STUDIES INTO TRAIN SUICIDE THE CONTRIBUTION OF PSYCHOPATHOLOGY, RAILWAY PARAMETERS AND ENVIRONMENTAL FACTORS

Transcription

1 STUDIES INTO TRAIN SUICIDE THE CONTRIBUTION OF PSYCHOPATHOLOGY, RAILWAY PARAMETERS AND ENVIRONMENTAL FACTORS Cornelis AJ van Houwelingen

2 Studies into train suicide The contribution of psychopathology, railway parameters and environmental factors

3 Cornelis AJ van Houwelingen Studies into train suicide The contribution of psychopathology, railway parameters and environmental factors Thesis in English with a summary in Dutch Cornelis AJ van Houwelingen, s Hertogenbosch, the Netherlands All rights reserved. No part of this thesis may be reproduced, stored or transmitted in any form or by any means without the written permission of the copyright owner. caj.van.houwelingen@ggze.nl ISBN: Cover photo: Straga; portrait pg 129: Charlotte Bogaert, Haarlem, the Netherlands Cover lay out & design: Lidia Palumbi Lay out: Simone Vinke, Ridderprint BV, Ridderkerk, the Netherlands Printed by: Ridderprint BV, Ridderkerk, the Netherlands

4 VRIJE UNIVERSITEIT Studies into train suicide The contribution of psychopathology, railway parameters and environmental factors ACADEMISCH PROEFSCHRIFT ter verkrijging van de graad Doctor aan de Vrije Universiteit Amsterdam, op gezag van de rector magnificus prof.dr. L.M. Bouter, in het openbaar te verdedigen ten overstaan van de promotiecommissie van de faculteit der Psychologie en Pedagogiek op donderdag 19 mei 2011 om uur in de aula van de universiteit, De Boelelaan 1105 door Cornelis Arie Johannes van Houwelingen geboren te Gorinchem

5 promotoren: prof.dr. A.J.F.M. Kerkhof prof.dr. D.G.M. Beersma The studies presented in this thesis were conducted within the EMGO + Institute ( nl). The EMGO + Institute participates in the Netherlands School of Primary Care Research (CaRe) which was re-acknowledged in 2000 by the Royal Netherlands Academy of Arts and Sciences. The studies were supported by the Integrated Mental Health Services Eindhoven (GGz Eindhoven). The publication of this thesis was financially supported by GGz Eindhoven, ProRail and NS (Netherlands Railways).

6 Suicide may thus not be seen solely as a private problem for the individual but as a form of human interaction for which society has to take responsibility Beskow et al, 1994 Reading committee: prof.dr. BL Mishara, Université du Québec à Montréal prof.dr. K-H Ladwig, Technische Universität München prof.dr. C van Heeringen, Universiteit Gent prof.dr. ATF Beekman, Vrije Universiteit Amsterdam prof.dr. M van der Gaag, Vrije Universiteit Amsterdam dr.ir. A Veenman, former CEO of NS (Netherlands Railways)

7

8 To Lidia and Dario

9

10 Contents Chapter 1 Introduction 11 Chapter 2 Train suicides in the Netherlands 17 Chapter 3 Mental healthcare status and psychiatric diagnoses of train suicides 31 Chapter 4 Psychopathology and suicide method in mental health care 39 Chapter 5 Seasonal changes in 24-h patterns of suicide rates: 51 a study on train suicides in the Netherlands Chapter 6 Difference in train suicide mortality between the Netherlands and 63 Germany: the impact of availability of trains Chapter 7 Train suicide in the Netherlands: the impact of railway traffic 75 intensity re-examined Chapter 8 Concluding remarks 85 References 99 Summary 107 Samenvatting 115 Acknowledgements 123 About the author 127

