We enrolled a total of 558 healthcare workers into the study, of which 96% (537/558) had ≥1 follow-up blood sample; 6 participants were excluded because of missing follow-up review questionnaires, leaving 531 persons for analysis. Of these, 35 (6.6%) seroconverted. Table 1 compares selected characteristics of seroconverters and nonseroconverters. Seroconverters were sampled earlier than nonseroconverters (49% vs. 38% in the first week of enrollment), and 86% of seroconverters had both follow-up samples compared with 81% of nonseroconverters, but these differences were not significant (p = 0.20 and p = 0.73, respectively). Seroconverters were slightly more likely to have received seasonal influenza vaccine than were nonseroconverters (97% vs. 91%), but this difference was not significant. There were no also significant differences between seroconverters and nonseroconverters by age or gender. HI titers in baseline samples from nonseroconverters were higher than in baseline samples from seroconverters (GMT 7.8 vs. 5.9; p = 0.02). Among seroconverters, 63% and 51% reported having an ARI and FRI episode, respectively, and only 15% and 8% of nonseroconverters reported having an ARI and FRI episode, respectively (p<0.01 for both).

Most of our participants were nurses (290/531, 55%; Table 2); allied health staff, which included mostly participants from paramedical professions such as pharmacists, laboratory medicine technicians, physiotherapists and occupational therapists, formed the second largest group (116/531, 22%); ancillary and support staff, which included mainly hospital attendants and patient service associates, formed the next largest group (69/531, 13%); and administrative support staff (35/531, 7%) and doctors (21/531, 4%) made up the rest. Seroconversion rates were highest in nurses (28/290, 10%) and lowest in allied health staff (2/116, 2%). To facilitate interpretation, allied health staff were designated the reference group for computing ORs; only nurses had a significantly higher odds of infection compared with allied health staff (OR 6.1, 95% CI 1.4–26.0; p = 0.02). Compared with those working in non–patient care areas, participants whose primary work area was an inpatient ward had higher odds for seroconversion (OR 1.4, 95% CI 0.5–3.5; p = 0.54), while those in other patient care settings had lower odds for seroconversion (OR 0.5, 95% CI 0.2–1.5; p = 0.21), but neither result was significant. Significantly higher odds for seroconversion were also observed for participants whose primary work area was in pandemic (H1N1) 2009 isolation wards (OR 4.8, 95% CI 1.5–15.6; p<0.01).

Participants who had contact with patients who had pandemic (H1N1) 2009 had marginally but not significantly increased odds of seroconversion (OR 1.8, 95% CI 0.9–3.7; p = 0.10). Those who reported having contact with a sick colleague(s) whose illness was subsequently diagnosed as pandemic (H1N1) 2009 had significantly increased odds of seroconversion (OR 2.9, 95% CI 1.2–6.9; p = 0.02).

Figure

Figure. Univariate analysis for nonoccupational exposures to pandemic (H1N1) 2009 among healthcare workers, Singapore. Error bars indicate 95% confidence intervals (CIs) for odds ratios (ORs). †n/N, no. of seroconverters/no. in strata. HH,...

Results of the univariate analysis for nonoccupational exposures are presented in the Figure. Healthcare workers from larger households had increased odds of seroconversion (OR 1.2 per additional household member, 95% CI 1.0–1.4; p = 0.04), but no discernible association was seen between seroconversion and having another healthcare worker in the same household or reporting another household member with ARI or FRI symptoms during the study. However, having a child or adolescent in the household increased the odds of seroconversion. In particular, significantly higher ORs were observed if the healthcare workers reported a child 5–12 years of age in the household (OR 2.1, 95% CI 1.0–4.4; p = 0.05), with the ORs being even higher if a child 5–12 years of age in the household had FRI symptoms (OR 4.1, 95% CI 1.1–15.6; p = 0.04).

The variables included in the final multivariate analysis are shown in Table 3. Being a nurse remained significantly associated with increased odds of infection (OR 4.5, 95% CI 1.0–19.6; p = 0.05), as did having pandemic (H1N1) 2009 isolation wards as a primary work area (OR 4.5, 95% CI 1.3–15.6; p = 0.02). Contact with colleagues with pandemic (H1N1) 2009 (OR 2.5, 95% CI 0.9–6.6; p = 0.06) and coming from a larger household (OR 1.2 per additional household member, 95% CI 1.0–1.4; p = 0.06) were of borderline significance. Having higher HI titers in the baseline serum sample was protective (OR 0.5 per unit of increase, 95% CI 0.3–1.0; p = 0.05).

In Table 4, allied health participants are compared with nurses, the groups with the lowest and highest seroconversion rates respectively; the latter was further stratified by whether they worked in inpatient wards (ward based vs. non–ward based). The proportion who seroconverted was slightly, but not significantly, higher in ward-based nurses than in non–ward-based nurses (11% vs. 8%; p = 0.53). Ward-based nurses were from significantly larger households than the other 2 groups (p<0.01 vs. allied health, p = 0.01 vs. non–ward-based nurses). No significant difference was found in the proportion who reported using face masks all or almost all of the time in patient care, but ward-based nurses were significantly more likely to have had seasonal influenza vaccine than were allied health workers (p<0.01). Key differences were found in the mean number of contacts and occupational related factors. Non–ward-based nurses mostly worked in large areas, including operating theaters, the emergency department, and outpatient clinics, and hence had significantly higher numbers of contacts than either ward-based nurses or allied health workers (p<0.01 on all measures). On the other hand, ward-based nurses were significantly more likely to be in contact with patients with confirmed pandemic (H1N1) 2009 (p<0.01 vs. non–ward-based and allied health), and allied health staff were significantly less likely to be in contact with a sick colleague who had pandemic (H1N1) 2009 (p<0.01 vs. either nursing group).

