Clinical Characteristics and 30-Day Outcomes for Influenza A 2009 (H1N1), 2008-2009 (H1N1), and 2007-2008 (H3N2) Infections FREE

The pandemic 2009 influenza A(H1N1) virus caused widespread transmission in the United States and other countries. The Centers for Disease Control and Prevention (CDC) estimates that 43 million to 89 million infections occurred in the United States from April 2009 through April 10, 2010, with mid-range estimates of 274 000 H1N1-related hospitalizations and 12 470 deaths (http://www.cdc.gov/h1n1flu/estimates_2009_h1n1.htm). Children, young adults, pregnant women, and individuals with underlying chronic medical conditions appear to have a higher risk of hospital admission and critical illness when infected with the pandemic virus.^1- 6 Serologic studies suggest that most children and young adults do not have preexisting cross-reactive antibodies against 2009 H1N1 and that they are highly susceptible to infection.⁷

It is difficult to compare the spectrum of illness and outcomes for 2009 H1N1 and seasonal influenza A infections because most reports of 2009 H1N1 influenza have been based on surveillance reports, particularly those for hospital admissions and fatalities.^5,8- 11 These reports have provided valuable descriptive information, but differing criteria for influenza testing by season and the lack of uniform standards for data collection and reporting limit comparisons with other influenza viruses. The objective of this study was to compare the characteristics of pandemic and seasonal influenza A infections occurring in a defined population, identified using consistent enrollment and laboratory methods.

The source population included approximately 50 000 persons living in 14 zip codes surrounding Marshfield, Wisconsin. In this area, nearly all residents receive their inpatient and outpatient care from Marshfield Clinic facilities, which use an electronic medical record that captures 90% of outpatient visits, 99% of deaths, and 95% of hospital discharges for the population.^12- 13 This population and the enrollment procedures described herein have been used to estimate the effectiveness of influenza vaccines for several seasons.^14- 15 Major types of insurance coverage for the population include private commercial (54%), Medicaid (10%), Medicare (10%), and mixed public/private insurance (15%).

For each season, we restricted enrollments to community-dwelling individuals with at least 12 months of continuous residency (or since birth if <12 months old) to ensure availability of relevant data in the electronic medical record. The cohort of individuals who met residency criteria was identified at the beginning of each enrollment period, and only those individuals were eligible to be screened and enrolled during a clinical encounter. We classified individuals as having a chronic medical condition if they had 2 or more visits to the Marshfield Clinic during the prior calendar year with an International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis code for an underlying chronic disease conferring elevated risk of influenza complications.¹⁴ Receipt of seasonal influenza vaccine and 2009 H1N1 vaccine was determined by a real-time, Internet-based immunization registry (http://www.recin.org) that captures 95% of all influenza vaccinations in the study population.¹⁶

Among persons in our source population meeting residency criteria, those presenting for an outpatient or inpatient health care visit with at least 1 of the symptoms of subjective fever, chills, or cough were screened and enrolled by trained research coordinators. Physicians providing care had no role in identifying or testing patients for inclusion in this study. Patients with illness duration of 8 days or longer were excluded because of the potential for false-negative influenza test results after prolonged illness.^17- 18

Eligible patients were recruited at the main Marshfield Clinic campus, the Colby satellite outpatient facility, and an acute care hospital contiguous with Marshfield Clinic (St Joseph's Hospital). This hospital provides nearly all inpatient care for the source population. Some patients were contacted at home on the following day if they were not approached during the clinical encounter. Those who consented were screened for eligibility and tested for influenza infection. Each participant (or parent) completed a short interview to assess symptoms, self-reported severity, and onset date.

A Dacron-tipped nasopharyngeal swab (adults and children ≥13 years old) or nasal swab (children 6 months to 12 years old) was obtained from all participants. Nasal swabs were obtained from children aged 12 years or younger because they were more acceptable to parents and children and because studies have suggested that nasal and nasopharyngeal samples have comparable sensitivity for influenza detection in children.^19- 20 During 4 seasons, we found that the proportion of children with a positive influenza real-time reverse transcriptase polymerase chain reaction (rRT-PCR) result was the same for participants aged 7 to 12 years (nasal swab) and those aged 13 to 19 years (nasopharyngeal swab), suggesting that using the nasal swab did not adversely affect influenza detection in children.

