In the fall of 2009, the reported incidence of patients infected with the pandemic influenza A(H1N1) virus in France exceeded the usual incidence of seasonal flu 12. Notification of all patients infected with A(H1N1) virus became mandatory from 1 July to 2 November 2009 but later was restricted to intensive care unit (ICU) admissions because of a large and rapid increase in the number of cases 3.

Taking into account observations made during the early stage of the pandemic in other countries 456 (especially in the Southern hemisphere 7891011), the French Ministry of Health organized the response to the pandemic according to the possible needs of seven regions, so-called 'defense areas', in order to regulate the supply of equipment (Figure 1). The purpose of this plan was to distribute resources equitably across regions while avoiding any shortage in ICU beds. Extracorporal membrane oxygenation (ECMO) devices and ICU ventilators were distributed in the reference centers of each defense area to cover the French territory. A 15% rate of ICU bed occupancy by flu patients in a region was indicated as a critical threshold to consider cancellation of scheduled surgical activities.

At the same time, we recorded data corresponding to the cohort of ICU patients through a large research network on mechanical ventilation (REVA-SRLF registry) 12. In addition, we designed a dedicated website screening registry to prospectively assess the specific burden related to the A(H1N1) pandemic in ICUs recruited on a voluntary basis. We proposed that a relatively simple screening registry would be able to give a much more exact picture of the respective burden on the different ICUs, geographical areas, and university versus non-university centers. Here, we report the exact rate of ICU bed occupancy by flu-infected patients (Flu-OR) during the pandemic in a representative subset of French ICUs.

The French REVA-SRLF registry was a multi-center prospective observational survey based on a website registry, and several results of this registry have been published elsewhere 131415. In brief, from November 2009 (week 09-45) through January 2010 (week 10-03), 89 out of 108 ICUs participating in the general registry (representing one fourth of all French ICUs) accepted an invitation to participate in a website screening registry to do a daily assessment of the rate of ICU bed occupancy by influenza-infected and non-infected patients. Sixty-nine ICUs belonging to either referral university hospitals (31 ICUs) or general hospitals (38 ICUs) eventually completed the daily screening for at least three consecutive weeks and admitted at least one A(H1N1)-infected patient and were kept for the present analysis. Twenty ICUs that, for organization reasons or for lack of A(H1N1) patients, did not complete three consecutive weeks were excluded from the analysis.

Recording of patients' data in the registry was approved by the national commission for protection of patients' rights and electronic data recording. The study was approved by the ethics committee of the French Society of Intensive Care (SRLF). Informed consent was waived in agreement with the observational design of the study.

In this prospective observational study using a dedicated online screening registry designed to assess the daily burden of the A(H1N1) influenza pandemic on French ICUs, we found important variations in the actual influenza burden between geographical areas, university and non-university hospitals, and time. In several individual ICUs, our screening registry has permitted us to observe peak occupancy rates greatly exceeding those calculated by averaging data observed in the largest defense areas. Our findings suggest that the organization of future pandemic response plans can greatly benefit from online data obtained in almost real time 1920. A dedicated online registry able to assess the week-by-week Flu-OR in each ICU may help to better distribute resources according to the actual needs. Even if ICUs were encouraged to do a daily assessment of the presence of patients with A(H1N1), we chose to report the calculation per week first to be consistent with the French organization and the National Institute for Public Health Surveillance (NIPHS), which displayed the time course of the pandemic weekly, and also to simplify data notification for participating centers. In fact, in collaboration with the NIPHS, we developed a strong communication strategy via the REVA website to simplify and therefore encourage rigorous notification.

Flu-ORs calculated per week in the present study were comparable to and often higher than those observed in the Southern hemisphere. Over the 3-month pandemic period in Australia and New Zealand 11, the ANZICS (Australian and New Zealand Intensive Care Society) investigators reported that an average of 5.2% of available ICU bed-days were occupied by patients with H1N1 infection, whereas the peak percentage ranged from 8.9% to 19% 11. In our study, 7.6% of the beds of enrolled ICUs were occupied by influenza-infected patients during the whole study period, and a maximum Flu-OR recorded per week in individual ICUs reached 35%. Inclusion of the 10 ICUs that had not admitted any patients with A(H1N1) would have lowered global Flu-OR but would not have changed the maximal Flu-ORs that we observed. The ICU length of stay, however, may have impacted the Flu-OR since a large difference between the two reports was observed: the median durations of ICU stay were 7.4 days (IQR of 3 to 16) in the ANZICS study and 10 days (IQR of 5 to 24) in French ICUs. Uneven time and regional distribution may also impact the Flu-OR calculations.

