|Home | About | Journals | Submit | Contact Us | Français|
Background.Wisconsin was severely affected by pandemic waves of 2009 influenza A H1N1 infection during the period 15 April through 30 August 2009 (wave 1) and 31 August 2009 through 2 January 2010 (wave 2).
Methods.To evaluate differences in epidemiologic features and outcomes during these pandemic waves, we examined prospective surveillance data on Wisconsin residents who were hospitalized ≥24 h with or died of pandemic H1N1 infection.
Results.Rates of hospitalizations and deaths from pandemic H1N1 infection in Wisconsin increased 4- and 5-fold, respectively, from wave 1 to wave 2; outside Milwaukee, hospitalization and death rates increased 10- and 8-fold, respectively. Hospitalization rates were highest among racial and ethnic minorities and children during wave 1 and increased most during wave 2 among non-Hispanic whites and adults. Times to hospital admission and antiviral treatment improved between waves, but the overall hospital course remained similar, with no change in hospitalization duration, intensive care unit admission, requirement for mechanical ventilation, or mortality.
Conclusions.We report broader geographic spread and marked demographic differences during pandemic wave 2, compared with wave 1, although clinical outcomes were similar. Our findings emphasize the importance of using comprehensive surveillance data to detect changing characteristics and impacts during an influenza pandemic and of vigorously promoting influenza vaccination and other prevention efforts.
During the first year of 2009 H1N1 influenza A virus (pandemic H1N1) circulation, the United States experienced distinct pandemic waves of infection occurring during the period 15 April through 30 August 2009 (wave 1) and 31 August 2009 through 2 January 2010 (wave 2), resulting in an estimated 41–84 million pandemic H1N1 infections, 183,000–378,000 hospitalizations, and 8330–17,160 deaths . Epidemiologic characteristics, clinical spectrum of illness, and risk factors for severe illness among patients hospitalized with pandemic H1N1 infection in the United States during wave 1 have been well defined [2–4]. However, to our knowledge, epidemiologic and clinical characteristics of patients hospitalized with pandemic H1N1 infection during wave 2, and comparisons of patients hospitalized during wave 2 with those hospitalized during wave 1, have not been published.
Wisconsin was severely affected during both pandemic waves. To better understand epidemiologic and clinical features of this pandemic, the Wisconsin Division of Public Health (WDPH) expanded its existing influenza surveillance program in April 2009 to include all hospitalizations and deaths from pandemic H1N1 infection. Using these surveillance data, the WDPH determined that wave 1 in Wisconsin disproportionately affected Milwaukee residents, racial and ethnic minorities, and children .
In this study, we compared epidemiologic and clinical features of patients who died of or were hospitalized with pandemic H1N1 infection during wave 2 with those of persons who died or were hospitalized during wave 1.
Since 15 April 2009, all Wisconsin acute care hospitals and local health departments have conducted prospective surveillance to detect patients who died of or were hospitalized with confirmed or probable pandemic H1N1 infection. All laboratories, hospitals, and health care providers were required to report pandemic H1N1 infections to the WDPH. WDPH staff regularly contacted local health departments and infection preventionists at Wisconsin hospitals to ensure complete reporting.
