In total 3,840 respondents participated in the seven surveys; per survey the number of participants varied between 467 and 650. Overall response was 64% and varied between 55% and 77% (see Table ). In all but one, (i.e. the August 2006 survey) the majority of respondents were women, in total 54%. Compared to the general Dutch population, where 51% is female, women are slightly overrepresented in our surveys. The proportion of women in the first 4 surveys was lower compared to the last 3 surveys (51% vs. 58%, χ2
(df) = 16.5 [1
]; p < 0.001) (Table ). Mean age was 45 years (range 18-86, SD 14.8), with respondents in survey 1 being significantly younger (41.3) than respondents in the other surveys (46.0), (F(df) = 51.0[1
]; p < 0.001). Compared to the general Dutch population, people aged 45-59 years were somewhat overrepresented in the survey (31% versus 27%), and people aged 60 years and older underrepresented (20% versus 24%). Over all surveys, 26%, 39% and 35% of respondents had received lower, intermediate and higher education respectively. This distribution is 27%, 44%, and 29% in the general population, which shows overrepresentation of participants with high educational level and underrepresentation of those with intermediate educational level. 11% of the participants were of non-Dutch origin and 23% had been vaccinated against influenza. Children under 12 years lived in 22% of the households. Only 4% kept chicken or poultry at home.
Participation rates and distribution of general characteristics in the study population.
Infection with AI was perceived as a (very) severe health problem by 91.8% of the study population with a mean score of 4.57 (95% CI 4.55 - 4.60; scale 1 - 5); 0.7% reported (very) high perceived vulnerability, mean = 1.69 (95% CI 1.66 - 1.71, scale 1 - 5) (see Table ). Comparative vulnerability for AI was 2.59 (95% CI 2.56 - 2.61, scale 1 - 5, whereby 3 stands for an equal chance to others of the same age and gender). Compared with other diseases, getting infected with AI was perceived as very serious with a score of 4.57 on a scale from 0-5 (see Table ). Only HIV, 4.92 (95% CI 4.91 - 4.93), and heart disease, 4.82 (95% CI 4.80 - 4.83), had a significantly higher perceived severity. Perceived severity of other diseases and conditions varied from 4.18 (95% CI 4.16 - 4.21) to 1.87 (95% CI 1.85 - 1.89) for a common cold. In contrast perceived vulnerability for common cold was the highest with 3.64 (95% CI 3.61 - 3.67), while for HIV it was the lowest at 1.23 (95% CI 1.21 - 1.24). Comparative vulnerability of HIV was also low at 1.88 (95% CI 1.85 - 1.91), while for a common cold it was 3.07 (95% CI 3.05 - 3.09).
Perceived severity, vulnerability and comparative vulnerability (mean and 95% CI).
All four knowledge questions were answered correctly by 10.2% of the respondents; 16.5% had all answers wrong. The mean score for knowledge was 1.87 (95% CI 1.83 - 1.91, scale 0-4). Of all respondents 49% reported to have received a reasonable amount of information about AI and 7% (very) much, with a mean score for information received of 2.63 (95% CI 2.61 - 2.65, scale 1 - 5); 3% of respondents thought often or always about AI and 30% sometimes, resulting in a mean score of 2.19 (95% CI 2.17 - 2.22). 33% of respondents were not sure that they could do anything to prevent themselves from getting infected with AI, while 20% felt reasonably sure or very sure. The mean score for self efficacy was 2.23 (95% CI 2.20 - 2.27)
Almost half of the respondents (46%) reported taking one or more preventive measures, with 36% reporting to have stayed away from (wild) birds or poultry, 26% not going to areas where AI was present and 2% buying antiviral drugs (see Table ). 54% of respondents did not take any measures, 13% took one measure, 14% took two measures, 10% took three and another 10% took four or more measures.
Proportion of respondents that took preventive measures, overall and by measurement.
