Effects of the Canadian Suicide Prevention Service's Text Interventions on Texters' Emotions, Distress Relief, Perceived Abilities, and Practices Associated With Better Outcomes

ABSTRACT

Aims

To describe users of the Canada Suicide Prevention Service textline (now “988”), explore their perceived impact of the service and identify characteristics of interventions associated with a greater likelihood of positive effects of exchanges.

Methods

Data from 146 transcripts were analyzed using quantitative content analysis, and data were associated with counselor assessments and pre‐ and post‐intervention questionnaire responses. Suicide risk was assessed using the Suicidal Ideation Attributes Scale (SIDAS).

Results

78.8% of texters exhibited “severe” suicidal ideation on SIDAS, with 26.7% reporting specific plans for suicide. Complete risk assessments were often not conducted, but counselors extensively explored texters' resources and discussed potential solutions. Positive emotional changes were associated with counselors' thorough exploration of resources. Only one technique, “Reinforcing a strength or positive action of the texter,” was significantly associated with positive outcomes.

Limitations

Low response rates to post‐intervention survey questions may affect the representativity of participants compared to all textline texters.

Conclusion

A large proportion of texters reported they were less upset and were better able to cope with their problems after the text exchange. However, there is a need for more training and supervision to ensure that adequate suicide risk assessments are conducted, or the development of shorter assessment procedures.

1. Introduction

1.1. Background

Text and chat intervention services for suicide prevention are becoming increasingly popular. They are more accessible to young people than telephone services (Mokkenstorm et al. 2017; Gould et al. 2021; Gould et al. 2022) and are generally appreciated for their sense of anonymity and the control of the interaction they provide to users (Drexler 2013; Predmore et al. 2017). In the absence of scientific data on best practices for text‐based interventions with suicidal people, services have adapted telephone intervention methods to their written communications. To date, there is little empirical data to corroborate the efficacy of using the same intervention techniques as in oral communications, nor do we have substantial evidence indicating what constitutes best practices in text and chat interactions with individuals at risk of suicide.

1.2. The Challenges of Text Interventions

According to interviews with helpline directors, text interventions generally begin by exploring the texters' situation and assessing their suicide risk, and continue by identifying texters' resources and possible solutions to their problems (Drexler 2013; Predmore et al. 2017). However, written exchanges often do not allow enough time to apply conventional intervention models in their entirety (Bambling et al. 2008; Chardon, Bagraith, and King 2011; Drexler 2013). The slowness of written communication requires that text exchanges take longer than telephone interventions (Bambling et al. 2008; Drexler 2013). Since organizations usually need to limit the duration of interventions to provide greater access to their services, practitioners often have difficulty completing their interventions within the allotted time (Mokkenstorm et al. 2017).

It may be more difficult to understand the texter's state of mind due to the absence of auditory cues (Drexler 2013; Predmore et al. 2017). It may also be more difficult to focus on what counselors consider relevant, to engage texters, and to get answers to their questions. For all these reasons, suicide risk assessment could be particularly difficult to complete (Drexler 2013; Predmore et al. 2017). Furthermore, the slowness of communication could make it difficult to get to the development of an action plan (Bambling et al. 2008; Chardon, Bagraith, and King 2011; Drexler 2013).

1.3. Current Evidence on Efficacy and Best Practices

The effectiveness of written interventions compared to telephone help in suicide prevention has not yet been established. Studies have found that telephone interventions are effective in reducing callers' distress between the beginning and the end of calls (Gould et al. 2007; Mishara et al. 2007). However, evaluative research on written communications is scarce. Research to date indicates that chat and text interventions are effective in reducing distress, hopelessness, and helplessness in suicidal individuals (Côté and Mishara 2022; Gould et al. 2021; Gould et al. 2022; Mokkenstorm et al. 2017; Sindahl et al. 2019; Williams et al. 2020).

Also, it has yet to be determined whether best practices in telephone interventions are as effective as written exchanges. Studies of practices associated with positive impacts on callers show that interventions that are longer, where the counselor takes the time to establish a bond of trust at the start of the call, assesses suicidal risk, explores resources and reasons for living, and discusses possible solutions, are more effective than entirely non‐directive interventions (Gould et al. 2013; Mishara et al. 2007).

