FluCAN 2009: initial results from sentinel surveillance for adult influenza and pneumonia in eight Australian hospitals

The emergence of pandemic (H1N1) 2009 influenza posed challenges to the Australian health system.1,2 Soon after the first locally acquired cases were recognised in Australia in May 2009, it became clear that, despite the presence of pandemic plans and a variety of surveillance systems for influenza,3 there was a gap in surveillance at the severe end of the clinical spectrum.

Data on severity are important because an increase in severe respiratory complications such as pneumonia in adults could signal a change in the virulence of pandemic viral strains. Although some newly emerged influenza strains, such as H5N1, have been associated with direct viral-associated lung disease,4 it has been suggested that mortality in previous pandemics was associated with secondary bacterial pneumonia,5,6 emphasising the importance of surveillance for pneumonia as well as influenza during a pandemic. This gap in surveillance was common in many countries, and a variety of systems were rapidly set up or augmented during 2009 to gather information on hospitalisations.1,7,8,9,10,11,12

In Australia, existing systems were rapidly scaled up to meet the challenge, while state and territory health departments provided data from hospitals to the Australian Government Department of Health and Ageing.13 Other sources of influenza surveillance data were limited to patients in intensive care units (ICUs) or hospitalised children.14,15 A variety of systems were set up in jurisdictional health departments, but these relied on personal networks of public health physicians, laboratory data and automated reporting systems.13,16

Thus, at the onset of the H1N1 pandemic in 2009, there were no reliable, comprehensive, consistent and rapidly available data on adult acute respiratory hospitalisations outside the ICU setting in Australia. The Influenza Complications Alert Network (FluCAN) was designed to fill this gap by providing data from sentinel hospitals on acute respiratory hospitalisations, including ICU admissions. Here, we present its initial findings for the first 6 months, and compare FluCAN data on pandemic (H1N1) 2009 influenza with publicly available data from the national surveillance system13 and published data from the Australian and New Zealand Intensive Care Society (ANZICS) surveillance system.14,17

In 2009, FluCAN had two aims:

• Rapid collection and reporting of data on adult patients with laboratory-confirmed influenza hospitalised at 10 sentinel sites (nine in Australia and one in New Zealand);

• Detailed data collection on adult patients admitted to the same hospitals with laboratory-confirmed influenza or community-acquired pneumonia (with or without influenza).

The network of sentinel hospitals was established based on a pre-existing coalition of thoracic physicians who were members of the Thoracic Society of Australia and New Zealand Pandemic (H1N1) 2009 Task Force and representatives of the Australasian Society for Infectious Diseases. While the network predominantly represented tertiary care adult hospitals in capital cities, two non-urban hospitals with specifically high caseloads of Indigenous patients were also included.

One of the non-urban hospitals was unable to commence data collection due to logistical issues, and the Waikato Hospital in New Zealand is excluded from this report, leaving eight Australian hospitals providing data for this analysis.

The eight public acute care hospitals in six Australian jurisdictions reporting FluCAN data in 2009 (Box 1) represented 10% of all Australian hospitals with over 200 beds (n = 79). The 3461 beds in the sentinel hospitals represent 6.4% of national bed capacity (n = 54 338).19 All but one hospital was in a major urban area. From 1 July to 4 December 2009, 190 hospital-weeks of observation were undertaken (136 [71.6%] prospectively).

A total of 538 patients were recorded as being hospitalised with PCR-confirmed influenza; of these cases, 465 (86.4%) were due to pandemic (H1N1) 2009 influenza (representing 9.3% of the 4992 admissions for the pandemic strain notified nationally), with the remainder due to seasonal strains of influenza A. There were also 1468 patients hospitalised with radiologically confirmed community-acquired pneumonia, 718 (48.9%) of whom were tested for influenza; 194 cases (13.2%) were found to be associated with influenza, mostly the pandemic strain (163, 11.1%) (Box 1).

There was marked geographic variation in the distribution of the 538 influenza cases, with lower than expected caseloads in Victoria (which had 25.1% of total beds in the sentinel hospitals, but only 8.7% of the total influenza cases) and New South Wales (13.0% of beds and 6.9% of influenza cases), and higher than expected caseloads in the other states, especially Western Australia (16.7% of beds and 29.6% of influenza cases) (Box 1). This was mainly due to the timing of data collection, which commenced after the peak of cases in Victoria in mid to late June 200920 but before the majority of cases in WA (Box 2). This timing issue was also evident in the proportion of pandemic (H1N1) 2009 influenza cases in each state, which was as low as 64.9% in NSW and as high as 97.0% in South Australia (Box 1).

The epidemiology of pneumonia was less affected by the timing of data collection, except for higher than expected numbers of admissions in Tasmania and lower than expected numbers of admissions in NSW (Box 1 and Box 3). Compared with influenza patients, a similar proportion of patients with pneumonia were men, but patients were older, with a median age of 69 years (interquartile range, 18–101 years). There were 107 Indigenous patients admitted with pneumonia (7.3%).

Box 4 compares pandemic (H1N1) 2009 influenza cases recorded in FluCAN with national notifications13 and intensive care admissions.14,17 As a proportion of nationally notified hospitalisations, the number of FluCAN-reported cases was higher than expected (465/4992, 9.3%), based on the number of beds in the sentinel hospitals and the limited period of observation. FluCAN reported 102 ICU admissions (15.0% of the nationally reported number), again more than expected given the proportion of national ICU bed capacity in the sentinel hospitals.

