Risk Factors For COVID-19 Positivity in Hospitalized Patients in A Low Prevalence Setting

Background: Identifying hospitalized patients with Coronavirus disease 2019 (COVID-19) in a low prevalence setting is challenging. We aimed to identify differences between COVID-19 positive and negative patients. Methods: Hospitalized patients with respiratory illness, or fever, were isolated in the emergency room and tested for COVID-19. Patients with a negative PCR and low probability for COVID-19 were taken out of isolation. Patients with a higher probability for COVID-19 remained in isolation during hospitalization and were retested after 48 hours. Risk factors for COVID-19 were assessed using logistic regression. Results: 254 patients were included, 37 COVID-19-positive (14.6%) and 217 COVID-19-negative (85.4%). Median age was 76 years, 52% were males. In a multivariate regression model, variables signicantly associated with COVID-19 positivity were exposure to a conrmed COVID-19 case, length of symptoms before testing, bilateral and peripheral inltrates in chest X-ray, neutrophil count within the normal range, and elevated LDH. In an analysis including only patients with pneumonia (N=78, 18 positive for COVID-19), only bilateral and peripheral inltrates, normal neutrophil count and elevated LDH were associated with COVID-19 positivity. Conclusions: The clinical presentation of COVID-19 positive and negative patients is similar, but radiographic and laboratory features may help to identify COVID-19 positive patients and to initiate quick decisions regarding isolation.


Introduction
In December 2019, a novel coronavirus, severe acute respiratory syndrome related to coronavirus 2 (SARS-CoV-2), emerged in Wuhan, Hubei, China, causing severe pneumonia [1]. Since then, corona virus disease 2019 (COVID-19) became prevalent worldwide and in 11th March 2020, the World Health Organization declared a pandemic caused by this virus. As of May 30 th, 2020, more than 5.9 million people were infected worldwide, and more than 360,000 people have died. In Israel, 17,000 patients were diagnosed with COVID-19, and 284 patients died.
COVID-19 has a wide range of severity, and most patients have a mild illness and a good prognosis some require hospitalization. Usually at the beginning of the illness patients present with symptoms compatible with a u like illness or an upper respiratory tract infection (URTI) and as disease progresses after several days about 20% of patients will develop frank pneumonia. Few will have additional complications including acute respiratory syndrome (ARDS), high levels of cytokines such as Interleukin-6 (IL-6), thromboembolic events, and multiorgan failure. The overall reported mortality is in the range of 1-10% [2][3][4].
The clinical presentation of COVID-19 is not speci c and can mimic other viral infections. Symptoms such as fever, cough, malaise, and frank pneumonia are common for most viral respiratory tract infections including In uenza, Respiratory Syncytial Virus (RSV), Human Metapneumovirus (HMP) and many others [5]. This poses challenges to correct identi cation of COVID-19 patients. Laboratory diagnosis of SARS-CoV-2 is performed by real-time polymerase chain reaction (RT-PCR) on nasopharyngeal samples. Diagnosis is complicated by false-negative results that may reach 30% and false-positive results in the range of 1-2% [6][7][8].
COVID-19 is a highly contagious disease. Transmission to other persons occurs primarily by respiratory secretions through droplets, although other modes of transmission have been described. These include aerosol formations in patients on mechanical ventilation and transmission by contact with contaminated surfaces [9][10]. Isolation of patients with COVID-19 and treating these patients with appropriate personal protection equipment is important for reducing COVID-19 transmission.
Nonspeci c clinical presentation and di culties in interpreting PCR results pose challenges in differentiating between patients with and without COVID-19 for prompt isolation. This is particularly di cult in areas where the incidence of COVID-19 is low to moderate, as is the current situation in Israel. In places with low background transmission most patients with respiratory symptoms attending emergency rooms and patients hospitalized due to respiratory illness have other diagnoses. To date, there is limited data regarding the differences between hospitalized patients with con rmed COVID-19 and those who eventually were considered not to have COVID19 [11][12].
Wolfson Medical Center is a 700-bed hospital located in central Israel. During the rst wave of the corona pandemic in Israel our hospital catered for approximately 100 COVID-19 patients. Many of these patients were referred for hospitalization after they were diagnosed in the community. In this study, we investigated the epidemiologic, clinical, radiographic and laboratory features of con rmed COVID-19 cases diagnosed upon admission to Wolfson Medical Center, or during their hospitalization, and compared them to hospitalized patients that were suspected of having COVID-19 but eventually were considered to have other diagnoses.