In the regression analyses, time was not significantly associated with perceived severity (β = -0.002, p = 0.772) and for perceived vulnerability the regression coefficient of time was just short of being statistically significant (β = -0.011, p = 0.08). Time was significantly associated with amount of information received (β = -0.065, p < 0.001), knowledge (β = -0.127, p < 0.001) and taking preventive measures (Table ). In the pairwise comparisons, some more significant differences in variables of interest were found between different surveys. Perceived severity was stable over the seven surveys, ranging between 90% (August 2006) and 94% (March 2007). Perceived vulnerability decreased slightly between March 2006, when 2% perceived a (very) large chance of getting infected the coming year, and September and December 2006, with 0.4% and 0.4% respectively (ANOVA September and December 2006 vs. March 2006 p = 0.005 and p = 0.040 respectively). Perceived vulnerability was increased in February 2007 compared to September 2006 (ANOVA p = 0.04) (Figure ). There was a significant decrease in the amount of information received about AI, from 2.88 in March 2006 to 2.47 in March 2007, in March 2006 14% had received (very) much information, in March 2007 this had decreased to 5% (Figure ). Knowledge about AI also showed a significant decrease from 2.33 in March 2006 to 1.51 in March 2007, with in March 2006 43% of respondents answering three or four questions correct while in March 2007 this was 22%. The level of knowledge was positively associated with the amount of information received (Pearson r = 0.24, p < 0.001).
Mean severity and vulnerability for AI with 95% confidence intervals. * additional surveys
Mean knowledge score and mean amount of information with 95% confidence levels. * Additional surveys.
There was a significant increase in the percentage of respondents who had taken preventive measures between March and June 2006 from 38% to 50% (χ2
(df) = 18.4 [1
], p < 0.001), while there was no difference among the second until the last survey, ranging from 50% in the second survey to 43% in the fourth survey (χ2
(df) = 8.2 [5
], p = 0.147). Avoiding contact with (wild) birds or poultry was reported most often, by 36% of respondents (ranging between 33% in September 2006 and 39% in August 2006). For the specific preventive actions, an increase over time was observed for not going to areas with AI (March 2006 18%, February 2007 33%), paying more attention to hygiene (9% March 2006, 18% in March 2007), cancelled or did not book a holiday to an area with AI (6% in September 2006, 11% in March 2007), getting oneself vaccinated against influenza (2% March 2006, 8% December 2006), avoiding gatherings of people (2% March 2006, 4% March 2007), buying antiviral drugs (1% March 2006, 4% March 2007), and buying a mouth mask (1% March 2006, 3% March 2007) (Table ; Figure ). No differences in risk perceptions, precautionary actions or information received were observed related to the specific events (August 2006 and February 2007) when compared to the previous and consecutive surveys.
Proportion of respondents that took preventive measures, overall and by measurement. * Additional surveys
Most demographic factors and knowledge and information determinants were significantly associated with perceived vulnerability (Table ). Perceived vulnerability was higher for women, for elder respondents, for respondents without children below 12, for those with a lower education, for those who thought more about AI, for those with a lower level of knowledge and for those vaccinated against influenza. Ethnicity and amount of information were not significantly associated with perceived vulnerability.
Determinants of vulnerability for avian influenza.
In univariate logistic analysis of precautionary behaviour as the dependent (outcome) variable, the demographic variables, the specific surveys and variables from the Protection Motivation Theory were included as independent variables (Table ). Of the demographic variables all, apart from gender and keeping chicken or poultry, were significantly associated with precautionary behaviour. Respondents with a higher age, a lower education, without children below 12, of non-Dutch ethnicity, and those who had been vaccinated against influenza were more likely to take preventive measures. Furthermore preventive measures were taken more often by those respondents who considered AI very severe, who had a higher perceived vulnerability, who had a higher self efficacy, who had less knowledge, who had received more information about AI and thought more about AI. Compared to the first survey, respondents in the subsequent surveys reported to have taken precautionary measures more often.
Proportions of respondents that reported to have taken any preventive measures and results from logistics regressions analyses (Odds ratio (OR and 95% confidence intervals (95%CI)) exploring correlates of preventive measures.
The results of the multivariate logistic regression analysis are also shown in Table . As the odds ratio's of variables already in the model did not change substantially after inclusion of variables in subsequent steps we present the full model. The only statistically significant interaction term in the model was between self efficacy and thinking about AI. In the final model the time of the survey, a higher age, a lower level of education, a non-Dutch ethnicity, being vaccinated against influenza, a higher perceived severity, a higher perceived vulnerability, a higher self efficacy, a lower level of knowledge, more information about AI, and thinking more about AI were all associated with taking preventive measures. Self efficacy was a stronger predictor of precautionary behaviour for those who never or seldom think about AI (OR 2.3; 95% CI 1.9 - 2.7), compared to those who think about AI more often (OR 1.5; 95% CI 1.2 - 1.9). The discriminative value of the final model, expressed as the area under the curve (AUC) with 95% confidence limits is 0.69 (0.67-0.71).