Predicting Need for Hospitalization of Patients with Pandemic (H1N1) 2009, Chicago, Illinois, USA

In the absence of established guidelines for hospitalization of patients with pandemic (H1N1) 2009, we studied emergency department patients to identify clinical parameters that predict need for hospitalization. Independent predictors of hospitalization include multiple high-risk medical conditions, dyspnea, and hypoxia. These findings are easily applicable, with a 79% positive predictive value for hospitalization.

variables that vary with age (respiratory rate, blood pressure, hematologic counts) were regrouped as normal or abnormal by using age-specifi c normal ranges (4,5). Obesity (body mass index >30) for adults and children 2-19 years of age and for those with high-risk medical conditions was defi ned according to CDC guidelines (6,7).
The Mann-Whitney U test and Pearson χ 2 test or Fisher exact test were used to compare continuous and categorical variables, respectively. p values <0.05 were considered signifi cant. Backwards stepwise logistic regression was performed for factors associated with hospitalization and intensive care unit (ICU) admission. Goodness-of-fi t was determined with the Hosmer-Lemeshow statistic. Data were analyzed by using SPSS version 16.0 (SPSS Inc., Chicago, IL, USA). The study was approved by the institutional review board of Rush University Medical Center.
A total of 189 cases that were identifi ed by review of microbiology records were considered eligible for the study. However, only 83 patients who were examined in the ED were included in the study; the remaining 106 patients were seen at outpatient clinics and private doctors' offi ces. Demographic, clinical, and laboratory data of hospitalized patients (32 [39%]) were compared with data from those discharged from the ED (51 patients) (Tables 1, 2).
Hospitalized patients were signifi cantly more likely to report a high-risk medical condition than were nonhospitalized patients (p<0.0001). Univariate analysis showed that the following high-risk medical conditions were also signifi cantly associated with hospitalization: history of prematurity, hemoglobinopathy, and chronic neurologic disease (p<0.05). A trend toward a higher incidence of chronic pulmonary disease was seen in hospitalized patients (41% vs. 22% of nonhospitalized patients; p = 0.06). Obesity was not found to be a signifi cant risk factor for hospitalization (p = 0.18).

Conclusions
We sought to identify predictors of hospitalization in patients with confi rmed pandemic (H1N1) 2009 infection. Univariate analysis showed that presence of high-risk medical conditions, age <5 years, dyspnea, and fi ndings of tachypnea, hypoxia (SpO 2 <92%), chest radiograph infi ltrate, and acute renal failure were signifi cant risk factors for hospitalization. Notably, headache, rhinorrhea, sore throat, and cough were inversely associated with hospital admission. We hypothesize that treating physicians perceive these symptoms as more suggestive of upper respiratory tract disease and hence are less likely to hospitalize such patients.
Multivariate analysis showed that only a higher number of high-risk medical conditions (including age <5 years), dyspnea, and a lower median oxygen saturation level were predictive of hospitalization. We found that dyspnea and a low median oxygen saturation level were also associated with ICU admission. These fi ndings suggest that clinicians' decision to hospitalize was not infl uenced by mere perception of illness severity, but rather it accurately refl ected the risk for complicated or severe disease. Our fi ndings are also consistent with CDC alerts on emergency warning signs of pandemic (H1N1) 2009 infl uenza (8) and a recent report by Echevarría-Zuno et al. (9) from Mexico in which dyspnea, tachypnea, and cyanosis were prognostic factors for admission and death. Although the presence of any 1 underlying high-risk medical condition has been previously described as a risk factor for complication with seasonal infl uenza including hospitalization (10), a higher number of high-risk medical conditions is a stronger predictor of hospitalization (median number 2 vs. 0; p = 0.01). Although excellent clinical prediction rules for hospitalization of patients with community-acquired pneumonia are available (e.g. CURB-65 or pneumonia severity index), few data exist for infl uenza admissions (10,11). We propose a simple clinical guide for hospitalization of patients with pandemic (H1N1) 2009 infection by using results of our multivariate analysis. The presence of any of 3 predictors->2 high-risk medical conditions (including age <5 years), dyspnea, or hypoxia-has sensitivity, specifi city, positive predictive value, and negative predictive value of 72%, 88%, 79%, and 83%, respectively, for hospitalization. Dyspnea or hypoxia was also predictive of ICU admission, with sensitivity, specifi city, positive predictive value, and negative predictive value of 94%, 91%, 71%, 98%, respectively. Identifi cation of these risk factors is widely applicable as a triage tool, especially in settings such as physician's offi ces where laboratory and radiologic data are not immediately available.