Patients
The study was performed in Wolfson Medical Center and approved by the hospitals Internal Review Board. We included all adult patients, who were hospitalized between March 15 th 2020 and May 15 th 2020, and had con rmed COVID-19 diagnosed at admission or during hospitalization, and a convenient sample of patients who were suspected to have COVID-19 but this diagnosis was ruled out. In some cases, the question of COVID-19 infection was raised only several days after admission. For these patients, the data was collected at the day the rst SARS-CoV-2 test was performed (index date).
Suspected cases were tested for SARS-CoV-2 by RT-PCR at presentation. If the result was negative, cases were reviewed by a physician from the Infectious Diseases Unit to determine the probability of COVID-19 infection. Cases that were de ned as high probability were tested again after 48 hours and remained in isolation. In cases that were de ned as low probability isolation measures were stopped. Criteria for remaining in isolation were peripheral lung in ltrates on chest X-ray or chest cat-scan, fever with no alternative explanation, unexplained hypoxemia, and lymphopenia below 1.0 cellx10 3 /liter.
Patients who were diagnosed with COVID-19 before admission to the hospital were excluded.

Diagnosis of COVID-19
Diagnosis of COVID-19 was done by detection of SARS-CoV-2 RNA by RT-PCR obtained from the nasopharynx or in patients on mechanical ventilation by a sample obtained from deep suction.
Kits for RT-PCR varied according to availability. All tests were approved by the Israeli Ministry of Health. We used the following kits: Seegene, GeneExpert and BGI. Tests were performed and interpreted according to the manufacturer's instructions. Results were positive, negative, or inconclusive. In all inconclusive cases a second test was performed, and the nal interpretation was based on the pretest probability in discussion with physicians from the Infectious Disease Unit.

Data collection
We reviewed patients' clinical charts, laboratory exams and imaging results on the day in which the rst COVID-19 test was performed, at admission, or during hospitalization prior to receiving the results of the PCR tests. Patients' data was recorded in a tabular manner. Information included epidemiologic data, clinical presentation, laboratory results and radiological characteristics. Uncertainties were discussed with the attending physician. For each patient we determined if the clinical presentation was pneumonia, upper respiratory tract infection (URTI), other fever syndromes or isolated dyspnea.
URTI was de ned as having respiratory symptoms, fever, and on chest X-ray no new lung in ltrates compatible with an infectious process.
Pneumonia was de ned as having respiratory symptoms, fever, and on chest X-ray new lung in ltrates compatible with an infectious process.

Statistical analysis
Statistical analyses were performed using IBM SPSS® version 25. Categorical variables were presented as absolute number and percent, and continuous variables were presented as median and intra-quartile range (IQR). Con rmed and non-con rmed COVID-19 patients were compared using Pearson Chi-Square test and Fisher Exact test for categorical variables and Student T-test and Mann-Whitney test for continuous variables. We assessed risk factors for con rmed COVID-19 using logistic regression. Independent variables found to be signi cantly associated with the dependent variable in a univariate analysis (p<0.1) were entered into multivariate binary logistic regression analysis, (backwards conditional), with results presented as odds ratio (OR) with a con dence interval (CI) of 95%. Statistical signi cance was set at p<0.05.
Since we assumed the clinical factors might differ between patients with or without pneumonia, we performed another logistic regression for the subgroup of patients with pneumonia, excluding other diagnoses. We also performed a logistic regression analysis including only non-COVID-19 cases with two negative tests (patients with a higher pretest probability for COVID-19).