11

12 Chapter 1 Introduction

13

14 Introduction Train suicides belong to the most violent and lethal forms of self-destructive behaviour. They take place in the public space, involve other people and cause considerable delays to transport services. These characteristics set train suicides apart from other suicides. Although the earliest official recordings of train suicides date back as far as 1852 (Clarke M, 1994), the interest in the study of train suicide as a distinct phenomenon is of more recent date. Besides some earlier forensic reporting, it is only 60 years ago that the first systematic study on this subject appeared. Swiss physician Franz Baur wrote a dissertation describing 62 train suicides in (Baur, 1951). He noticed that the majority of the cases had had mental health problems. Almost all suicides occurred at night at places where one can wait for trains without being disturbed or seen by train drivers. Some twenty years later Guggenheim and Weisman pioneered with a study on underground train suicides (Guggenheim and Weisman, 1972), which was followed by a substantial body of international research on underground train studies, and since 1985 more studies on overground train suicides in different countries have been published (Symonds, 1985; Lindekilde and Wang, 1985). A high prevalence of diagnosed mental illness and a history of mental healthcare appeared to be a consistent finding among underground and overground train suicide studies (Mishara, 2007; Ratnayake et al, 2007; Krysinska and De Leo, 2008; Ladwig et al, 2009). As the type of psychopathology reported varied considerably, the question remained whether certain types of psychopathology had a special link with this suicide method. Also, the questions of whether train suicides are characterised by an overrepresentation of psychopathology or whether those who die by this suicide method have a different mental healthcare usage compared to other suicides had not been sufficiently studied. Some studies demonstrated that persons who committed train suicide lived or resided at short distances from the suicide location (Mishara, 1999; Abbott et al, 2003). This residential link would explain the existence of observed high-risk locations near psychiatric hospitals in some countries (Emmerson and Cantor, 1993; Erazo et al, 2004). At the same time this implies that high-risk locations may not be exclusively linked to psychiatric hospitals. An important question is whether train suicides are related to the tendency for suicide in a nation or rather are autonomous events. Only two studies have addressed this issue. One cross-sectional study on underground train suicides found no association with general population suicide rates (Lester, 1995a). Another, longitudinal study on overground train suicides, covering a 10-year period, found a negative association (Baumert et al, 2005). This scarce and equivocal evidence seemed insufficient to draw solid conclusions from, however. A study by Clarke convincingly demonstrated the contributory role to suicide frequency of the availability of the railway system in terms of railway track length (Clarke M, 1994). It should be mentioned, though, that this observation was made in the era of railway expansion. Later studies have questioned the relevance of railway density (Symonds, 1985; Baumert et al, 2005). Railway system availability can be divided into two components: a. railway density and Chapter 1 13

15 Chapter 1 b. the number of trains actually passing. The effect of the latter on train suicide incidence has hardly been explored and deserves further attention. The timing of train suicidal behaviour is an interesting study subject as different factors can be expected to play a role with various implications for train suicide prevention, such as a tendency to hide in the dark, as described by Baur, clock-time related factors like train traffic intensity or duty shifts in psychiatric hospitals, and environmental factors like seasonality (Maes et al, 1993). It seems important to distinguish underground and overground train suicides, as the majority of underground train suicides happen in artificially lit stations, whereas the majority of overground train suicides happen outside, on open tracks. While underground train suicides seem to be restricted to daytime hours, overground train suicides may manifest a different pattern. Schmidtke, who used 3-hr intervals, found a shift in male overground train suicides from an early evening peak in winter towards later hours in summer, demonstrating a possible influence of the light-dark cycle on train suicidal behaviour (Schmidtke and Ober, 1991). This finding needed to be replicated with a powerful dataset allowing for a finer resolution of time units in order to study time patterns better. In the early nineties of the previous century no comprehensive body of knowledge regarding train suicide existed in the Netherlands. Anecdotal reports fuelled a general awareness of train suicide as being a nasty but somehow inevitable problem occurring on the tracks near psychiatric hospitals. It was uncertain whether the observations reported from other countries were also representative of the Netherlands. In this period the Netherlands Railways started to develop a growing interest to mitigate this problem and improve aftercare for train personnel affected. Thus, with the help of the Netherlands Railways, it became possible to set up a large national database of suicidal behaviour resulting in train-person collisions. This database, which includes incidents of jumping or lying in front of a moving train or crashing a motor vehicle against a moving train, formed the backbone of the majority of the studies of this thesis. The size of the database and its relative completeness regarding demographic and location characteristics of each incident of suicidal behaviour made it possible to precisely depict the magnitude of the problem in the Netherlands and make it more tangible in our society. Moreover, the database allowed addressing the above-mentioned research questions for which databases used in previous research had been insufficient. 14

16 Introduction The main research questions investigated in this thesis and the chapters corresponding to these questions are: - What is the frequency of train suicides in the Netherlands in ? (Chapter 2) - Are trends in train suicide related to trends in national suicide figures? (Chapter 2) - Is there a relationship between train suicide and railway parameters? (Chapters 2, 6 and 7) - What is the distribution pattern of train suicides over the railway network? (Chapter 2) - What is the psychopathology involved? (Chapter 3) - Is there a relationship between psychopathology and certain suicide methods? (Chapter 4) - Are there seasonal and 24-hour patterns in train suicides? (Chapter 5) Chapter 1 In Chapter 8 the main findings of the studies will be discussed and recommendations for prevention and further research will be formulated. 15

17

18 Chapter 2 Train suicides in the Netherlands Cornelis AJ van Houwelingen a, Ad JFM Kerkhof b, Domien GM Beersma c a GGz Eindhoven, Eindhoven, the Netherlands b Vrije Universiteit Amsterdam, Department of Clinical Psychology, EMGO + Institute, Amsterdam, the Netherlands c Rijksuniversiteit Groningen, Groningen, the Netherlands Journal of Affective Disorders 2010; 127: 281-6