Most swabs were obtained by research coordinators, but an aliquot was obtained from physician-collected swabs when available to avoid collection of multiple swabs. Swabs were placed in M4 viral transport media and refrigerated or stored on ice until they were delivered to the laboratory on the same day. Samples were routinely tested within 1 day after collection; samples obtained on weekends were processed on Mondays. Total nucleic extractions were performed using the Roche MagNA Pure Total Nucleic Acid Kit (Roche Diagnostics, Indianapolis, Indiana) on 200 μL of clinical sample. Real-time RT-PCR was performed on nucleic acid extracts using the Roche LightCycler 480 Real-Time PCR System. All rRT-PCR protocols and probe and primer sequences were provided by the CDC (protocols available from the CDC on request).²¹ Real-time RT-PCR subtyping was performed on all samples with a positive influenza A result (crossing threshold of <40 cycles) during each season.

Patients were recruited for a 10-week period beginning January 21, 2008, for the 2007-2008 season and for a 12-week period beginning January 19, 2009, for the 2008-2009 season. Recruitment and testing was restarted on May 4, 2009, for pandemic influenza surveillance and continued until November 6, 2009 (27 weeks). Enrollment was then suspended to complete a previously scheduled software update of the research application that is used for screening, enrollment, and sample tracking. During the pandemic enrollment period, the number of recruitment days per week varied from 2 to 5 based on influenza activity as documented by rRT-PCR testing, with fewer days of recruitment per week when influenza activity was absent. During the seasonal enrollment periods, patients were recruited and tested 6 days per week. Patients could be reenrolled with a new respiratory illness after an exclusion period to allow time for recovery from the first illness episode.

We compared the number of influenza A cases in study participants with the weekly incidence of all medically attended influenza-like illness in the source population. International Classification of Diseases, Ninth Revision, Clinical Modification diagnosis codes were used to estimate the incidence of influenza-like illness. For adults, these codes included 487 (influenza), 078.8 (other specified viral and chlamydial diseases), 780.6 (fever), 079.9 (unspecified viral and chlamydial infections), or 465.9 (acute upper respiratory tract infection). In multivariable analyses of adult enrollments in 4 prior seasons, each of these codes was significantly associated with influenza detection by rRT-PCR performed independent of physician evaluation and diagnostic coding. The following codes were used to define a pediatric encounter for influenza-like illness: 487, 079.9, 078.8, 780.6, 490 (bronchitis), 462 (acute pharyngitis), or 786.2 (cough).

Medical records were reviewed for each participant with rRT-PCR–confirmed influenza A to ascertain treatment, hospital admission, and complications during the 30-day period after illness onset. For this analysis, an episode of influenza-related pneumonia required a physician diagnosis, antimicrobial treatment for pneumonia, and an opacity or infiltrate on chest radiograph that was not known to be chronic.²² Diagnoses without radiographic confirmation were excluded to limit misclassification due to lack of standardized clinical criteria for pneumonia diagnosis. A physician (E.A.B.) reviewed all cases of pneumonia and all hospital admissions. A serious outcome was defined as a pneumonia diagnosis or hospital admission within 30 days after onset.

A subjective severity score was calculated for adults and children aged at least 24 months based on self-reported (or parent-reported) severity of 12 symptoms: cough, subjective fever, chills, fatigue, nasal congestion, wheezing, vomiting, headache, muscle aches, sore throat, ear pain, and nausea. Each symptom was scored on a 4-point scale (0 = absent; 1 = mild; 2 = moderate; and 3 = severe). The severity score was calculated by summing the points for the 12 individual symptoms. Possible scores ranged from 1 (a single mild symptom) to 36 (all symptoms severe). The score has not been validated against objective measures of influenza severity, but a similar severity score (based on 7 symptoms) was significantly reduced by oseltamivir treatment in a randomized clinical trial.¹⁸

We estimated the minimum cumulative incidence of medically attended influenza A in the source population for each strain and time period, as well as the cumulative incidence of hospital admission. Separate incidence estimates were generated for children and adults. Incident cases included all participants with rRT-PCR–confirmed influenza A during each enrollment period (for medically attended influenza incidence) and all hospital admissions within 30 days after onset of influenza A illness (hospital admission incidence). The denominators included all individuals in the 14–zip code area who met the age and residency criteria. These data underestimate the cumulative incidence because some patients declined to participate or were not approached by research staff during the clinical encounter, particularly during fall 2009, when the volume of patients with respiratory illness exceeded the capacity of research staff to perform outpatient enrollments.

