Immediately after the epidemic, the National Institute of Public Health and the Environment of the Netherlands sent a questionnaire to 1,747 persons, of whom 872 (49.9%) responded. Response was lowest among persons who actively handled the culling of poultry (9). Of the 872 persons who responded, 450 could be linked to a farm-visits database kept during the epidemic that contained information about who had visited which farm on which date for what reason. Of these 450 persons, 194 had been actively involved in hands-on culling during the depopulation; activities included catching live poultry and picking up dead poultry. Because this group had the highest exposure (4), it was used to analyze the effect of personal preventive measures.

The questionnaire asked for information about symptoms of infection of the eyes, from which a self-reported conjunctivitis result was derived as an outcome measure for subtype H7N7 infection. The presence of >2 of the following eye symptoms was classified as conjunctivitis: redness, tearing, itchiness, pain, burning, purulence, or sensitivity to light. Blood samples were collected from survey respondents 3 weeks after possible exposure and tested by a hemagglutination-inhibition (HI) assay for antibodies by using a modified cutoff based on validation studies of persons known to be infected and of nonexposed controls (10). The serologic result was available in addition to the self-reported conjunctivitis information.

We used a case definition that combined both outcomes, i.e., a person needed to self-report conjunctivitis and have a positive HI assay result. For sensitivity analysis, both the conjunctivitis and the serologic result were used in separate case definitions.

The survey also contained information about prophylaxis with a neuraminidase inhibitor (oseltamivir [Tamiflu; Roche, Basel, Switzerland]) and the use of personal protective equipment (PPE [safety glasses and respirator]). Beginning on March 14, prophylaxis with oseltamivir (75 mg daily) was prescribed to persons in contact with potentially infected poultry (2 weeks after the first infection was diagnosed and 10 days after culling began). Continuation of oseltamivir treatment was recommended until 2 days after possible exposure. The subtype H7N7 strain that circulated was susceptible to oseltamivir (2). To determine oseltamivir use, persons were asked the following questions: 1) Were you prescribed oseltamivir by a medical doctor? (yes/no). 2) If oseltamivir was prescribed, when was it prescribed and when did you stop using it? 3) How often did you fail to take a capsule? Persons who were not prescribed oseltamivir were classified as “did not use.” On the basis of the period of use stated in the survey, persons who were prescribed oseltamivir were classified per visit as “used” or “did not use.” Within the used category, persons who missed <3 capsules were classified as consistent users; if more capsules were missed, they were classified as inconsistent users.

PPE (respirator and safety glasses) were provided during the entire epidemic to persons involved in the depopulation. The Dutch Food and Consumer Product Safety Authority provided the PPE and supervised its use. For each visit, workers received new PPE. They were instructed how to use the PPE but received no extensive training. For both respirators and safety glasses, workers were asked the following questions: 1) Did you use respirators/safety glasses (yes/no)? 2) How often did you use respirators/safety glasses (always, almost always, sometimes, almost never, never)? 3) How often did you not use respirators/safety glasses? From responses to these questions, we classified persons into 3 categories: used, sometimes used, and did not use. Persons who had either always or almost always used respirators or safety glasses and who stated that they had not missed using them more than twice were placed in the “used” category. Persons who had not used them were classified as did not use, and all remaining persons with answers were classified as sometimes used. For safety glasses and respirators, information was available only about their use during the whole epidemic and not per visit. We therefore assumed that a person’s use of a respirator and safety glasses did not change during the epidemic. The respirators provided were type FFP2 (US equivalent N95, Figure 1), which protected both nose and mouth and were all the same size. Safety glasses covered only the front of the eyes and were open above, below, and on both sides of the eyes (Figure 1). They were effective against splashes but not against dust. No information was available about the overall health of the workers.