Results
The study included 254 patients, 37 COVID-19-positive (14.6%) and 217 COVID19-negative (85.4%) patients (Fig. 1). Thirty-ve (94.6%) of the positive patients were tested positive on the index date (the date the rst RT-PCR was performed). Two patients (5.4%) tested negative on the index date and positive in the following test performed 48 hours later. Of the 217 COVID-19-negative patients, 107 (49.3%) were considered to have a low probability for COVID-19 infection and were tested once. One hundred and ten patients (50.7%) had a higher probability of having COVID-19 and thus were tested twice. None of the patients that were taken out of isolation were later found to have COVID-19 infection, so our algorithm proved to be safe.
The median age was 76 years, 52% were male, and 28% were debilitated and suffered from dementia (Table 1). Two hundred and thirty three patients (91.7%) were suspected to have COVID-19 at admission with the rst test performed in the emergency department, while for the remaining 21 (8.3%) patients suspicion for COVID-19 was raised 72 hours or more after admission to the hospital. Of these, 13 were hospitalized in the internal medicine departments, four in general surgery, and four in the orthopedics department. were the most common presentation. Information regarding length of symptoms before presentation was available in 195 patients (76.8%) -COVID-19 positive patients presented later in the course of their illness as compared to COVID-19 negative patients (5 versus 2 days, p < 0.001) (Fig. 2).
Pulmonary in ltrates were more common in COVID-19 positive patients, and patients with in ltrates had a higher rate of bilateral in ltrates and of peripheral in ltrates (Table 1). COVID-19-positive patients had a higher rate of leukopenia and lymphopenia and lower rate of neutrophilia. They also had higher rates of elevated lactate dehydrogenase (LDH) (> 500 U/L), although LDH elevation in most cases was moderate (in the range of 500-1,000 U/L). Liver enzymes were elevated in a minority of patients in both groups, and there was no difference in C-reactive protein (CRP) levels.
In a multivariate analysis (Table 2), variables signi cantly associated with COVID-19 positivity included exposure to a known COVID-19 case, longer duration of symptoms before COVID-19 testing, bilateral pulmonary in ltrates, peripheral pulmonary in ltrates, an absolute neutrophil count lower than 7.5 × 10 3 /L and LDH levels > 500 U/L. Next, we performed another analysis including only patients with pneumonia (N = 78, 18 COVID-19positive patients and 60 COVID-19-negative patients). Median age was 78 years, and 53.8% were male (Table 1). A third of pneumonias patients were identi ed 72 hours or more after admission (similarly in both groups). Clinical features were mostly similar between COVID-19 positive and negative patients, but diarrhea was present in two COVID-19-positive patients and in none of the COVID-19-negative patients (P = 0.051). Bilateral and peripheral in ltrates were more common in the COVID-19-positive group. COVID-19positive patients with pneumonia had a higher rate of leukopenia, lower rate of neutrophilia and a higher rate of elevated LDH. Multivariate analysis (Table 3) showed that bilateral in ltrates, peripheral in ltrates, lack of absolute neutrophilia and elevated LDH were associated with COVID-19 positivity.