In this study, we used paired serum samples to assess infection rates and risk factors for infection in healthcare personnel during an influenza pandemic in an acute care hospital in Singapore. We observed surprisingly lower seroconversion rates in healthcare personnel than in the rest of the community, as was emphasized in another publication (15), and found that a mixture of occupational and nonoccupational exposures were associated with risk for infection.

When the study was planned, we had expected the healthcare workers cohort to have a higher seroincidence than a group of community-dwelling adults, given previous reports of pandemic and nonpandemic influenza outbreaks in hospitals (19,20), our own experience with SARS (1), and recent work showing the intensity of work-related contacts in the healthcare setting (3). Instead, we found that only 7% of our healthcare workers seroconverted, compared with 13% of participants in the community cohort (15). Our study corroborates case-reporting data in the United States, which suggest that healthcare workers did not have a higher incidence of infection than the general community, without being subject to biases that might arise from underreporting or differential case ascertainment (9).

Although definitively attributing the low infection rates in healthcare workers to improved infection control practices is difficult without the appropriate control groups, much evidence supports the efficacy of the common bundle of measures used in hospitals to reduce spread of respiratory viruses (21). Notably, there was a high level of preparedness and widespread implementation of airborne and respiratory droplet precautions and other pandemic (H1N1) 2009 infection control practices in healthcare institutions in the United States, Singapore, and elsewhere (11,13,22–24).

However, our study also suggests that the risk to healthcare staff should not be underestimated. We found some occupational factors associated with seroconversion. The higher seroconversion rates in nurses posted to designated pandemic (H1N1) 2009 isolation wards should be interpreted with some caution in view of the small number of seroconversion events (2 of those infected had symptoms and 2 did not) and participants (n = 17) from these wards. Since masks (either surgical or N95 masks) were widely used in all clinical areas around the hospital, this group essentially had the same level of protection as other staff while being far more intensely exposed to pandemic (H1N1) 2009. The higher risk for seroconversion for nurses on multivariate analysis also deserves notice. Nurses had higher seasonal influenza vaccination rates and were more compliant than other occupational subgroups in following preventive measures such as mask use. However, they also were more likely to be exposed to patients as well as to have colleagues with confirmed pandemic (H1N1) 2009; the latter factor was significantly associated with seroconversion by univariate analysis (and of borderline significance on multivariate analysis), and staff-to-staff transmission was also implicated in TTSH and elsewhere (10,13). Non–ward-based nurses also had higher contact rates than the other main occupational subgroup (allied health staff), a factor that we could not account for in multivariate analysis (as questions on contact rates were not answered by all participants). We also could not account for the nature of patient contacts, which might be more prolonged and intense in nurses (3). We suggest that our finding of the higher seroconversion risk in nurses is the result of residual confounding by the sum of these factors, many of which are an integral part of the nursing profession.

Lastly, our study suggests that nonoccupational exposures should not be forgotten as a potential source of healthcare worker infections. Other studies based on case investigations have also attributed some infections to community sources, and in our study, we found that having a child of primary school age was a risk factor on univariate analysis, particularly if that child had an FRI during the study period, although the direction of transmission in the latter could not be ascertained. Studies on nonpandemic influenza have found that index cases from pediatric age groups were more likely to generate secondary cases (25,26), although the same was not observed with pandemic (H1N1) 2009 (27). The effect of having children in the household was superseded in multivariate analysis by overall household size, which was unsurprising since households with children also tended to be larger. In any case, the significance of such nonoccupational exposures should be taken into account in any hospital-level pandemic preparedness plan.

We do acknowledge several limitations in our study. First, our findings are based on data from healthcare workers from just 1 hospital. Moreover, the unexpectedly low seroconversion rates in our cohort reduced the power of the study to investigate exposures more weakly associated with the outcome. We were also unable to assess the usefulness of personal protective equipment due to the lack of appropriate control groups. The resolution of exposure data from what was a self-administered questionnaire survey was also lower than insights that may be gained from detailed case investigations or exposure diaries that have been used in the healthcare setting (3,11). Finally, the lack of randomization also leaves scope for bias in our results.

An effective vaccine for pandemic (H1N1) 2009 has now been introduced, and this will likely reduce intrahospital risk of infection from this particular strain of influenza until significant genetic drift occurs, provided healthcare institutions can overcome the challenges to achieving high vaccine coverage rates in healthcare personnel (28,29). Seasonal influenza vaccination rates may be atypically high in TTSH because of its designated status as a first-line screening and referral center; 1 study on healthcare workers from 2 other hospitals in Singapore found that only 39% of participants were vaccinated (30). Although the low incidence of healthcare workers infections provides some suggestion that measures in place during the pandemic were effective, our findings suggest that some occupation-related risk factors remain. Nurses, particularly those working in pandemic (H1N1) 2009 isolation wards, were disproportionately affected, possibly because their higher levels of protective behaviors inadequately compensated for their increased occupational risk. This situation should be recognized when planning for future pandemics.