Relative risk of hospital-acquired influenza-like illness (HA-ILI) in patients according to exposure to source individuals at Edouard Herriot Hospital, 2004-2007. HCW indicates health care worker. Vertical bars represent 95% confidence intervals.
Vanhems P, Voirin N, Roche S, Escuret V, Regis C, Gorain C, Pires-Cronenberger S, Giard M, Lina B, Najioullah F, Barret B, Pollissard L, David S, Crozet M, Comte B, Hirschel B, Ecochard R. Risk of Influenza-Like Illness in an Acute Health Care Setting During Community Influenza Epidemics in 2004-2005, 2005-2006, and 2006-2007A Prospective Study. Arch Intern Med. 2011;171(2):151–157. doi:10.1001/archinternmed.2010.500
The person-to-person transmission of influenza-like illness (ILI) and influenza has been described mostly in long-term care units. Studies in acute hospital settings are rare and mostly retrospective.
We prospectively estimated the relative risk (RR) of hospital-acquired (HA) ILI during hospitalization according to in-hospital exposures to contagious individuals. Surveillance of ILI and laboratory-confirmed influenza was undertaken at Edouard Herriot Hospital (1100 beds) during 3 influenza seasons. A total of 21 519 patients and 2153 health care workers (HCWs) from 2004 to 2007 were included. The RR of HA-ILI in patients was calculated according to exposure to other contagious patients and HCWs.
For patients exposed to at least 1 contagious HCW compared with those with no documented exposure in the hospital, the RR of HA-ILI was 5.48 (95% confidence interval [CI], 2.09-14.37); for patients exposed to at least 1 contagious patient, the RR was 17.96 (95% CI, 10.07-32.03); and for patients exposed to at least 1 contagious patient and 1 contagious HCW, the RR was 34.75 (95% CI, 17.70-68.25).
Hospitalized patients exposed to potentially infectious patients and HCWs with ILI inside the hospital are at greater risk for HA-ILI. Such results identify priorities regarding preventive measures for seasonal or pandemic influenza.
Nosocomial outbreaks of influenza-like illness (ILI) or influenza are reported regularly.1- 3 Patients and health care workers (HCWs) may become infected themselves and, in turn, be sources of infection for others. Attack rates greater than 60% have been found in nursing homes during seasonal epidemics, with 10% lethality from influenza-related complications.4 The incidence rates in short-stay units are less well known. Seasonal influenza remains a frequent cause of hospitalization, and it was observed that the hospitalization rate increased during the influenza A/H1N1 pandemic.5- 7 In the United States, the community attack rates of seasonal influenza ranged from 10% to 20% by population groups and the hospitalization rate from 20 to 1000 per 100 000 inhabitants.8 Therefore, the risk of the virus spreading in health care settings is a major concern, especially in the elderly and individuals with immunosuppression or severe chronic disease.
The person-to-person transmission of influenza or ILI has been described mostly in long-term care units.9,10 Relevant studies in acute care facilities are infrequent and mostly retrospective.3,11,12 In particular, the relative risk (RR) of hospital-acquired (HA) ILI, compared with ILI in the community, is unknown, as is the effect of in-hospital exposures to infected patients. Regarding HCWs, 2 other issues should be considered. First, an excess of absenteeism among HCWs13- 17 is observed during community influenza outbreaks. Second, up to 75% of HCWs with influenza17 continue to work, increasing the risk of hospital outbreaks. In addition, quantifying the risk of HA-ILI might help in pandemic preparedness.18
The main objective of this prospective study was to estimate the risk of HA-ILI for hospitalized patients in short-stay units considering the levels of exposure to contagious patients and HCWs with ILI in the hospital. The risk of HA-ILI for hospitalized patients was also compared with the risk of ILI, assuming patients would have stayed in the community.
A prospective surveillance study was performed by the infection control team between October 15, 2004, and April 15, 2007, at Edouard Herriot Hospital. This university hospital is a tertiary medical center with 1100 beds and 102 units located in 32 buildings. Three investigations lasted from October 15 to April 15 in 2004-2005, 2005-2006, and 2006-2007. A total of 36 adult short-stay units participated on a voluntary basis, 12 with 224 beds in 2004-2005, 29 with 493 beds in 2005-2006, and 30 with 537 beds in 2006-2007 (Table 1). All the data were collected prospectively. No specific intervention related to the study was proposed, and the usual infection control measures were implemented. The Edouard Herriot Hospital Institutional Review Board approved the study. All the ILI cases (patients and HCWs) received printed information and signed an informed consent form.
