Spectrum of clinical and radiographic findings in patients with diagnosis of H1N1 and correlation with clinical severity

Background The aim of this study was to evaluate the correlation between clinical and imaging findings with a worse clinical outcome in patients with a confirmed diagnosis of H1N1 influenza A virus. Methods Patients with a positive viral test for influenza A H1N1 in 2016 and chest radiography (CR) and/or computed tomography (CT) results had clinical and imaging data reviewed. Hospitalization, admission to the intensive care unit or death were defined as worse clinical outcomes. The association between clinical and imaging features and the worse outcome was calculated in a logistical regression model. Results Eighty of 160 (50%) patients were men, with a mean age of 43 ± 19 years. The most common symptoms were as follows: flu-like symptoms 141/160 (88%), dyspnea (25/160, 17%), and thoracic pain (7/160, 5%). Abnormalities on CR were detected in 8/110 (7%) patients, and 43/59 (73%) patients had an abnormal CT. The following variables were associated with worse clinical outcomes: the presence of diabetes mellitus (DM), hypertension, dyspnea, thoracic pain, abnormal CR or CT regardless of the type of finding, CT with consolidation or ground glass opacity. Conclusions The presence of DM, hypertension, dyspnea, thoracic pain, or an abnormal CR or CT on admission were associated with worse clinical outcomes in patients with H1N1 influenza A virus infection. Thus, the use of readily accessible clinical and imaging features on admission may have a role in the evaluation of patients with H1N1 infection.


Background
Influenza A H1N1 virus is the subtype of influenza virus that typically leads to a more severe infection than the usual seasonal influenza virus, and its viral strains were historically responsible for the major outbreaks worldwide, similar to the Spanish flu in the twentieth century. It was the cause of the last influenza pandemic in 2009, which resulted in thousands of deaths worldwide, principally among young individuals and pregnant woman. Since then, the H1N1 influenza virus has assumed a seasonal spreading similar to other influenza types and continues to cause substantial morbidity, although it is controlled with the immunization programs [1]. The immunization campaign for Influenza A H1N1 in our country recommends the vaccination for the entire population annually from April to May. The public healthcare system offers all year round the vaccine for risk groups, including pregnant women, elderly patients, children, healthcare professionals and people with any chronic disease. Nevertheless, an atypical outbreak of Influenza A H1N1 infection occurred in South America in 2016, which started before the vaccination campaign, and caused approximately 800 deaths and 500 intensive care unit admissions in a single country [2].
Since the first outbreak of H1N1, many studies have been conducted to identify an imaging pattern [3,4] or clinical features that indicate a worse prognosis [1,5,6]. The identification of clinical and radiographic findings in patients with H1N1 infection that correlate with the clinical severity is of key importance in the management of these patients. Previous studies have demonstrated that all patients with H1N1 admitted to an intensive care unit had at least an abnormal conventional radiograph (CR) at diagnosis. However, there is a relative lack of studies in the literature that have evaluated predictive tools in the evaluation of patients with H1N1 [7,8].
In this scenario, the purpose of this study was to evaluate the correlation between clinical and imaging findings with a worse clinical outcome in patients with a confirmed diagnosis of H1N1 influenza A virus.

