key: cord-0991632-pemva816
authors: Shi, Shu Jing; Li, Hui; Liu, Meng; Liu, Ying Mei; Zhou, Fei; Liu, Bo; Qu, Jiu Xin; Cao, Bin
title: Mortality prediction to hospitalized patients with influenza pneumonia: PO(2)/FiO(2) combined lymphocyte count is the answer
date: 2015-08-11
journal: Clin Respir J
DOI: 10.1111/crj.12346
sha: 90876ed8556ef5e9f97b4d39db2bfa875d8514b5
doc_id: 991632
cord_uid: pemva816

INTRODUCTION: Community‐acquired pneumonia (CAP) severity scores perform well in predicting mortality of CAP patients, but their applicability in influenza pneumonia is powerless. OBJECTIVES: The aim of our research was to test the efficiency of PO(2)/FiO(2) and CAP severity scores in predicting mortality and intensive care unit (ICU) admission with influenza pneumonia patients. METHODS: We reviewed all patients with positive influenza virus RNA detection in Beijing Chao‐Yang Hospital during the 2009–2014 influenza seasons. Outpatients, inpatients with no pneumonia and incomplete data were excluded. We used receiver operating characteristic curves (ROCs) to verify the accuracy of severity scores or indices as mortality predictors in the study patients. RESULTS: Among 170 hospitalized patients with influenza pneumonia, 30 (17.6%) died. Among those who were classified as low‐risk (predicted mortality 0.1%–2.1%) by pneumonia severity index (PSI) or confusion, urea, respiratory rate, blood pressure, age ≥65 year (CURB‐65), the actual mortality ranged from 5.9 to 22.1%. Multivariate logistic regression indicated that hypoxia (PO(2)/FiO(2) ≤ 250) and lymphopenia (peripheral blood lymphocyte count <0.8 × 10(9)/L) were independent risk factors for mortality, with OR value of 22.483 (95% confidence interval 4.927–102.598) and 5.853 (95% confidence interval 1.887–18.152), respectively. PO(2)/FiO(2) combined lymphocyte count performed well for mortality prediction with area under the curve (AUC) of 0.945, which was significantly better than current CAP severity scores of PSI, CURB‐65 and confusion, respiratory rate, blood pressure, age ≥65 years for mortality prediction (P < 0.001). The scores or indices for ICU admission prediction to hospitalized patients with influenza pneumonia confirmed a similar pattern and PO(2)/FiO(2) combined lymphocyte count was also the best predictor for predicting ICU admission. CONCLUSION: In conclusion, we found that PO(2)/FiO(2) combined lymphocyte count is simple and reliable predictor of hospitalized patients with influenza pneumonia in predicting mortality and ICU admission. When PO(2)/FiO(2) ≤ 250 or peripheral blood lymphocyte count <0.8 × 10(9)/L, the clinician should pay great attention to the possibility of severe influenza pneumonia.

Three global influenza pandemics of the 20th century, ranging from Spanish flu of 1918 to the 1968 Hong Kong pandemic, proved to be disasters in the history of mankind and caused death toll more than that of the First World War (1) . The recent swine origin 2009 pandemic influenza A H1N1 virus led more than 60 million laboratory confirmed cases in 214 countries and over 18 449 deaths until August 2010, resulting in substantial human admission rate and mortality (2) . Most patients with severe influenza infections present with pneumonia (3) (4) (5) (6) .

As we know that, at the time of a pandemic, one of the most important decisions is to decide the appropriate site of care, which needs not only physician's clinical judgment, but also the objective severity scores or indicators to help the physicians, especially the physicians in the emergency department, to make a right decision.

A variety of pneumonia severity scores have been used to help the physicians to evaluate the mortality or severity of patients with community-acquired pneumonia (CAP), including pneumonia severity index (PSI) (7) , confusion, urea, respiratory rate, blood pressure, age 65 year (CURB-65) (8) , confusion, respiratory rate, blood pressure, age 65 years (CRB-65) (9) , systolic blood pressure, multi-lobar chest radiography involvement, albumin level, respiratory rate, tachycardia, confusion, oxygenation, arterial pH (SMART-COP) (10) and lung injury score (LIS) (11) . PSI and CURB-65 are valid scores in predicting 30-day mortality with CAP patients (12) . SMART-COP can predict patients who might need vasopressoror ventilatory support interventions with more than 90% accuracy (10) . LIS is a useful tool to diagnose acute lung injury/ adult respiratory distress syndrome (13) .