For each patient, sex, age at hospital admission, number of diagnoses, and major diagnostic categories were provided by the hospital. For HCWs, work timetables were made available by the hospital department in charge of staff management.
Once a day, the research nurses contacted participating units to determine whether any patients or HCWs had presented with ILI. An ILI case was defined as a patient or HCW with a rectal or axillary temperature of at least 37.8°C in the absence of antipyretic drug therapy, with cough or sore throat.18,19 All patients with ILI were included whatever the timing of ILI onset. Cases of ILI were defined as index cases (ICs) if no exposure to other individuals with ILI was reported. The HCWs were included and observed identically as patients. Prevalent cases of ILI were defined as individuals with ILI at hospital admission, whereas incident cases of HA-ILI were individuals free of ILI at admission who acquired ILI during hospitalization.
The following information was collected for all patients and HCWs with ILI: dates of hospital admission and discharge (only for patients), underlying diseases, start and end dates of ILI clinical features, potential sources of exposure to influenza (such as relatives or colleagues at work), treatments during hospital stay, outcome, influenza vaccination status, and date of vaccination. Symptoms of ILI were collected daily during 5 days for patients and HCWs. For HCWs with ILI, this clinical information was completed by detailed work timetable. Incubation of ILI considered in the study was up to 5 days, and the duration of contagiousness was up to 6 days (1 day before onset, the day of onset, and the 4 following days).20,21
When an IC (patient or HCW) was reported to the infection control team, a research nurse visited the unit twice a day to document follow-up of the case and to detect secondary patient or HCW cases (SCs). The SCs were included in the study according to clinical criteria similar to those for ICs. Each SC was followed up as an IC, and the whole unit was followed up by a research nurse for 10 days after IC onset to detect SCs. If there was a third case in the same unit, the unit was again observed for 10 days, etc. No SC was contacted after hospital discharge, and no patient exposed to an IC or an SC was observed after discharge to identify potential HA-ILI starting outside the hospital. For each SC, a list of IC exposures in the previous days was documented prospectively based on patient recall and data reported in the nursing logbook (same bedroom, same restroom, same radiologic examination room, etc). Information was collected on the clinical status of HCWs in charge of SCs regarding ILI. As soon as an SC was identified, he or she was considered to be a potential source for other individuals and was observed similarly as ICs.
Concurrent with clinical data collection, the research nurse performed a nasal swab of the case (patient or HCW) for influenza laboratory confirmation. The nasal swab was sent to the national influenza reference center for the south of France located in Edouard Herriot Hospital. Influenza A and B viruses, respiratory syncytial virus, adenovirus, and rhinovirus were tested using an immunocapture enzyme-linked immunosorbent assay, immunostaining, and tissue cell culture. Sample suspensions were inoculated onto Madin-Darby Canine Kidney cells, epithelial Hep-2 cells, and fibroblastic MRC-5 cells. Influenza A and B viruses and rhinovirus were detected by polymerase chain reaction analysis. Among the 166 cases of ILI, 2 patients (1.2%) refused the nasal swab, and for 1 patient (0.6%) it could not be performed owing to the severe medical conditions of the patient (ie, severe pneumopathy with renal and cardiac failure). An influenza case was an ILI case with a nasal swab that tested positive for influenza.
Community ILI cases were individuals detected by the Sentinelles general practitioner network in the Rhône-Alpes region (almost 6 million inhabitants in southeastern France) during the study periods.22 The ILI incidence data were provided weekly and stratified by age, as described elsewhere.22 Regional population figures were obtained from estimations made on January 1 of each observation period by the National Institute of Statistics and Economic Studies (http://www.insee.fr).
Population characteristics are described by numbers and percentages for categorical variables and by medians and ranges for continuous variables. The HA-ILI attack rate was the number of incident cases of HA-ILI in patients divided by the number of hospitalized patients at risk during the study period. The incidence rate of HA-ILI was the number of incident cases of HA-ILI divided by the number of person-weeks for hospitalized patients, defined as the length of stay for noncases and the length of stay until onset of HA-ILI for cases. We did not calculate incidence rates for HCWs.