We compared the clinical and demographic characteristics of patients with 2009 H1N1 infection with those with seasonal H1N1 and H3N2 infection. Cases of influenza B were identified during the 2007-2008 season and the 2008-2009 season, but there were no influenza B infections during the pandemic enrollment period. Influenza B cases were not included in this analysis because our objective was to compare 2009 H1N1 with other influenza A strains. Most subgroup comparisons were preplanned, although there were some post hoc comparisons such as that of severity score across virus types stratified into 2 different pediatric age groups (5-17 years vs <5 years). Medians and interquartile ranges (IQRs) were calculated for continuous variables. Univariate comparisons were made using the χ² test for categorical variables and the nonparametric Wilcoxon rank sum test for continuous variables. The significance level was set at P <.05 based on a 2-sided test. Analyses were performed using SAS software, version 9.1 (SAS Institute Inc, Cary, North Carolina); confidence intervals (CIs) for proportions were calculated using WINPEPI.²³

The study procedures were reviewed and approved by the Marshfield Clinic Institutional Review Board. All participants or parents provided oral consent for influenza testing and received written information on study procedures. Documentation of written consent was waived by the institutional review board because the research presented no more than minimal risk and did not involve procedures for which written consent is normally required outside of the research context.

Eighty-six percent of eligible patients enrolled during the 2009 pandemic enrollment period (May 4–November 6), 74% during the 2008-2009 influenza season, and 70% during the 2007-2008 season. We identified 2009 H1N1 influenza in 545 (24%) of 2280 enrolled patients (2136 unique individuals) with subjective fever, chills, or cough from May 4 through November 6, 2009. Illness due to 2009 H1N1 was clustered in October and early November. The temporal distribution of influenza A cases in study participants was consistent with the weekly incidence of medically attended influenza-like illness in the source population (Figure). By early December, influenza activity had declined to normal preseason levels. No seasonal H1N1 or H3N2 infections were identified during the pandemic enrollment period.

In the 2008-2009 season, we identified 221 seasonal influenza A H1N1 infections among 262.2 participants enrolled before the pandemic period. During the 2007-2008 influenza season, we identified 632 cases of influenza A H3N2 among 1972 participants. Real-time RT-PCR subtyping of all positive samples demonstrated that seasonal H1N1 was the only influenza A subtype in study participants during the 2008-2009 season and H3N2 was the only subtype in study participants during the 2007-2008 season.

Visits for pandemic 2009 H1N1 influenza occurred in a narrow time window relative to seasonal influenza A infections: 98% of 2009 H1N1 cases were enrolled in the 5-week period of highest incidence (ie, 2 weeks before and after the week with the greatest number of rRT-PCR–confirmed cases). In contrast, 61% of seasonal H1N1 cases and 63% of H3N2 cases were enrolled during the 5-week period of highest incidence.

The minimum cumulative incidences of medically attended influenza A were 34.9 cases per 1000 children for 2009 H1N1, 12.0 for seasonal H1N1, and 21.7 for H3N2. The risk ratio for medically attended influenza A in children vs adults was 8.7 (95% CI, 7.2-10.6) for 2009 H1N1, 5.2 (95% CI, 3.9-6.8) for seasonal H1N1, and 2.2 (95% CI, 1.9-2.6) for H3N2. The median ages were 10 years (IQR, 6-20 years), 11 years (IQR, 6-36 years), and 25 years (IQR, 9-45 years) for participants with 2009 H1N1, seasonal H1N1, and H3N2 infection, respectively (P < .001).

The distribution of symptoms was similar for patients with 2009 H1N1 infection compared with those infected with seasonal H1N1 or H3N2 strains (Table 1). Children and adolescents infected with the 2009 H1N1 strain were less likely to have sore throat compared with those infected with seasonal H1N1 (P = .03) or H3N2 (P < .001) strains. Nausea and vomiting occurred with similar frequency by strain and time period. We did not record diarrheal symptoms. Among participants older than 24 months, the median symptom severity score was 14 (IQR, 10-19) for 2009 H1N1, 16 (IQR, 13-20) for seasonal H1N1, and 17 (IQR, 13-20) for 2007-2008 H3N2 infections. The severity score was significantly lower (ie, milder perceived illness) in children with 2009 H1N1 infection relative to those with H3N2 infection (P = .009). We found a significant association between virus type and severity score for children 5 to 17 years old but not for children younger than 5 years. Among children 5 to 17 years old, the median symptom severity score was 14 (IQR, 10-18) for 2009 H1N1 infections and 15 (IQR, 13-18) for H3N2 infections (P < .001). Among adults, the symptom severity score was also significantly lower for 2009 H1N1 infections (median, 16; IQR, 12-20) compared with H3N2 infections (median, 18; IQR, 14-21; P = .04).