A case was defined as a hospitalization with duration ≥24 h or death in a Wisconsin resident with laboratory-confirmed or probable pandemic H1N1 virus infection with illness onset occurring during the wave 1 or wave 2 intervals. Confirmed infection was defined as detection of pandemic H1N1 in an oropharyngeal or nasopharyngeal swab specimen using a real-time reverse-transcriptase polymerase chain reaction (RT-PCR) assay conducted at a laboratory certified by the Centers for Disease Control and Prevention (CDC) to conduct confirmatory testing . Probable infection was defined as detection of influenza A using an RT-PCR assay with subtype not determined or available or as using another testing method (culture or direct fluorescent antibody or rapid antigen test) without subtyping results available. During both waves, the Wisconsin and Milwaukee public health laboratories provided confirmatory testing of hospitalized patients for free. During wave 2, most major Wisconsin hospital and health care system laboratories were certified to conduct confirmatory testing. Throughout both waves, the WDPH recommended that clinicians order confirmatory tests for all patients hospitalized with severe respiratory illness, patients who died of an acute illness suspected to be influenza, and pregnant women with signs and symptoms of influenza. During peak wave 2 activity, subtyping was not conducted on all influenza A–positive specimens, because surveillance data revealed that >99% of subtyped specimens were pandemic H1N1. Wave 2 cases included 6 deaths among nonhospitalized persons with autopsy or medical examiner's evaluations determining pandemic H1N1 infection to be the cause of death.
Each patient's medical records were reviewed by an infection preventionist, initially using a 16-page case report form developed by CDC staff  and later using an abridged version developed by WDPH staff. Both forms captured data on age, sex, race, ethnicity (Hispanic or non-Hispanic), residential address, clinical signs and symptoms at presentation, underlying medical conditions, radiographic findings, treatment course, and dates of hospitalization, discharge, onset of symptoms, and initiation of antiviral therapy. This surveillance study was approved by the WDPH as a public health response to a novel influenza virus and did not require institutional review board approval.
Data were analyzed using SAS, version 9.2 (SAS Institute). Incidence rates were calculated using United States Census Bureau 2008 population estimates . Ninety-five percent confidence intervals for rate ratios were calculated using Poisson regression. For time calculations, date of illness onset or date of hospital admission was considered to be day 0. Because some patients died outside the hospital or soon after admission, all patients who died were excluded from calculations of hospital length of stay. Body mass index (BMI) was calculated for nonpregnant patients aged ≥2 years for whom height and weight data were obtained from medical records. Obesity was classified as nonmorbid obesity (BMI = 30.0–39.9 kg/m2 for patients aged ≥18 years or BMI percentile ≥95% for patients aged 2–17 years) or morbid obesity (BMI ≥40.0 kg/m2 for patients aged ≥18 years).
For all variables, bivariate analysis was conducted to determine statistically significant differences between waves 1 and 2, and stratified analyses were conducted by age, race or ethnicity, and geography. Differences in proportions were evaluated using the Fisher exact or Pearson χ2 test, and analyses of trend were conducted using the Cochran-Mantel-Haenszel test. The Wilcoxon-Mann-Whitney test was used to compare distributions of continuous variables (ie, length of stay) between 2 independent samples. All reported P values were 2 sided and not adjusted for multiple testing. A P value of <.05 was considered to indicate a significant difference.
Multivariate logistic regression analysis was used to identify geographic, demographic, clinical, or hospital course–related characteristics that were independently associated with hospitalization during wave 2 compared with wave 1. Variables included were statistically significant in bivariate analyses. Stepwise selection was used to exclude collinear variables and choose the final model. Model fit was tested using the Hosmer-Lemeshow goodness-of-fit test.
During wave 1, a total of 252 hospitalizations (case classification, 232 confirmed cases and 20 probable cases, all RT-PCR positive for influenza A but not subtyped) and 9 deaths (8 confirmed cases and 1 probable case that was RT-PCR positive for influenza A but not subtyped) were reported to the WDPH. During wave 2, a total of 1077 hospitalizations (1034 confirmed cases and 43 probable cases [subtyping not determined or results unavailable: 16 RT-PCR assay, 9 direct fluorescent antibody assay, 8 culture, and 5 rapid antigen assay A positive, and 5 test method unknown]) and 46 deaths (all confirmed H1N1 cases) were reported. During wave 1, cases were primarily centered in Milwaukee, with low rates of hospitalization (Figure 1), intensive care unit (ICU) admission, and mortality among Wisconsin residents residing outside Milwaukee (non-Milwaukee) (Table 1). During wave 2, rates of hospitalization (Figure 1), ICU admission, and mortality among Milwaukee residents were similar to corresponding wave 1 rates (Table 1); however, among non-Milwaukee residents the hospitalization rate increased >10-fold, and ICU admission and mortality rates increased about 7- and 8-fold, respectively (Table 1).
