Introduction
The availability of crisis lines is one of the core components of many suicide prevention strategies worldwide (Platt et al., Reference Platt, Arensman and Rezaeian2019). The main goals of crisis lines are to promote efficient, prompt access to counselling for those at risk of suicide or those experiencing distress (Joiner et al., Reference Joiner, Kalafat, Draper, Stokes, Knudson, Berman and McKeon2007; Gould et al., Reference Gould, Munfakh, Kleinman and Lake2012). In a recent systematic review, crisis lines were demonstrated to be effective at improving several proximal outcomes (e.g., changes in level of distress immediately after the call) and distal effects (e.g., changes in emotional states from one week after the service use, to up to four years) (Hoffberg et al., Reference Hoffberg, Stearns-yoder and Brenner2020).
Suicides that occur at high risk locations – commonly referred to as ‘suicide hotspots’ – have been a focus of particular concern (Reisch and Michel, Reference Reisch and Michel2005; Sinyor et al., Reference Sinyor, Schaffer, Redelmeier, Kiss, Nishikawa, Cheung, Levitt and Pirkis2017; Kõlves et al., Reference Kõlves, Leske and De Leo2023) and immediate access to crisis lines for people at imminent risk at these sites is one intervention that has been proposed (Glatt et al., Reference Glatt, Sherwood and Amisson1986; Zarkowski, Reference Zarkowski2008; Lockley et al., Reference Lockley, Cheung, Cox, Robinson, Williamson, Harris, Machlin, Moffat and Pirkis2014; Stack, Reference Stack2015). This typically takes the form of fixed phone boxes with landlines that enable distressed individuals to make a priority call to a crisis line where they receive support from a crisis counsellor. Access is immediate and direct, and the phones cannot call any other number apart from the crisis line (except sometimes an emergency services number).
It is an open question as to whether these fixed landline crisis phone boxes are effective in reducing suicides at high-risk locations. Only two studies have been published on this topic, both in the United States. These studies compared the change in suicide numbers at bridges in a pre-intervention period with those in a post-intervention period and suggested that fixed landline crisis phone boxes on the bridge may be associated with an increase in suicides. In particular, at the Aurora Bridge in Seattle, there were six jumping suicides in the first four months after installation of a fixed landline crisis phone box, which was two suicides more than the yearly average (Zarkowski, Reference Zarkowski2008). At the Sunshine Skyway Bridge (Skyway Bridge hereafter) in St. Petersburg, there was an increase of 4.46 additional suicides per year after installation of crisis phone boxes (Stack, Reference Stack2015).
Further work is needed to determine whether the results of these studies are anomalous and whether fixed landline crisis phone boxes may in fact be effective in preventing suicides. The aim of the present study was two-fold: 1) to compare suicide rates before- and after-installation of fixed landline crisis phone boxes in six public places; and 2) to measure the magnitude of the overall effect of the fixed landline crisis phone boxes by synthesising results from each site to an overall estimate.
Methods
Identifying sites for inclusion
Public places were identified where fixed landline crisis phone boxes had been installed and where pre- and post-intervention data were available to calculate suicide rates. All sites included in our study have been anonymised.
For the sites in Australia, information was obtained on the installation of the boxes from Lifeline Australia, which provides nationwide telephone crisis support. In Australia, there are now three sites in different states. These were all included in the study (sites A, B and C).
To identify other sites in other countries, a search was conducted. Modifying a search query developed by Pirkis et al. (Reference Pirkis, Too, Spittal, Krysinska, Robinson and Cheung2015) and Hoffberg et al. (Reference Hoffberg, Stearns-yoder and Brenner2020), a Google search was undertaken. Combined search terms referred to suicide, crisis lines and sites where jumping suicides could occur: Suicide AND (hotline OR ‘crisis line’ OR ‘crisis phone’ OR ‘crisis phones’ OR helpline OR lifeline) AND (bridge OR overpass OR cliff OR building OR high-rise OR multi-storey OR viaduct). The All, News and Images sections of Google were checked to identify sites, and sites identified through Google Street View were cross-checked to see if the phone boxes on the sites were actually installed. An investigation was then conducted to determine whether it would be possible to access relevant suicide data from these sites (either by request or through an open data platform). After this process, three US sites were included: sites D, E and F.
