OUP user menu

Ten-year incidence and time trends of railway suicides in Germany from 1991 to 2000

J. Baumert, N. Erazo, K. H. Ladwig
DOI: http://dx.doi.org/10.1093/eurpub/cki060 173-178 First published online: 10 August 2005

Abstract

Background: Railway suicides strongly impact psychological and socioeconomic aspects of the railway company, its employees and possible eye witnesses. We aimed to assess the incidence of railway suicides compared with the total burden of suicides in Germany and its time trend over a 10-year observation period. Methods: From 1991 to 2000, a total of 8653 fatal railway suicides were recorded by the national central registry of all passenger accidents on the German railway net. To assess time trends, the average annual percentage change (AAPC) of the number of suicides was estimated using Poisson regression. Adjustments were made for two age groups and measures of availability. Results: Railway suicides accounted for 7.0% (annual range 5.0–7.9%) of all suicides in Germany. The time trend of the number of suicides by all means declined, with a significant stronger decline for older subjects [AAPC −2.2; 95% confidence intercal (CI) −3.0 to −1.5; and −4.2; 95% CI −5.2 to −3.0]. An AAPC of 1.8% (95% CI 0.1–3.4) indicated an increasing time trend of the number of railway suicides for subjects aged ≤65 years, whereas, in contrast, fatal incidences for subjects aged >65 years exhibited a strong decrease (AAPC −6.2%; 95% CI −10.9 to −1.3). Conclusions: The present study demonstrates different time trends for railway suicides compared with all suicides. It underlines the urgent need for amplified research efforts in the particular field of railway suicides. Further examinations are recommended, in particular concerning possible reasons for this suicidal method.

  • availability
  • public burden
  • railway suicides
  • time trends

The first railway suicide ever officially recorded was a male subject in the year 1852 about 22 years after the beginning of the railway era.1 Since then, corresponding with the build-up of the railway system worldwide, a steady increase in the number of railway suicides have been registered, accounting for a serious mortality issue.

However, only a few studies about railway suicides on a population-derived basis have been carried out so far.18 As can be seen in table 1, the evidence of the contribution of railway suicides compared with suicides by all means relies mainly on restricted and mostly dated figures. Data on long-term trends of railway suicides are even more scarce. Railway suicides are a suicidal behaviour that strongly impacts psychological and socioeconomic aspects of the railway company, its employees and possible eye witnesses.

View this table:
Table 1

Population-based studies on railway suicides (overview)

StudyPopulationSample sizeStudy period% of all suicidesRailway suicide ratesaTrend analysesb
Veress and Szabó (1980)2Hungaryn = 4361965–19692.74Decrease
Attempts: n = 50
Lindekilde and Wang (1985)3County of Fyn, Denmarkn = 161979–19823.1Increase
Emmerson and Cantor (1993)4Brisbane, Australian = 231980–19862.5Stability
Symonds (1994)6South Englandn = 801979–19800.31
Clarke (1994)1England/Walesn = 50371850–1949Males: <1–>6Males: 0.1–1.1Strong increase among males
Females: <1–3Females: <0.1–<0.2Small increase among females
Schmidtke (1994)5Federal Republic of Germanyn = 60901976–1984Males: 5.65Males: 1.86
Attempts: n = 391Females: 4.16Females: 0.66
Deisenhammer et al. (1997)7Austrian = 5521990–19945.73Increase
Attempts: n = 65Males: 5.32
Females: 6.80
van Houwelingen and Beersma (2001)8The Netherlandsn = 301980–199410–14
  • a: Per 100 000 inhabitants

  • b: Trend over the study period

Based on the stable grounds of a 10-year representative critical incidence dataset for the German railway system, the major aim of the present study was to assess the mean proportion of railway suicides in Germany and to investigate the time trend of railway suicide incidences compared with national suicide incidence over a 10-year observation period.

Methods

Material

The database for the present study is derived from the STABAG (Statistik der Bahnbetriebsunfälle und Gefährlichen Unregelmäßigkeiten, Statistics of Railway Accidents and Dangerous Irregularities) registry, which is the national central registry of all person accidents for the German railway company covering the entire German railway track system except municipal subway providers. The mean track length of the German railway system during the 10-year observation period was 39 696 km. The mean total train mileage (in km) was 936 million and the mean total passenger mileage (in km) was 69 920 million (based on data from 1993 to 2000). For annual figures see table 2.