In the ANZICS investigation, the number of days that ICU beds were occupied by infected patients per region was calculated by multiplying the total number of patients by their length of stay. This approach allowed us to estimate a peak Flu-OR per region but not per individual ICU. The results of our registry show that this can greatly underestimate true peak activities in some ICUs. This is of particular importance since H1N1 burden changed rapidly over time and from one ICU to the other. The web-based screening registry specifically developed for our study allowed a daily assessment of the ICU occupancy rate in individual ICUs with an accuracy similar to that in larger regions. For example, observations reported from the Paris area (Figure 3) showed that the maximal Flu-OR significantly differed from the weekly Flu-OR calculated within all ICUs belonging to that region. Overall, our observations suggest that an accurate estimation of the influenza burden on ICUs requires a daily and real-time Flu-OR assessment, which seems feasible in view of our findings.

In anticipation of the flu surge, health authorities decided that, in order to alleviate the pressure on ICU beds, planned surgery activity should be cancelled when Flu-OR exceeded 15%. We contacted each participating ICU and confirmed that no canceling of scheduled surgery had been decided although Flu-OR exceeded 15% in many ICUs. Such peak activities occurred for more than 103 weeks among 32 units. Future analysis may help to refine this threshold and thus may have to take into account how long it is exceeded. The magnitude of a pandemic and its consequences on the organization of a health-care system depend on several parameters and are notoriously difficult to predict 2122. Therefore, experiences reported during the first waves throughout the world were useful to better elaborate forecasts taking into account different attack rates and epidemic wave durations 2324. Data gathered in Canada during the first wave were applied to a variety of second-wave models to determine its impact on ICU and ventilator demand 2125. An attack rate of greater than 20% can result in significant shortages in ICU beds and ventilators 21. Retrospective estimations showed that attack rates in Europe 26, North America, and Australia did not exceed 10% during each of the 2009 pandemic waves 27. The overall impact also greatly depends on the durations of mechanical ventilation and ICU stay. High attack rate combined with a short epidemic duration and long expected duration of mechanical ventilation represents the worst scenario in terms of bed occupancy rate and thus the maximal burden. Given a 20% attack rate with a similar clinical presentation, French ICU resources in university hospitals would probably have been overwhelmed, according to our observations.

A weakness of our study is that only a subset of French ICUs participated in the screening. Our cohort represented 35% of the whole French cohort reported during the same period to the health ministry. The distribution of the subset of university and non-university hospitals participating in this study within the seven areas (except in the northern area) suggests that French ICU health-care resources were reasonably well represented (Figure 1). This is supported by the comparability of baseline characteristics of patients within the present cohort to those of the entire French cohort (Table 1) 3.

General inferences that can be made from this study are limited because of the difficulties in extrapolating from one health-care system to another. But regardless of the system, the primary goal here was to be able to assess, as closely to reality as possible, the related burden of an epidemic, and such an assessment was made possible by the specific screening registry.

Revisiting the pandemic plans in light of emergent findings is certainly a key issue to be able to cope with further pandemic waves. Our observations suggest that the specific activity related to critically ill H1N1-infected patients varied widely according to time, regions, and individual ICUs as well as within larger areas. We report maximal ICU Flu-ORs that significantly exceed the 15% predefined critical threshold. The website registry described here and tested during the first pandemic wave in France allowed a real-time awareness of bed utilization and capacity.

• A simple online registry permits us to accurately describe specific H1N1 ICU burden in real time.

• ICU burden related to H1N1-infected patients varied widely according to time, regions, and individual ICUs.

• In most ICUs, the maximal rate of bed occupancy by patients with flu exceeded the predefined 15% critical threshold.

• Online assessment of Flu-OR in the ICU may help to better match resources and needs in case of new H1N1 pandemics.

ANZICS: Australian and New Zealand Intensive Care Society; CI: confidence interval; ECMO: extracorporal membrane oxygenation; Flu-OR: rate of bed occupancy by flu-infected patients (in the intensive care unit); ICU: intensive care unit; IQR: interquartile range; NIPHS: National Institute for Public Health Surveillance; REVA: Recherche en Ventilation Artificielle (Research on Mechanical Ventilation); SRLF: Sociéte de Réanimation de Langue Française (French Society of Intensive Care).

The authors declare that they have no competing interests.

J-CMR helped to conceive of the study and participated in its design and coordination, helped to design the website registry and to coordinate data management, and participated in the editing of the manuscript. TP helped to conceive of the study and participated in its design and coordination, helped to coordinate data management, and participated in the editing of the manuscript. CB-B helped to conceive of the study and participated in its design and coordination and helped to design the website registry and to coordinate data management. AM helped to conceive of the study and participated in its design and coordination and helped to design the website registry. LB helped to conceive of the study and participated in its design and coordination, helped to design the website registry, and participated in the editing of the manuscript. GB helped to design the website registry and actively participated in the study as a REVA correspondent from the six most active centers. JR and BR (president and member of the SRLF, respectively) helped and encouraged participating centers to complete the screening registry. BM, CG, MC, AC, and YLT actively participated in the study as REVA correspondents from the six most active centers. All authors read and approved the final manuscript.