All six sites included in the study were sites where the method of suicide was jumping, and with the exception of site A (a cliff site), all were bridges.
Intervention description by site
The structural descriptions of the six sites are presented in Table 1. At site A, a clifftop landscape facing the ocean, a fixed landline crisis phone box was installed at each of two main entrances in March 2010, as part of a program to minimise self-harm at the location. The logo of Lifeline, the largest crisis line service in Australia, and 000, which is the national emergency telephone number, are displayed on the phone boxes.
Profiles of the sites
Table 1 Long description
Columns list Site, Fixed phone booth or boxes, and Other measures for sites A through F. Site A has 2 booths, each one at two main entrances, with 1.3 m high fences marked with footnote 1 and signages. Site B has 6 boxes, each one at four end posts of trusses, plus 2 at both sides of a bridge entrance, with 1.4 m high rails marked with footnote 2 and signages. Site C has 6 boxes, 2 at both sides at apex and 4 at both sides of bridge entrances, with 1.59 m high rails marked with footnote 2 and signages. Site D has two boxes, one box at each entrance, with 0.81 m high rails marked with footnote 2 and no signages listed. Site E has two boxes, one box at each entrance, with waist high rails marked with footnote 2 and signages. Site F has two kiosk type boxes, one box at each entrance, with waist high rails marked with footnote 2 and signages.
a Fences at two locations at site A were upgraded in tandem with installation of the booths, while the height of the fence remained nearly unchanged. At the time, the remaining most part of fences at site A remained unchanged.
b Rails or fences on the sites existed at the time of installation of phone box
At site B, a bridge built in 1940, six yellow phone boxes were installed on 22 May 2012. Four were located at each of the end posts of trusses on the bridge, and the remaining two were installed on both sides of the northern bridge entrance. The bridge is located in the middle of the capital city of a state and has footpaths on both sides. It originally had 1.4 m-high steel rails on both sides, and subsequently a 3-metre-high curved fence was erected on 22 December 2015.
At site C, six phone boxes were installed on 11 January 2016. Two of them were installed on each side of the apex on the bridge, and the other four were installed on each side of both entrances. The bridge has footpaths on both sides.
At site D in the United States (a bridge that has won awards for its design), two crisis line boxes were installed at each entrance of the bridge on 29 August 2019. The bridge does not have a footpath as it carries a two-lane highway.
Site E in the United States is a high twin-arch bridge located over a creek, where a yellow phone box was installed at each entrance of the double spans some time in 2015. The bridge has a shoulder lane at each span but does not have pedestrian walkway.
Lastly, site F (located over a gorge) had two kiosk-type phone boxes installed at each entrance in January 2017. The bridge has pedestrian access on each side.
Defining the pre- and post-intervention periods
It was not possible to identify the exact date the fixed landline crisis phone boxes were installed at site A. However, because some information had been published on an Australian Government website indicating the phone boxes started operating in March 2010, the date 1 March 2010 was used as the installation date. 1 January 2001 to 28 February 2010 was set as the pre-intervention period, and 1 March 2010 to 31 December 2018 as the post-intervention period.
For site B, the observation window was limited to the period between 1 January 2001 and 20 December 2015, as construction of 3 metre-high safety barriers was completed on 21 December 2015. Fixed landline crisis phone boxes were installed on 22 May 2012, so 1 January 2001 to 21 May 2012 was defined as the pre-intervention period and 22 May 2012 to 20 December 2015 as the post-intervention period.
The exact date of installation of fixed landline crisis phone boxes at site C was identified through a search of newspaper articles; this was 11 February 2016. Using this date, the pre-intervention period was defined as 1 January 2001 to 11 February 2016 and the post-intervention period as 12 February 2016 to 31 December 2018.