But current CAP severity scores (PSI, CURB-65 and CRB-65) in predicting mortality in patients with influenza pneumonia is unpersuasive. Several studies (14) (15) (16) indicated that current CAP severity scores failed to predict mortality in patients due to influenza pneumonia. Only one research pointed out that SMART-COP presented the best performance to indicate intensive care unit (ICU) admission in patients with H1N1 pneumonia (17) . To our knowledge, there is no relative research about LIS in predicting mortality or severity with influenza pneumonia. The current CAP severity scores are mainly developed for CAP patients with the major pathogens of Streptococcus pneumonia, Haemophilus influenza and atypical bacterial pathogens (9, (18) (19) (20) , which are different from pneumonia caused by influenza virus. As a kind of interstitial pneumonia, pathologic characteristics (21) of influenza pneumonia includes interstitial inflammation, hyaline membranes formation, intra-alveolar hemorrhage, edema and necrotizing bronchitis that affecting pulmonary diffusion function (22) .

Taking these into considerations, it may be more possible that severity indices reflecting pulmonary diffusion functions may perform well in predicting prognosis with hospitalized patients due to influenza pneumonia. According to our previous study (23) and those from Korea (24) , hypoxia (PO 2 /FiO 2 250) was an independent risk factor for death in patients with influenza A (H1N1) 2009 pneumonia.

In this study, we want to test the efficiency of PO 2 / FiO 2 and CAP severity scores in predicting mortality and ICU admission with influenza pneumonia patients.

During the 2009-2014 influenza seasons (from November to the next February in Beijing area of China), we screened patients with positive influenza virus RNA detection of respiratory specimen from the microbiology Lab in Beijing Chao-Yang Hospital. Then we excluded outpatients and inpatients with no pneumonia. Cases with unavailable data to calculate severity scores were also excluded from this analysis.

Patients with influenza pneumonia: during the influenza seasons, patients with respiratory symptoms and a new pulmonary infiltrate on the chest radiograph, combined with positive influenza virus A RT-PCR testing and negative for other kinds of viruses as below.

Patients with bacterial co-infection: a patient who fulfilled the above criterion of influenza pneumonia, at the same time, with culture positive for bacterial pathogen from blood/sputum/bronchoalveolarlavage fluid or positive urinary antigen or positive Mycoplasma pneumonia, Chlamydia pneumonia RT-PCR detection before or within the initial 48 h of hospital admission.

Microbiological evaluation was performed according to our previous reports (25, 26) , (i) Qualified sputum (defined as an adequate quality sputum sample with >25 leukocytes and <10 epithelial cells per 3 100 magnification field) were sent for Gram staining and culture according to standard methods; (ii) urinary antigen for Legionella pneumophila (Binax Now L. pneumophila urinary antigen test; Trinity Biotech, Bray, Ireland) and urinary antigen for S. pneumoniae Neoplastic disease: defined as any cancer except of basilar or squamous cell cancer of the skin that was active at the time of presentation or diagnosed within 1 year of presentation.

Liver disease: defined as a clinical or histological diagnosis of cirrhosis or another form of chronic liver disease, such as chronic active hepatitis.

Cerebrovascular disease: defined as a clinical diagnosis of stroke or transient ischemic attack or stroke documented by magnetic resonance imaging or computed tomography.

Renal disease: defined as a history of chronic renal disease or abnormal blood urea nitrogen and creatinine concentrations documented in the medical record.

Cardiac disease: defined as systolic or diastolic ventricular dysfunction documented by history, physical examination and chest radiograph, echocardiogram, multiple gated acquisition scan or left ventriculogram.

Obesity: a patient with the body mass index more than 30 kg/m 2 was considered obesity.

Mortality: the outcome variable of mortality was defined as all-cause mortality at the time of hospital discharge. According to PSI and CURB-65, we also calculated the number of deaths and actual mortality in each class or score. Predicted mortality rate was acquired from the original publications of the severity score (7, 8) .

Severity of influenza pneumonia was evaluated using PO 2 /FiO 2 , PSI, CURB-65, CRB-65, SMART-COP and LIS. The severity scores or indices were calculated within 48 h of hospital admission, which were compared between deceased and survival groups.

Statistical analysis was performed using statistical software package SPSS (version 17.0). Patients were grouped as dead vs alive. Continuous variables were described as means (6standard deviation). vChisquared 2 test or Fisher exact test was used to compare categorical variables and t-test for continuous variables. We used multivariate logistic regression to identify independent predictors of mortality. Receiver operating characteristic curves (ROCs) were generated to compare the total predictive accuracy of PO 2 /FiO 2 , PSI, CURB-65, CRB-65, SMART-COP and LIS for mortality and ICU admission, and the area under the curves (AUCs) were calculated. A P value < 0.05 was considered statistically significant, 95% confidence intervals were calculated.

