Participants and Setting

The Influenza Cohort Study, initiated by the Working Adult Influenza Cohort Study Group, a research team based in Toronto, Ontario, Canada, was started in May 2009. The purpose of the study was to examine incidence, clinical features, and epidemiology of infection caused by A(H1N1)pdm09 among HCWs and other working adults in Canada. For this analysis, participants were enrolled during May 29–September 27, 2009. Participants were eligible if they were 18–75 years of age and either worked >8 hours per week in 1 of 5 acute care hospitals (HCW) or in an office-based setting in Toronto (non-HCW). Non-HCWs were intended to provide a sample of working adults at low occupational risk for influenza, so as to bias the study toward the ability to identify an occupational risk in health care. Details of the recruitment of these control participants are included in the Technical Appendix. The study was approved by the Research Ethics Boards of all participating hospitals and universities and by the human resources departments of participating employers.

Upon enrollment, participants received a collection kit, an illustrated guide, and instruction from a nurse for mid-turbinate nasal swab sample self-collection. They also completed a Web-based questionnaire detailing influenza vaccination history, underlying medical conditions, demographic data, potential work- or school-related risk factors for respiratory virus infection, and potential community risk factors. Blood samples were taken from consenting participants at enrollment and again in April or May of 2010.

Participants were asked to complete weekly Web-based diaries from enrollment until March 31, 2010, detailing respiratory symptoms and acute respiratory illness (ARI) or febrile illnesses and documenting time-dependent risk factors (e.g., contact with persons with ARI symptoms). Per the study protocol, if any signs or symptoms suggestive of an ARI developed, participants provided a self-collected mid-turbinate nasal swab sample as soon as possible after onset to be tested for influenza by using PCR. ARI was defined as 1) fever without another obvious source; or 2) new symptoms, including >2 of the following: runny or stuffy nose, sneezing, sore or scratchy throat, hoarseness, or cough; or 3) one local (runny/stuffy nose, sneezing, sore/scratchy throat, hoarseness, or cough) and 1 systemic symptom (fever, malaise, myalgia, headache, or fatigue).

Participants whose specimens tested positive for influenza were offered treatment in accordance with public health recommendations (8). All participants with undetermined A(H1N1)pdm09 vaccine status as of March 31, 2010, were contacted again to confirm whether they had received it and, if so, when. For logistical reasons, participants with unconfirmed 2009–2010 seasonal influenza vaccine status could not be contacted again; instead, these participants were assumed not to have received it. In Canada, vaccine for A(H1N1)pdm09 became available for HCWs and patients at high risk for complications of influenza during calendar week 43 (starting October 25, 2009) and was available for healthy adults during calendar week 47 (starting November 22).

Definitions

For this study, HCW were defined as persons working in an acute care hospital. A non-HCW was defined as a person working in an office-type environment not associated with the provision of health care. The first and second waves of the influenza pandemic in Ontario were defined as the periods for which the weekly proportion of respiratory specimens that were positive for A(H1N1)pdm09 was >5%, as reported by the Ontario Agency for Health Protection and Promotion. Similarly, seasonal influenza waves were defined as periods for which >5% of weekly specimens tested positive for seasonal influenza. By this definition, the first pandemic wave occurred during calendar weeks 21–31 of 2009 (May 17–August 8); the second wave occurred during calendar weeks 39–48 (September 27–December 5). Peak weeks were defined as weeks during which positivity rates were >15% and comprised calendar weeks 21–27 (May 17–July 11) during wave 1 and calendar weeks 41–46 (October 11–November 21) during wave 2. As expected, few cases of seasonal influenza were identified during the study period.

Aerosol-generating medical procedures were defined as any of the following: administration of nebulized therapy or humidified oxygen at >40%, use of bag-valve mask, manual ventilation, noninvasive ventilation, open airway suctioning, bronchoscopy or other upper airway endoscopy, tracheostomy, endotracheal intubation, cardiopulmonary resuscitation, oscillatory ventilation, or any procedure that involved manipulation of open ventilator tubing in a mechanically ventilated patient or sputum induction or other deliberate induction of coughing.

Adherence to hand hygiene and facial protection recommendations was defined as the self-reported proportion of situations during which hand hygiene and facial protection were performed according to infection control recommendations (9). Symptomatic influenza infection was defined as influenza-positive PCR results for a participant-collected mid-turbinate nasal swab sample.