For site D, it was established that the phone boxes were installed on 26 August 2019. For this site, 1 January 2006 to 25 August 2019 was defined as the pre-intervention period and 26 August 2019 to 28 February 2021 was defined as the post-intervention period.
Despite Internet searches and contacts with the organisation involved in the installation work at site E, it was only possible to identify the year in which the fixed landline crisis phone boxes were installed – 2015. This year was therefore excluded from the analysis. For this site, 1 January 2009 to 31 December 2014 was defined as the pre-intervention period and 1 January 2016 to 31 December 2021 was defined as the post-intervention period.
Kiosk-type fixed landline phone boxes have been operational since January 2017 at site F. Therefore, for this site, 1 January 2008 to 31 December 2016 was set as the pre-intervention period and 1 January 2017 to 26 October 2018 as the post-intervention period.
Suicide data
For each site, information was gathered on suicides occurring pre- and post-installation of the fixed landline crisis phone.
For the three Australian sites, information was retrieved from the National Coronial Information System (NCIS) on suicides that occurred between 1 January 2001 and 31 December 2018. NCIS is the database source of data managed by the Department of Justice and Community Safety, Victoria, Australia. 2018 was selected as the endpoint of the data extraction as there were cases in 2019 that were not completely closed at the time of extraction in January 2021. Closed intentional self-harm cases with the mechanism code ‘Blunt force’ or ‘Threat to breathing’ that occurred at postcodes where each site was located were initially extracted. Suicides where the precise incident location could not be identified, and cases where the date of death and sex were unknown, were excluded. Information about the fixed landline crisis phone boxes at site D was gathered from a newspaper article that included such information, as provided by a County Sheriff’s Office, the organisation responsible for the safety of the area surrounding the site. For sites E and F, information was obtained from a County Sheriff’s Office and a state Department of Health, respectively.
Statistical analyses
To estimate the effectiveness of the fixed landline crisis phone boxes at each site, defined here as their effectiveness in reducing suicide mortality, we employed a two-stage analysis. First, a Poisson regression model was used to calculate an incidence rate ratio (IRR) and 95% confidence interval (95% CI) for each site. Rates of suicide in the pre- and post-intervention periods were calculated by dividing the number of suicides by the number of days in the pre- or post-intervention period. Second, to calculate a pooled IRR, the DerSimonian–Laird random-effects estimator was used to combine the IRRs from the six sites. Effectiveness was inferred when an IRR for suicide mortality below 1.0 and a statistically significant p-value (p < 0.05) were observed. All analyses were conducted using package rmeta in R (version 4.0.5).
Results
A total of 291 suicides were recorded over 88.9 study-years. In total, there were 169 suicides in 64.3 study-years in the pre-intervention period (2.63 suicides per study-year) and 122 suicides in 24.6 study-years in the post-intervention period (4.95 suicides per study-year). The numbers of suicides and rate ratios per site are shown in Table 2.