19 Chapter 2 ABSTRACT Background Little is known about train suicide and factors influencing its prevalence. This study tests the hypotheses that railway density, railway transportation volume, familiarity with railway transportation and population density contribute to train suicide. It also tests the relationship between train suicide and general population suicide and examines the prevalence and the characteristics of high-risk locations and their contribution to the grand total of train suicides. Methods Trends in train suicides were compared with trends in railway track length, train kilometres, passenger kilometres and national suicide figures over the period The geographical distribution over the national network over the period was studied. Data were obtained from the Netherlands Railways, ProRail and Statistics Netherlands. Results 1. The incidence of train suicides is unrelated to railway parameters. 2. Being familiar with railway transportation as a passenger is not a contributory factor. 3. Train suicide rates are unrelated to regional population density. 4. The incidence of train suicides parallels that of general population suicides. 5. Half of the train suicides took place at a limited number of locations, the most important of which were situated within a village or town and were close to a psychiatric hospital. Limitations Most conclusions are based on correlational relationships between variables. Conclusions 1. Train suicide trends reflect trends in general population suicides. 2. Increased train transportation does not lead to more train suicides. 3. The prevention of train suicide at high-risk locations (HRLs) in built-up areas and near psychiatric hospitals deserves first priority. 18

20 Train suicides in the Netherlands 1. INTRODUCTION Technological developments are often accompanied by undesirable side-effects. The development of rail transportation at the beginning of the nineteenth century led to the first officially recorded train suicide in England in At present, train suicides account for % of all suicides in the Netherlands, which is about twice as high as in Germany, Austria, England and Sweden (Van Houwelingen and Beersma, 2001b; Symonds, 1985; Rådbo et al, 2005). Train suicides lead to high costs as a result of driver and bystander trauma, and delays to the service (O Donnell et al, 1994), aspects that have been extensively reviewed elsewhere (Krysinska and De Leo, 2008; Ladwig et al, 2009). It can be hypothesized that the number of train suicides depends upon: a. availability of railways and trains. This hypothesis was supported by the work done by Clarke M (1994), who observed a positive correlation between the number of train suicides and track length. Contrary to this, Baumert et al. (2005) observed that a reduction of track length was not followed by a reduction of train suicides. Subsequently, no time-trend related to mileage covered by trains was found in suicides aged 65 or under (Baumert et al, 2005). b. familiarity with trains as a means of suicide. Clarke s observation of a strong correlation between the number of train suicides and the number of railway passengers carried supports this hypothesis. He considered the number of passengers an indication of public familiarity with this suicide method (Clarke M, 1994). c. population density. In England, clusters of train suicides were found around major cities and towns, with a median distance of 1.8 km between residential location and place of death in 95 % of the cases (Abbott et al, 2003). This combination of urban and proximity factors suggests that the most densely populated areas are at the highest risk. d. the prevalence of general population suicides. The expectation that train suicides depend on the general suicide trend in a country was not supported by the study by Baumert, who found opposite trends: an increasing number of train suicides went together with a substantial decrease of suicides by all others means (Baumert et al, 2005). e. the presence of high-risk populations near railways. Suicide risks of psychiatric inpatients have been described as being times higher than those in the general population (Brunenberg et al, 1991). Several publications have indicated that high-risk locations (HRLs) are found in the vicinity of psychiatric hopitals (Emmerson and Cantor, 1993; Erazo et al, 2004). Accessibility and acceptibility of the means of suicide may play complementary roles (Clarke and Lester, 1989). So far, these aspects of overground train suicides have not been systematically evaluated. Studies on train suicide dealing with national population and railway variables are rare (Clarke Chapter 2 19

21 Chapter 2 M, 1994; Baumert et al, 2005) and results are contradictory at times. For this reason we set out to test hypotheses a - e in the Netherlands, a country with one of the most intensively used railway networks in the world and the highest percentage of train suicides. We were able to test the familiarity hypothesis in a natural experiment, as a spectacular increase of passengers (28.9 %) and passenger kilometres (37.4 %) was observed in 1991 after the introduction of free public transport for students aged 18 or over. Although the existence of high-risk locations has been well-documented (Abbott et al, 2003; Emmerson and Cantor, 1993; Erazo et al, 2004) the proportional contribution of risk locations with varying degrees of severity is poorly understood. For this reason the frequencies at all risk locations were examined. As male/female ratios of suicide methods may change over time (Centraal Bureau voor de Statistiek, 1995), the gender specificity of trends in the period was investigated. 2. METHODS Demographic data and information regarding the location of suicidal behaviour in were obtained from the Department of Corporate Communication of the NV Nederlandse Spoorwegen (the Nether lands Railways), who keep records of all suicidal behaviour on the national railway network, with the exception of the underground, light rail and tram systems. Records are based on statutory investigations of every unnatural death by the local police and coroner. The time window for this study was expanded to 58 years by adding the annual frequencies of transportation suicides (train and underground, unspecified) of , from Centraal Bureau voor de Statistiek (Statistics Netherlands). The component of underground suicides in this period is considered very small, as underground systems only started to function on a limited scale in two cities (Amsterdam and Rotterdam) in 1968 and ProRail and the Netherlands Railways provided data on the length of the national railway system, national and international passenger train and freight train kilometres as well as on passenger kilometres by all carriers on Dutch territory. Passenger kilometres by other companies were estimated by the Netherlands Railways and included in the dataset. National suicide statistics and national population figures were obtained from Statistics Netherlands. Annual train suicide and general population suicide rates (per 100,000 inhabitants) were calculated over the period , based on the size of the population on January 1. Age was divided into 10-year bins. Location coordinates were used to investigate geographical clustering and distribution of the suicides over 4 Dutch railway regions with different population densities: North East, Randstad North, Randstad South and South. 20

