Comparison of Pandemic (H1N1) 2009 and Seasonal Influenza, Western Australia, 2009

TOC summary: Infections were similar in terms of symptoms, risk factors, and proportion of patients hospitalized.

There are little data directly comparing confi rmed pandemic (H1N1) 2009 with contemporaneous seasonal infl uenza over the same infl uenza season (4)(5)(6). Many of the reports on the epidemiology of infl uenza in 2009 to date have focused exclusively on pandemic (H1N1) 2009 or have used limited laboratory-based surveillance data on isolation rates for seasonal and pandemic (H1N1) 2009 infl uenza viruses (7)(8)(9)(10)(11)(12)(13). Other reports have compared pandemic (H1N1) 2009 and seasonal infl uenza infections that occurred outside the usual infl uenza season (14). Still other investigators have compared various indicators of infl uenza severity during the current pandemic with historical data from previous annual infl uenza epidemics (15,16). Interpretation of such comparisons is challenging because of variation in infl uenza activity from season to season. Furthermore, heightened awareness surrounding the current pandemic may have affected patient care-seeking behavior or physician diagnostic practices, thus potentially creating bias in year-to-year comparisons. Examining confi rmed pandemic (H1N1) 2009 and seasonal infl uenza infections occurring in the same population during the 2009 infl uenza season enables a more straightforward comparison.
We interviewed persons with laboratory-confi rmed pandemic (H1N1) 2009 or seasonal infl uenza infection over a 10-week period encompassing the peak of the winter infl uenza season. This effort enabled us to directly compare the clinical illness and predisposing medical risk factors associated with pandemic (H1N1) 2009 and seasonal infl uenza infections diagnosed contemporaneously from the general population of Western Australia, which has a population of 2.2 million persons.

Methods
All clinical laboratories report positive infl uenza test results to the Communicable Disease Control Directorate  Figure 1) (18). From the study inception through July 13, 2009, attempts were made to interview every patient with confi rmed infl uenza illness reported to CDCD. On July 14, 2009, we instituted a sampling framework because of increasing numbers of reported infections. The sampling scheme entailed identifying the last digit of a sequentially-assigned identifi cation number from the fi rst patient reported each day with seasonal or pandemic (H1N1) 2009 infl uenza, then selecting all patients reported that day with the same last digit. If <20 patients were identifi ed for interview, we added 1 to the digit (n + 1) and selected additional patients by using the same procedure. This process was repeated until up to 20 persons with seasonal infl uenza and 20 with pandemic infl uenza were chosen each day. If <20 infl uenza infections were reported on a given day for either seasonal or pandemic (H1N1) 2009 infl uenza, we attempted to interview all patients reported on that day. Study participants were interviewed by a trained nurse who used a standard questionnaire. If a patient was unable to answer questions or was <18 years of age, the nurse interviewed a parent or other family member familiar with the patient's situation. We made 6 attempts to contact the patient or a proxy, after which the patient was considered not contactable. Diagnostic specimens of participating patients were collected a median of 2 d after illness onset (interquartile range [IQR] 1-3 d), and patients were interviewed a median of 6 d after onset (IQR 5-8 d).
The patient's self-reported symptoms, treatment with antiviral medications, presence of underlying medical conditions, and disease disposition were recorded. Information on hospitalization was obtained at the time of the interview and by retrospectively querying a hospital discharge database that covers all public hospitals in the state and 1 major private metropolitan facility. A cross-check with the hospital discharge database was performed for every infl uenza notifi cation received at CDCD.
For our analysis, we fi rst characterized all patients with pandemic (H1N1) 2009 or seasonal infl uenza infection reported to CDCD during the study period (the target population) in terms of age, sex, and hospitalization status by using univariate Mantel-Haenszel χ 2 tests for proportions and t tests for population means. We then compared patients who were interviewed (the study population) with the remaining patients not interviewed in the target population in terms of age, sex, and hospitalization status; if a signifi -Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16 cant difference was identifi ed, we weighted the interview responses from the study participants to refl ect the target population (19). Finally, by using the data obtained during interviews, we performed univariate analyses to compare patients with pandemic infl uenza with those with seasonal infl uenza with respect to reported symptoms, underlying medical conditions, and treatment. Because the age structure of the population with pandemic infl uenza differed from that with seasonal infl uenza, we also computed odds ratios (ORs) for individual symptoms or underlying medical conditions by using logistic regression to control for age. In each of the regression analyses, the dependent variable was defi ned as infl uenza type (pandemic/seasonal), and the independent variables were limited to age in years and 1 dichotomous variable representing the presence or absence of a single patient characteristic (e.g., a symptom or underlying medical condition).
To assess whether antiviral medications might have infl uenced the symptoms reported, we performed a subanalysis restricted to patients who were treated with antiviral agents within the fi rst 2 days of illness onset (early use of antiviral drugs) and compared those patients with patients who were never treated, controlling for age, infl uenza type, and the presence of underlying medical conditions. Infl uenza-like illness was defi ned as documented fever >38°C or a history of fever when the temperature was not known, and cough or sore throat, or both. Risk difference was defi ned as the absolute difference in the proportion of pandemic and seasonal infl uenza patients reporting a given parameter. A p value <0.05 was considered signifi cant. Statistical analyses were performed by using Epi Info 2000 (Centers for Disease Control and Prevention, Atlanta, GA, USA).