Statistical results on the effectiveness of fixed phone boxes installation on suicides
Table 2 Long description
Rows list Sites A through F with total days and suicides with percent female, plus pre intervention and post intervention days and suicides, and an incidence rate ratio with a 95 percent confidence interval. Site A totals 6,574 days and 134 suicides with 41.0 percent female; pre intervention is 3,346 days and 59 suicides, post intervention is 3,228 days and 75 suicides, with an incidence rate ratio of 1.32 and a 95 percent confidence interval from 0.94 to 1.85. Site B totals 5,467 days and 69 suicides with 18.9 percent female; pre intervention is 4,159 days and 39 suicides, post intervention is 1,308 days and 30 suicides, with an incidence rate ratio of 2.45 and a 95 percent confidence interval from 1.52 to 3.94. Site C totals 6,574 days and 35 suicides with 37.4 percent female; pre intervention is 5,520 days with suicides reported as more than 5, post intervention is 1,054 days with suicides reported as fewer than 5, with an incidence rate ratio of 0.32 and a 95 percent confidence interval from 0.08 to 1.33. Site D totals 5,538 days and 30 suicides with 36.6 percent female; pre intervention is 4,985 days and 22 suicides, post intervention is 553 days and 8 suicides, with an incidence rate ratio of 3.28 and a 95 percent confidence interval from 1.46 to 7.35. Site E totals 4,383 days and 11 suicides with percent female not reported; pre intervention is 2,191 days with suicides reported as more than 5, post intervention is 2,192 days with suicides reported as fewer than 5, with an incidence rate ratio of 0.57 and a 95 percent confidence interval from 0.17 to 1.95. Site F totals 3,952 days and 12 suicides with percent female not reported; pre intervention is 3,288 days with suicides reported as more than 5, post intervention is 664 days with suicides reported as fewer than 5, with an incidence rate ratio of 1.65 and a 95 percent confidence interval from 0.45 to 6.10.
Note: If fewer than five suicide deaths occurred in either the pre- or post-intervention period, the frequencies for both periods were suppressed.
At site B in Australia and site D in the US, the number of suicides during the post-intervention period was higher than the number during the pre-intervention period (IRR for site B: 2.45, 95% CI: 1.52–3.94; IRR for site D: 3.28, 95% CI: 1.46–7.35). At the remaining four sites, there was no evidence that the numbers in the post-intervention period were different from the pre-intervention period (IRR for site A: 1.32, 95% CI: 0.94–1.85; IRR for site C: 0.32, 95% CI: 0.08–1.33; IRR for site E: 0.57, 95% CI: 0.17–1.95; IRR for site F: 1.65, 95% CI: 0.45–6.10).
When these results were combined, the pooled IRR was 1.46 (95% CI 0.86–2.49), indicating no overall change in suicide rates between the two periods. The I 2 statistic was 66.5%, indicating large heterogeneity.
Discussion
The main finding from this study is that the impact of the fixed landline crisis phone boxes varied substantially by site. At sites B and D, suicide numbers were higher in the post-intervention period than they were before the phone boxes were installed. At the other sites, there was no evidence of a change in suicide numbers between the pre- and post-intervention periods. When the data were pooled, effectiveness measured by IRRs showed a wide range of uncertainty, and the pIRR did not provide firm evidence of any impact. These findings suggest that, contrary to expectations, fixed landline crisis phone boxes, on their own, may not be sufficient to reduce suicides at high-risk locations. If anything, caution should be exercised in installing them as an isolated preventive measure. Instead, these phone boxes may need to be combined with other interventions to maximise their usefulness as a suicide prevention initiative.
There has been little investigation into what factors may influence the effectiveness of fixed landline crisis phone boxes. In the study of the installation of phone boxes on the Skyway Bridge, there was an increase in suicides after the phone boxes were installed. The authors examined potential reasons for this and noted that the installation of the phone boxes had coincided with a controversial website and interactive forum about suicides on the bridge. They noted that the website/forum had significant reach and may have had a contagion effect that counteracted any potential positive impact of the phone boxes (Stack, Reference Stack2015).
Even in the absence of factors that may work against the effectiveness of fixed landline crisis phone boxes in reducing suicide mortality, these phone boxes do not appear to be as effective – at least on their own – as other proven interventions at high-risk locations. Notably, restricting access to means at these sites has been shown to be extremely effective (Pirkis et al., Reference Pirkis, Too, Spittal, Krysinska, Robinson and Cheung2015). Phone boxes may need to be installed in conjunction with interventions like barriers that interrupt a suicide attempt and ‘buy time’ for the individual to reconsider their actions or for others to intervene. Many sites known for suicides, including sites B, C, D and F included in the current study, initially had phone boxes installed and subsequently ended up with barriers being constructed. Other considerations may also be important to maximise the effectiveness of fixed landline crisis phone boxes. One obvious consideration is where they are located. They are often installed at the entrance to the high-risk location. In the case of a bridge, this means that the driver in crisis, who gets out of their car in the middle of the bridge span, might not be aware of the presence of the phone boxes. Careful location of the phone boxes might go further to prevent suicides. Although the result was not statistically significant, the site C showed the lowest IRR among the six sites. Two of the six phone boxes installed at this site were located at the apex of the curved arch bridge, which is the most lethal jumping spot. Phone boxes at the apex might increase the likelihood of people calling the crisis line at the critical moment and could enhance the effectiveness of the intervention.