View this table:
Table 2

Track length, train and passenger mileage (in km) of the German railway system during the observation period from 1991 to 2000

YearTrack kma (absolute)Train kmb (millions)Passenger kmc (millions)
199141 112.37
199240 826.82
199340 384.92919.9062 741
199441 401.04927.1064 539
199541 718.26924.7070 334
199640 826.42925.8071 028
199738 450.23930.1071 630
199838 126.32941.6071 853
199937 525.02950.2072 846
200036 587.60965.3074 387
  • a: Total length of the German railway system

  • b: Kilometres covered by all trains, data for 1991 and 1992 was not available

  • c: Kilometres covered by all passengers, data for 1991 and 1992 was not available

The present analysis covers the 10-year observation period from 1991 to 2000. During this time period, all cases from the STABAG registry were included that satisfied the operational definition of an act of suicidal behaviour according to the ICD-10 category ‘intentional self-harm by jumping or lying in front of a moving object’ (code X81).9 Fatal suicide was defined as ‘death within 30 days’. The dataset allowed only a rough age distribution with two categories (subjects aged ≤65 years and aged >65 years).

The national suicide rates were taken from the German Report on Health Statistics provided by the Federal Statistical Office of Germany and the Robert-Koch-Institute, which were currently available until the year 2000.10 The total population in Germany increased from 80.275 million to 82.260 million inhabitants during the 10-year observation period.11

Statistical analysis

Estimation of time trend. Time trends in the 10-year observation period from 1991 to 2000 were analysed by Poisson regression models.12 The regression equation (1): Math contained basically the year of suicide as linear term, the age group (two categories) and an offset term ‘log(population number per year and age group)’ to adjust for annual changing of population figures and age structure. The average annual percentage change (AAPC) of the number of suicidal events estimated by equation 1 was then defined as: Math A 95% confidence interval (CI) for the AAPC was calculated with Poisson regression in a corresponding manner. In case of over-dispersion of the Poisson model, the dispersion parameter was estimated by the ratio of the deviance to its associated degrees of freedom.12

Time trend for railway and all suicides. First, to assess an overall time trend during the 10-year observation period, a main effect model was calculated for railway and all suicides using regression equation 1. To assess the time trend regarding different age groups, an AAPC was estimated by calculating an interaction model with the inclusion of the interaction term ‘year * age group’ as an additional covariate in the regression equation 1.

Owing to fixed documentation rules unalterable by the authors, the age groups were slightly different for railway suicides (≤65 years, >65 years) and all suicides (<65 years, ≥65 years).

Time trend regarding measures of availability. To assess the time trend regarding the location of suicide, we repeated the regression analysis stratified for suicides located at the station area or on the open track. Moreover, to investigate the availability as a possible reason for a time trend in railway suicides, the AAPC with the same covariates was calculated using each railway system parameter per year instead of the population number per year and age group as offset term in the Poisson regression equation 1. An interaction model was calculated with the inclusion of an interaction term ‘year * age group’ as an additional covariate in the regression equation 1.

For all statistical analyses a P value <0.05 was considered to be statistically significant. Operations were performed with the statistical software package SAS for Unix, version 8.2.

Results

Railway suicides and all suicides

During the 10-year observation period from 1991 to 2000, a total of 9510 suicidal incidences occurred on the German railway net. In 8653 (91.0%) cases, the outcome was fatal, indicating a mean number of fatal incidences of ∼17 suicidal events per week. The absolute number of completed suicide incidences in the German population covering the same time period was 124 124 cases. Thus, fatal railway suicides accounted for 7.0% of all suicides over the 10-year observation period, ranging from 5.0% in 1991 to 7.9% in 1998 (table 3). With ∼10% aged >65 years, railway suiciders were on average younger than suiciders by all means, with a percentage of ∼30% of subjects aged ≥65 years.

View this table:
Table 3

Absolute numbers and rates (per 100 000 inhabitants) of railway suicides and all suicides in Germany during the observation period from 1991 to 2000

YearNumber of railway suicidesNumber of all suicidesProportion of all suicidesRailway suicide rateGeneral suicide rate
199170314 0115.00.917.5
199278413 4585.81.016.6
199390112 6907.11.115.6
199486712 7186.81.115.6
199598612 8887.71.215.8
199689512 2257.31.114.9
199794112 2657.71.114.9
199892411 6447.91.114.2
199984311 1577.61.013.6
200080911 0687.31.013.5
Total8653124 1247.0

Time trends for railway and all suicides

As can be seen in table 3, the number of suicides by all means in the general population declined overall significantly from 14 011 in 1991 to 11 068 in 2000, accounting for an AAPC of −2.8% (95% CI −3.6 to −2.1; P < 0.0001) (table 4). In contrast, the number of railway suicides remained stable (P = 0.303). In the interaction models, the interaction term ‘year * age group’ was revealed to be significant for all suicides (P = 0.006) and railway suicides (P = 0.003), indicating different time trends for the age groups. For all suicides, the same time trend direction with a stronger decline in the number of suicides was observed for older subjects. In contrast, for the number of railway suicides, a qualitative interaction with age group was observed: for subjects aged ≤65 years, an AAPC of 1.8% (95% CI 0.1–3.4) indicated an increasing time trend, whereas fatal incidences for subjects aged >65 years exhibited a strong decrease (AAPC −6.2%; 95% CI −10.9 to −1.3).