Exposure to ILI compatible with a transmission was defined as the presence in the unit, within 5 days before onset, of at least 1 person with ILI (patient or HCW) considered contagious from 1 day preceding ILI onset to 4 days thereafter. The risk of HA-ILI was calculated by stratification of exposures in the unit as follows: no exposure to an individual with ILI, exposure to at least 1 patient with ILI without exposure to HCWs with ILI, exposure to at least 1 HCW with ILI without exposure to patients with ILI, and exposure to at least 1 HCW and 1 patient who both have ILI. The RR of HA-ILI was calculated with its 95% confidence interval (CI) as the incidence rate of HA-ILI of an exposure stratum divided by the incidence rate of HA-ILI in the reference stratum (ie, where no exposure to ILI was documented). We did not perform the analysis for laboratory-confirmed influenza because only 13 of 64 HA-ILI cases (20%) were positive for influenza.
We calculated weekly ILI incidence rates in the Rhone Alpes region per age group by dividing the weekly number of new ILI cases in the Rhone Alpes region made available by the Sentinelles network22 by the Rhone-Alpes population figures made available by the National Institute of Statistics and Economic Studies. Incidence rates of ILI from the community population were then applied week by week to the age structure of the patient population (ie, using indirect standardization). With this calculation, the expected number of cases in the community was defined as the number of cases in the same patients if they had stayed at home. This allowed the observed incidence of HA-ILI in the hospital and the expected incidence of ILI in the community to be compared to determine whether exposure to the hospital may be a risk factor for HA-ILI. For this analysis, a standardized incidence ratio (SIR) of HA-ILI was estimated with its 95% CI by dividing the number of observed cases of HA-ILI by the number of expected cases of ILI in the community, assuming patients would have stayed at home. An SIR greater than 1 favored increased risk in the hospital, an SIR of 1 indicated a similar risk in the hospital and in the community, and an SIR less than 1 represented a protective effect of the hospital compared with the community. All statistical analyses were performed using a software program (SAS, version 9.1.3 for Windows; SAS Institute Inc, Chicago, Illinois).
The characteristics of hospital units included in the surveillance period are given in Table 1. The characteristics of the total patient population hospitalized during the 3 surveillance periods are given in Table 2.
A total of 166 patients with ILI were observed during the 3 study periods, of which 37 (22.3%) had confirmed influenza (32 influenza A virus and 5 influenza B virus). Viral co-infection occurred in 1 patient (influenza A and respiratory syncytial virus). The HA-ILI attack rate in hospitalized patients at risk during the 3 seasons was 3.0 (95% CI, 2.3-3.8) per 1000 patients. It varied by season (P = .006) and major disease category (P < .001) and increased with the number of diagnoses at admission (P < .001), with age at admission (P < .001), and with length of stay (P < .001) (Table 2). The attack rate of laboratory-confirmed influenza was 0.6 (95% CI, 0.4-1.0) per 1000 patients at risk.
After exclusion of 102 patients (61.4%) with ILI at admission, the median delay between onset and admission was 3 days (range, 1-67 days) for the 64 remaining patients (38.6%) (ie, incident cases of HA-ILI). Thirty-five patients (21.1%) presented with HA-ILI more than 72 hours after admission.
Of 2153 HCWs included in the study, 93 (4.3%) with ILI were observed, of which 20 (21.5%) had laboratory-confirmed influenza (17 influenza A virus and 3 influenza B virus). Two HCWs presented with influenza A and rhinovirus co-infection. Among HCWs, the attack rate of ILI was 4.3% (95% CI, 3.5%-5.3%) and that of laboratory-confirmed influenza was 0.9% (95% CI, 0.6%-1.4%).
The HA-ILI incidence rate by exposure category is reported in Table 3, and the RR (95% CI) of HA-ILI by exposure category compared with patients with no exposure in the ward is illustrated in the Figure. The data indicate a gradient of risk according to the number of exposure sources. With exposure in the same unit to at least 1 contagious HCW, the HA-ILI risk for patients was 5.48 (95% CI, 2.09-14.37) and reached 34.75 (95% CI, 17.70-68.25) when a patient was exposed to both contagious patients and HCWs compared with patients with no exposure.
The HA-ILI incidence rate was 3.24 (95% CI, 2.53-4.14) per 1000 patient-weeks of hospitalization, and the ILI incidence rate estimated in the community for similar age structure was 1.66 (95% CI, 1.18-2.33) per 1000 patient-weeks (Table 4). The SIR was 1.95 (95% CI, 1.53-2.50, P < .001), meaning that the risk of ILI for patients was almost 2 times higher in the hospital than in the community.
Analysis of the SIR heterogeneity of ILI by ward specialty is given in Table 4. The SIR of ILI was significantly different between hospital units (P < .001). No evidence of increased risk was apparent in surgical units (SIR, 1.01; 95% CI, 0.54-1.88) compared with in the community. A significantly higher ILI incidence was observed in medical units (SIR, 4.91; 95% CI, 3.75-6.43), and a significantly lower ILI incidence was noted in obstetrical units (SIR, 0.09; 95% CI, 0.01-0.65) compared with in the community.