Differences in pandemic H1N1 hospitalization rates between Milwaukee and non-Milwaukee residents were observed within sex, age, and race or ethnicity subpopulations (Table 1). Notably, non-Milwaukee hospitalization rates increased significantly among all subpopulations examined, with the largest rate increases observed among patients aged ≥65 years and among non-Hispanic whites (whites). In contrast, Milwaukee hospitalization rates increased significantly during wave 2 only among whites and substantially, although not significantly, among patients aged ≥65 years—groups that experienced the lowest rates during wave 1. Milwaukee hospitalization rates decreased significantly during wave 2 among patients aged <18 years and non-Hispanic blacks (blacks) and decreased substantially, although not significantly, among Asians—groups that experienced the highest wave 1 rates. Despite observed decreased hospitalization rates among Milwaukee residents and increased hospitalization rates among non-Milwaukee residents during wave 2, hospitalization rates among Milwaukee residents remained higher than those among non-Milwaukee residents for most subpopulations.
During wave 2, the proportion of hospitalized patients residing in Milwaukee decreased significantly, and the number of Wisconsin counties reporting pandemic H1N1 hospitalized cases or deaths increased 3-fold (Table 2). Furthermore, during wave 1, 48% of patients were black and 29% were white; this significantly changed during wave 2, when 15% of patients were black and 68% were white. Other significant demographic differences included an overall increase in median age and greater proportions of patients aged ≥50 years or reporting American Indian/Alaskan Native race or ethnicity. Significantly lower proportions of patients during wave 2 were aged <18 or 18–49 years or were black, Hispanic, or Asian.
The most common signs and symptoms reported during both waves were fever, cough, and influenza-like illness (fever plus cough or sore throat) (Table 3). Significant increases in proportions of patients reporting fever, cough, influenza-like illness, or myalgia during wave 2 were noted. The most common underlying medical conditions during both waves were asthma and diabetes. About one-fourth of patients during both waves reported no underlying conditions; among these patients, 59% were aged <18 years, whereas among patients with underlying conditions, only 25% were aged <18 years (P<.001). During both waves, approximately 27% of women of childbearing age were pregnant. Significant differences between waves included increased proportions of patients with chronic obstructive pulmonary disease (COPD) and neurologic conditions and a decreased proportion of patients with hematologic conditions during wave 2. Patients with hematologic conditions (primarily sickle cell disease) were more likely than other patients to report black race or ethnicity (72% vs 21%; P<.001) and age <18 years (47% vs 33%; P=.02); and patients reporting COPD were more likely than other patients to report white race or ethnicity (76% vs 56%; P<.001) and age ≥50 years (76% vs 26%; P <.001). The proportion of patients with nonmorbid obesity was significantly greater during wave 2, although no significant difference was observed in the proportion with morbid obesity.
Changes in frequency of reported signs and symptoms from wave 1 to wave 2 were age and race specific. Significant increases in frequency of fever (74% vs 85%; P=.01), cough (67% vs 84%; P<.001), and influenza-like illness (63% vs 76%; P=.01) were observed only among patients aged 18–49 years. A significant increase in myalgia occurred only among patients aged <18 years (3% vs 12%; P=.01). Among black patients, frequencies of cough (60% vs 82%, P<.001) and myalgia (19% vs 31%; P =.03) increased. Wave 1 to wave 2 differences in underlying conditions were age specific. Significant decreases in the proportion of cases involving patients with hematologic conditions occurred among patients aged <18 years (9% vs 3%; P=.04) and 18–49 years (7% vs 2%; P=.02); a significant increase in the proportion of cases involving patients with COPD occurred among patients aged ≥50 years (14% vs 31%; P=.005).