Characteristics of the Target Population
A total of 3,313 notifi cations of laboratory-confi rmed infl uenza were received at CDCD during the study period. Of these notifi cations, 117 (3.5%) were excluded because information on the viral strain and/or subtype was incomplete, and 18 (0.5%) were excluded because pandemic and seasonal infl uenza viruses were detected ( Figure 2). Of the remaining 3,178 infl uenza infections reported, 2,794 (87.9%) were pandemic (H1N1) 2009 infl uenza and 384 (12.1%) were seasonal infl uenza. The proportion of each infl uenza subtype identifi ed is shown in the inset in Figure  1. The mean age of patients with pandemic infl uenza was signifi cantly lower than that for patients with seasonal infl uenza, 27 and 35 years, respectively (p<0.005).
Most of the pandemic and seasonal infl uenza patients were female, but the proportion of female patients among the seasonal infl uenza patients was signifi cantly greater than that among patients with pandemic infl uenza, 57.3% (220/384) and 51.2% (1,431/2,794), respectively (p = 0.03 by χ 2 test). However, the association between female sex and seasonal infl uenza was not signifi cant when we controlled for age (p = 0.09).
A total of 415 (14.9%) of the 2,794 patients with pandemic infl uenza and 48 (12.5%) of the 384 patients with seasonal infl uenza were hospitalized. The difference between proportion of patients hospitalized with pandemic and seasonal infl uenza was not signifi cant on univariate analysis (p = 0.22, by χ 2 test). However, when we controlled for age, the odds of hospitalization were signifi cantly greater for persons with pandemic infl uenza (OR 1.53, 95% confidence interval [CI] 1.10-2.13; p = 0.011).

Selection and Representativeness of Study Participants
A total of 984 patients with pandemic (H1N1) 2009 infl uenza and 356 patients with seasonal infl uenza were selected for interview, and 871 (88.5%) and 288 (80.9%) of selected patients completed the interview, respectively ( Figure 2). Of the 181 patients selected but not interviewed, 168 were not able to be contacted because they did not have a working telephone number or did not answer after 6 attempts, and 13 declined to participate.
Patients who completed interviews were very similar to the remaining notifi ed infl uenza patients who were Hospitalized persons were underrepresented among the study participants compared with the remaining infl uenza case-patients who were not interviewed, i.e., 11.9% (138/1,159) of the interviewed patients had been hospitalized compared with 16.1% (325/2,019) of the patients not interviewed (p<0.05). The interview data were therefore weighted to refl ect the hospitalization rate in the target population for both pandemic and seasonal infl uenza.