Sindalh and colleagues (2018) studied 100 suicidal young texters to the Danish Bornetelefoon youth helpline to investigate best practices. They found that expressing empathy, discussing emotions, and assisting the young person in identifying someone to confide in were associated with better results. Cox et al. (2021) analyzed 269 crisis chats with suicidal adults. They suggested that the main difference between successful and unsuccessful chats lay in the counselors' ability to make a gradual transition between active listening and problem‐solving during the intervention. Côté and Mishara (2022) analyzed 112 text intervention transcripts from the Canada Suicide Prevention Service, which has since changed to become the 988 Suicide Crisis Helpline. They found that counselors almost always explored texters' resources and discussed potential solutions, and that reinforcing texters' strengths and positive actions was significantly associated with positive outcomes, and longer exchanges were more often comprehensive. Salmi et al. (2024) reached similar conclusions employing artificial intelligence to analyze 6903 chat exchanges from the Dutch 113 suicide prevention service. Their machine learning model found that exchanges incorporating positive affirmations—such as showing empathy, offering praise, and expressing happiness for the chatter; and when counselors actively collaborated with chatters to find concrete solutions, were significantly associated with reducing suicidality. Peart et al. (2023) conducted an in‐depth thematic analysis of 35 interventions with youths via Instagram. They found that counselors were able to employ a structured approach to investigate suicidality. However, they had to move beyond conventional procedural risk assessments and embrace more conversation‐oriented, meaningful discussions about suicidality.

1.4. Objectives

This exploratory study examined under‐researched dimensions of text interventions in suicide prevention. The main objective was to evaluate the impact of the Canada Suicide Prevention Service text message service on users. Secondary objectives were to describe the characteristics of texters, describe the content of interventions, examine how the duration and completeness of interventions affect outcomes, and identify practices associated with positive outcomes.

2. Methodology

We combined three methodologies: quantitative content analysis of intervention transcripts, pre‐post measurements and comparisons, and a cross‐sectional analysis of the relationships between session characteristics and outcome measures. Due to a constrained timeline for conducting the evaluation study, the questionnaires used in this study were not pilot tested beforehand to assess items reliability and validity.

2.1. Sample

CSPS provided us with 150 complete transcripts of exchanges, collected between 14 August and 28 September 2020 (during the Covid pandemic), that all included at least some suicide content and lasted at least 20 min. They also provided the associated responses to pre‐and post‐session questions answered by texters, and the reports completed online by the counselors after each exchange. We retained 146 exchanges for analyses after eliminating three that were from a relative or friend of the texter and one where the texter was under age 14.

2.2. Pre‐Intervention Questionnaire

Pre‐intervention questionnaires were completed before texters were directed to the virtual waitroom where they waited to be connected to the next available counselor. They were asked their age, gender, level of distress (How upset are you now?: 1 = OK– 5 = Extremely upset), their perceived ability to cope (Do you feel able to cope with your situation?: 1‐ Not at all to 5‐ Very well), presence of suicidal ideation (Do you have thoughts of suicide? Yes or No) and about suicide plans (Do you have a specific plan to kill yourself?: Yes or No). These questions are modified versions of the default questions used by the helpline software iCarol (https://www.icarol.com) and are similar to those used in previous research on text‐based crisis lines (Williams et al. 2020; Gould et al. 2022).

2.3. Post‐Intervention Questionnaire

Texters received an automated text message immediately after their intervention, inviting them to evaluate their satisfaction with the help received. The text hyperlink directed them to an online consent form explaining the study. After providing consent, participants were redirected to the post‐intervention questionnaire.

This questionnaire was based on the scales used in studies conducted by Mishara et al. (2007) and Mokkenstorm et al. (2017). Although these scales were originally designed as observational tools, we created self‐report items to assess the same constructs. The questionnaire assessed again texters' level of distress, perceived ability to cope, level of satisfaction concerning various aspects of the intervention, using a 5‐point scale ranging from ‘Not at all true’ to ‘Completely true’ (My responder seemed warm, supportive, and concerned; I felt heard, understood, and respected; I was able to express my feelings; We talked about what I wanted to talk about; I figured out possible ways of coping with a particular situation or problem; The session was helpful to me; Overall, I was satisfied with the session), and the impact of the intervention on their emotions (As a result of the interaction, I feel more: 1‐ Sad to 10‐ Happy, 1‐ Tense to 10‐ Relaxed, 1‐ Angry to 10‐ Calm, 1‐ Afraid to 10‐ Confident, 1‐ Uncertain to 10‐ Definite, 1‐ Helpless to 10‐ Capable, 1‐Hopeless to 10‐Hopeful). The questionnaire concluded with the Suicidal Ideation Attributes Scale (SIDAS; Spijker et al. 2014), a validated five‐item self‐report measure assessing suicidal ideation over the past 30 days on a 0–10 Likert scale. The SIDAS evaluates risk based on frequency, controllability, proximity to attempt, distress, and interference with daily activities, with scores ranging from 0 to 50, higher scores indicating greater suicide risk. The SIDAS has demonstrated a strong internal consistency across various community samples (Gauvin, Bardon, and Côté 2022; Han et al. 2017; van Spijker et al. 2014), as well as among users of a crisis chat service (Gauvin and Côté 2023).