During the study period, 1067 patients with positive influenza virus RNA detection were screened. Five hundred fifty-eight outpatients, 322 inpatients without pneumonia and 17 patients with incomplete data were excluded. At last, a total of 170 hospitalized patients with influenza pneumonia were enrolled in the study (Fig. 1) . Among 170 patients, 114 cases of H1N1, one case of H1N11H3N2, one of H3N2 and 54 of influenza A, but not typed. Bacterial co-infection was documented in seven (4.1%) patients, with three cases of Pseudomonas aeruginosa, two with Acinetobacter baumannii, one with Staphylococcus aureus and one of Klebsiella pneumonia. All the other pathogens (including M. pneumonia and C. pneumonia, L. pneumophila) were negative. Mortality was 17.6% (30/170). Baseline characteristics of the hospitalized patients with influenza pneumonia were shown in Table 1 .

The variables associated with mortality in hospitalized patients with influenza pneumonia were decreased lymphocyte count, decreased hemoglobin, decreased platelet, decreased albumin, decreased PO 2 /FiO 2 , elevated respiratory rate, elevated blood urea nitrogen, elevated lactate dehydrogenase, elevated serum glucose, multilobar infiltrates and pleural effusion. Multivariate logistic regression indicated that hypoxia (PO 2 / FiO 2 250) and lymphopenia (peripheral blood lymphocyte count <0.8 3 10 9 /L) (27) were independent risk factors for mortality, with OR value of 22.483 (95% confidence interval 4.927-102.598) and 5.853 (95% confidence interval 1.887-18.152), respectively. The result of multivariate logistic regression was showed in Table 2 . Current CAP severity scores underestimate mortality with low-risk patients

The deceased patients and the total number of patients stratified by PSI and CURB-65 with their actual and predicted mortality rates were shown in Table 3 . Predicted mortality in the range of 0.1%-2.1% was defined as low risk. According to PSI and CURB-65, low risk was determined in 130 (76.5%) patients and 136 (80%) patients, respectively, which accounted for a majority part of hospitalized patients with influenza pneumonia. The actual mortality in these patients ranged from 5.9 to 22.1%. That was to say, PSI and CURB-65 underestimated the mortality in a significant number of hospitalized patients with influenza pneumonia.

As we mentioned above, lymphopenia was also an independent risk factor for mortality. So, we finally used eight severity scores or indices (PSI, CURB-65, CRB-65, SMART-COP, LIS, PO 2 /FiO 2 , lymphocyte count and PaO 2 /FiO 2 combined lymphocyte count) to compare AUCs for mortality prediction in hospitalized patients with influenza pneumonia ( Table 4 ). The ROCs for mortality prediction was shown in Supporting Information e- Fig. 1 . PaO 2 /FiO 2 combined lymphocyte count performs best in mortality prediction with hospitalized patients due to influenza pneumonia with AUC of 0.945 (95% confidence interval, 0.910-0.979), and the AUC of PO 2 /FiO 2 was 0.916 (95% confidence interval, 0.869-0.962), which was lower than PaO 2 / FiO 2 combined lymphocyte count. The AUCs of PSI, CURB-65 and CRB-65 were less than 0.70. PaO 2 /FiO 2 combined lymphocyte count was significantly better than current CAP severity scores of PSI, CURB-65 and CRB-65 (P < 0.001). The AUCs of LIS, SMART-COP and lymphocyte count were respectively 0.892 (95% confidence interval, 0.823-0.961), 0.808 (95% confidence interval, 0.728-0.889) and 0.827 (95% confidence interval, 0.750-0.905), which were all less than the AUC of PaO 2 / FiO 2 combined lymphocyte count.

The AUCs of the eight severity scores or indices for ICU admission prediction in hospitalized patients with influenza pneumonia confirmed a similar pattern to mortality prediction ( Table 5 ). The ROCs for predicting ICU admission was shown in Supporting Information e- Fig. 2 . PaO 2 /FiO 2 combined lymphocyte count was also a good predictor for ICU admission prediction with AUC of 0.857 (95% confidence interval 0.797-0.917), which was higher than PaO 2 /FiO 2 alone. The AUC of PaO 2 /FiO 2 combined lymphocyte count was also greater than current CAP severity scores of PSI, CURB-65 and CRB-65 (P < 0.05). The AUC of LIS, SMART-COP and lymphocyte count were all less than that of PaO 2 /FiO 2 combined lymphocyte count.