For each influenza A strain, antiviral agents were prescribed more often for adults than children. Both children and adults infected with 2009 H1N1 viruses were significantly more likely to receive antiviral therapy relative to those infected with 2008-2009 H1N1 viruses, a strain known to be resistant to oseltamivir.²⁴ Antiviral use was not significantly different for 2009 H1N1 infections compared with H3N2 infections in children or adults.

Twelve patients with 2009 H1N1 infections were hospitalized within 30 days after symptom onset. Influenza A was a known or suspected contributing factor for all 12 admissions. The most common pediatric discharge diagnosis was dehydration, present in 4 of 6 hospitalized children and adolescents. Among adults, discharge diagnoses included pneumonia (n = 2), exacerbation of chronic obstructive pulmonary disease with hyponatremia (n = 1), and influenza (n = 3). No children or adults with 2009 H1N1 were admitted to the intensive care unit and no deaths occurred.

The cumulative incidence of hospital admission within 30 days of onset (per 1000 residents) was 0.25 (95% CI, 0.1-0.4) for 2009 H1N1, 0.15 (95% CI, 0.1-0.3) for seasonal H1N1, and 0.5 (95% CI, 0.3-0.7) for H3N2. In children, 2009 H1N1 infection was not associated with either hospital admission or pneumonia compared with H1N1 or H3N2 (Table 2). Among adults, pneumonia occurred in 4.0% (95% CI, 1.6%-8.1%) of those with 2009 H1N1 infection and 1.1% (95% CI, 0.3%-2.7%) of those with H3N2 infection (P = .03). There were no significant differences by strain in the proportion of children or adults with any serious outcome (pneumonia or hospital admission) during the 30 days after onset. Among adults with any serious outcome, the median age was 31.9 years (IQR, 24.8-37.6 years) for 2009 H1N1, 30.5 years (IQR, 29.5-43.7 years) for seasonal H1N1, and 63.0 years (IQR, 28.1-71.2 years) for H3N2 (P = .02 for 2009 H1N1 vs H3N2). The median ages of children with any serious outcomes were 5.8 years (IQR, 4.1-8.1 years), 4.9 years (IQR, 2.5-6.4 years), and 8.3 years (IQR, 3.1-12.5 years), respectively.

We compared the characteristics of patients with 2009 H1N1 infections and those with noninfluenza respiratory illness (negative rRT-PCR result) during the pandemic enrollment period (Table 3). The median age was 10.4 years (IQR, 5.8-20.2 years) for those with 2009 H1N1 infection and 18.1 years (IQR, 5.7-42.2 years) for those with noninfluenza respiratory illness (P < .001). Among children younger than 18 years, those infected with 2009 H1N1 were significantly older than those with a noninfluenza respiratory illness. Children with 2009 H1N1 infections were also more likely to have had a prior diagnosis of asthma compared with those testing negative for influenza. For both children and adults, symptom severity scores were significantly higher among those infected with the 2009 H1N1 strain relative to those with noninfluenza respiratory illness.

Thirty-one children (23%) with seasonal H1N1 infection and 61 (24%) with H3N2 infection were fully immunized with seasonal influenza vaccine at the time of enrollment (vaccine received at least 14 days before enrollment). Among adults, 25 (29%) with seasonal H1N1 and 117 (31%) with H3N2 infection had received seasonal influenza vaccine at least 14 days before enrollment. Distribution of monovalent H1N1 vaccine began as the 2009 H1N1 outbreak was reaching its peak in our population; no participants infected with the 2009 H1N1 strain and only 4 with a negative rRT-PCR result had received monovalent vaccine 14 or more days before enrollment.

We were able to directly compare the epidemiologic and clinical characteristics of 2009 H1N1 and seasonal influenza A infections in the same population using identical enrollment, laboratory diagnostic, and follow-up methods. Quiz Ref IDWe found that participants with the 2009 H1N1 strain were similar in age to those with seasonal H1N1 infection, but participants with H3N2 infection were significantly older. This age differential between H1N1 and H3N2 is consistent with sentinel surveillance reports from Australia, although the patients with 2009 H1N1 infection in Australia were older (median, 18-22 years) than those identified in this study (median, 10 years).²⁵ The minimum cumulative incidence of medically attended influenza in the population was higher in children than adults for every influenza A strain, and the relative increase in children was highest for the 2009 H1N1 strain.