2.4. Intervention Reports

Counselor intervention reports identified issues raised by the texters (e.g., mental health diagnoses and difficulties, interpersonal and social challenges) using a computerized checklist. These reports are generated using the Icarol software employed by the CSPS. This software enables counselors to compile issues discussed with the texters. The issues listed are not mutually exclusive, allowing counselors to select multiple issues for each texter.

2.5. Quantitative Content Analysis

We developed scales to describe intervention behaviors by counselors, adapted from Mishara et al. (2007) evaluations of telephone helpline interventions in the U.S., using silent monitoring. The scales indicate the extent to which risk assessment questions were asked, how information about suicide risk was provided, and the frequency intervention techniques were used. Inspired by the work of Chardon, Bagraith, and King (2011), we also developed four ordinal scales to describe the quality and extent of completion of different stages: suicide risk assessment, exploration of situation, exploration of texter resources, discussion of possible solutions, and agreement on a plan of action (see Supporting Information). These scales align with CSPS standards and training, and intervention models commonly used by helplines in suicide prevention.

Two research assistants (RAs) with experience working at a suicide prevention helpline, who worked on our previous study of text messaging (Côté and Mishara 2022), were extensively trained over 3 months to use the scales, using 30 additional intervention transcripts. RAs were required to code three transcripts per week. Once a week, the RA, the first author (LPC), and a suicide prevention trainer met to review the coded transcripts and discuss each person's coding decisions. Each RA received over 100 h of training. After training, 40 randomly selected transcripts were independently coded by the two RAs to verify interrater reliability, assessed by Fleiss Kappa for categorical variables and weighted Kappa for ordinal variables. Interrater reliabilities were high for all items (see Table 1).

TABLE 1.

Interrater reliabilities (N = 40).

Suicide risk assessment	Kappa	C.I. 95%
Suicidal ideation	1	0.74 1.26
Asked question directly (Are you thinking of suicide?)	0.76	0.4 1.12
Suicide plans	1	0.71 1.12
Prior suicide attempts	0.94	0.61 1.26
Are you alone?	1	0.71 1.3
Substance abuse	1	0.75 1.25
Intervention techniques
Validation	0.88	0.82 0.96
Moral support	0.95	0.91 0.1
Reinforcing a strength or a positive action	0.94	0.89 0.1
Reformulation	0.93	0.88 0.98
Questions about feelings	0.85	0.72 0.99
Psychoeducation	0.91	0.83 0.98
Mettre des limites	0.93	0.67 1.18
Questions about past and present resources	0.93	0.87 0.99
Empowerment: resources & solutions	0.91	0.81 1.01
Suggestions: resources & solutions	0.94	0.89 0.1
Degree of completion of steps in the intervention
Exploration of suicide risk	1	1.00 1.00
Exploration of situation	1	1.00	1.00
Exploration of resources and solutions	0.94	0.82 1.1
Elaboration of an action plan	0.93	0.8 1.1

3. Results

3.1. Characteristics of Texters

Texter ages ranged from 14 to 58 (M = 24.95, SD = 9.26; Mdn = 22) (see Table 2). Of the 146 texters, 71.23% (N = 105) identified as female; 20.54% (N = 30), 4.79% (N = 7) transgender, and 2.73% (N = 4) as “other.” In the post‐intervention questionnaire, 78.76% had SIDAS scores indicating severe suicidal ideation (M = 30.08, SD = 10.83, Mdn = 32), and 26.71% said they had a specific suicide plan.

TABLE 2.

Texters' characteristics.