22 Train suicides in the Netherlands 2.1. Statistical analysis Chi-squared analysis for independence was used to assess the impact of free transport on the frequency of train suicides by youngsters. 3. RESULTS 3.1. Characteristics of the population During this period, 5695 cases of suicidal behaviour were registered by the Netherlands Railways, comprising 5178 train suicides and 517 train suicide attempts, i.e. train contact with a non-fatal out come. This makes the lethality of this suicide method 90.9 %. The mean annual number of train suicides was 185 (SD=18). The train suicide/general population suicide proportion over the period had a mean value of 11.5 % (SD=1.4, median 11.4, min max 14.3). The mean rate for train suicide is 1.21 per 100,000 inhabitants per year (SD=0.13, median 1.17, min-max ), for general population suicides 10.6 (SD=1.3, median 10.3, min-max ). The annual rates are shown in Table 1. Chapter Age and gender Information about age was missing in 271 cases (5.2%), information about gender in 125 cases (2.4%). Mean age = 40.0 (SD=15.6, median 38, min-max 11-89). Maximum values for male train suicides were found in the age group 20-29, for female train suicides in the age group (data not shown). The proportion of train suicides of the general population suicides by age group revealed that almost a quarter of the suicides aged 10-19, boys (23.8 %) and girls (23.9 %), committed suicide by jumping before a train. The proportion of train suicides was seen to decrease steadily for both genders in subsequent age groups. The M/F ratio over the years was 3334/1719 (1.9). The M/F ratios of train suicides and general population suicides have both risen since the nineties of the previous century (Table 1). 21

23 Chapter 2 Table 1. Absolute numbers, rates (per 100,000 inhabitants) and M/F ratios of train suicides and general population suicides in the Netherlands in M/F ratio train suicides M/F ratio general suicides Year Train suicides a (n) General suicides (n) Train/general suicides in % Train suicide rate General suicide rate a Including suicides of non-residents (n=31). Rates and ratios are exclusive of non-residents. 22

24 Train suicides in the Netherlands Chapter 2 Fig.1. Annual numbers of suicides and parameters of overground railway transportation in the Netherlands during Includes: train suicides, general population suicides (divided by 10), train kilometres in millions, passenger kilometres in millions (divided by 100), length of the railway network (divided by 100) and the number of Dutch inhabitants (divided by 100,000). Train suicide data were incomplete in Association with railway parameters The number of train suicides started to rise significantly in 1970, kept increasing till 1989 and then gradually declined in the following years (Fig. 1). Over the period passenger kilometres and train kilometres (passenger and freight trains) showed an increase in railway mobility parallel to the national population growth, with the railway network remaining almost the same size (Fig. 1). Neither increase nor decrease of the numbers of train suicides would seem to bear any apparent relationship to the railway parameters presented. The relationship between train suicide frequency and railway density was also examined by means of a cross-sectional analysis of train suicides in 4 railway regions. It appeared that regional train suicide rates did not correlate with regional railway densities (Table 2) The association with familiarity In order to study the impact of free transport for students, the frequencies of train suicides in the age-group and all other ages in two 5-year periods before and after the introduction of free transport for students in 1991 ( and ) were compared. Contrary to expectations it was found that in the age-group the number of train suicides had decreased slightly in the second period. These differences were significant for men and women taken together (Chi 2 =4.498; df=1; P= 0.034) and for men separately (Chi 2 =4.749; df=1; P=0.029), but not for women (Chi 2 =0.499; df=1; P= 0.48). 23

25 Chapter The association with population density In similar rates of train suicides (1.3 per 100,000 inhabitants) were found in three railway regions with different population densities (Table 2). Remarkably enough, a much lower rate (0.7) was found in the most densely populated region, Randstad South The association with general population suicides Figure 1 shows trends in train suicides from 1950 to General population suicides showed a similar pattern, reaching a maximum in Train suicides, however, showed a much larger growth-rate in In this period, general population suicides increased by 63 %, whereas train suicides increased by 311 %. In order to understand the contribution of gender in the period of decline, rates differentiated by gender were calculated for the period (Fig. 2). The decrease in general population suicides rates after 1984 is more pronounced in women, resulting in increased M/F ratios. Female train suicide rates decreased at a pace similar to that of female suicides in the general population. Male train suicide rates showed a large annual variation and no apparent downward trend. Fig.2. Rates of train suicides (multiplied by 10) and general population suicides during the period by gender (per 100,000 inhabitants). 24

26 Train suicides in the Netherlands Table 2. Suicide statistics and other measures for 4 Dutch railway regions. Train suicides per km Rail density m/km2 Track-length in km Train rate/gen rate in % General suicide rate c (n/100,000) Population Train suicide density a rate b (n/100,000) Train suicides Region NorthEast Rural RndstdN Urban RndstdS Urban South Semi-rural a On 1/1/2007. b Average number of train suicides per year per 100,000 inhabitants of the railway region on 1/1/1997. c Suicides per 100,000 inhabitants in 2005 of the provinces (roughly) matching the railway regions. Chapter 2 25

