Statistics from Altmetric.com
Commentary on the paper by Beard et al (see page 20)
Influenza has long been recognised as a disease which affects children; however, it is only fairly recently that the literature on this subject has switched focus from community settings towards the burden of hospitalisations. This issue carries an article by Frank Beard and colleagues which draws attention to the issue in Sydney, Australia and addresses the issue in a quantitative as well as a qualitative way.1 It follows on from, and replicates the methodologies employed by similar pivotal studies in the USA and Hong Kong.2,3
Most experienced commentators would agree that the foundations of our understanding of the burden of influenza in children, are based on data generated by a series of prospective community studies which took place in the 1960s and 1970s in the USA, all of which combined clinical surveillance with attempts at virus isolation and serological studies, to a greater or lesser extent. They are probably too large and too expensive to ever contemplate repeating in the present era. In Tecumseh, Michigan, between 100 and 300 families with at least one child were studied continuously for six years from 1966 to 1971—a period which included the emergence of influenza A/H3N2, the last pandemic virus, in 1968.4 In Seattle, Washington, a similar study took place between 1965 and 1969 and again from 1975 to 1979, involving over 215 families with young children.5 In Houston, Texas, similar observations were made over the period 1976 to 1984, including two influenza B epidemics.6 The findings of these major studies have consistently shown that the highest serological attack rates for influenza each season occur in children (typically 15–40%) compared with adults (12–20%), although there is far less consensus on whether the peak rate occurs in teenagers, primary age children, or preschoolers.7,8,9,10 From the above studies,5,7,8,9,10 and others,11,12 there are also convincing data that children act as introducers and spreaders of influenza infection in communities and individual households (presumably because they mix more often and shed viruses in higher titre than adults, and for longer). Glezen and Couch showed an upwards shift in the age distribution of influenza positive patient specimens (from children aged 5–19 years towards adults aged 20–44 years) as two influenza A epidemics progressed over time; similarly, in the same setting, school absenteeism preceded industrial absenteeism by about two weeks and paediatric admissions for pneumonia preceded those in adults by a similar period.13 Another notable finding from these studies was the disparity between serological and clinical attack rates. Fox et al showed that among teenagers and children with serologically proven infection, between 83% and 69% suffered a clinical illness;9 other studies have estimated that only 58% of infections in teenagers are accompanied by symptoms.14 What seems clear is that asymptomatic infections in children are not uncommon.
The relative importance of influenza compared with other childhood respiratory viruses, notably RSV, has also been well described in hospital based studies. In Kawasaki, Japan, a study of paediatric admissions over seven winter seasons from 1991 to 1998 revealed that 14% of admissions were attributable to serologically confirmed influenza, compared with 17.5% due to RSV; other respiratory viruses accounted for a much lower proportions of the total burden.15 In Kiel, Germany, a similar study of acute respiratory admissions over four winter seasons using PCR testing of nasopharyngeal specimens, revealed that 8.3% of such admissions were attributable to influenza infection, compared with 12.7% due to RSV; while RSV was the dominant pathogen in children aged 0–3 months, from the age of 2 years upwards, influenza predominated.16 Similar contemporary data are also available from Spain17 and Italy.18 In Leicester, UK a similar prospective study was carried out over one winter season in children <6 months, hospitalised for a broad variety of acute illnesses (not just acute respiratory illness), and again based on PCR testing of nasopharyngeal specimens. This revealed that 5.4% overall had an influenza infection at the time of admission, but more surprisingly, the proportion of influenza infected children among those presenting with acute respiratory illness (5.0%) was similar to the proportion among children presenting with non-respiratory illness (6.0%).19 These data suggest that the burden of hospitalisation due to influenza is underestimated by considering only acute respiratory admissions. Further evidence from Lyon, France suggests that during defined influenza epidemic periods the rate of virus isolation from small children who are unwell is probably many times higher.20
Attempts have been made to quantify hospital admissions in children due to influenza, since the 1980s. Mullooly and Barker attempted this in 1982, by estimating the excess hospitalisation rate seen in epidemic compared with non-epidemic years in Oregon, using retrospective data.21 These researchers found an overall excess rate of 9/104 admissions with influenza related diagnoses, but this rose to 32/104 in children with at-risk conditions. Broadly comparable results also emerged from a similar study in Harris County, Texas.22 However neither of these studies used methods which adequately adjusted for the potential influence of RSV infections. In contrast, the cluster of recent studies, including the current one by Beard et al, have paid much closer attention to the possible confounding effects of RSV infection.1–3,23,24
The featured study in this issue1 illustrates that the estimation of excess hospitalisation in children due to influenza is highly sensitive, not only to the precise mathematical method used, but also to the selection of either a summer baseline or one comprised of winter periods when influenza was not circulating (“periseasonal baseline”).1 Nevertheless, the study findings illustrate a consistent trend towards the highest rates of influenza related hospitalisation in children <12 months old. The excess hospitalisation rates observed in Sydney were also far higher than those recorded in recent US studies,2,23,24 but distinctly lower than those calculated by Chiu et al in Hong Kong.3 All three US studies focused on healthy children only, and might therefore be expected to have produced somewhat lower excess rates than in Sydney; but, on its own, this is unlikely to account for such dramatic differences. It is equally as unclear why the rates of excess hospitalisation in Hong Kong are so much higher than in Sydney. Aside from confounding, other possible explanations include true differences in the incidence of influenza in children in different countries, and genuine differences in thresholds for hospital admission of children with acute respiratory illness in different health systems.
Notwithstanding the uncertainties in the data regarding the true burden of influenza related hospital admissions in children, two consistent messages appear to be emerging. First, that the magnitude of hospital admission due to influenza in children is far from trivial and, at times, on a par with excess admission rates in adults at-risk. Second, that the burden of influenza related hospitalisation in children is concentrated among those <5 years, and especially those <12 months. However, in relation to the second point, the same is also true for RSV.25
There have been many calls for individual countries to emulate the US policy for routine annual vaccination of young children against influenza.26 One of the critical factors in deciding on the cost effectiveness of such a policy is being able to establish reliable country specific data on the burden of hospitalisation. The current study and others like it will assist national governments in this task. Nevertheless even if the likelihood of cost effectiveness seems high, there are other equally important considerations which have to be factored in before scientific evidence becomes official policy. Not least of these are the availability of a suitable vaccine; space within the national childhood immunisation schedule to accommodate delivery of an annual seasonal programme (first time vaccinees may require two doses); and finally, parental acceptability.
Note in Proof
Commentary on the paper by Beard et al (see page 20)
Competing interests: none declared