During both waves, most patients had abnormal chest radiographic imaging findings, the median duration of hospitalization was 3.0 days, approximately 20% of patients required admission to an ICU, approximately 15% required invasive mechanical ventilation, and 4% died (Table 4).
The proportion of patients experiencing acute respiratory distress syndrome (ARDS) was significantly lower during wave 2 than during wave 1 (Table 4). Among persons aged <18 years, ARDS occurrence (16% vs 4%; P<.001), length of hospital stay (median, 3 vs 2 days; P=.02), and death (4% vs 1%; P=.01) decreased significantly during wave 2. Among race and ethnicity groups, ARDS occurrence decreased among whites (21% vs 9%; P=.007) and length of stay decreased among Hispanics (median, 3 vs 2 days; P=.04). Other differences in outcomes observed between waves 1 and 2 were not statistically significant. Mechanical ventilation and ICU admission increased by 3% among Asians and blacks, respectively. All other outcomes decreased in frequency or did not change among race and ethnicity groups.
Most patients received antiviral and antibiotic medications; the proportions of patients treated with these medications were similar during waves 1 and 2 (Table 4). However, there was a significant wave 2 increase in the proportion of patients aged ≥50 years treated with antiviral medications (76% vs 87%; P=.04). Furthermore, among all patients, times from illness onset to receipt of antiviral medication, hospital admission to receipt of antiviral medication, and illness onset to hospitalization were significantly shorter during wave 2 compared with wave 1 (Table 4).
This regression analysis included 802 cases with complete data available to identify independent predictors of hospitalization during wave 2 compared with wave 1. All variables included were significant in bivariate analysis (Tables 2–4) and remained in the model after stepwise selection. Symptoms were excluded from analysis because of high correlation with age. Times from onset to hospitalization and from a hospitalization to receipt of antivirals were excluded because these variables were collinear with time from onset to receipt of antivirals. Obesity status was excluded because obesity data were missing for 40% of patients. The Hosmer-Lemeshow statistic indicated no lack of fit for the model (P=.48).
This regression analysis revealed that persons hospitalized during wave 2 were 4.6 times more likely to be non-Milwaukee residents than those hospitalized during wave 1 (Table 5). Wave 2 patients were also more likely to be aged ≥50 years and were 3 times less likely to be black, 2 times less likely to be Hispanic, and 5 times less likely to be Asian than were wave 1 patients. During wave 2 compared with wave 1, patients were less than half as likely to develop ARDS or to start receiving antiviral medications >96 h (vs <48 h) after symptom onset.
The 2009 H1N1 influenza pandemic occurred in Wisconsin as 2 distinct waves during the first year of virus circulation. Wave 1 was driven by an intense outbreak in Milwaukee, where rates of hospitalization, ICU admission, and death were 7–15-fold greater than elsewhere in Wisconsin. We found that children aged <18 years and members of racial and ethnic minorities were disproportionately affected during wave 1, with the highest hospitalization rates observed among non-Hispanic black, Hispanic, and Asian Milwaukee residents. In contrast, the impact of wave 2 was experienced broadly and associated with 4-fold more hospitalizations and 5-fold more deaths statewide than during wave 1. Hospitalization rates during wave 2 remained higher in Milwaukee and among members of racial and ethnic minority groups. However, the subpopulations most severely affected during wave 1 in Milwaukee, particularly patients aged <18 years and black patients, experienced relatively reduced hospitalization rates during wave 2. Accordingly, frequencies of underlying conditions highly associated with specific ages, races, or ethnicities, particularly COPD and hematologic conditions, were significantly different between waves 1 and 2.