View this table:
Table 4

AAPC of the number of railway suicides and all suicides related to the population number per year and age group during the observation period from 1991 to 2000

ModelRailway suicidesAll suicidesa
AAPC (95% CI)P valueAAPC (95% CI)P value
Main effect modelb
    Overall1.0 (–0.9 to 2.9)0.303−2.8 (−3.6 to −2.1)<0.001
Interaction modelc
    Year * age group0.0030.006
    ≤65 years1.8 (0.1 to 3.4)−2.2 (−3.0 to −1.5)
    >65 years−6.2 (−10.9 to −1.3)−4.2 (−5.2 to −3.0)
  • a: Age groups were <65 years and ≥65 years

  • b: Model covariates: year of suicide, age group with offset term log(population number per year and age group)

  • c: Model covariates: year of suicide, age group, year of suicide * age group with offset term log(population number per year and age group)

Time trends concerning measures of availability

The track length of the German railway system during the 10-year observation period from 1991 to 2000 decreased from 41 112.37 km to 36 587.60 km in the time period observed. In contrast, the total mileage (in km) run by the trains and the passenger volume (per km) increased in the observation period, as displayed in table 2.

We repeated the calculations as before, stratified for the location of suicides. The interaction term ‘year * age group’ was significant for the subgroup of suicides located on the open track (P < 0.001), but not for suicides located at the station area (P = 0.216). An AAPC of 2.4 (95% Cl 0.9–3.9) indicated an increasing time trend for suicides on the open track for subjects ≤65 years (table 4).

Moreover, three different models were calculated using track length, mileage covered by all trains or mileage covered by all passengers as an offset term for the Poisson regression model. Again, the interaction term ‘year * age group’ introduced in the Poisson model was highly significant in each of the three models (P < 0.01).

For subjects aged >65 years the AAPC was significantly lower than zero for train and passenger mileage, indicating a strong decrease of the numbers of railway suicides. In contrast, for subjects aged ≤65 years, a significant increase of the number of railway suicides was observed related to the track length (AAPC 3.2; 95% CI 1.7–4.7). No significant time trend related to mileage covered by all trains but a significant decrease in the number of suicides related to passenger volume was observed by the AAPC (table 5).

View this table:
Table 5

AAPC of all railway suicides stratified for the location of suicide related to the population number per year and age groupa,b and to the railway system parameters track length, train and passenger mileage (in km) of the German railway systemc during observation period from 1991 to 2000

Age groupAAPC95% CI
≤65 years
    Station areaa0.52.6 to 3.7
    Open trackb2.40.9 to 3.9
    Track lengthc3.21.7 to 4.7
    Train mileagec,d−1.1−2.8 to 0.6
    Passenger mileagec,d−2.7−4.0 to −1.4
>65 years
    Station areaa−5.9−6.3 to 5.4
    Open trackb−6.5−6.6 to −6.4
    Track lengthc−3.8−8.2 to 0.7
    Train mileagec,d−8.8−13.8 to −3.5
    Passenger mileagec,d−10.4−14.2 to −6.4
  • a: Model covariates: year of suicide, age group, year of suicide * age group with offset term log(population number per year and age group), stratified for suicides at the station area

  • b: Model covariates: year of suicide, age group, year of suicide * age group with offset term log(population number per year and age group), stratified for suicides on the open track

  • c: Model covariates: year of suicide, age group, year of suicide * age group with offset term log(railway system parameter)

  • d: Data were only available from 1993 to 2000

Discussion

Main findings

The first major result of the present analysis is to demonstrate that train suicide incidences over a 10-year observation period with a fatal outcome on the German railway net account for 7.0% of all suicides in the German population. This proportion, reflecting 17 fatal railway suicidal events per week, indicates that railway suicides are of public concern. To date, comparable epidemiological evidence from other populations in Europe and elsewhere are scarce. However, the limited data point to the conclusion that the proportion of 7.0% lies within, but at the outer end of, an international range, as can be seen in table 1. Only van Houwelingen and Beersma8 recently reported a substantially higher annual ratio of railway-to-all-suicides of 10–14% for The Netherlands.