Methods
The institutional review board approved this retrospective study and waived the requirement for informed consent. The database from our institution, which is a private hospital, was searched for consecutive patients with a diagnosis of influenza A H1N1 in 2016 as assessed via the upper airway by nasopharyngeal swabbing or a washing test with polymerase chain reaction. The inclusion criteria were a positive test for influenza A H1N1 and CR and/or chest computed tomography (CT) within 48 h before or after the positive test. The exclusion criteria were another confirmed concomitant pulmonary infection and the absence of follow-up data in the medical records. The patient accrual is summarized in Fig. 1.
Clinical and laboratorial data were obtained from a detailed medical record review conducted by two radiologists with 2 and 3 years of experience, respectively, using a standardized form. The following clinical data of the patients were assessed: gender, age, symptoms at admission (such as flu-like symptoms, dyspnea, thoracic pain, hemoptysis, and sepsis), and presence of comorbidities (including systemic hypertension, diabetes mellitus [DM], tobacco smoke, asthma, heart disease, chronic obstructive pulmonary disease [COPD], immunodeficiency, and other). Information regarding the physical examination at admission was also evaluated, including the heart rate (HR), temperature, oxygen saturation (O 2 ), and blood pressure (BP). Regarding the laboratory data, the white blood count, lymphocytes, and C-reactive protein (CRP) obtained at admission were assessed.
The chest CR were performed in a digital CR device in the posteroanterior and lateral incidences during maximal inspiration.
CT scans were performed on two of the following devices: a dual-source 256 row or a 128 row detector CT system. Contrast imaging was performed with 1.0-1.5 ml/kg injected at a flow rate of 4.0-5.5 ml/s − 1 . Images were reconstructed in the axial view using a slice thickness of 1 mm and an increment of 0.7 mm.
Two board-certificated chest radiologists (with 6 and 2 years of experience) blinded to the clinical data independently reviewed the radiological. The radiologists analyzed the imaging features according to the Fleischner Society glossary [9]. For discordant imaging findings (6/ 146 images; 5%), an agreement was reached by consensus between the readers. Figures 2, 3 and 4 summarize the imaging features assessed by the radiologists.
The following imaging features were evaluated on CR: pulmonary infiltrate (which was defined as any alterations, Fig. 1 Flowchart of patient selection including both ground glass and reticular opacities), consolidation, and pleural effusion [10,11]. The CT imaging features were fully assessed and the following findings were highlighted: ground glass opacity (GGO), consolidation, bronchial wall thickening, septal interlobular thickening, lymph node enlargement, tree-in-bud pattern, and pleural effusion. The distribution of the features was defined as follows: none, 1 field, 2 fields unilateral, 3 or more fields bilateral, and diffuse.
The patients were classified into two groups: good clinical outcome and worse clinical outcome. A good clinical outcome was defined as hospital discharge after evaluation in the emergency department without the need for hospitalization. Patients who were hospitalized, admitted to the intensive care unit (ICU) or died were determined to have a worse clinical outcome.

Statistical analysis
Continuous variables were tested for normality with the Kolmogorov-Smimov and Shapiro Wilk tests. The values are expressed as median and percentiles 25 and 75 or as the mean and standard deviation for parametric and non-parametric data respectively. The categorical data are presented as absolute values and percentages and were tested using Pearson χ2 test and Fisher exact test, if applicable. The Mc'Nemar test was used by compare two methods. Non-parametric data was compared using the Mann-Whitney U test for two independent samples. The logistic regression models included calculations of corresponding crude and adjusted odds ratios (ORs) and 95% confidence intervals (CI). Two-tailed p-values ≤0.05 were considered significant. Statistical analyses were carried out using SPSS for Windows version 19.0 (Chicago, IL, USA). Hierarchical cluster analysis was performed to identify individual similarities in the diagnostic model through analysis of clusters to try to determine from each of the groups to which they belong. Statistical significance was considered with p ≤ 0,05. Analyses were performed using SPSS 19.0 and GraphPad Prism 5.0.

Results
Our study population consisted of 160 patients, including 80/160 (50%) men and 80/160 (50%) women, with a median age of 43 years (IQR: 32-56). None of the patients   Considering outcome, 37/160 (23.1%) patients were hospitalized and classified as having a worse clinical outcome. Four of these patients (4/37, 10.8%) required intensive care unit (ICU) admission, and none of these patients died. The mean age of this subpopulation was 50 years old (IQR: 34-66). All patients who were hospitalized had an abnormality on CR and/or CT. Table 1 summarizes the patients' characteristics and the radiological findings.
Thirteen patients underwent both CR and chest CT ( Table 2). There was a significant correlation between an abnormal CR and abnormal chest CT (p = 0.031) and between the presence of pulmonary infiltrate on CR and GGO on CT (p = 0.016). In contrast, there was no significant correlation between consolidation and pleural effusion on CR and CT (p = 0.501 and p = 0.999, respectively).

Correlation with worse clinical outcome
Considering the clinical data, the presence of DM, systemic hypertension, dyspnea, thoracic pain, temperature levels and CRP levels were significantly different between the groups ( . Furthermore, the distribution of the imaging features on CR and CT were correlated with outcome, in which a diffuse pattern was more related to worse clinical outcomes. In the logistic regression, the following data pre- In the hierarchical cluster analysis, three clusters were identified (Fig. 5). The cluster that better correlated with worse clinical outcomes contained the following variables: GGO on CT or CR, consolidation on CT or pulmonary infiltrate on CR and the extension of the disease.

