The German Federal Police comprises a total of approximately 41,000 members. Of those, about 33,000 persons are uniformed officers. Among them, 13% are women leading to a male/female ratio of 6.69:1 (all data were obtained by the press office of the German Federal Police on June 23rd, 2010). No official figures on the total number of officers working in the police unit formerly known as "Bundesbahnpolizei" (Federal Railway police) were available, but an estimated total of 5,500 officers was communicated to us through personal communication with the press office of the German Federal Police. However, their number may vary greatly due to special time dependent police operations.
Based on peer review and professional commitment, an investigator derived questionnaire (see additional file 1
: German Federal Police questionnaire on railway suicides) was developed to assess knowledge and experience of German Federal Police officers with railway suicide in a quantified manner. Previous versions of the questionnaire were tested at the Federal Police Academy Lübeck and critically evaluated by administration staff of the German Federal Ministry of the Interior. The authors took the protection of personal data very seriously and adhered strictly to the regulations of the world medical association declaration of Helsinki and of the German Data Protection Acts (§ 3). For the present study, particulars about personal or factual circumstances of a defined or definable natural person were factually anonymised. The German Federal Ministry of the Interior acted as the ethics committee for the study and questionnaire and approved both. The questionnaire consisted of 27 questions in six modules. The first module contained questions related to the German Federal Police officers' socio-economic status, the number of suicides they had experienced on duty and the counselling or debriefing after such a suicide. In the second module, items were related to the suicide that officers had experienced most intensely
(source of information, impact of suicide on the officer and whether they had ever prevented suicide). The following modules three and four were related to the suicide victim (age, sex) and the time, date and place (open track, station, eminent spot in station) of the suicide. Finally, in the fifth module, questions were aimed at the observed suicide victims' behavioural patterns such as erratic gesture, mimic or movement, out of the ordinary clothing, aimlessly wandering about, erratic communication pattern, dropping of personal belongings, confused impression, indication of alcohol or the avoidance of eye contact. In a final section, officers were given the possibility to write comments. The questionnaire was made available on the Federal Police Intranet. Officers who had experienced a suicide on duty were asked to fill in the questionnaire. In the questionnaire, we explicitly asked for the suicide which the officer remembered best. Filled questionnaires were returned to the Institute of Epidemiology II, Helmholtz Zentrum München, directly via email or postal mail. Descriptive and multivariate analyses were applied.
A total of 209 subjects returned their filled questionnaires from August to October 2008 and from February to April 2010. The two waves were mainly due to practical and administrative reasons. There was no significant difference between the waves in police officers' sex (p = 0.16), age group (p = 0.27), suicides experienced (p = 0.21) and years of service (p = 0.49). Of the returned questionnaires, seven were excluded from the analysis because the data was unreadable. The remaining 202 subjects (mean age: 41 years, ranging from 19 to 59 years of age) were included in the present analysis (Table ). However, information on officers' sex was missing in two questionnaires. The majority of the study population was male (n = 169; 84.9%), leading to a male/female ratio of 5.45:1 which was comparable to the overall male/female ratio among police officers of 6.69:1 as described above.
General description of the German Federal Police Officers' sample (n = 202)
To analyse the age distribution of the German Federal Police officers, "age" was divided into four groups (in years: < 30, 30-39, 40-49, ≥ 50). To analyse a possible influence of years of service, years of service were divided into three groups (in years: ≤ 10, 11-20, >20). Furthermore, the number of suicides experienced on job was divided into three groups (≤ 5, 6-20, >20), following roughly the tertiles of the distribution.
Bivariate statistical associations were assessed by χ2 test or Fisher's exact test. A multivariate logistic regression analysis with backward elimination variable selection was performed for predicting the prevention of suicide as the outcome. In this model, age, number of suicides experienced on the job, years of service, sex, counselling/debriefing, information about suicide obtained by witnesses and the fact that officers had observed a suicide by themselves were used as predictors.
A multivariate logistic regression analysis with backward elimination variable selection was performed for predicting demands for counselling as the outcome. In this model, age, number of suicides experienced on the job, years of service, sex, whether officers had prevented a suicide, information about suicide obtained by witnesses and the fact that officers had observed a suicide by themselves were used as predictors. For both multivariate logistic regression analyses an entry criterion of a p value < 0.25 in the bivariate analysis and a stay criterion of p value < 0.05 in the end model of the logistic regression was used, following recommendations of Hosmer and Lemeshow [13
]. To assess the goodness of fit of the multivariate logistic regression models, we used the c statistic and the Hosmer-Lemeshow Goodness of fit test. For all statistical analyses, a p value less than 0.05 was considered to be statistically significant. All evaluations were performed with the statistical software package R and SAS Version 9.2.