The successive waves of pandemic H1N1 infection in Wisconsin resembled those observed during the 1918 and 1968 pandemics, when second waves were associated with higher rates of morbidity and mortality than first waves [7–9]. Evidence that the pandemic H1N1 virus was genetically stable, with no increased virulence between waves 1 and 2 , suggests that the increased magnitude of wave 2 was attributable to broader geographic spread of pandemic H1N1 to immunologically naive populations throughout Wisconsin. In contrast to relatively limited geographic spread of pandemic H1N1 in wave 1 that occurred during warm weather and as the school year ended, the broader geographic spread during wave 2 may partially be related to colder temperatures and more time for widespread transmission to occur in schools, factors previously noted to increase influenza virus transmission [11, 12].
Although we observed extensive spread of pandemic H1N1 throughout Wisconsin during wave 2, hospitalization and mortality rates within Milwaukee remained disproportionately greater than in other Wisconsin areas during both waves. These findings resemble those during the 1918 influenza pandemic, when higher mortality rates were observed among urban populations compared with rural populations [13, 14]. A possible contributing factor is Milwaukee's high proportion of racial and ethnic minority residents, who are more likely than white residents to reside in densely populated areas, to have lower socioeconomic status, and to have medical conditions that are risk factors for complications from seasonal influenza [15–17]. The hospitalization rate among Milwaukee's white residents was much lower than rates among racial and ethnic minority residents, particularly during wave 1. This disparity might also reflect patterns of social mixing, because Milwaukee is among the most segregated large metropolitan areas for blacks and Hispanics in the United States .
Minority populations that were severely affected during wave 1 likely developed high levels of infection-acquired immunity, which consequently provided some degree of protection during wave 2. Accordingly, although hospitalization rates among white Milwaukee residents increased 2.5-fold from wave 1 to wave 2, rates remained constant among Hispanic Milwaukee residents and decreased significantly among black Milwaukee residents. Similarly, during the 1918 pandemic, army camps comprising troops exposed during the first wave had significantly lower rates of clinical illness and mortality during the second wave, compared with camps with higher proportions of previously unexposed troops .
In Wisconsin, hospitalization rates among black, Hispanic, Asian, and American Indian/Alaskan Native residents were substantially greater than among white residents. These disparities occurred during both waves and were not restricted to urban Milwaukee; they were found throughout Wisconsin and much of the United States [20–22]. Studies suggest that racial, ethnic, and socioeconomic disparities in mortality also occurred during the 1918 pandemic , demonstrating a need for additional study regarding origins of these disparities. Regardless of cause, these findings underscore the importance of promoting influenza vaccination among racial and ethnic minority populations, particularly considering evidence that these groups have lower rates of vaccination against seasonal influenza .
Planning for continued pandemic H1N1 transmission and for future influenza pandemics should consider the vulnerability of immunologically naive urban and rural populations. To identify these populations, surveillance systems must provide sufficient coverage and geographic detail to detect local and regional outbreaks and changes in influenza activity. Identifying communities and subpopulations that escaped substantial impact during a pandemic wave should be as important to public health planning as identifying those that were severely affected.
In Wisconsin and elsewhere, the 2009 H1N1 pandemic disproportionately affected children [2, 25, 26]. Children are important drivers of influenza virus transmission and were found to be highly susceptible to pandemic H1N1 infection, compared with older persons, particularly those aged >65 years, many of whom might have had partial immunity related to exposure to previously circulating influenza viruses [27, 28]. We also noted that the disease burden in Wisconsin among these older populations varied geographically and temporally, similar to characteristics of the 1918 pandemic in 2 Mexican cities . Although children had the highest hospitalization rates in Wisconsin during both waves, hospitalization rates among older persons significantly increased during wave 2. This trend resembles those noted during successive waves in previous influenza pandemics  and, coupled with increased case-fatality ratios associated with pandemic H1N1 infection among older age groups, reinforces the need for sustained vaccination efforts targeting all age groups .