Compared with national notifications, FluCAN-reported hospitalised patients had a higher median age, a similar proportion of women, and a higher proportion of pregnant women (Box 4). Most pregnant women (29, 72.5%) were in their third trimester. Of note, 40% of pregnant women (16/40) admitted to FluCAN hospitals with pandemic (H1N1) 2009 influenza were over the age of 45 years; these women are not included in the national notifications of women of childbearing age (age range, 15–44 years). FluCAN reported a lower proportion of admitted Indigenous patients but a higher ascertainment rate for Indigenous status (96.8% v 80.7%), and a higher rate of at least one comorbidity. Prominent comorbidities included asthma, obesity and immunosuppression. Almost 30% of admitted patients for whom information was available were current smokers. Median length of stay and death rate were slightly higher in FluCAN, although the latter can be partly explained by FluCAN’s inclusion of data on post-discharge deaths.

Compared with the national notifications, FluCAN reported a higher proportion of patients with pandemic (H1N1) 2009 influenza admitted to ICUs (Box 4). For ICU patients, demographic and risk factor data were generally similar between patients reported in FluCAN and both the national notifications and ANZICS data — median age, proportion of women, and the proportion of patients with at least one comorbidity were very similar. FluCAN reported a higher proportion of pregnant women admitted to ICUs than both other systems. Specific comorbidities available from ANZICS data were very similar to those reported by FluCAN, as were the proportion of patients with pneumonia, length of stay and death rate. This level of detail was not available from the national notification data.

The FluCAN system was established during the peak of the 2009 influenza pandemic and provided real-time surveillance data on patients with influenza and pneumonia admitted to Australian hospitals. Reported case numbers for influenza were higher than expected, given the number of beds in the sentinel hospitals. The demographic, comorbidity and outcome data for FluCAN patients were broadly similar to those from national notifications and ANZICS ICU admissions data, demonstrating the representative nature of the sentinel surveillance. It should be noted that both of the non-FluCAN datasets collected relatively minimal data from all hospitals and all age groups throughout the 2009 influenza season, whereas the FluCAN dataset was more detailed but restricted to adults aged over 18 years in sentinel hospitals and commenced 2 months after diagnosis of the first Australian pandemic cases. The high ascertainment rate, accuracy of diagnosis and level of detail in FluCAN reinforce the utility of a strategically established sentinel surveillance system for hospital admissions.

Initial information from Australia and overseas on the severity of pandemic (H1N1) 2009 influenza infection suggested that the illness was relatively mild and mainly affected children and young adults. Pregnancy, obesity, asthma, chronic obstructive pulmonary disease and some other non-respiratory comorbidities appeared to be associated with more severe disease in a minority of patients.21,22,23 Subsequent reports have shown a range of estimates for underlying comorbidities and particular populations at greater risk of severe disease. A recent report reviewed hospitalisation data globally,8 and there have been Australian reports of hospitalisations in single hospitals in Sydney24 and Darwin,25 or clusters of hospitals in specific geographic regions.16,26 FluCAN data are broadly in agreement with the estimates of frequency of vulnerable groups, proportion admitted to ICUs, and mortality in these other studies. Only FluCAN has thus far included estimates of post-discharge health service use and deaths.

Any differences between the data reported by FluCAN and those in Australian Government reports can mainly be explained by FluCAN being confined to adult hospitalisations (for median age, proportion of pregnant women), the location of the Flu CAN hospitals (lower proportion of Indigenous patients due to a concentration in southern metropolitan hospitals), and more accurate and detailed ascertainment of clinical status in Flu CAN data (higher proportion of comorbidities and detailed information on a wider range of underlying diseases and other risk factors). ICU admissions captured by FluCAN were broadly similar to population-level ANZICS data and national data, demonstrating the representative nature of FluCAN for this more severely affected group.

Data on pneumonia admissions in Australia are provided from routinely collected sources that are not timely and generally lack specificity in diagnosis and details of comorbidities. The Australian Community-Acquired Pneumonia Study (ACAPS) reported more detailed information on pneumonia from six hospitals (including two that contributed to FluCAN) in three states.18 Compared with patients with pneumonia in FluCAN, ACAPS patients were similar in age and the proportion of female patients was similar. There were large differences in Indigenous status and underlying comorbidities, perhaps because almost half of all ACAPS patients were admitted to a single tertiary referral hospital in Melbourne.18

There were several limitations to the FluCAN surveillance system. For a variety of reasons, influenza PCR tests were not done routinely for all patients with respiratory illnesses in some hospitals, despite national guidelines encouraging testing for hospitalised and high-risk patients with influenza-like illness.27 Thus, an unknown number of admissions during this period may have been due to influenza but not diagnosed and therefore not included in the data presented here. In its first year, FluCAN attempted to collect as much relevant information as possible, but there was a surprising lack of consistent recording of important variables such as smoking, vaccination status, weight and height. Ongoing surveillance will provide a comparison during a non-pandemic season and hopefully address some of the limitations of data collection in 2009.1,8,14

Accurate, rapidly available and country-specific data on severity are essential for health authorities to appropriately respond to pandemic threats. Due to difficulties in collecting such information from hospitals, particularly in a pandemic situation, sentinel surveillance systems are an alternative to whole-of-system approaches. Despite the increased workload inherent in the 2009 pandemic, our group of clinicians and public health practitioners was able to establish a hospital-based sentinel surveillance system relatively quickly. FluCAN was able to capture representative information comparable to more comprehensive surveillance systems in a more timely manner and with high-quality clinical data. This suggests that a strategically selected sentinel network can augment whole-of-system surveillance and provide data that are representative of the wider hospital system in Australia.