The probability of becoming infected during a visit was calculated from the number of visits each person had made and whether each person was infected (according to the case definitions). The analysis comprised only visits to farms with infected poultry. During each visit, a person could either escape infection or become infected. The probability of each sequence of events could be added, from which the probability of infection per visit could be estimated by maximum likelihood (Technical Appendix). The dependent variable in this analysis was the case definition, i.e., whether a person was infected. The probability of infection was calculated according to the categorical level of personal protection used, e.g., use of oseltamivir or not, as independent variables. We calculated confidence intervals (CIs) using profile likelihood methods (11). Statistical testing was performed by using likelihood ratio tests (11).

For oseltamivir, risk for infection was calculated per group and then aggregated over the groups of persons who had and had not used oseltamivir (Table 1, Table 2). Use of respirators, safety glasses, and oseltamivir were classified into the earlier described groups. We then calculated the risk for infection for each combination of groups, e.g., always used respirators/sometimes used safety glasses (Table 3, Table 4). Although onset date of infection was unknown, the likelihood function took into account that a person could become infected only once and that infection could occur at any farm visit.

Oseltamivir had been prescribed for 159 persons; 35 stated that oseltamivir had not been prescribed. Forty-five persons for whom oseltamivir had been prescribed did not use it during some of their visits. For the group of 130 (combined case definition), oseltamivir was prescribed for 114 and not prescribed for 16; for 34 persons, oseltamivir was prescribed but not used.

The estimated risk for infection per visit without use of oseltamivir was 0.145 (95% CI 0.078–0.233). This risk dropped significantly to 0.031 (95% CI 0.008–0.073; p = 0.005) per visit when oseltamivir was used (Table 2; Figure 2). When calculated over the infection probabilities, oseltamivir use had a protective effect of 79% (95% CI 40%–97%; relative risk [RR] 0.21, 95% CI 0.03–0.60). The risk for infection seemed to increase when oseltamivir was used inconsistently, but this risk did not reach significance.

Results with case definitions that used either the serologic results or self-reported conjunctivitis differed slightly. With the serologic result, the estimated risk for infection without oseltamivir use was 0.513 (95% CI 0.370–0.656). This risk dropped to 0.274 (95% CI 0.157–0.418; p = 0.01) with oseltamivir use (Table 5, Table 6), which resulted in a protective effect of 46% (95% CI 19%–68%; RR 0.54, 95% CI 0.32–0.81). With self-reported conjunctivitis used as case definition, the estimated risk for infection without oseltamivir use was 0.115 (95% CI 0.064–0.180). This risk dropped to 0.073 (95% CI 0.032–0.135; p = 0.08) with oseltamivir use (Table 5, Table 7), which resulted in a protective effect of 53% (95% CI 2%–82%; RR 0.47, 95% CI 0.18–0.98).

We could not analyze interactions between use of oseltamivir and PPE because the resulting groups would have had too few visits. In the group that always used respirators and protective glasses and the group that sometimes used respirators and protective glasses, oseltamivir use was relatively equal (63%–74%). In the group that sometimes used respirators and did not use safety glasses, oseltamivir was used in 38% of the visits. In the group that did not use safety glasses or respirators, oseltamivir was used in 25% of the visits. These findings indicate no strong correlation.

Persons generally were more inclined to use respirators than safety glasses (Table 3, Table 8). Persons who always used safety glasses also used a respirator. Only a small number of persons stated they had used no respirator and no safety glasses.

We gauged the effect of using safety glasses by comparing the risk for infection in persons with different safety glasses use within the group who always used respirators (Table 4). Within this group, risk for infection decreased with use of safety glasses; however, this trend was not significant. We gauged the effect of respirator use by comparing the risk for infection in persons with equal levels of safety glasses use but different respirator use (Table 4). Risk for infection was lower in persons who sometimes used a respirator than in persons who always used a respirator (within the groups that sometimes used or did not use safety glasses). Risk was higher for persons who had not used respirators than for those who had sometimes or always used them (within the group that had not used safety glasses). However, the group that had used neither respirators nor safety glasses was small. In both comparisons, the protective effect of respirators was not statistically significant. In the sensitivity analysis, results for which conjunctivitis and the serologic result were used as case definitions were similar to results with the combined case definition (Table 9).