This study demonstrated that PO 2 /FiO 2 combined lymphocyte count is simple and reliable severity predictor of hospitalized patients with influenza pneumonia, which is significantly better than current CAP Commons and Denholm (28) investigated 105 patients of H1N1 influenza infection and found that the common used CAP severity scores (PSI and CURB-65) had insufficient predictive ability to lowrisk patients in ICU admission. Other researches (29) (30) (31) also showed that routine prediction rules underestimated severity of influenza A 2009 (H1N1) Pneumonia, but no effective severity score had been put forward. In our study, PSI and CURB-65 underestimated a significant number of hospitalized patients. But the influenza pneumonia patients with high risk (PSI class of V and CURB-65 score of 3-5) are scarce, so the conclusion is inadequate and we need to increase the sample size in our further research.

The severity scores of PSI and CURB-65 heavily weight on advanced age and complication (7, 8) , but 2009 H1N1 influenza A virus infection has been reported to occur in young, previously healthy individuals (32) (33) (34) (35) , which may be one reason that the scores fail to predict mortality. Another possible explanation may be due to the facts that current CAP severity scores are mainly developed for CAP patients with the typical bacterial and atypical bacterial pneumonia.

Concerning on PaO 2 /FiO 2 and the prognosis of influenza pneumonia, a study in Washington (36) identified different mean PaO 2 /FiO 2 values in survivors (230) and non-survivors (91) (P 5 0.005) in patients with influenza pneumonia. Ho et al. (37) indicated that PaO 2 /FiO 2 was a useful parameter in predicting mortality with influenza pneumonia, but the study only evaluated 38 cases and there was no comparison with other severity scores.

The pathologic changes of influenza pneumonia are characterized by diffused alveolar damage and altered pulmonary diffusion function (21, 22) . The clinical and radiological characteristics of CAP caused by influenza A 2009 (H1N1) differed markedly from CAP caused by bacterial agents, with high frequency of dyspnea, hemoptysis and bilateral interstitial infiltrate (29) . These above features of influenza pneumonia are closely associated with low PO 2 /FiO 2 . Therefore, there was no surprise that the index of PO 2 /FiO 2 performed so well in predicting mortality of hospitalized patients with influenza pneumonia. And it is also understandable that severity prediction scores including the variable of PaO 2 /FiO 2 , such as SMART-COP and LIS, presented better performance than PSI, CURB-65 and CRB-65, which have no variables associated with PaO 2 /FiO 2 . But the severity scores of LIS and SMART-COP are more complex to calculate, and some variables (e.g. respiratory system compliance) in the scores are difficult to obtain, which confine its application.

Our study also found that decreased lymphocyte count was an independent predictor of mortality. Previous studies pointed out that lymphopenia were an early and reliable laboratory finding of adult severe influenza A infection (27, 38, 39) . Concerning on the mechanism of lymphopenia, it needs further research.

Some limitations of this study should be noted. First, our study only included 170 patients from one centre, the conclusion generalized to be widely used still need a large-scale clinical validation. Second, our study consists of only hospitalized patients with influenza pneumonia, we are not able to define severity scores or indices in ambulatory patients.

Despite the above limitations, we believe that our study has shown important and novel findings about the death prediction in hospitalized patients with influenza pneumonia. To our knowledge, this is the first study that compared PO 2 /FiO 2 and lymphocyte count with current CAP severity scores. Our findings that PO 2 /FiO 2 combined lymphocyte count is a useful predictor for severity of influenza pneumonia may help clinicians more accurately predict prognosis, and triage place to improve outcome.

In conclusion, we found that PO 2 /FiO 2 combined lymphocyte count is simple and reliable predictor of hospitalized patients with influenza pneumonia in predicting mortality and ICU admission. When PO 2 / FiO 2 250 or peripheral blood lymphocyte count <0.8 3 10 9 /L, the clinician should pay great attention to the possibility of severe influenza pneumonia.

Additional Supporting Information may be found in the online version of this article at the publisher's website: Figure 1(e-Fig 1) . The ROCs for mortality prediction in hospitalized patients with influenza pneumonia PaO 2 /FiO 2 combined lymphocyte count performs best in mortality prediction with hospitalized patients due to influenza pneumonia with AUC of 0.945 (95% confidence interval, 0.910-0.979), and the AUC of PO 2 / FiO 2 was 0.916 (95% confidence interval, 0.869-0.962), which was lower than PaO 2 /FiO 2 combined lymphocyte count. The AUCs of PSI, CURB-65 and CRB-65 were less than 0.70. PaO 2 /FiO 2 combined lymphocyte count was significantly better than current CAP severity scores of PSI, CURB-65 and CRB-65 