27 Chapter High-risk locations (HRLs) Geographical distribution In 4683 out of 5178 train suicide cases, the exact location of the incident, expressed in kilometre and hectometre coordinates was either known or could be inferred (90.4%). A 1-kilometre resolution, corresponding to the railway grid, was chosen to analyse the geographical distribution. Below, every kilometre is referred to as a location. Table 3. High-risk locations (HRLs) of train suicides in the Netherlands over the period , stratified by categories of suicide frequency per kilometre. HRL Number of suicides per km Number of locations a Percentage of mean track length b Number of suicides on all locations Percentage of total number of train suicides, n=4683 c Category I >= Category II Category III Category IV Category V Category VI Total a Each location measures 1 km. b Mean track length over period is 2797 km. c 495 train suicides of which the exact location could not be determined were not included. All locations were divided into categories according to the number of train suicides taking place in These categories range from very high: 28 or more suicides in 28 years (category I (> =1/km/yr)) to very low (1-3 cases in 28 years, category V) and ultimately to category VI with no incidents at all of fatal suicidal behaviour (Table 3). It became evident that train suicides were clustered on relatively small sections of the national railway network. The majority of train suicides (54.2%) took place on 341 km (12.2%) of the network and, even more extremely, 16.6 % took place on 1.2 % of the tracks (categories I and II, Table 3). Furthermore, it was noted that a considerable number of suicides (45.9%) were spread over locations of very low frequency (category V, Table 3) Location characteristics A further investigation revealed that all locations of category I and 15 locations of category II were situated near 13 different psychiatric hospitals, which were within 800 meters walking distance. All locations of category I and 20 of category II were situated within built-up areas. 26

28 Train suicides in the Netherlands Type of location The totalized data over the years showed the following distribution: platform 18.7 %, level crossing 25.7 % and open track 55.6 %. 4. DISCUSSION 4.1. Major findings Train suicide frequency does not correlate with railway density or with the intensity of railway transportation. Familiarity with trains through travelling would not seem to be a contributing factor either. High population density in urban settings would not seem to contribute to the risk of train suicide. However, the prevalence of train suicides does correlate with that of general population suicides. About half of all train suicides are clustered on small sections of the railway network, the others take place at scattered locations. Chapter Relationship with railway parameters During the study period track length remained fairly stable, some tracks were lifted, some new ones were constructed. The observed rise and fall in train suicides did not depend on railway density. Nor did regional railway densities correlate with the corresponding train suicide rates. Therefore, the relationship between railway density and train suicide, as predicted by Durkheim (1897) and observed by Clarke M (1994) in times of railway expansion, does not apply to the Netherlands. No relationship was found between the number of train suicides and the number of train kilometres or passenger kilometres. Given the fact that train suicide is a feasible option for most Dutch inhabitants as 75 % of them live within a 5-kilometre distance from a railway station (the Netherlands Railways, 2007, personal communication), this finding stands out. From this it can be inferred that it is unlikely that intensification of railway transportation in itself will lead to an increased number of train suicides Familiarity with railway transportation In the present study, the hypothesis was tested that the more contact the public has with trains, the more train suicides will take place (Clarke M, 1994). We did not find this type of relationship after a massive increase of railway transport among students. This negative finding should be interpreted with some caution, as students may not be representative of all railway passengers. 27

29 Chapter Population density Contrary to what was expected, population size (density) and train suicide rates were not related. Highly densely populated areas do not necessarily generate more train suicides. Since general population suicide rates do not depend strongly on urbanisation, the reduction in train suicides in the most highly urbanized region is not easily explained. Maybe this is because noise barriers and a high building density along tracks make overground tracks less easily accessible. While the lowest rate was found in the most urbanized region, the number of suicides per kilometre remained high, causing high levels of disturbance in areas with the highest traffic density Relationship with general population suicide figures In the seventies, train suicides showed a disproportional increase compared to general population suicides. This was preceded by a decrease in domestic gas suicides from 24.5 % to 0.5 % due to the detoxification of gas (Clarke and Lester, 1989), which may have resulted in a shift towards train suicide. Furthermore, in the period , substantial decreases in self-poisoning and drowning and increases in hanging and jumping from a high place were observed (Centraal Bureau voor de Statistiek, 1995). It is possible that the strong increase of train suicides in the seventies is a reflection of a substitution of non-violent towards more violent means High-risk locations (HRLs) Although regional population density does not contribute to train suicide, a relationship exists with communities on a local level. In % of the suicides took place on just 33 (1.2%) kilometres of the railway track. Most of these locations were within built-up areas of villages or towns and were close to psychiatric hospitals. Housing psychiatric patients near railways and the vicinity of infrastructural elements like railway stations and multiple level crossings, constitute a synergy of risk factors. The data of this study did not allow for differentiating between psychiatric patients and non-patients among the suicides. However, combined studies showed that 53 % of train suicides received mental healthcare at the moment of suicide, with half of them being inpatients (Van Houwelingen and Kerkhof, 2008). These high-risk locations illustrate that collaboration between hospital staff, railway company and local authorities is needed in order to create appropriate solutions. This study also demonstrated that during a 28-year period more than one-third of the national railway network has never been exposed to fatal suicidal behaviour. Therefore, if priorities have to be indicated, investments that make railway tracks less easily accessible for suicidal persons could be limited to selected areas with higher incidences. 28

