Responses were received from one law professor respectively from Italy, Denmark, Germany, and from two law professors from the UK. The responses of European law professors did not differ significantly from those of their Swiss colleagues, except for case 4 (Appendix 1
A total of 508 physicians sent in completed questionnaires. The participation rate was 56% for community physicians (n = 378) and 52% for hospital physicians (n = 130). The mean age of community physicians was 51 ± 8.4 years, compared to 36 ± 6.7 years for hospital physicians (P<0.01). The mean total years of practice as a physician was 24 ± 8.7 for community physicians and 9 ± 6.3 for hospital physicians. Community physicians had been working in private practice for 14 ± 8.8 years, whereas only a minority of hospital physicians were working in private practice (mean 0.4 ± 1.9 years, P<0.01). Community physicians had practised for a considerable number of years in the hospital before starting a private practice (11 ± 6.0 years) a mean of 2 years more than physicians working in the university hospital: 9 ± 5.8 years, P<0.01. The majority of participants were internists: 86% of hospital physicians and 68% of community physicians (P<0.01).
Among both groups of physicians, the same percentage of participants (40%, P = 0.97) had a subspecialty (such as, oncology, infectious diseases, or other) in addition to the specialty exam in internal or general medicine (in Switzerland, general medicine is considered a specialty requiring 5 years of training and a board certification).
Participants' responses were analysed with respect to two questions: did participants recognise that a breach of confidentiality took place; and did participants attribute scores that were similar to the gold standard?
The percentage of physicians who thought that no violation occurred varied in an important way according to the type of case; for example 4% in case 6, compared to 57% in case 2. Community physicians' attitudes did not differ significantly from the attitudes of hospital physicians except for case 3: 14% (community physicians) versus 32% (hospital physicians, P<0.01) were convinced that no violation took place.
The means scores of law professors, physicians, and law and medical students who answered the same questionnaire in a previous study are compared in and .43
In case 2, physicians identified the violation of confidentiality and their severity more correctly than young law students in Geneva did. In case 3, hospital physicians did not differ from students, but physicians in private practice identified the violation more often than hospital physicians and medical and law students. For the comparison of the results in case 1 it should be noted that physicians had two versions of the case, whereas students only had one that was similar but not identical to case 1b: in the student's version, it was assumed, but not explicitly stated, that the physician does not mention the patient's name, but only refers to him as one of the five state councillors.
Table 1 How many times the score 0 (no violation of confidentiality) has been attributed among physicians, medical,43 and law students.43
Participants who thought that no violation (0), a non-important violation (1), a violation for which the physician should receive a blame (2), or a serious violation (3) had taken place in the case scenarios 1–6.
A total of 11% of physicians, 9% of medical students, and 7% of law students thought that a violation of confidentiality had taken place in all cases, in line with the gold standard (). In only two cases was the proportion of physicians who mistakenly believed that no violation occurred below 10%: in case 1a (politician's name is mentioned) and case 6 (computer specialist).
A total of 83% of physicians identified the violation of confidentiality in case 5 (squat). A similar percentage (74% versus 80%, P = 0.200) of hospital and community physicians thought that the patient is identifiable. Those who thought that the patient is identifiable expressed significantly different opinions about the case (they judged that an important violation occurred: score 1.9 ± 0.8) than those who thought that the patient is not identifiable (score 0.6 ± 0.6, P<0.001).
In two cases, the majority of physicians (76% in case 4 and 57% in case 2) considered that no violation had taken place. In most cases, mean scores were 1 to 2 points lower than the gold standard. Only in case 6 was the mean score of physicians very close to the gold standard (2.7 compared to 3).
Responses were influenced significantly by the type of case. The study analyses did not show a trait towards more or less respect of confidentiality; that is, the hypothesis that some physicians respect confidentiality in general less than others, due to education, ethical choices, or other factors, was not confirmed. Reliability analysis showed a Cronbach's α of 0.34 for students and physicians for all six cases (using only variant b for case 1). It was highest among community physicians (α = 0.57 for seven cases including case 1a, and 0.48 without case 1a), followed by hospital physicians (α = 0.48 for seven cases and 0.34 without case 1a) and medical students (α = 0.13 for six cases). Reliability was lowest for law students (α = 0.06 for six cases).
Factor analysis revealed one factor explaining 34% of the variance of responses to questions 2, 3, 4, and 5, and one factor explaining 51% of the variance of the responses to questions 1b and 6. One factor explained 59% of the variance of the responses to question 2 and 4. The results of the factor analyses were similar in the subgroups of physicians and students.
Multivariate analysis entering all subgroups as dependent variables showed that the group variable that best explained the variances of the dependent variable ‘non-identified violations among 6 cases’ () was the binary variable ‘all physicians and medical students’ versus ‘law students’. However, this variable explained only 1.5% (adjusted R2) of the variance of the number of non-identified violations. More detailed results concerning the influence of demographic and other variables on the responses to the cases are reported elsewhere.44