We estimated that during the first wave in Hong Kong, 43.4% of school-age children and 10.7% of individuals aged 5–59 were infected by pdmH1N1. A serologic survey in England found similar IARs in London and the West Midlands [8
]. Both studies highlight the importance of including serologic surveys as a component of pandemic surveillance. While our core results are based on data from serial cross-sectional samples, we found similar infection attack rates inferred from participants for whom paired sera were available. The geographically compact and homogeneously mixing population in the urban environment of Hong Kong permits some degree of confidence in the validity of our infection attack rate and severity estimates. The detailed pdmH1N1 reporting system, the wide coverage of the public healthcare system (which includes >90% of all local inpatient days [11
]), and the resource investments since SARS have led to routine laboratory testing for all patients hospitalized with fever or pneumonia. This should allow identification of the majority of hospitalizations, ICU admissions and deaths directly associated with pdmH1N1 infection. Thus the completeness of the pdmH1N1 surveillance system, well-defined population denominator and our large-scale serologic survey provide accurate numerators and denominators for the severity measures.
We based severity estimates for pdmH1N1 on the infection attack rate as the denominator. Some previous studies of pdmH1N1 severity used clinical illness attack rate as the denominator which depends on the probability of symptoms as well as medical care seeking behavior of the population [3
]. Our estimated CIRs and CFRs are broadly consistent with Presanis et al’s “Approach 2” severity estimates but around 7–9 times lower than their “Approach 1” estimates [3
]. Our estimates of CHR are 2–10 times higher than their Approach 2 estimates of symptomatic CHR. However, the hospitalization-death ratio was 4,253/27 = 164 as of November 15 in Hong Kong but 996/53 = 19 as of June 14 in New York [3
], implying that the clinical threshold for admission in terms of disease severity at presentation may have been lower in Hong Kong.
We estimated that around 0.4 in 100,000 infections in school-age children led to mortality, while the risk of ICU admission and death per infection was 9.5 and 66 times higher in older adults aged 50–59 (). These estimates for mortality rates in children are consistent with data from the UK, where one study estimated a mortality rate of 11 per 100,000 symptomatic cases in children aged 5–14 [12
], while a serologic study suggested that the attack rates had been underestimated by a factor of 10 [8
]. Our estimates are lower than early estimates of the case fatality rate but the denominators may not have been well estimated in those studies [1
]. Previously a statistical model was used to estimate that the excess number of deaths due to seasonal influenza in Hong Kong was 11.8 (3.8–20.1) per 100,000 population for those aged 40–64 [15
]. Assuming an annual infection attack rate of 20%, this estimate would translate into 44(14–81) per 100,000 infections which is similar to our pdmH1N1 CFR estimate of 18(9–74) per 100,000 infections for the 40–59 yo. This supports the prevailing view that pdmH1N1 is not more severe than seasonal influenza in terms of morbidity and mortality.
Simulation studies have suggested that administering pandemic vaccines to school-age children provides substantial benefits to the community [16
]. However since 43.4% of school-age children were infected in the first wave and likely immune to reinfection, and infections in school-age children are rarely severe (), there may be less justification to include this age group as a target group for vaccination after the first wave of the pandemic in Hong Kong. Furthermore, given the substantial attack rate in children in the first wave, we speculate that a large second wave may be unlikely to occur unless there is substantial antigenic drift. So far there has been no evidence for the emergence of antigenically-drifted pdmH1N1 viruses [18
Our study has a number of limitations. First, we have used antibody titers of ≥1:40 by viral microneutralization as an indicator of recent infection, correcting for baseline (pre-existing) seroprevalence levels, but this may lead to underestimation of the IAR if not all infections led to antibody titers ≥1:40, or if some individuals with baseline titers ≥1:40 were infected. Another study found that around 5% of laboratory-confirmed H1N1 cases did not develop convalescent antibody titers above 1:40 by hemagglutinin inhibition tests [8
]. Second, our estimates of the IAR would be biased upwards if infection with other circulating influenza viruses led to cross-reactive antibody responses resulting in antibody titers ≥1:40. However between August and October 2009, 83% of influenza A viruses detected in Hong Kong were pdmH1N1 and only 3% of isolated viruses were seasonal H1N1 viruses which are more likely to be associated with cross-serological cross reactions with pdmH1N1 (appendix
]. Third, a minority of severe illnesses associated with pdmH1N1 infection might not be identified by molecular detection methods, for example if admission occurred after cessation of viral shedding associated with the primary infection, and thus we may have underestimated the pandemic disease burden. We did not have sero-epidemiology data from those older than 60 years of age and consequently cannot comment on attack rates or complication rates in this important age group. Finally, our analyses are primarily based on seroprevalence among blood donors to the Hong Kong Red Cross, who may not be representative of the whole population. We do not have detailed data on donors to compare their risk of infection with the general population, but we did observe very similar seroprevalence rates across the three groups of subjects in our study, i.e. blood donors, hospital outpatients and participants in a community cohort ( and Appendix Figure 2
We chose to use microneutralization tests rather than hemagglutination inhibition tests following preliminary studies which showed that microneutralization was more sensitive than hemagglutination inhibition for detection of antibody responses in pdmH1N1 infection [7
]. There is only limited cross reactivity between pandemic and recent seasonal H1N1 viruses by the microneutralization test (used in this study) but there is some cross-reaction in individuals, increasing with age and particularly noticeable in those over 65 years of age[8
]. Thus in a given individual, current serological methods do not conclusively distinguish between antibody resulting from pdmH1N infection from cross-reactive antibody arising from prior infections with seasonal H1N1 influenza, especially in those over 60 years of age, which is one reason why we did not address the infection rates or disease severity in the elderly. Our study design is based on the difference in age-stratified seroprevalence in the pre vs post first-wave period, a time-period when there was minimal seasonal H1N1 virus circulation in Hong Kong. Thus, our conclusions for individuals <60 years of age are unlikely to be confounded by issues pertaining to serological cross-reactivity. While immune-senescence could potentially lead to an underestimation of attack rates in the elderly, especially those over >65 yo, this is unlikely to affect our study which investigates those younger than 60 years.
In conclusion, around 10.7% of the population aged 5–59 and half of all school-age children in Hong Kong were infected during the first wave of pandemic H1N1. Compared to school-children aged 5–14, older adults aged 50–59, though less likely to acquire infection, had 9.5 and 66 times higher risk of ICU-admission and death if infected. Thus the apparently low morbidity and mortality burden of 2009 pandemic influenza (H1N1), despite an infection attack rate in the first wave similar to a seasonal epidemic, appears to be due to low infection rates in older adults who faced a much greater risk of severe illness if infected. The reasons why older adults appear relatively resistant to pdmH1N1 infection even though they appear to lack neutralizing antibody remains unclear. If antigenic drift or other adaptation of the pdmH1N1 virus allows these older age groups to be infected more efficiently, the morbidity and mortality of subsequent waves of the pandemic could yet become substantial.