30 Train suicides in the Netherlands 4.7. Other findings A remarkable finding is that male and female train suicide rates have diverged since around It would seem that the male population at risk for train suicide, for reasons we do not understand, does not benefit from factors causing the overall reduction of general population suicides and female train suicides. A rather alarming finding is the high percentage of youngsters that make use of this violent method. This finding is in agreement with reports from Germany and Austria (Schmidtke, 1994; Deisenhammer et al, 1997). High levels of impulsivity at this age and not having access to alternative means, like medication, may play a role in the choice for this method. Chapter Strengths and limitations The strength of this study is that it is based on a very large, almost complete dataset, covering an entire national railway network. Data on age, gender and location characteristics were obtained over a period of 28 years, data regarding incidence, rail parameters and general population suicides cover 58 years. Therefore, the conclusions that railway parameters have little impact on train suicide frequency, that population density is not a major factor either, and that high-risk locations (HRLs) are situated within built-up areas, and near psychiatric hospitals, are quite strong. A limitation of this study is that most conclusions are based on correlational relationships between variables Future research In our study, easy access to the railway tracks could not be investigated as an independent variable. Our own observations throughout the country made it clear that the Dutch tracks were easily accessible during the period under study. On Jan a new Railway Law, replacing a law dating from 1875, laid down that railway tracks must be inaccessible to the public (Staatsblad, 2003). Consequently, the construction of fences along the tracks has recently begun. The impact of this general intervention on the incidence of train suicides should be monitored carefully, including a possible shift towards railway-station platforms and level crossings Conclusion The overall picture emerging from this study is that train suicide does not depend on the availability of or familiarity with trains. Undoubtedly, easy access to the tracks makes the railway system vulnerable to non-railway contributory factors such as the general suicide trend in a society. Consequently, the problem of train suicide could benefit from successful national suicide prevention programs and long-term investments in rail systems that reduce the possibility of train-person collisions. 29

31

32 Chapter 3 Mental healthcare status and psychiatric diagnoses of train suicides Cornelis AJ van Houwelingen a, Ad JFM Kerkhof b a GGz Eindhoven, Eindhoven, the Netherlands b Vrije Universiteit Amsterdam, Department of Clinical Psychology, EMGO + Institute, Amsterdam, the Netherlands Journal of Affective Disorders 2008; 107: 281-4

33 Chapter 3 ABSTRACT Background The objective of this study is to investigate mental healthcare status and psychiatric disorders in train suicides. Methods Data of 4 published train suicide studies were combined with a study of 57 train suicides in the Netherlands. Results 53 % of all train suicides received psychiatric care at the time of suicide, with 49 % of them being inpatients. These values are higher than those found in general suicides. When compared to general suicides, functional non-affective psychoses are overrepresented by 25 % vs 14 %. The percentage of affective disorders approximates that of general suicides namely: 39 % vs 42 %. Other diagnoses are underrepresented by 23 % vs 40 %. Limitations All psychiatric diagnoses were based on clinical data, which may not accurately reflect the patient s psychiatric condition at the time of suicide. Conclusions Train suicides receive mental healthcare more often than general suicides and are more often characterized by severe psychopathology. The study indicates that patients with affective and psychotic disorders in particular should be targeted in order to prevent train suicide. 32

34 Mental healthcare status and psychiatric diagnoses of train suicides 1. Introduction Relatively little is known about train suicides. In the Netherlands, train suicides account for 12 % of all suicides. Each train suicide disrupts train traffic to an extent comparable to that of a bomb-alert (Griffioen and Van den Tweel, 1996). Knowing more about the psychopathology of this population might contribute to a better understanding of this type of suicidal behaviour. A literature search was performed to analyse information on mental healthcare and psychiatric diagnoses of train suicides. In addition, the connection between train suicide and mental disorder was investigated in Drenthe, a Dutch province. 2. Materials and methods Chapter 3 Over the period 1966 to May 2006 the databases Embase, PsycINFO and Medline were screened for literature on overground train suicides by means of the keywords rail, railway, railroad and suicide. Publications were selected that contained quantitative information on mental healthcare status and mental disorders. For the second part of the study demographic data on train suicides in the Netherlands over the period were obtained from the archives of the Department of Corporate Communications of the Netherlands Railways. The archives contained 2503 cases, 57 (2.3%) of which could be identified as residents of the province of Drenthe. These 57 cases were characterized on the basis of place of residence, date of birth, gender, date of death and the first letter of the suicide s last name (if available). Subsequently, these characteristics were used to match these cases with identical cases registered by the Groningen Psychiatric Case- Register (PCR). The Groningen PCR has registered all contacts of residents of the province of Drenthe with mental health services since by means of a probability record linkage method (Brook, 1996). The treatment history and principal diagnoses (ICD-9) were investigated. Any cases not recognised by the PCR were concluded not to have received mental healthcare after Drenthe covers 6.5 % of the Netherlands. In the period of study Drenthe had a population of about 450,000 inhabitants, or 2.96 % of the national population. The province can be characterized as a rural area. In terms of mental healthcare consumption, Drenthe is representative of the Netherlands (Ten Have et al, 1996). The railway tracks in Drenthe measure 100 km or 3.6 % of the total Dutch track length. The province has three psychiatric hospitals, two of which are situated immediately next to the railway track. 33