It is widely acknowledged that suicides on the railway track are inappropriate to serve as ‘gesture’ or ‘cry for help’, because no control over the effects of the attempt is available and it is not possible to solicit help after initiating the suicidal act. It is most likely that survival of a railway suicide attempt is accidental in the majority of cases.2,13

A second major result of the present study is to show that for subjects aged ≤65 years, an increase of the numbers of railway suicide events on the German railway net was observed in the 10-year observation period from 1991 to 2000. Contrary to these results, a substantial decrease of suicides by all means for both age groups was observed in the same time period, which is in line with findings in the European Union, Japan and the USA.1416

To the best of our knowledge, no time trend study relying on statistical analysis for railway suicides has been published so far. However, the finding that railway suicide trends contrast with general suicide trends is confirmed by subway suicide studies: Lester17 correlated the subway suicide rates of 17 cities around the world with the respective national suicides and also found differing rates. Recently, Ladwig and Baumert18 confirmed an almost stable annual incidence of subway suicides in Munich over a 20-year period despite a pronounced decline in total suicide mortality in the Munich population.

Measures of availability

Data on the suicide location and on various railway system parameters including track length, mileage covered by all trains and mileage covered by all passengers could be analysed to mirror the density and accessibility of this specific suicide method. Availability has been traditionally accused of be a strong promoter of selecting the means of suicide.1,19 In the present study, availability yielded no major influence on the time trend analysis; on the contrary, although track length was substantially reduced over the observation period, the number of railway suicides increased continuously. This finding may be in part due to a complex interaction between a substantial reduction of track length and a concomitant increase in passenger figures.

Regarding the suicide location for subjects aged ≤65 years, there was an increase in the number of suicides on the open track, in contrast to a stable number at the station area. In a recent British study based on a project for prevention of railway suicides (Suicides and Open Verdicts on the Railway Network, SOVRN), the distance between the nearest track to the last residence of the suicider (possibly a psychiatric clinic) was shown to be the main factor to explain a railway suicide.20 An open track is easily reachable by a possible suicide victim.

Railway suicides constitute a part of the total number of suicides, and since the total number of suicides decreased and railway suicides increased in the observation period, the net effect must be that non-railway suicides decreased with a steeper decent than the total amount. The contrast between railway and non-railway suicides was therefore more pronounced at the end of the observation period than at the beginning.

Limitations of the study

The dataset for the present study is derived from the STABAG registry, which basically serves routine documentation purposes to record all intentional and unintentional injuries on the railway track system in Germany except municipal subway providers. The present study encompasses only self-inflicted, intentional injuries. Every unnatural death is investigated by the local police and the local coroner by law, which makes misclassification and incomplete records unlikely.

The dataset for the present study did not provide sufficient information to explain possible reasons for the differences in the observed time trends. Mainly, the rough age distribution, with only two categories, and the lack of gender information seems to be dissatisfactory. The national central registry of all person accidents of the German railway company uses slightly different cut-off points for age groups than the Federal Statistical Office of Germany; however, we consider this difference to be only marginal.

The strength of the present analysis is that it relys on a complete and valid dataset of incident railway suicide cases covering the total German railway net over a 10-year observation period.

Implications of the study

Undoubtedly, railway suicides represent an important, however a still minor, compared with suicide means like hanging or poisoning, proportion on suicides by all means. Nevertheless, a mean of 17 cases occurred each week in the 10-year observation period on the German railway system. Suicides on the track often represent a ‘public death’, which is likely to distress and traumatize the train driver and may also severely affect an uncounted number of bystanders who witness the mutilated body.18 In some cases, particularly occurring on the open track, breakdowns of railway circulation for several hours are implicated, which may entail after-effects for large parts of the system. Therefore, prevention campaigns aiming to reduce railway suicide incidences are an important issue for both psychological and economical aspects of a society. The finding of an increasing time trend for railway suicides for subjects aged ≤65 years may help to promote the development of prevention campaigns to reverse this inauspicious trend.

Key points

  • Railway suicides strongly impact psychological and socioeconomic aspects of the railway company, its employees and possible eye witnesses.

  • Time trends of the incidence of railway suicides in Germany over a recent 10-year observation period were investigated.

  • The number of railway suicides for subjects aged ≤65 years increased significantly on average by 1.8% per year.

  • In contrast, fatal incidences of railway suicides for subjects aged >65 years exhibited a strong decrease on average by −6.2% per year.

  • The increase in railway suicides for subjects aged ≤65 years indicates the urgent need for prevention campaigns to reverse this inauspicious trend.

Acknowledgments

This study was supported by a grant from the Deutsche Bahn AG (to K.H.L.). We are indebted to Fritz Schröder, Jürgen Bedau and Holger Senzel (Deutsche Bahn AG) for initiating the research project and providing the data.

References

View Abstract