The main objective of this prospective study was to estimate the risk of HA-ILI in hospitalized patients according to exposure to contagious patients and HCWs with ILI in the hospital. The A/H1N1 influenza pandemic7 strengthens the need for data on the spread of influenza in health care settings because nosocomial transmission of this new virus is known to have occurred.23,24 We found a relationship between HA-ILI risk and exposure to potentially infectious patients or HCWs inside the hospital (Figure). The results also suggested a 2-fold greater risk of ILI in the hospital compared with in the community (Table 4). These results support the public health decision in France to encourage patients with influenza to stay at home. Thus, spread of the virus in the hospital could be high because the risk of HA-ILI increases rapidly as soon as there is exposure to contagious patients or HCWs.
Nosocomial influenza and nosocomial ILI are reported most often when there are outbreaks,25 and prospective surveillance data are sparse. There are some advantages to this study's prospective design: data quality is better compared with that of retrospective designs, RR can be calculated with enough statistical power, and 3 years of investigation provides opportunities to estimate the RR of nosocomial ILI during periods with various ILI epidemics in the community, assuming, nevertheless, that the characteristics of the hospitalized population did not change dramatically during the observation period. The patients hospitalized at Edouard Herriot Hospital come from a part of the Rhone-Alpes region, and the selection bias regarding demographics is limited. The adjustment on age took account partially for that point.
However, some limitations should be addressed. A detection bias might be related to differences regarding the detection of ILI in the community compared with in the hospital. In the community, cases need to consult a general practitioner to be diagnosed as having ILI and, therefore, not all ILI cases are observed, providing an underestimation of the true incidence of ILI in the community. Then, the patients identified through the community network could be the more severe cases of the community. In a different way, the hospital active surveillance leads to better detection of ILI cases whatever the severity, reflecting the actual ILI incidence in hospitalized patients. The underestimation of cases in the community might have contributed to marginally overestimate the 2-fold higher risk of ILI in the hospital compared with in the community. However, it is less likely that these differences between the 2 surveillance systems have affected the intrahospital analysis and the relationship between patients' HA-ILI risk and exposure to potentially infectious patients and HCWs. A limitation of this study was the potential confounding effect of vaccination. We have no information on vaccination coverage among non-ILI cases, avoiding any adjustment. We reported the vaccination status of cases only and, therefore, could not estimate the potentially protective effect of immunization in patients. However, in 1 unit where 75% of HCWs were vaccinated against influenza, 9 incident cases were found in patients and 1 case was observed in HCWs, but no transmission occurred among HCWs (data not shown). In another unit where 5.3% of HCWs were vaccinated against influenza, 1 case of influenza was found among patients and 5 cases of influenza were diagnosed among HCWs, and we identified 3 potential transmissions between HCWs (data not shown). Finally, the present study was conducted in symptomatic patients, and the proportion of patients infected by influenza without any symptoms may be estimated to range from 30% to 50%. The analysis was performed on HA-ILI cases because only 13 of 64 patients (20%) with HA-ILI had laboratory-confirmed influenza. However, a strong correlation between ILI incidence and the epidemic activity of influenza has been reported.26
The definition of nosocomial ILI or influenza remains an open issue. Salgado et al17 suggest an incubation period of 72 hours after hospital admission, whereas a recent systematic review27 proposes less than 48 hours. For bacterial nosocomial infection, a delay of 48 hours after admission is common.28 We considered incident cases to estimate the risk of HA-ILI without taking into account the delay until onset. However, sensitivity analysis assuming that patients were not at risk for HA-ILI within 48 hours after admission obtained similar results (data not shown). The issue of delay remains open for viral nosocomial diseases, which are studied less than bacterial nosocomial infections. Viral nosocomial infections receive less attention than do bacterial infections, and the magnitude of morbidity and mortality from nosocomial influenza,17,25 respiratory syncytial virus,29,30 or rotavirus infections31,32 highlights the need for appropriate preventive action to control them in health care settings. In addition, emerging threats, such as influenza A/H1N1, severe acute respiratory syndrome, or avian influenza might arise or reemerge.