Discussion
Our study population was obtained from our database throughout the year of 2016; however, all patients selected had the H1N1 infection diagnosed between the 3 months before winter in the south hemisphere, which is unusual of a typical seasonal respiratory infection and configured an atypical and early outbreak of H1N1 infection [12].
The major clinical findings among our population were flu-like symptoms (97.9%) as expected for a respiratory infection [13]. We found in our population clinical and radiological changes associated with worse clinical outcomes. DM, systemic hypertension, dyspnea, thoracic pain, temperature and CRP levels were the clinical features associated with worse clinical outcomes. Moreover, the radiological findings were the presence of an abnormality on CT or CR, pulmonary infiltrate, consolidation, and pleural effusion on CR, presence of GGO and consolidation on CT, and diffuse pattern of imaging abnormality on CT or CR. In the logistic regression, the presence of DM, thoracic pain, levels of body temperature, an alteration on CR or CT, diffuse pattern of distribution of any abnormality on CR or CT; pulmonary infiltrate on CR, GGO on CT, and pleural effusion on CT or CR were the variables that predicted worse clinical outcomes. In the hierarchical cluster analysis, the cluster that better correlated with worse clinical outcomes contained GGO on CT, pulmonary infiltrate on CR, consolidation and the diffuse pattern of distribution on CT or CR.
Regarding the clinical features, our results are in line with similar studies that evaluated symptoms associated with a worse outcome, as they demonstrated that fever and dyspnea may predict hospitalization [13]. Studies that have analyzed only patients with severe H1N1 admitted to an ICU have also demonstrated that symptoms were similar to typical seasonal H1N1, such as cough, dyspnea, fever, myalgia and headache [1]. The results are also comparable to studies that have analyzed mild and severe infections [7].
Our study also demonstrated chest pain as a predictor of worse outcome, a feature that was not found in previous studies. However, another study also identified particularities that may be related to the institution sample or type of H1N1 virus. Rohani et al. [1] found gastrointestinal symptoms associated with severe infection with a worse  [14].
Our study also demonstrated that body temperature was significantly lower in patients with worse clinical outcomes. Although the underlying reason of hypothermia in these patients remains unclear, some other studies also have demonstrated an association between low body temperature and worse outcomes in critically ill patients with and without infection [15,16]. With regards to imaging features, the CR and CT findings in our study population are comparable to the most representative studies over the years since the first H1N1 outbreak in 2009 [5,13,17]. All 37 patients with worse outcomes had at least a pulmonary abnormality in an imaging exam, and we identified significant differences among the groups with and without hospitalization.
The imaging features detected in our population are in line with previous data of viral pneumonia, including pulmonary infiltrate, GGO, consolidation and pleural effusion [10,18]. Comparable to other studies, we found that diffuse distribution on an imaging abnormality was associated with worse clinical outcomes [5,13,17].
The clinical relevance of our results essentially relies on the management of patients with H1N1 infection. Informing the multidisciplinary team of the potential for a worse clinical outcome based on easily accessible information, such as the presence of DM, thoracic pain, body temperature, and an abnormality on CR or CT, may add value in the setting of emergency departments. If clinicians know in advance the possibility of a worse outcome, they may consider a closer evaluation, and the emergency discharge decision may be postponed. This type of guidance may be valuable, particularly at noncomprehensive centers and during epidemic situations.
There are several potential limitations of this study. First, it is a single-center retrospective study performed at a private hospital focused on adult nonpregnant patients, which is thus subjected to selection bias. Second, the study population seemed to have a milder infection than previous studies, without deaths or sepsis at admission. These results may be due to an earlier outbreak of 2016 in the early autumn or due to a less severe H1N1 genotype. Although our study population was relatively large compared to other imaging studies, future prospective multicenter studies with a more diverse population, including pediatric and pregnant patients are needed to overcome the limitations of this study and to provide a better generalization of our results.

Conclusions
In conclusion, our study performed at a private hospital focused on adult non-pregnant patients demonstrates that the presence of DM, hypertension, dyspnea, thoracic pain, or an abnormal CR or CT on admission were associated with worse clinical outcomes in patients with H1N1 influenza A virus infection. Consolidation and ground glass opacities on CT were also associated with the clinical severity in these patients. Thus, the use of readily accessible clinical and imaging features on admission may have a role in the evaluation of patients with H1N1 infection.