], it is crucial to identify clues that indicate pneumonia at the rst visit. Thus, physicians can differentiate pneumonia and ensure appropriate treatment and disposition at an earlier time.
Previous studies identi ed some predictors of pneumonia, such as higher temperature and higher C-reactive protein (CRP) level. In a cohort study of 28 883 adult participants with acute cough, 720 had a chest radiograph within the rst 7 days [10]. In total, 115 patients were radiographically diagnosed with pneumonia. The results showed that a temperature > 37.8 °C, crackles on auscultation, oxygen saturation < 95% and pulse > 100 beats per minute were predictors of pneumonia. [10] A recent meta-analysis evaluating the accuracy of biomarkers such as CRP and procalcitonin in patients with acute cough or suspected CAP showed that CRP was the preferred biomarker for the diagnosis of outpatient CAP with a positive likelihood ratio (+ LR) and a negative likelihood ratio (-LR) of 2.08 and 0.32, respectively, for a cutoff value of 20 mg/L; 3.64 and 0.36, respectively, for a cutoff value of 50 mg/L; and 5.89 and 0.47, respectively, for a cutoff value of 100 mg/L. [11] However, the best indicator of pneumonia in patients with normal or low leukocyte counts is still unknown.
With the emergence and spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) since December 2019, [12][13][14] several patients presenting with fever or respiratory symptoms have come to hospitals. To strictly control the spread of COVID-19 (coronavirus disease 2019), all patients with acute fever (axillary temperature ≥ 37.3 °C) or respiratory symptoms were required to go to the fever clinic of Peking Union Medical College Hospital. Additionally, patients with an epidemiological history of COVID-19 (de ned as: 1) a history of travel to or residence in Wuhan and its surrounding areas and communities with reported cases within 14 days before the onset of disease; 2) contact with suspected or con rmed COVID-19 patients within 14 days before disease onset; 3) contact with patients with fever and respiratory symptoms from Wuhan and its surrounding areas and communities with reported cases within 14 days before disease onset; and 5) clustering onset of disease) [15]were also required to go to the fever clinic, regardless of their temperature. Most of those patients had normal or low leukocyte counts, which may mislead physicians, resulting in the misdiagnosis of their condition as an upper respiratory tract infection (URTI). The aim of our study was to investigate the characteristics and predictors of pneumonia in patients with normal or low leukocyte counts.

Study design and setting
In this retrospective study, we screened patients who visited the fever clinic of Peking Union Medical College Hospital with acute fever or respiratory symptoms between 26 January 2020 and 10 March 2020. Adult patients (age ≥ 18) with normal or low leukocyte counts (white blood cell count ≤ 9.5 × 10 9 /L) con rmed by the laboratory were included. Demographic, clinical and laboratory data were extracted from electronic medical records. Patients were categorized into the pneumonia or URTI group according to chest CT scans.
Acute fever or respiratory symptoms were de ned as follows: 1) illness onset in the community within 2 weeks; 2) fever with axillary temperature ≥ 37.3 °C; 3) respiratory symptoms presenting as sore throat, cough, sputum, or shortness of breath; and 4) no other obvious cause.
Pneumonia was diagnosed in our study when the above criteria were met, and new pulmonary in ltrate was con rmed on chest CT.
Patients were excluded if 1) they were aged < 18 years; 2) they had been hospitalized during the last 28 days; 3) they had taken antibiotics before admission; and 4) they had respiratory symptoms were de nitely attributed to diseases such as tuberculosis, chronic obstructive pulmonary disease, asthma, interstitial lung disease, lung cancer, pulmonary embolism, and heart failure.

Data Collection
Demographic, clinical, laboratory data, treatment and outcome data for all patients were extracted from electronic medical records using formatted case record forms by physicians and checked by another researcher.
Complete blood cell counts and chest CT were performed in all patients. The results of routine blood examinations, including renal and liver function tests and the measurements of the levels of creatine kinase, lactate dehydrogenase, and hsCRP, were collected.
Pathogens in respiratory specimens were examined by real-time polymerase chain reaction (RT-PCR) for common respiratory viruses (such as in uenza and RSV) in most patients and in sputum culture for bacteria or fungi in a few patients. Blood serum was tested for Mycoplasma pneumoniae and Legionella pneumophila in patients who were suspected of having these diseases. Given the ongoing COVID-19 epidemic, all patients were tested for SARS-CoV-2 test using next-generation sequencing or RT-PCR methods.

Statistical analysis
The results are presented as the median (IQR) and n (%). The differences between the pneumonia and URTI groups were assessed with the Mann-Whitney U test, the χ² test, or Fisher's exact test, as appropriate. To explore the predictors of pneumonia, we used logistic regression, and variables signi cant at a two-tailed p value of < 0.001 in the univariate analysis were included in the multivariable regression model. A two-tailed p value < 0.05 was considered statistically signi cant. Statistical analyses were performed using SPSS software (version 19.0).
To determine the accurate cutoff values of predictors of pneumonia, receiver operating characteristic (ROC) curve analysis was performed with an online statistical calculator (www.medcalc.org).