This is the largest study to date to statistically evaluate the effectiveness of the fixed phone boxes at the public sites following a study on the Skyway Bridge in the United States (Stack, Reference Stack2015). This study added to the evidence of the direct effectiveness of fixed phone boxes on suicides, allowing stronger conclusions to be drawn.
Some potential confounding variables were not adjusted for in this study. For example, one of the factors that may have influenced suicides after the installation of the fixed phone boxes might be media coverage (as suggested by the Skyway Bridge study (Stack, Reference Stack2015)). Further studies should evaluate the effect of any potential media coverage about the sites where fixed phone boxes have been installed.
In a similar vein, it was not possible to determine whether other interventions at the identified sites may have exerted an influence. For example, at all of the sites except for site D, the phone boxes were installed alongside signs that indicated sources of help. Others have considered the impact of signage alone at high-risk locations, with mixed results (King and Frost, Reference King and Frost2005).
Also, ratios of the length of the post-intervention period to that of the pre-intervention period are heterogeneous by site and were less than 1. As other similar intervention studies with quasi-experimental design have indicated (Pirkis et al., Reference Pirkis, Too, Spittal, Krysinska, Robinson and Cheung2015), the long-term effect may differ from the current shorter-term findings. In addition to the cases mentioned above, at Aurora Bridge in Seattle (Zarkowski, Reference Zarkowski2008) and Skyway Bridge in Tampa, the United States (Stack, Reference Stack2015), barriers were eventually erected after the installation of fixed phone boxes. It is expected that many other public sites will follow a similar process, eventually resulting in the introduction of means restriction if the initial preventive strategy fails to lower suicides at sites.
Conclusion
Fixed landline phone box interventions do not consistently mitigate suicide mortality, as evidenced by significant heterogeneity and observed rate increases at two sites in the post-installation period. The absence of a demonstrable aggregate benefit implies that these measures might be insufficient as standalone deterrents. This suggests that additional or alternative interventions are necessary to achieve meaningful reductions in suicide mortality at high-risk locations.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S2045796026100924.
Availability of data and materials
We concealed the names of included sites to mitigate the imitation of suicidal behaviour (use the sites for suicide) from readers who are at risk of suicide. However, we will provide the names of sites and data providers, as well as information on data access availability, upon reasonable request.
Author contributions
S.S., J.P. and L.S.T. conceptualised, designed and led the study. S.S. prepared the first draft of the manuscript. S.S., J.P., M.S., A.J.C. and L.S.T. interpreted data and made critical intellectual revisions to the manuscript. A.J.C. is a guarantor. All authors approved the final version for submission.
Financial support
S.S. and A.J.C. are supported by a National Health and Medical Research Council (NHMRC) Partnership Project Grant (No. 1191874). J.P. is supported by an NHMRC Investigator Grant (No. 1173126) and holds an NHMRC Partnership Project Grant (No. 1191874). M.S. is supported by an NHMRC Investigator Grant (No. 2025205). L.S.T. is supported by an NHMRC Early Career Fellowship (No. 1156849).
Competing interests
We declare no competing interests.
Ethical standards
Ethical approval for the study was received from the University of Melbourne Human Research Ethics Committee (2022-23756-35566-4), and the Justice Human Research Ethics Committee of the Victorian Department of Justice and Community Safety (CF/14/22880) for the purpose of ethical clearance of Australian data.
Open access