(P < 0.001). The AUCs of LIS, SMART-COP and lymphocyte count were, respectively, 0.892 (95% confidence interval, 0.823-0.961), 0.808 (95% confidence interval, 0.728-0.889) and 0.827 (95% confidence interval, 0.750-0.905), which were all less than the AUC of PaO 2 /FiO 2 combined lymphocyte count. Figure 2(e-Fig 2) . The ROCs for predicting ICU admission in hospitalized patients with influenza pneumonia PaO 2 /FiO 2 combined lymphocyte count was also a good predictor for ICU admission prediction with AUC of 0.857 (95% confidence interval 0.797-0.917), which was higher than PaO 2 /FiO 2 alone. The AUC of PaO 2 /FiO 2 combined lymphocyte count was also greater than current CAP severity scores of PSI, CURB-65 and CRB-65 (P < 0.05). The AUC of LIS, SMART-COP and lymphocyte count were all less than that of PaO 2 /FiO 2 combined lymphocyte count.

Mortality prediction to hospitalized patients with influenza pneumonia Shi et al. 

World Health Organization (WHO) Weekly Epidemiological Record. Ten things you need to know about pandemic influenza

Swine influenza A (H1N1) infection in two children-Southern California

A report of influenza of pneumonia

Pneumonia complicating Asian influenza

Severe influenza virus pneumonia in the pandemic of 1968-1969

Complications of viral influenza

A prediction rule to identify low-risk patients with community-acquired pneumonia

Defining community acquired pneumonia severity on presentation to hospital: an international derivation and validation study

BTS guidelines for the management of community acquired pneumonia in adults: update

SMART-COP: a tool for predicting the need for intensive respiratory or vasopressor support in community-acquired pneumonia

An expanded definition of the adult respiratory distress syndrome

Severity assessment tools for predicting mortality in hospitalized patients with community-acquired pneumonia. Systematic review and meta-analysis

Computed tomography in established adult respiratory distress syndrome correlation with lung injury score

Predicting mortality in hospitalized patients with 2009 H1N1 influenza pneumonia

Severity assessment tools in ICU patients with 2009 influenza A (H1N1) pneumonia

Mortality and severity evaluation by routine pneumonia prediction models among Japanese patients with 2009 pandemic influenza A (H1N1) pneumonia

The role of pneumonia scores in the emergency room in patients infected by 2009 H1N1 infection

Infectious Diseases Society of America/American Thoracic Society consensus guidelines on the management of communityacquired pneumonia in adults

Chinese Society of Respiratory Diseases. Guide for diagnosis and treatment of community-acquired pneumonia

The JRS guidelines for the management of community-acquired pneumonia in adults: an update and new recommendations

Fatal diffuse influenzal pneumonia: premortem diagnosis by lung biopsy

Acute respiratory distress syndrome: the Berlin definition

Clinical features of pneumonia caused by 2009 influenza A (H1N1) virus in Beijing

A prediction rule to identify severe cases among adult patients hospitalized with pandemic influenza A (H1N1) 2009

Emergence of communityacquired adenovirus type 55 as a cause of community-onset pneumonia

Accuracy of IgM antibody testing, FQ-PCR and culture in laboratory diagnosis of acute infection by Mycoplasma pneumonia in adults and adolescents with community-acquired pneumonia

Diagnostic importance of relative lymphopenia as a marker of swine influenza (H1N1) in adults

Triaging pandemic flu: pneumonia severity scores are not the answer

Severity of influenza A 2009 (H1N1) pneumonia is underestimated by routine prediction rules. Results from a prospective, population-basedstudy

Pandemic influenza (h1n1) 2009 pneumonia: CURB-65 score for predicting severity and nasopharyngeal sampling for diagnosis are unreliable

Usefulness of CURB-65 and pneumonia severity index for influenza A H1N1v pneumonia

Pneumonia and respiratory failure from swine-origin influenza A (H1N1) in Mexico

Intensive-care patients with severe novel influenza A (H1N1) virus infection-Michigan

Intensive care adult patients with severe respiratory failure caused by influenza A (H1N1)v in Spain

Critically ill patients with 2009 influenza A (H1N1) in Mexico

Influenza pneumonia: a descriptive study

Prognostic factors for fatal adult influenza pneumonia

Depletion of lymphocytes and diminished cytokine production in mice infected with a highly virulent influenza A (H5N1) virus isolated from humans

Clinical features of the initial cases of 2009 pandemic influenza A (H1N1) virus infection in China

All work was performed at the Beijing Chao-Yang Hospital, Capital Medical University. The authors thank all the authors that contribute to the study. The authors also thank Ning Chen and Yang Li Gao in the Department of Radiology for their help with radiologic evaluation. This work was supported by a grant from the National Science for Distinguished Young Scholars