Study population
A total of 3128 patients were screened at the fever clinic of Peking Union Medical College Hospital between 26 January 2020 and 10 March 2020. Forty-three percent (1345) presented with acute fever or respiratory symptoms. A total of 866 patients underwent both complete blood cell counts and chest CT scans, of whom 392 had normal or low leukocyte counts. A total of 197 were excluded because they met one of the exclusion criteria, as described in the methods. A total of 195 patients were included in the nal analysis. Of these patients, 32% (63) were diagnosed with pneumonia by chest CT (Fig. 1).

Clinical and Laboratory Characteristics
The demographic and clinical characteristics are provided in Table 1. The median age of all patients was 35 years (IQR 29-42), ranging from 18 years to 86 years, and females accounted for more than half (59%). Comorbidities were more commonly seen in patients with pneumonia than in those with URTIs. Compared with patients with URTIs, patients diagnosed with pneumonia had a lower pulse oxygen saturation level (98% (97-100%) versus 99% (98-100%), p < 0.001) and higher maximum body temperature (38.5 °C (38.0-38.8 °C) versus 37.5 °C (37.4-37.8 °C), p < 0.001), especially temperatures higher than 37.8 °C (p < 0.001). Eight patients were found to have pneumonia without fever in the study. The incidences of cough and sputum production were higher in patients with pneumonia than in those with URTIs (73% versus 32%, 41% versus 14%, p < 0.001). The median duration from fever onset to clinic visit was 2 days, with a signi cant difference between the two groups. Similarly, the duration from illness onset to clinic visit was longer in the pneumonia group than in the URTI group (1 (1-4) versus 3 (2-4) days, p < 0.001). The median white blood cell count was 6.45 × 10 9 /L (IQR 5.39-7.86) in all patients ( Table 1). The lymphocyte count and albumin level were lower in patients with pneumonia than in those with URTIs, whereas the neutrophil-to-lymphocyte ratio Pathogens were identi ed in 19 patients (Fig. 2), and 19% (12 patients) of the pneumonia group had identi able pathogens. SARS-Cov-2 was tested for in the respiratory specimens of all patients by using next-generation sequencing or RT-PCR methods; 7 patients with pneumonia and 1 patient with a URTI tested positive. PCR was performed on nasopharyngeal swabs from 171 patients to detect in uenza A or B virus and respiratory syncytial virus (RSV). Twenty-three patients underwent antigen testing for in uenza A or B virus. The results showed that 7 patients had common viral infections (2 patients with in uenza A and 1 with in uenza B in each group and 1 with RSV in the URTI group, Fig. 2). Sputum culture was performed for only one patient, revealing K. pneumonia. None of the patients had coinfections with multiple pathogens.

Treatment and Outcome
Ninety-one patients initially received antibiotics (38 patients in the URTI group, 53 in the pneumonia group), and 41 underwent antiviral treatment (20 patients in the URTI group, 21 in the pneumonia group). Nine patients were transferred to the designated COVID-19 hospital (8 with con rmed cases and 1 with a strongly suspected case). Seven patients with pneumonia were admitted to our hospital. Mechanical ventilation was implemented in one of these inpatients. All patients with pneumonia in this study survived.

Predictors of Pneumonia
We used logistic regression analysis to explore the predictive values of pneumonia (Table 2). In univariable analysis, age, pulse oxygen saturation level, maximum body temperature, cough, sputum production, duration from illness onset to visit, lymphocyte count, NLR, albumin level, lactate dehydrogenase level, and hsCRP level showed an increased association with pneumonia. We chose variables with p < 0.001 in the univariable analysis for inclusion in the multivariable regression model to investigate predictors of pneumonia. Our study found that older age ( (Fig. 3). A cutoff value of 37.8 °C was predictive for pneumonia, with a sensitivity of 81.8% and speci city of 75.8%. A hsCRP level of 1.64 mg/L had a negative likelihood ratio of 0.07, with a sensitivity of 95.0% and a speci city of 70.4% (Table 3). OR = odds ratio, CI = con dence interval, NLR = neutrophil-to-lymphocyte ratio, hsCRP = high sensitivity C-reactive protein. AUC = area under the receiver operating characteristic curve, +LR = positive likelihood ratio,-LR = negative likelihood ratio, hsCRP = high sensitivity C-reactive protein.

