The study population comprised 9,978 healthcare workers working at 7 tertiary-level hospitals distributed throughout Japan; 4 of the hospitals are university-affiliated, 2 are municipal, and 1 is private. The study participants held a wide range of jobs in each institution. The questionnaire was administered from July through September 2003. Overall, 7,463 healthcare workers responded to the questionnaire (crude response rate 74.8%). After missing or invalid responses for sex, age, or job category were excluded, 7,282 were finally analyzed (valid response rate 73.0%) (Table 1).

This study formed part of an international collaborative study involving healthcare workers in Singapore, China, Taiwan, Canada (Toronto), and Japan. The questionnaire was developed in English at the National University of Singapore, translated into Japanese, and adapted to accommodate background conditions (i.e., no outbreak). The questionnaire was anonymous, and procedures involving human participants were approved by the institutional review board of the University of Occupational and Environmental Health, Japan.

The questionnaire included 24 items regarding knowledge of preventive measures (15 items), concept of (opinion regarding) institutional measures (4 items), and perception of risk (5 items) (Appendix). These 24 items were measured on a 7-point scale for responses (strongly agree, agree, probably agree, probably disagree, disagree, strongly disagree, and not applicable). In the statistical analyses, we dichotomized this scale into positive response (strongly agree, agree, and probably agree) and negative response (strongly disagree, disagree, and probably disagree) after excluding "not applicable."

To assess knowledge of preventive measures, we analyzed responses to questions regarding the effectiveness of measures to avoid contracting SARS (personal protective equipment, personal hygiene, environmental measures). The 15 items are shown in Appendix. The correct response to each item was designated on the basis of World Health Organization (WHO) guidelines (2) and other findings. The correct responses for the 15 items were a positive response for all items except "paper mask" and "gauze mask," which required a negative response. To assess concept (opinion) of institutional measures, we used 4 items regarding "clear policies and protocols," "specialist available," "adequate training," and "effectiveness." To assess perception of risk, we used 5 items regarding "avoidance of patient," "acceptance of risk," "little personal control," "fear," and "job change," as indicators (Appendix).

We quantified the degree of concept of institutional measures and that of knowledge of preventive measures by calculating the institutional (I) and knowledge (K) scores. The I-score was defined as the total number of positive answers to the 3 specific questions regarding "clear policies and protocols," "specialist available," and "adequate training"; the maximum possible I-score was 3 points. "Effectiveness" was excluded from the calculation of the I-score because it could be looked upon as a combined, general concept of institutional measures. The K-score was defined as the total number of correct (either positive or negative) answers to the 15 questions regarding the knowledge of preventive measures; thus the maximum possible K-score was 15 points. The K-score was categorized as high (11–15 points), middle (6–10 points), or low (0–5 points). Cronbach's α was 0.87 for the K-score and 0.76 for the I-score, which indicated a high degree of internal consistency for each score.

The chi-square test was used to evaluate differences in the proportion of respondents according to job category (physician, nurse, and other), sex, age, and type of facilities. The Student t test was used to evaluate differences in the mean value between 2 groups, and analysis of variance was used to evaluate differences in the mean value among 3 groups. Logistic regression analyses were used to identify factors associated with the overall concept of the effectiveness of institutional measures ("effectiveness") and perceptions of risk ("avoidance of patient" and "acceptance of risk") as the dependent variables. The independent variables were the I-score (0, 1, 2, and 3 points), the K-score (low, middle, and high), age (<35 years old, >35 years old), sex (men, women), fear (–, +), and type of facility (nonuniversity hospital, university hospital). The logistic regression model was applied to each of the 3 job categories and to all participants. Spearman's correlation coefficients among 6 independent variables were <0.26, and no strong correlations were seen among them. Data were analyzed by using SPSS, version 11.5J (SPSS Inc., Chicago, IL, USA) for Windows and SAS V8 (SAS Institute Inc., Cary, NC, USA). All reported p values are 2-tailed, and p < 0.05 was considered statistically significant.

Regarding knowledge of preventive measures, most respondents assigned relatively high importance to hand hygiene and area isolation but saw personal protective equipment as being of relatively low importance. This finding may be partly due to healthcare workers' not having previously used some of the protective equipment recommended for use with SARS patients (3). The use of personal protective equipment as countermeasures for SARS has been rightly advocated by various authors (10,14,15). Thus, adequately training healthcare workers in the use of personal protective equipment is an important aspect of reducing the incidence of SARS infection.

Regarding the concept of institutional measures, 40% of respondents believed that they had received adequate training; for example, less than half felt that they had adequate training in the use of masks. During the SARS epidemic, medical institutions were required by authorities to provide adequate training to healthcare workers in affected countries (5,6,10,11). Because no outbreaks were in Japan, however, Japanese institutions have not been forced to implement sufficient measures to adequately cope with future outbreaks of SARS and other emerging diseases.