Comparison of Pandemic and Seasonal Infl uenza in the Study Population
The age distribution for study participants, by infl uenza type, is shown Figure 3. As in the target population, the mean age of study participants with pandemic (H1N1) 2009 infl uenza was signifi cantly younger than the mean age of study participants with seasonal infl uenza, 26 and 36 years, respectively (p<0.005). Only 6% (49/871) of the study participants with pandemic infl uenza were >55 years of age compared with 23% (65/288) of those with seasonal infl uenza (p<0.005).
The distribution of the total number of symptoms reported by each patient with pandemic or seasonal infl uenza is shown in Figure 4. Patients with pandemic infl uenza and seasonal infl uenza reported a median of 6 symptoms (IQR 5-8 symptoms and 4-8 symptoms, respectively). When we controlled for age, no signifi cant association was found between infl uenza type and the total number of symptoms that patients reported (p = 0.19).
The number and proportion of patients reporting specifi c symptoms are presented in the Table. The difference in the proportion reporting a given symptom between patients with pandemic and seasonal infl uenza was <10% for all symptoms. Univariate analyses showed that fever and diarrhea were signifi cantly more common for patients with pandemic infl uenza. Controlling for age added cough and myalgia/arthralgia to the symptoms signifi cantly associated with pandemic (H1N1) 2009 infl uenza compared to those with seasonal infl uenza. Rhinorrhea was signifi cantly associated with seasonal infl uenza on univariate analysis, and this association persisted when controlling for age.
The distribution of the total number of underlying medical conditions reported by individual patients with pandemic (H1N1) 2009 or seasonal infl uenza is shown in Figure 5. Just over half of the patients with seasonal or pandemic infl uenza had no underlying medical condition(s).
The proportion of patients who reported a specifi c underlying medical condition is presented in the Table. The absolute difference in the proportion of patients that reported a given medical condition between those with pandemic (H1N1) 2009 and seasonal infl uenza was greatest for pregnant patients but still <5% for all underlying conditions queried. None of the 11 underlying medical conditions we queried were signifi cantly associated with pandemic or seasonal infl uenza in univariate analysis. When controlling for age, we found that only the odds of reporting a history of diabetes were signifi cantly greater among patients with pandemic infl uenza (OR 1.93, 95% CI 1.07-3.51; p = 0.03).
Having >1 underlying medical condition was not signifi cantly associated with pandemic infl uenza in the univariate analyses. However, after we controlled for age, we observed that the odds of reporting >1 underlying medical conditions were signifi cantly greater among patients with pandemic infl uenza (Table).
By design, the analysis that used weighted data shown in the Table mirrors   Of the 10 patients hospitalized for >14 d, all had >1 underlying medical condition, and 6 had >2 conditions. Two patients in the study population died; both had pandemic (H1N1) 2009.
The proportion of patients reporting treatment with antiviral medication was signifi cantly greater among those with pandemic infl uenza (Table). Information on the type of antiviral drug received was recorded for 427 (94.9%) of 450 patients treated with antiviral drugs; 426 reported taking oseltamivir and 1 reported taking zanamivir. The mean lag time between illness onset and starting antiviral treatment was 2.7 d (median 2 d, IQR 1-3 d) for patients with pandemic infl uenza and 2.3 d (median 2 d, IQR 1-3 d) for patients with seasonal infl uenza (p = 0.39).
Comparing patients given antiviral medications with those who were not, we found a signifi cant inverse relationship (i.e., a protective effect) between early antiviral drug use and reported rhinorrhea for patients with pandemic infl uenza (OR 0.5, 95% CI 0.4-0.8; p = 0.005). We also observed a positive association between early antiviral drug use and nausea and vomiting; this association was robust and persisted when the analysis was simultaneously controlled for age, infl uenza type, underlying conditions, as well as other symptoms commonly associated with infl uenza and/or gastrointestinal illness (i.e., fever, cough, sore throat, diarrhea) (OR 1.6, 95% CI 1.2-2.1; p = 0.02). No other symptoms were associated, positively or negatively, with antiviral drug use in the fi rst 2 days of illness onset.