Discussion
The rapid and accurate diagnosis of patients with respiratory symptoms is a common challenge for outpatient clinicians. When patients present with fever or respiratory symptoms and a normal leukocyte count, the diagnosis of a URTI is usually considered rst. Our study revealed that 32% of those patients were ultimately con rmed to have pneumonia by chest CT. Therefore, it is necessary for physicians to appropriately order chest radiography and even chest CT. In a multicenter prospective study, researchers found that 3% of patients were found to have pneumonia on CT but not on concurrent chest radiography. [16] These patients had similar clinical characteristics to those with pneumonia on chest radiography, [16] emphasizing the importance of chest CT.
It is crucial for physicians to recognize pneumonia in patients who present with fever or respiratory symptoms and order a chest CT appropriately, especially during the epidemic of COVID-19. To our knowledge, this is the rst study to explore the predictors and calculate the associated cutoff values for the identi cation of pneumonia in patients with normal or low leukocyte counts. There were 63 patients diagnosed with pneumonia among 195 patients in our study. Older age, cough, higher temperature and higher hsCRP level were found to be associated with pneumonia.
The median age of patients with pneumonia was 40 years (IQR 34-60) in this study, ranging from 20 years to 86 years. With increasing age, the risk for pneumonia increased. As a previous study revealed that advanced age is one of the independent prognostic predictors of mortality in patients with CAP, [17,18] it was also associated with readmission [19].
Cough is a common symptom in patients presenting at the emergency department [1] and represents infectious or noninfectious lung disease. It was identi ed in 73% of the patients with pneumonia and 32% of the patients with URTI in this study. Pneumonia caused by some pathogens is characterized by a dry cough and even the absence of a fever [8]. Therefore, even without fever, cough was an indication of pneumonia.
It had been demonstrated in a previous study that temperature is a good predictor of pneumonia. [10] In our study, the median maximum body temperature was 37.5 °C (IQR 37.4-37.8 °C) in the group of patients with URTI and 38.5 °C (IQR 38.0-38.8 °C) in patients with pneumonia (p < 0.001). There were 8 patients diagnosed with pneumonia without a fever. The etiology in these 8 patients was SARS-CoV-2 in 4 patients and K. pneumonia in one patient, leaving 3 patients in whom no pathogen was detected. Eighty-seven percent patients experienced a temperature higher than 37.3 °C. We used a cutoff value of 37.8 °C based on the ROC analysis to predict pneumonia. This was similar to a prospective cohort study conducted by Moore M et al. [10]. Therefore, the diagnosis of pneumonia and more examinations should be considered once the temperature exceeds 37.8 °C.
CRP is an acute-phase protein that indicates acute in ammatory responses. It was identi ed as a predictor of pneumonia in several studies. [11,[20][21][22] A CRP level ≤ 10 mg/L was useful for ruling out CAP in most patients. [11]Steurer et al found there was no pneumonia diagnosed when the CRP levels were < 10 mg/L or if the CRP levels were between 11 and 50 mg/L and the patient did not complain of dyspnea and fever associated with cough. [23] However, some researchers found that the CRP level did not improve the prediction of pneumonia; instead, it can help clinicians make a decision regarding the use of antibiotics. [24,25] In contrast to previous studies [11,20], we found that a hsCRP level of 1.64 mg/L had a negative likelihood ratio of 0.07 in the ROC analysis, which means it was a strong predictor of the absence of pneumonia in patients with normal or low leukocyte counts. Further investigations showed that three patients had a hsCRP level less than 1.64 mg/L in the pneumonia group, two of whom were diagnosed with COVID-19 according to the Chinese management guidelines (version 7.0) [15]. Most cases with pneumonia had a higher hsCRP.
The overall detection rate of pathogens in this study was 10% in all patients and 19% in patients with pneumonia, which was lower than that previously reported [26,27]. A population-based study showed the detection of pathogens in 853 (38%) patients, of which viruses accounted for 22%. [26] In adults, viruses, particularly in uenza, rhinovirus, and coronavirus, cause one-third of cases of pneumonia. [28] In this study, the prevalence of viral infection was 14% in patients with pneumonia, with SARS-Cov-2 being the most common virus, followed by in uenza virus. One important reason was that all patients screened were outpatients, and few of them had the opportunity to undergo further etiology detection. It is worth noting that the COVID-19 pandemic is still a crisis. [29,30]A prospective study is needed with more examinations of pathogens in the future.
There are several limitations of our study. First, due to the retrospective study design, there may be some selection bias. The interpretation of our ndings might be limited by the study design and sample size. Second, there were no data about physical examinations owing to the thick protective suits worn by medical staff, which may have resulted in missing risk factors such as crackles that were observed in a previous study [10]. Third, this study was performed in a single-center setting, and patients were included during the epidemic of COVID-19, which may have affected the results. However, by including adult patients from the fever clinic of a Class A tertiary comprehensive hospital in China, we believe our study population is representative of cases presenting with acute fever and respiratory symptoms.

Conclusions
Nearly one-third of the patients who presented with acute fever or respiratory symptoms with normal or low leukocyte counts were con rmed to have pneumonia. The predictive values of pneumonia were older age, cough, higher temperature and higher hsCRP level. A temperature higher than 37.8 °C was a predictor of pneumonia, and chest radiography or CT is recommended. However, a hsCRP level less than 1.64 mg/L may help clinicians rule out most cases of pneumonia. For patients with a de nite exposure history of COVID-19, both the detection of SARS-CoV-2 and chest CT are suggested, regardless of body temperature and hsCRP level.
Abbreviations CAP community-acquired pneumonia CI con dence interval COVID-19 coronavirus disease 2019 CT computed tomography hsCRP high-sensitivity C-reactive protein IQR interquartile range LR Figure 1 Flow chart of patient selection