35 Chapter 3 3. RESULTS Only four studies have dealt quantitatively with clinical aspects of overground train suicides. Symonds reported on the diagnostic characteristics of 82 cases in 1979 and 1980 (Symonds, 1985). A Danish study focused on 16 train suicides in a delimited area (population approx. 44,000), in the period of (Lindekilde and Wang, 1985). In Brisbane, Australia, a survey was done of 23 train suicides between 1980 and 1986 (Emmerson and Cantor, 1993). Most recently, the British SOVRN Project has investigated psychiatric diagnoses of 84 train suicides at their time of death in the period of (Abbott et al, 2003) Mental healthcare status Four studies Table 1 shows the numbers of suicides with a known psychiatric history as well as those receiving mental healthcare at the time of suicide. The average percentages are 64 % and 50 % respectively. 44 % of the suicides that received mental healthcare at the time of suicide were inpatients Dutch study 40 cases (70%) out of the Dutch sample of 57 cases were identified by the Groningen Psychiatric Case-Register as having had psychiatric treatment in the period starting on 1 January Demographic characteristics are presented in Table 2. m/f ratios of the identified and unidentified cases are not significantly different (Fisher s Exact Test (2 sided; P = 0.576), but the unidentified cases are younger (Mann-Whitney test statistic = 177; P = 0.004). Table 2 also presents age and gender characteristics of the entire sample as well as those of the national population of train suicides. The m/f ratios of the populations differ (1.1 and 1.8 respectively). However, no significant relationship between m/f ratio and the two populations could be found (Fisher s Exact Test (2 sided; P = 0.069)). Nor did we find any significant relationship between populations and age (Mann-Whitney test statistic = ; P = 0.31). The same applied when age was classified in 4 age groups (Chi 2 = 0.696; df = 3; P = 0.87). 36 out of the 57 cases (63%) received mental healthcare at the time of suicide. 23 (64%) of them received inpatient care. 11 (31%) received outpatient care, one received daycaretreatment and one person lived in community-staffed housing. 34

36 Mental healthcare status and psychiatric diagnoses of train suicides Table 1. Care status and main diagnostic categories of train suicides (4 + 1 studies) n Care status at time of death Diagnoses Other psychiatric diagnoses Affective disorders Nonaffective psychoses No psychiatric diagnosis Valid diagnostic data In-patient status of those receiving care Received mental healthcare b Known psychiatric history (81%) 8 (50%) 7 (88%) 16 3 (19%) 5 (31%) 4 (25%) 4 (25%) Lindekilde and Wang, Denmark (60%) 40 (49%) 12 (30%) 75 9 (12%) 12 (16%) 34 (45%) 20 (27%) Symonds, UK (83%) 17 (74%) 13 (76%) 19 0 (0%) 13 (68%) 4 (21%) 2 (11%) Emmerson and Cantor, Australia Abbott et al, (60%) 37 (44%) 13 (35%) (24%) 12 (18%) 26 (39%) 12 (18%) UK a Subtotal (64%) 102 (50%) 45 (44%) (16%) 42 (24%) 68 (39%) 38 (22%) (70%) 36 (63%) 23 (64%) 36 0 (0%) 12 (33%) 14 (39%) 10 (28%) Present study, the Netherlands Total (65%) 138 (53%) 68 (49%) (13%) 54 (25%) 82 (39%) 48 (23%) a SOVRN Project. b Mental health care provided by a general practitioner not included; percentage of study population. Chapter 3 35