Antibody Assays and Interpretation

Serum specimens were extracted from blood samples and 1 mL aliquots frozen at −70°C. Aliquots were tested by hemagglutination-inhibition (HAI) assay to determine antibody titers against the A(H1N1)pdm09 strain (A/California/07/2009-like) and the 2008–09 seasonal A(H1N1) strain (A/Brisbane/59/07) to identify potential cross-reactivity by using a protocol adapted from World Health Organization methods (10). Two HAI assays were performed per aliquot by using 0.5% turkey erythrocytes and 4 hemagglutination units per 25 µL of virus. For discordant pairs, the higher of the 2 geometric mean titers was used. Serum specimens were tested at the Queen Elizabeth II Health Sciences Centre, Halifax, Nova Scotia, Canada. Seroprotection was defined as having HAI antibody titers of >40. Seroconversion was specifically defined as a prevaccination HAI titer of <10 and a postvaccination titer of >40 or a 4-fold change in titers for participants with a prevaccination titer of >10 (11,12).

Data Management and Statistical Analyses

Data were entered online by the participants, then cleaned and manually inspected for errors and outlying values. Differences in group proportions were assessed by the χ² or Fisher exact test, as appropriate, and differences in means (for normally distributed data, on the basis of the Shapiro-Wilk test for normality) and medians (for non-normally distributed data) were calculated by using Student t test and Wilcoxon rank-sum test, respectively.

The analysis for the primary objective (i.e., to determine whether the risk for laboratory-confirmed symptomatic influenza was higher in HCWs than in non-HCWs) included all participants who were enrolled by the start of the second wave of the 2009 H1N1 influenza pandemic (calendar week 39, starting September 27, 2009). Multivariable generalized estimating equation logistic regression analysis was used to determine adjusted odds ratios with 2-sided 95% CIs for constant and time-dependent risk factors for symptomatic influenza infection on the basis of information from baseline questionnaires and weekly diaries. Model construction was performed on the basis of the method proposed by Harrell (13) including A(H1N1)pdm09 vaccination status and changing risk for influenza infection over time (community influenza activity). Our a priori approaches to adjust for changing risk for influenza infection over time were to 1) adjust for weekly percentage of specimens positive for influenza reported to the Ontario Agency for Health Protection and Promotion (continuous variable) and 2) adjust for peak weeks (defined as weeks during which >15% of specimens were positive for influenza; [dichotomous variable]). Vaccine failure among participants was defined as acquiring A(H1N1)pdm09 infection after receipt of A(H1N1)pdm09 vaccine >7 days before symptom onset. Participants who acquired A(H1N1)pdm09 within 7 days after vaccination were considered not fully protected. To evaluate the validity of this assumption, we performed sensitivity analyses by calculating lags of 0 days and 14 days, respectively. The same criteria were used in the analysis of the secondary objective (i.e., to determine risk factors for laboratory-confirmed symptomatic influenza among HCWs). The models with the lowest quasi-likelihood under the independence model criterion were preferred.

Data were analyzed in SAS, version 9.1 for PC (SAS Institute, Cary, NC, USA). We considered p values <0.05 as statistically significant.

Sample Size

This study was initiated at the onset of the 2009 influenza pandemic; because the expected incidence of infection was unknown, a formal sample size was not established. Details of the sample size estimate for the planned seasonal study can be found in the online Technical Appendix.

Study Population, Symptomatic Influenza Case-patients and Community Influenza Activity

Figure

Figure. . . . Flowchart of 732 persons enrolled in the Influenza Cohort Study, Toronto, Ontario, Canada

The first participant was enrolled in the study on May 28, 2009 (calendar week 21). By October 11 (calendar week 41), at the start of the second wave of the pandemic, 732 participants were enrolled in the Influenza Cohort Study: 563 (76.9%) were HCWs who worked in 1 of 5 community and teaching acute care hospitals in the Toronto area and 169 (23.1%) were non-HCWs who worked in an office environment not associated with the provision of health care (Table 1; Figure). Of the 2 cohorts, HCWs were younger and were more likely to have been vaccinated against seasonal and pandemic influenza, to work with children, to have children <5 years of age in their households, and to use public transportation >8 times per week. Of 422 HCWs who were vaccinated against A(H1N1)pdm09, 403 (95.5%) received vaccine within 2 weeks after its availability; of 61 non-HCWs, 28 (45.9%) were vaccinated during the same time period (p<0.001).

