This aim of this study was to determine the burden of A/H1N1/09 at a tertiary care hospital in south India. In the preliminary approach, detection of viral RNA in respiratory samples as tested by real time RT-PCR, was used as a marker of exposure. In this analysis we found that patients of all age groups presenting with ILI and SARI were infected with A/H1N1/09. The nature of the epidemic curve and the age distribution of cases is typical of pandemic influenza 
. Similar findings of widespread transmission have been reported from other parts of India 
and other regions of the world 
. Laboratory confirmed A/H1N1/09 infection was documented in 27.4% (384/1403) of ILI cases and 14.6% (173/1185) of SARI cases. Laboratory confirmed pandemic influenza contributed a minor proportion of hospitalizations (123/2588,4.7%) and severe disease including ICU admission and death (50/2588, 1.9%) during the period. The risk factor analysis showed a lower risk of hospitalization, ICU admission among A/H1N1/09 positive cases. Taken together, the findings suggest that cases of pandemic influenza, confirmed by the detection of virus in respiratory specimens, were less often associated with hospitalization and ICU admission. Numerous studies of pandemic H1N1 disease have also reported similar findings 
. Though studies from animal models have demonstrated that A/H1N1/09 transmits efficiently and is able to cause severe disease 
, the latter was observed only in a minority of cases of human infection.
We also analyzed the data from a different perspective and found evidence suggestive of a higher pandemic burden at our center. The evidence in support this is as follows. Firstly, the pandemic caused a sudden increase in the number of persons with ILI and SARI presenting to the hospital. A concurrent surge in hospitalization occurred with majority (51%) of inpatient admissions being children under the age of 5 years. A significant increase and subsequent reduction of SARI admissions, that coincided with the rise and fall in detection rates of pandemic influenza, provide evidence of a direct impact of the pandemic on hospitalization. In addition, weekly number of ICU admissions and mortality rose significantly in the pre- and peak-pandemic periods suggesting a direct impact of the pandemic. Studies have also demonstrated a surge in hospitalization, especially among the under-5 age group similar to our study 
. Secondly, a good correlation of weekly numbers of total SARI admissions with A/H1N1/09 SARI admissions (r
0.793) and A/H1N1/09 mortality (r
0.610), suggests that the increase in SARI admission was very likely due to the pandemic. The negative correlation that was observed between A/H1N1/09 positive SARI cases and all-cause mortality must be viewed as an absence of increase in relation to increase in the former. This finding can be explained by a “replacement effect” which occurred at our center. In spite of the attempted scaling-up of health infrastructure at our center, the sudden surge in number of cases requiring hospitalization necessitated an additional reallocation of existing facilities in response to the pandemic. This led to the preferential admission of patients with SARI over cases with severe disease of other etiology. This may have resulted in a non-increase in the all-cause mortality, even though there was an increase in P(H1N1) positive SARI admissions. Thirdly, a higher detection rate of A/H1N1/09 seen among ILI as compared with SARI cases, can be explained by the difference in the level of virus shedding at presentation between the 2 groups. ILI cases seek healthcare early in the course of illness and so a sample is more likely to test positive, due to higher level of virus shedding which is in contrast to SARI cases who present later in the course of their illness. In addition, our study has shown that a longer duration of symptoms prior to presentation to healthcare is more likely to result in a negative PCR test result. A recent study has shown that laboratory confirmed cases only represented a minority of the actual burden of severe disease that occurred during the 3 waves of infection in Mexico 
. It seems plausible that many of the persons presenting with severe disease may have been infected but were found negative for viral RNA. If this indeed was the case in our setting, then a large proportion of people who tested negative should be considered positive for A/H1N1/09.
The paradoxical “protective” effect of pandemic influenza against hospitalization that we observed on our preliminary risk analysis is an important caveat of hospital-based studies in a developing country setting. Considering that hospital-based studies form the majority of the literature of the impact of the pandemic, we feel that data of this nature must be examined carefully from different perspectives to assess the true magnitude of the pandemic in our population.
We here present our attempts to estimate the burden of the pandemic using two methodologies –a) direct estimation based on real time RT-PCR and b) indirectly by comparing the relationship of weekly trends in hospitalization with all-cause mortality. The results of our analyses, clearly demonstrate that the burden of the pandemic at our center was higher than was estimated using a positive index test alone as a marker of exposure. Seroprevalence studies have also shown a high level of positivity for pandemic influenza in the general population 
as well as among healthcare workers 
. In our attempts to estimate the burden of pandemic influenza, it became clear that methodological issues highlighted above prevented an accurate estimation of disease burden in our setting. Hospital-based studies on influenza are skewed towards a severe disease perspective and when viral RNA alone is used as a marker of exposure to the virus, the actual burden of disease may be under estimated. In addition, a replacement effect of pandemic influenza on hospital admissions will further cause a paradoxical lowering of all-cause mortality. Overall, these factors lead to an underestimation of the impact of a pandemic in a developing country setting. Estimates from community-based studies using virus detection and seroconversion as markers of virus exposure, are likely to provide far more reliable estimates of burden of disease. Outbreaks of respiratory disease like pandemic influenza can have a far-reaching impact on the healthcare system in the developing countries and the findings of this study are key to estimation of the impact of future epidemics.
A few points about this study are noteworthy. The presence of an influenza surveillance program before the onset of the pandemic, ensured an early detection of cases presenting to the hospital. The triage facility established in the hospital was an efficient system for prioritization of resources, given the limited medical supplies and infrastructure available for the management of the pandemic. This experience of the management of the pandemic could be replicated in other resource poor settings with similar results.