Many countries prepared good responses against pandemic risk at the community level. However, attention must also be given to the management of cases in hospitals, and the risk of HA-ILI or influenza must be anticipated18 because it will affect patients and HCWs.23 Descriptive and analytical reports of the risk of HA-ILI or influenza transmission in hospitals would be helpful to identify primary interventions, including hand hygiene, the use of appropriate masks and gloves, patient isolation, and vaccination to reduce the likelihood of hospital outbreaks with pandemic and seasonal strains. These measures are intended to curtail the risk of transmission by respiratory droplets, air, and contact.33,34 Their implementation, including vaccination, remains challenging for patients and especially HCWs, for whom vaccination coverage is highly variable.35 Intuitively, high coverage of vaccination among HCWs could diminish the spread of viruses inside hospitals and have a protective effect on patients and HCWs. In addition, influenza vaccination of patients with regular contact for care in hospitals might have a protective effect on them and, additionally, against viruses spreading in hospitals.
In conclusion, these results indicate that the risk of HA-ILI in patients increased with the level of exposure to contagious patients and HCWs. Preventive actions, including hygiene measures and vaccination, might reduce HA-ILI and influenza incidence during hospital stays. Education on the risk of ILI and influenza transmission should be provided to patients,36 visitors, and HCWs for effective prevention of transmission inside hospitals in the contexts of seasonal and pandemic influenza.
Correspondence: Philippe Vanhems, MD, PhD, Service d’Hygiène, Epidémiologie et Prévention, Hôpital Edouard Herriot, 5 place d'Arsonval, Lyon F-69437, France (firstname.lastname@example.org).
Accepted for Publication: July 16, 2010.
Author Contributions: Dr Vanhems had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Vanhems, Pires-Cronenberger, Giard, Lina, Barret, and Pollissard. Acquisition of data: Escuret, Regis, Gorain, Giard, Lina, Najioullah, David, Crozet, and Comte. Analysis and interpretation of data: Vanhems, Voirin, Roche, Pires-Cronenberger, Giard, Lina, Hirschel, and Ecochard. Drafting of the manuscript: Vanhems, Voirin, Gorain, Pires-Cronenberger, Crozet, and Comte. Critical revision of the manuscript for important intellectual content: Vanhems, Voirin, Roche, Escuret, Regis, Pires-Cronenberger, Giard, Lina, Barret, Pollissard, David, Hirschel, and Ecochard. Statistical analysis: Voirin, Roche, Pires-Cronenberger, Giard, Hirschel, and Ecochard. Obtained funding: Vanhems, Voirin, Pires-Cronenberger, Lina, and Barret. Administrative, technical, and material support: Vanhems, Escuret, Regis, Gorain, Pires-Cronenberger, Giard, Najioullah, Pollissard, David, Crozet, and Comte. Study supervision: Vanhems, Pires-Cronenberger, Lina, Pollissard, Comte, and Hirschel.
Financial Disclosure: Dr Lina has received travel grants from Roche; research grants from Roche and Sanofi Pasteur; and consulting fees from Roche, GlaxoSmithKlein, Biocryst, Merck, and Sanofi Pasteur.
Funding/Support: This study was supported by a “Programme Hospitalier de Recherche Clinique (PHRC) Régional” from the French Ministry of Health and Sanofi Pasteur.
Role of the Sponsor: The sponsors had no role in the design and conduct of the study; in the collection, management, analysis, and interpretation of the data; or in the preparation, review, or approval of the manuscript.
Previous Presentation: This study was presented at the 49th Interscience Conference on Antimicrobial Agents and Chemotherapy; September 15, 2009; San Francisco, California.
Additional Contributions: Collaborative clinical units at Edouard Herriot Hospital: Rheumatology—Pierre-Dominique Delmas, Roland Chapurlat, and Aurélie Fontana; General and Digestive Surgery—Etienne Tissot, Xavier Barth, and Olivier Monneuse; Geriatry—Brigitte Comte and Henriette Paulet-Lafuma; Obstetrics-Gynaecology—Pascal Gaucherand; Federation of Digestive Specialities—Jean-Alain Chayvialle, Olivier Boillot, and Jean Boulez; Emergency and Resuscitation—Dominique Robert; Ophthalmology—Philippe Denis and Carole Burillon; Hematology—Mauricette Michallet; Burns Service—Fabienne Braye; Internal Medicine—Jacques Ninet and Brigitte Coppere; Nephrology—Maurice Laville, Jean-Pierre Fauvel, and Denis Fouque; Transplantation Medicine—Jean-Louis Touraine; Otorhinolaryngology—Éric Truy and François Disant; Urology—Xavier Martin; and Orthopaedic Surgery—Jean-Paul Carret and Jacques Bejui-Hugues.