Regarding perception of risk, although we did not compare healthcare workers with an external group, more than half (55%) of the healthcare workers surveyed indicated that they were afraid. Furthermore, a high proportion of healthcare workers preferred to avoid the patient (92%), although almost two thirds accepted the risk (64%). When these 2 items were cross-classified, 55% of respondents showed a mixed attitude (i.e., avoidance of patient [+] and acceptance of risk [+]), 32% showed a disloyal attitude (i.e., avoidance of patient [+] and acceptance of risk [–]), and 6% showed a loyal attitude (i.e., avoidance of patient [–] and acceptance of risk [+]). These results indicate a high level of fear and anxiety with complex psychology in Japanese healthcare workers, even in the absence of an epidemic.

Significant differences were seen in the level of knowledge and attitudes among the 3 job categories. Nurses showed the best knowledge of preventive measures and concept of institutional measures, while physicians showed the highest acceptance of risk. Both sex and job characteristics may have influences in this regard. Ninety-eight percent of nurses were women, whereas 84% of physicians were men. Quah et al. reported that, in Singapore, women showed better practice of SARS preventive measures than men among the general population (7). Similarly, our results indicated a higher level of knowledge regarding preventive measures for female physicians compared to male physicians. However, this trend was not observed within the nurse job category, although the number of male nurses was sufficiently small that separating the effect of sex was difficult. In terms of job characteristics, nurses may receive more official training in infection control than physicians, under the assumption that physicians are already knowledgeable. In fact, compared to physicians, nurses have higher levels of compliance with universal precautions (16) and hand-washing (17,18) in their respective countries. However, nurses tend to have higher job turnover rates than physicians, which reflect less stability or security in their profession. These factors directly and indirectly influence the response pattern among the 3 job categories.

In the logistic regression model, K-score, an indicator of knowledge of preventive measures, was not a significant predictor of perception of either risk or concept of institutional measures. This finding implies that professional knowledge has little, if anything, to do with positive perception of risk (in terms of accepting risk and not avoiding patients) and concept of institutional measures. However, the importance of providing accurate knowledge cannot be discounted solely on this ground. In contrast, I-score, an indicator of concept of institutional measures, was a significant positive predictor of concept of effectiveness and acceptance of risk and a significant negative predictor of avoidance of patient. In other words, a collective assertion of 3 specific institutional measures (clear policies and protocols, specialists available, and adequate training) had the greatest effect on a person's 2 different aspects of perception of risk and concept of the effectiveness of institutional measures. These findings corroborate earlier studies reporting that administrative support enhances compliance with universal precautions (19–21) and hand washing (17,18). Therefore, we infer that perception of institutional measures affects perception of risk and related behaviors.

Fear was a significant negative predictor of concept of effectiveness in 2 job categories (physicians and others) and a significant positive predictor of avoidance of patient in all 3 job categories. These findings were in line with our expectations and signal the need to reduce fear as a practical goal. Older age was a significant positive predictor for the concept of effectiveness in 2 job categories (nurses and others). Among nurses, older age was also a significant negative predictor for avoidance of patient and a significant positive predictor for acceptance of risk. Age has previously been shown to be a positive predictor for practicing SARS preventive measures among the general population (7). Hence, older age seems to correlate with an increased ability to cope with emergency situations related to infectious diseases. Type of facility (university hospital) was a significant negative predictor for the concept of effectiveness in 2 job categories (nurses and others). Although confined to the 7 facilities studied, university hospitals may have been less stringent in the formulating or implementing infection control measures, which in turn affected the overall concept of effectiveness of measures among healthcare workers.

Our study has several limitations. First, the cross-sectional nature of the study prevents assertion of cause and effect. Our conclusion, particularly on interrelationship among individual factors, is based on inferences. Second, responder bias may have been in play, i.e., only workers with a strong interest in SARS may have been motivated to respond, although the fairly high response rate counteracts this argument to an extent. Third, K-score may not accurately reflect knowledge of preventive measures. For example, workers who, in practice, had accurate knowledge about shoe covers as personal protective equipment may have answered incorrectly because they had been taught conflicting information. In fact, the Infectious Disease Surveillance Center (IDSC), Japan, categorizes shoe cover as an optional personal protective equipment (22). However, among the personal protective equipment considered in this study, only alcohol rubs (WHO) (2) and shoe cover (IDSC, Japan) (22) are considered optional, and the effect of conflicting information should not be strong. Fourth, we considered the difference in type of facility (university or nonuniversity) but did not consider differences by facility (hospital A or B) or type of unit (internal medicine, surgery, and others), which may be related to differences in job descriptions (even within the same job category) as well as the study variables. Such effects caused by affiliation constitute a separate theme worth further investigation, which will be pursued.

We found that the level of anxiety among healthcare workers in Japan was relatively high and that the implementation of preventive measures at the institutional level was not perceived to be sufficient. However, a collective assertion of 3 specific institutional measures stood out as the most important predictor for individual perception of risk, including avoidance of patient and acceptance of risk, as well as concept of general effectiveness of institutional measures. In view of the potential for future epidemics of SARS or other emerging infectious diseases, the planning and implementation of institutional measures should be given a high priority.

"Do you believe the following measures are useful in protecting you from contracting SARS?"

Concept of Institutional Measures

Perception of Risk