Discussion
In this study, we aimed to nd characteristics that may differentiate between patients with or without PCR con rmed COVID-19, in a population of hospitalized patients with suspected COVID-19. We found several parameters in the laboratory and radiographic results which are signi cantly more common among COVID-19 positive patients, including peripheral and bilateral in ltrates in chest X-ray, elevated LDH, and lack of absolute neutrophilia.
Only two studies [11][12] assessed the ability to differentiate between COVID-19 positive and negative patients in a hospital setting but they were performed in a high-incidence setting area (Wuhan, China) at the beginning of the pandemic, and thus included only a small number of COVID-19-negative patients.
Our study was performed in a low-incidence setting, thus having a different perspective on the pre-test probability of a patient having COVID-19. Also, both studies included only patients admitted with pneumonia, while our study included more diverse presentations, re ecting multiple stages in the natural history of COVID-19.
In a low COVID-19 incidence setting, one of the main challenges is identifying and isolating promptly COVID-19 positive patients to prevent hospital outbreaks. Because of the relatively low sensitivity of PCR tests, it is important to identify patients with high probability for COVID-19, and to continue isolation, until repeated testing is performed. Because in a low setting the number of COVID-19 negative patients is high, this poses a signi cant challenge in triaging patients with low and high probability for COVID-19 infection. Identifying features speci c for COVID-19-positive patients may help in performing this task.
The parameters assessed in our study can be divided to four groups, epidemiological, clinical, radiographic, and laboratory tests. Of the epidemiological parameters, working in the health care system was associated with COVID-19 positivity as shown previously [13]. Unsurprisingly, exposure to a con rmed case was also associated with COVID-19 infection.
Clinical symptoms were similar between COVID-19 positive and negative patients, as was shown in previous studies [11][12]. Clinical syndromes were different, with COVID-19 positive patients presenting mostly with pneumonia or URTI, and COVID-19 negative patients having more diverse and less speci c presentations, re ecting a high index of suspicion for COVID-19. We also showed that COVID-19 positive patients presented after a longer period of symptoms than COVID-19 negative patients. This is compatible with previous studies showing a mild disease in the rst week of COVID-19, not justifying hospitalization [3][4].
Presence of in ltrates in chest X-ray, and speci cally bilateral and peripheral in ltrates, were more common in COVID-19 patients, in accordance with the known radiographic features of COVID-19 [14]. It has been previously described that elevated LDH levels together with elevated liver enzymes can differentiate between COVID-19 positive and negative patients [11]. In our cohort only elevated LDH differed between COVID-19 positive and negative patients, and there was no difference in liver enzyme levels between the two groups.
Our approach proved to be safe. In none of the cases COVID-19 was identi ed after patients were taken out of isolation.
There are several limitations to our study. First, in some patients not all the laboratory data was available. Second, since in our hospital during the study period there was no strict de nition for a suspected COVID-19 patient, the COVID-19 negative group is heterogenous and includes patients with atypical presentation with a range of pretest probability for COVID-19. This re ects the real-life situation that health care personnel need to face with during this pandemic. We dealt with these issues in several ways. We did an additional analysis including only patients with a high pre-test probability for COVID-19. This yielded similar results. We also did another analysis that included only patients presenting with pneumonia. This again yielded similar results, but with less power due to the smaller sample included. Patients included in this study presented in different stages of COVID-19. The evolving nature of COVID-19 which starts as a viral URTI, progresses to pneumonia in various degrees of severity, and only later to multiorgan failure, cytokine storm, and need for mechanical ventilation makes it di cult to discuss speci c clinical features, as they are related to the stage of COVID-19. Another limitation is that this was a single-center study, performed at a secondary medical center. Thus, we did not have many immunocompromised patients, and we had a high proportion of elderly patients. This may limit the generalization of the results. Also, these results are relevant for a setup with a low to moderate incidence of COVID-19, where many patients hospitalized suffer from other medical problems. In a high incidence setting of COVID-19, decisions regarding rapid diagnosis and isolation may differ.

Conclusions
The clinical presentation of COVID-19 positive and negative patients is similar, but several radiographic and laboratory features can help to identify COVID-19 positive patients: bilateral and peripheral in ltrates on chest X-ray, a neutrophil count within normal limits, and elevated LDH. Identifying these features can help to initiate quick decisions regarding isolation of these patients.

Figure 1
Description of patients included in the study. Patients that had signs and symptoms compatible with COVID-19 were tested at the emergency room. If patients developed signs and symptoms suggestive of COVID-19 during hospitalization they were tested during hospitalization. Patients with negative test results and a low pretest probability for COVID-19 after a review of a physician from the Infection Disease Unit were taken out of isolation. Patients with a higher pretest probability for COVID-19 were tested again after 48 hours.