Quiz Ref IDOur results suggest that the clinical manifestations and risk of hospital admission are similar for 2009 H1N1 and other seasonal influenza A strains among those presenting for medical care and documented to have influenza infection. Our finding that adults with 2009 H1N1 infection had an increased risk of radiographically confirmed pneumonia compared with those with H3N2 infection requires confirmation in a different population. Other published studies have reported a higher-than-expected incidence of hospitalization and death associated with 2009 H1N1 infection, particularly in children.^3,10,26 This finding may be due to the greatly elevated incidence of 2009 H1N1 influenza in a highly susceptible population of children and young adults rather than increased virulence of 2009 H1N1 relative to seasonal influenza A viruses. The increased use of highly sensitive rRT-PCR testing for influenza in hospitalized patients during the pandemic period was another factor that may have contributed to improved case detection in severely ill patients during 2009 relative to prior seasons.

In this study, comparisons of severity by influenza A strain could have been influenced by antiviral therapy, vaccination, and differences in health care–seeking behavior. Quiz Ref IDAntiviral treatment was used more often for patients infected with 2009 H1N1 compared with seasonal H1N1, and this would be expected to reduce serious outcomes. However, antiviral therapy is unlikely to explain any differences between 2009 H1N1 and H3N2 outcomes, since the proportion of infected participants receiving antiviral therapy was similar. Influenza vaccination may have also reduced the occurrence of serious outcomes in vaccinated patients with seasonal H1N1 or H3N2 strains but not those with 2009 H1N1 infections, since monovalent H1N1 vaccine was first available in our population when H1N1 infections were peaking in mid October. Quiz Ref IDFinally, publicity surrounding the H1N1 pandemic may have caused more mildly ill individuals to seek health care compared with a normal influenza season. If this occurred, a smaller proportion of pandemic cases would be expected to be hospitalized or diagnosed as having pneumonia compared with seasonal cases. However, even if present, this effect is likely to be modest because we did not observe significant differences in symptom severity score or mean interval from symptom onset to swab in children or adults with 2009 H1N1 relative to 2008-2009 seasonal H1N1 infection.

Our study had several strengths, including prospective inpatient and outpatient enrollment from a defined population presenting for medical care, use of standardized screening criteria for rRT-PCR testing, collection of detailed data regarding clinical manifestations of illness, and medical record review to ascertain influenza-related pneumonia and hospitalizations. This study also has several limitations, including the aforementioned differences in antiviral use and vaccination. The cumulative incidence proportions underestimate the actual incidence of medically attended influenza in the population, particularly during October 2009, when patient volume exceeded the capacity of research staff to enroll outpatients. Although the source population was approximately 50 000, analyses of serious outcomes (pneumonia or hospital admission) had limited power because of the low number of events. We cannot conclude that there was no difference; however, if the clinical symptoms and severity were truly different, the actual magnitude of difference would be small and likely not of clinical relevance. We were also unable to compare outcomes and severity for high-risk individuals, including pregnant women. We did not enroll children younger than 6 months because data from the 2007-2008 and 2008-2009 seasons were originally obtained for a study to assess vaccine effectiveness and children younger than 6 months are not eligible for influenza vaccination. One study showed that children in this age group had a high rate of hospital admission for 2009 H1N1 infection.¹⁰ Finally, our source population was predominantly white and non-Hispanic, and we could not examine racial and ethnic subgroups across influenza subtypes. Other studies have suggested that certain racial or ethnic groups experienced higher rates of hospital admission or other complications because of 2009 H1N1 infection.^26- 27

In summary, we found that children were disproportionately affected by 2009 H1N1 infection, but the perceived severity of symptoms and risk of serious outcomes (pneumonia or hospital admission) were not increased in children with 2009 H1N1 infection relative to seasonal influenza A viruses. Quiz Ref IDThis study demonstrates the benefit of ongoing active influenza surveillance in a defined cohort with standardized testing criteria and uniform collection of clinical and epidemiologic data. The use of consistent enrollment and testing procedures offers the opportunity to directly compare illness patterns and outcomes over multiple seasons, and these results complement information obtained through traditional public health surveillance systems.