A total of 334 (45.6%) study participants submitted 436 nasal swab samples. More than half (52.1%) of these samples were collected on the day of symptom onset (day 1), 19.4% on day 2, 9.9% on day 3, and 12.1% on or after day 4. Among the 20 (4.6%) specimens yielding influenza, 12 (60.0%) were collected on day 1, four (20.0%) on day 2, three (15.0%) on day 3, and one (5.0%) on day 4 of illness. Thirteen (2.2%) of 563 HCWs and 7 (4.1%) of 169 non-HCWs submitted samples that tested positive for influenza. A(H1N1)pdm09 was detected in 19 (95%) of the 20 positive participants: 1 case during each of calendar weeks 24, 25, 31, 39, 40, and 47; two cases during each of calendar weeks 42, 44, and 45; and 7 cases during calendar week 43. Thus, 16 of 19 cases occurred during weeks of peak A(H1N1)pdm09 activity. Seasonal influenza A(H3N2) virus was isolated in a sample from 1 participant during calendar week 43.

Risk Factors for Symptomatic Influenza Infection

The probability of symptomatic influenza infection did not differ between HCWs and non-HCWs (p = 0.28) (Table 2). Study participants who had a child <18 years of age living in the household (36.2% of influenza negative/untested participants vs. 65.0% of influenza positive participants; p = 0.009), a child who attended day care living in the same household (12.5% vs. 30.0%; p = 0.03), and who were not vaccinated against A(H1N1)pdm09 >7 days before onset of infection (76.3% vs. 10.0%; p<0.001) were more likely to have respiratory illness with positive test results for influenza.

After adjusting for A(H1N1)pdm09 vaccination history and community influenza activity, we found no difference in the risk for influenza infection between persons working in an acute care hospital (HCWs) and other healthy adults (non-HCWs) (Table 3). Rather, contact with a family member with an ARI in the previous week was the main risk factor for symptomatic influenza infection, irrespective of the method of adjusting for changing risk over time. In general, quasi-likelihood under the independence model criterion statistics were lower in models adjusting for weekly percentage of specimens yielding influenza than in those adjusting for weeks of peak influenza activity (results not shown). A sensitivity analysis calculating lags of 0 and 14 days (vs. 7 days) from the time of receipt of A(H1N1)pdm09 vaccine did not alter these results.

Analyses restricted to HCWs and including potential occupational risk factors in health care are shown in Table 4. During the study period, 49.6% of HCWs worked in emergency departments, medical inpatient wards, intensive care units, or pediatric wards; 12.9% were present during >1 and 9.4% performed >1 aerosol-generating medical procedure per week. Approximately one quarter (26.5%) of HCWs reported providing direct care for >1 patient per week who had ARI. The analysis of risk factors for infection indicates that, similar to the combined study population, HCWs with symptomatic influenza infection confirmed by positive nasal swab sample were more likely to have children <18 years of age in their households (69.2% of HCWs who tested positive vs. 36.9% who tested negative or were untested; p = 0.02) and less likely to have been vaccinated against A(H1N1)pdm09 >7 days before onset of infection (15.4% vs. 86.3%; p<0.001) (Table 4). Compared with other HCWs, those with symptomatic influenza infection were more likely to be present during aerosol-generating medical procedures >1× per week (38.5% vs. 12.7%; p = 0.02) and reported lower adherence to hand hygiene recommendations (77.5% vs. 95%; p = 0.02). After adjustment for changing risks for influenza infection over time, risk factors for influenza infection among HCWs were: contact with a family member with ARI in the previous week, performing or assisting with aerosol-generating medical procedures, and lower adherence to hand hygiene recommendations (Table 5).

HAI Antibody Assays

Among the combined study population, 450 (61.5%) of 732 participants provided pre- and post-influenza season blood samples. Among those, 3.6% had protective HAI titers against A(H1N1)pdm09 at baseline. There was no association with workplace and baseline HAI titers. Of the 142 (31.6%) participants who tested positive after enrollment, 137 (96.5%) had received the A(H1N1)pdm09 vaccine, 2 (1.4%) submitted a nasal swab that tested positive by PCR, and 3 (2.2%) did not submit a swab for testing or report an ARI (consistent with asymptomatic infection).

Analysis of data collected during the period after vaccine became available for unvaccinated participants without known previous A(H1N1)pdm09 infection showed that 8 (16.3%) of 49 HCW and 3 (5.3%) of 57 non-HCWs seroconverted or had a positive mid-turbinate nasal swab sample. Although persons working in an acute care hospital were 3.1× as likely as other working adults to be infected with influenza, the results were not significant in this small unvaccinated group (95% CI 0.9–11.1). Influenza among unvaccinated participants was not associated with age, sex, or any of the other characteristics listed in Table 1.