Discussion
This comparison of >1,000 total confi rmed seasonal and pandemic (H1N1) 2009 infl uenza infections occurring contemporaneously over the peak of the traditional infl uenza season yielded several fi ndings. First, the spectrum of clinical illness due to pandemic infl uenza was similar to that caused by seasonal infl uenza. Although several symp-toms were more common in patients with pandemic infl uenza, the differences were modest and of limited clinical importance.
Our fi ndings generally parallel those from a recent comparative analysis in Singapore, with some differences. For example, in Singapore, the proportions of patients with seasonal and pandemic infl uenza who reported diarrhea were 0% and 4%, respectively; these fi gures are substantially lower than those found in our study (12% and 19%, respectively), despite the fact that in both settings most nonpandemic infl uenza viruses identifi ed were infl uenza A (H3N2). These differences highlight the need to consider data from diverse geographic, cultural, and healthcare environments when characterizing the clinical manifestations of infl uenza.
Second, we observed that the hospitalization rates for pandemic (H1N1) 2009 and seasonal infl uenza infections were similar. Our ability to use a comprehensive statewide database to identify hospital admissions in the broader target population permitted a robust analysis that found the overall proportion of confi rmed pandemic and seasonal illnesses hospitalized was not signifi cantly different when aggregated data were used in univariate analyses (p>0.05). However, if the analysis was controlled for age, the odds of being hospitalized were signifi cantly greater for the population with pandemic infl uenza. These seemingly dissonant results actually refl ect the fact that for many age groups there was a higher risk for hospitalization with pandemic (H1N1) 2009, but because patients with seasonal infl uenza were older relative to those with pandemic infl uenza and elderly patients are more likely to be admitted to hospital when ill with infl uenza, the cumulative hospitalization rate in the 2 patient groups was similar.
In addition, in this study, the mean duration of hospitalization was not statistically different between patients with pandemic (H1N1) 2009 and seasonal infl uenza even though other indicators suggested pandemic patients were hospitalized for longer periods. An analysis of a larger sample of hospitalized patients is under way.
Third, the underlying medical conditions associated with pandemic (H1N1) 2009 and seasonal infl uenza illnesses diagnosed in the community were nearly identical in terms of the type and number of conditions reported. Most patients in both groups reported no risk factors, and only when we controlled for age did we fi nd an association between having >1 underlying medical condition and pandemic infl uenza. Notably, the largest risk difference we observed was for was pregnancy (4%). Univariate analyses showed that the association between pregnancy and pandemic infl uenza approached statistical signifi cance (p = 0.08; analysis not shown). When we restricted our analysis to women 15-45 years of age, the risk difference nearly doubled, but signifi cance was still not attained, perhaps as a consequence of the smaller sample size. Seasonal infl uenza is a well-established cause of serious illness during pregnancy, and several reports indicate that the risk for severe illness from pandemic (H1N1) 2009 may be even greater (20)(21)(22)(23).
Obesity, newly recognized as a risk factor for severe infl uenza illness during the 2009 pandemic, was reported as often by patients with seasonal infl uenza as by those with pandemic infl uenza (11% vs. 9%; p>0.05). This fi nding suggests that obesity may be equally important as a risk factor for seasonal and pandemic (H1N1) 2009 (24).
Finally, because our study was not a randomized controlled trial, inferences about the effect of antiviral medications should be viewed with caution. For example, our observation that antiviral drug use was negatively associated with reported rhinorrhea may be due to the effect of treatment or may have resulted from a relative disinclination of providers to prescribe antiviral drugs for patients with rhinorrhea, on the basis of an assumption that nasal symptoms make infl uenza infection less likely (25). However, the robust positive association we observed in our population between antiviral agent use and nausea/ vomiting suggests that there was a causal relationship, a conclusion consistent with that of a recent metaanalysis on oseltamivir use (26).
The limitations of our study include the following: reported underlying medical conditions were not objectively 0.01 § *Totals respondents may not sum to 1,159 for all parameters because questions regarding underlying medical conditions and antiviral treatment were added shortly after the study was initiated, and there are intermittent missing values to individual questions for some respondents. RD, risk difference (absolute difference in the proportion of pandemic and seasonal influenza patients reporting a given parameter); OR, odds ratio; CI, confidence interval; NS, not significant; ILI, influenza-like illness (patient had fever and cough or sore throat). †ORs were computed by using logistic regression to control for age. Each row depicts data from a separate regression equation, where the dependent variable was defined as influenza type and age (in years) and a single patient characteristic, as listed in the first column of the row (coded as a dichotomous variable indicating the presence or absence of the respective symptom or underlying medical condition) were included as the predictor variables. In all of the logistic analyses performed, age remained significantly associated with influenza type, i.e., younger patients had a higher odds of having pandemic influenza compared with seasonal influenza. ‡Fever was defined as temperature >38 C or subjective fever if temperature was not measured. §Significant OR obtained using logistic regression. ¶ Patient reported >1 of the underlying medical conditions listed.
verifi ed, data on the duration of symptoms were not collected, and interviewers were not blinded to infl uenza type when administering the questionnaire. Also, because this was a public health evaluation of notifi ed infl uenza infections principally detected through routine healthcare practices in the community at large, we were unable to control for potential biases stemming from who was tested and who was not. However, because the healthcare provider could not be confi dent of whether the patient had pandemic (H1N1) 2009 or seasonal infl uenza at the time of testing, any bias in who was selected for testing should be approximately equal for pandemic (H1N1) 2009 and seasonal infl uenza patient groups. Lastly, a limitation inherent in the case-control study design we used was that we are unable to assess the extent to which the underlying medical conditions reported increased the risk for a diagnosis of infl uenza of either type, when compared with persons without underlying medical conditions. In summary, our head-to-head comparison of confi rmed pandemic (H1N1) 2009 and contemporaneous seasonal infl uenza infections found little to differentiate the 2 in terms of symptoms, underlying medical conditions, and the proportion of patients hospitalized. These results add to the growing body of knowledge about pandemic (H1N1) 2009 and are in general agreement with several studies that used different methods in other settings (27). These data are important because early in the pandemic some reports espoused different conclusions; 1 report estimated the lethality of pandemic (H1N1) 2009 to be ≈1 death per 10,000 infections, about 100× greater than that for regular seasonal infl uenza (28,29). Worldwide, unprecedented levels of resources have been expended to mitigate the impact of pandemic (H1N1) 2009. In the United States alone, the federal government appropriated $7.65 billion for this effort (30). This commitment to controlling pandemic (H1N1) 2009 is to be lauded, but we must not lose sight of the fact that seasonal infl uenza remains an important, albeit relatively uncelebrated, cause of illness and death each year. As the pandemic (H1N1) 2009 response draws to a close, it may be prudent to revisit the level of effort directed toward reducing the enormous effects, in terms of costs and health outcomes, associated with annually recurring infl uenza epidemics (31). Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16