Older patients also had a higher prevalence of underlying conditions and thus were more likely to experience more severe illness. Wave 2 in Wisconsin was associated with a dramatic increase in the presence of COPD as an underlying condition among hospitalized patients, a condition generally affecting adults . This increase likely was partially attributable to the increased proportion of older patients among those hospitalized during wave 2 and to other seasonal effects resulting in hospitalization among patients with COPD. This was also shown in our multivariate logistic regression model, in which the difference in COPD frequency between waves became statistically nonsignificant, likely revealing confounding by the change in age distribution. Although increased proportions of older patients and hospitalized patients with COPD should have resulted in higher proportions of severe outcomes during wave 2, these potential effects were countered by improved treatment, including increased proportions of patients aged ≥50 years receiving antiviral treatment and decreased times to hospitalization and receipt of treatment. In addition, improved treatment likely resulted in significantly improved outcomes involving ARDS among children and specific race or ethnicity groups.
Our study has several limitations. Because we used hospitalizations as the primary measure of the impact of pandemic H1N1 infection, and hospitals in Wisconsin vary in size, resources, case load, and influenza screening and hospitalization practices, some differences in rates according to geographic region, age, or racial or ethnic group might reflect differences between hospitals. In addition, because our data were limited to hospitalized patients, we were unable to conduct analyses and draw conclusions for the entire infected population, specifically regarding hospitalization and death rates. Also, without complete obesity and smoking data for all patients, we could not accurately estimate the prevalence of these factors or their association with disease severity. Furthermore, we were unable to obtain data regarding patient socioeconomic status, insurance type, or other information that could have enhanced our understanding of the demographic disparities identified in this analysis. Finally, our study describes the unique experience within Wisconsin, which was severely affected during both pandemic waves and experienced substantial geographic variation in impacts of the pandemic. Thus, our results might not apply to all areas.
Our study also has substantial strengths. Because we included all reported pandemic H1N1 hospitalizations during both waves in Wisconsin, we could calculate population-based hospitalization rates and describe geographic and demographic variations in disease incidence. In addition, consistent surveillance methods were used throughout the state of Wisconsin during the entire study period. This surveillance was strengthened by a WDPH recommendation to test all hospitalized patients with suspected influenza for pandemic H1N1 infection and by the availability of free RT-PCR confirmatory testing at Wisconsin public health laboratories, standardized case reporting, and an electronic, World Wide Web–based disease surveillance system that local health officials, hospital staff, and laboratories used to directly report cases. Also, regular communication by WDPH staff with Wisconsin local health departments and acute care hospitals resulted in consistent ascertainment of cases. Together, these efforts minimized detection bias and permitted us to accurately compare data from both waves.
Whether continued transmission of pandemic H1N1 will be associated with another wave of infection or with more typical seasonal transmission is currently unknown. Nonetheless, the disproportionate effect of pandemic H1N1 infection on many groups and regions in Wisconsin during both waves underscores the need to vigorously promote vaccination among all populations. In addition, because of the changing characteristics and impacts of successive influenza pandemic waves, comprehensive surveillance is necessary to guide influenza vaccination efforts and pandemic response planning, thereby reducing the morbidity and mortality associated with 2009 H1N1 and future influenza pandemics.
This work was supported by the Wisconsin Division of Public Health with funding from the Centers for Disease Control and Prevention Public Health Emergency Response and Epidemiology and Laboratory Capacity cooperative agreements.
We are indebted to Wisconsin clinicians, infection preventionists, and laboratory staff for diligently submitting case reports and responding to requests for information and to our local public health partners who coordinated local case investigations. We also recognize the contributions of the following WDPH surveillance staff: Susann Ahrabi-Fard, MS; Jean Druckenmiller; Steven Gilbert; Amanda Hardy; Kristin Hardy; Diep Hoang-Johnson; James Kazmierczak, DVM, MS; Katyelyn Klein; Rachel Klos, DVM, MPH; Carrie Nielsen, PhD; and Christopher Steward.