37 Chapter 3 Table 2. Age and gender of train suicides of the Dutch study n Age, mean Median Age m/f (SD) min-max ratio 1. Train suicides identified in Case Register (15.6) Train suicides not identified in Case Register (10.7) Train suicides of residents of Drenthe (1 + 2) (15.3) All train suicides in the Netherlands (15.5) between 1986 and Psychiatric diagnoses Four studies The diagnostic characteristics were grouped in 3 main diagnostic categories of principal diagnoses and showed the following distribution: I. functional non-affective psychoses 24 %, II. affective disorders 39 % and III. other diagnoses (including personality disorders) 22 %. In 16 % no psychiatric diagnosis was evident (Table 1) Dutch study In the Groningen PCR no diagnostic information was available on 4 cases. Of the remaining 36 cases, the last known psychiatric diagnosis was used. This dated back to the time of most recent referral for outpatient care or to the time of the most recent admission to hospital. In 7 cases the diagnosis at administrative discharge could be used. The time interval between diagnosis and suicide varied considerably, values being between 0 and 3064 days, with a median of 57. The last known diagnoses showed the following distribution: schizophrenic psychoses 10, other nonorganic psychoses 2, affective psychoses 9 (manic type 3, depressed type 4, unspecified 2), depressive disorder NOS 1, neurotic depression 4, neurotic disorders 5 (incl. anxiety states 3), personality disorder 1, alcohol dependence 1, drug dependence 1, acute reaction to stress 1 and mixed disturbance of conduct and emotions 1. The diagnoses were grouped in 3 main diagnostic categories for subgroup analyses: I. functional non-affective psychoses, II. affective disorders and III. other diagnoses. Results are presented in Table 1. The diagnostic distribution in main categories turned out to be dependent on age group, <= 39 years or >=40 years (Chi 2 = 15.8; df = 2; P = 0.000). 11/12 cases of category I (functional nonaffective psychoses) were younger than 40 and 12/14 cases of category II (affective disorders) were 40 or over. 36

38 Mental healthcare status and psychiatric diagnoses of train suicides 4. DISCUSSION Together, the Dutch study and the 4 reported studies demonstrate that a considerable proportion (65%) of train suicides has a psychiatric treatment history. This is slightly higher than the proportion of general suicides with a lifetime history of contact with mental health services (58%) reported by Foster (Foster et al, 1997). At least half of the train suicides (Table 1) received care from mental health services at the time of suicide. This proportion is substantially higher than in populations of general suicides (14-35%) (Booth and Owens, 2000; Appleby et al, 2001). The difference with general suicides is even more robust when the percentages of inpatient suicides are taken into account (Table 1). On average, nearly half of the train suicides receiving mental healthcare at the time of the incident are inpatients. This is 3 times higher than the 16 % reported on general suicides in the UK, who had been in contact with services the year before death (Appleby et al, 2001). It follows that, as a population, train suicides are more severely ill than general suicides. In the Dutch study, most cases with a functional non-affective psychosis belong to the age group of under 40. This is in accordance with the observation that patients with schizophrenia usually commit suicide before age 45 (Drake et al, 1985). When comparing the diagnostic categories of the 5 train suicide studies (Table 1) to that found in populations of general suicides, clear differences are found in the categories of functional non-affective psychoses and other psychiatric diagnoses. The combined data of two European psychological autopsy studies on known principal diagnoses (Foster et al, 1997; Henriksson et al, 1993) shows a distribution of 14 % of functional non-affective psychoses, 42 % of affective disorders and 40 % of other psychiatric diagnoses. A meta-analysis of 27 psychological autopsy studies from different regions of the world on average showed a distribution of 9.2 % of psychotic disorders, 43.2 % of affective disorders (including bipolar disorders), 25.7 % of substance abuse-related problems and 16.2 % of personality disorders (Arsenault-Lapierre et al, 2004). These findings all indicate that there is an overrepresentation of functional non-affective psychoses and an underrepresentation of other diagnoses in train suicides. The percentages of affective disorders in this study (39%) and that of the 5 train suicide studies (39%; Table 1) come close to that of general suicides, however (Foster et al, 1997; Henriksson et al, 1993). A limitation of the study is that, in the majority of cases, the diagnostic information describes the clinical picture at the time of the most recent referral for outpatient care or most recent admission to hospital. Consequently, the quality of the diagnostic descriptives in this study relies on the stability of the diagnoses over time. Nor do we know whether psychosocial factors or co-morbidity may have played a role at the time of suicide. Why is it that subjects with severe psychopathology have a preference for this method? Chapter 3 37

39 Chapter 3 There is some evidence that type of suicide method and mental disorders involved are interdependent. Patients with schizophrenia spectrum psychoses use methods that result in physical injuries more often than patients with affective psychoses (Radomsky et al, 1999). As this might explain for the relative overrepresentation of functional non-affective psychoses in train suicides, the majority however, have different diagnoses. In the SOVRN study the socalled proximity factor was investigated for various violent methods. The authors concluded that proximity is not a rail factor per se, but that people rather tend to die close to where they live with knowledge of local surroundings being important (Abbott et al, 2003). It might be that in severely ill patients, being familiar with and having easy access to a nearby railway track may compensate for the presence of disabling functional and cognitive deficits and therefore lead to using these means. Furthermore, it should not be overlooked that the supervised intake of psychopharmalogical drugs in a clinical setting and the use of depot medication may limit the access to less violent means, with the unfavourable outcome of patients resorting to more violent methods that are more easily accessible. This study reveals that at least 50 % of train suicides were in contact with mental health services at the time of suicide. When dealing with patients it is of vital importance to systematically and explicitly seek the dialogue about their suicide wishes and to discuss the negative consequences of a train suicide for themselves, the people close to them and numberless commuters. 38