Characteristics	Total sample (N = 146)
Age
14–18	39 (26.71%)
19–24	44 (30.13%)
25–44	56 (38.35%)
45–64	6 (4.09%)
Unknown	1 (0.68%)
Gender
Female	104 (71.23%)
Male	30 (20.54%)
Trans	7 (4.79%)
Other	4 (2.73%)
Suicide risk
Severe suicidal ideations (SIDAS)	115 (78.76%)
Do you have a specific plan to kill yourself? (Yes)	39 (26.71%)
Mental health issues (as reported by counselors)	112 (76.71%)
Depressed mood	68 (46.57%)
Anxious	31 (21.23%)
Prior suicide attempt	20 (13.69%)
Self‐esteem	25 (17.12%)
Mental Illness	22 (15.06%)
Self‐harm	19 (13.01%)
Medication Issues	7 (4.79%)
Barriers to accessing services/wait list	9 (6.16%)
Interpersonal Issues (as reported by counselors)	72 (49.31%)
Relationships	44 (30.13%)
Isolation – loneliness	33 (22.6%)
Abuse – violence	17 (11.64%)
Bereavement—loss	10 (6.84%)
Living condition (as reported by counselors)	13 (8.9%)
Financial – debt	10 (6.84%)
Housing	3 (2.05%)
Unemployment	3 (2.05%)
Issues related to the COVID‐19 pandemic	29 (19.86%)

The most prevalent reason given for contacting the textline (76.71%) was mental health problems, followed by interpersonal issues (49.31%), issues related to the COVID‐19 pandemic (19.86%), and difficult life circumstances (8.9%). Texters often had multiple issues and reported diagnoses. Texters aged 19–24 had significantly more interpersonal issues than texters aged 18 and under, as well as texters aged 25 and older (Chi² (2, N = 146) = 13.4, p = < 0.001).

3.2. Stages of the Intervention Completed During the Exchange

The average length of exchanges was 59.71 min (Mdn = 55.5, SD = 25.36, range = 20–164). Short interventions were generally ended by texters, while longer interventions were more often ended by counselors. Texters initiated the end of the exchange in 72.72% of sessions less than 30 min, 50% of sessions 31–60 min, 40.47% of sessions 61–90 min, and 23.52% of sessions 91–120 min.

The quality and extent of completion of different intervention stages are summarized in Table 3. Most risk assessments were classified as incomplete (56.84%). In these, we had information about the texter's suicidal ideation and/or if the texter had a plan, but no information or exploration of: if, when and where the plan would be executed. Complete assessment including exploration of all three characteristics were observed in 23.28% of sessions. Only 15.06% had a risk assessment classified as “exhaustive,” that is, with other risk factors also explored. Texters were generally collaborative with risk assessment and most answered all the counselors' questions. The degree of exploration of suicide risk was positively correlated with the length of the interventions (r _s = 0.26, p = 0.002).

TABLE 3.

Stages of the intervention (N = 146).

Degree of exploration	Stages of the intervention
	Suicide risk assessment	Exploration of situation	Exploration of resources and solution	Establishment of an action plan
Absent	7 (4.79%)	0 (0%)	4 (2.73%)	51 (34.93%)
Incomplete	83 (56.84%)	6 (4.11%)	9 (6.16%)	28 (19.17%)
Adequate	34 (23.28%)	19 (13.01%)	24 (16.43%)	34 (23.28%)
Exhaustive	22 (15.06%)	121 (82.87%)	109 (74.65%)	33 (22.6%)

The texter's situation was explored exhaustively in most interventions (82.87%), indicating they had in‐depth knowledge of the texter's problem, and about what the texter was doing, thinking, and feeling about their situation. The degree of exploration of the texter's situation was also significantly correlated with the intervention length (r _s = 0.33, p = < 0.001).

The degree of exploration of resources and solutions was measured by the number of resources and solutions discussed, and the extent to which these resources could be used. 74.65% of interventions were classified as exhaustive, with many of the texters' resources or coping strategies present. Exhaustive explorations included knowing whether they could use these resources to improve their current situation, with several ways to use these resources discussed. In 16.43% of the interventions, there was at least some discussion about how one resource could be of help. The degree of exploration of texter's resources was significantly correlated with the intervention length (r _s = 0.25, p = 0.002).

The extent of development of an action plan was measured by the presence of a discussion of what to do after the exchange and the level of detail of the steps to follow to implement the solutions identified. In 19.17%, information on the purpose of the action plan was present, but not on when and how to implement it. In 23.28%, information on the purpose and implementation of the action plan was present. In 22.6%, information on the purpose and implementation of the plan was present, as well as a discussion on the management of at least one potential barrier to the plan's realization. The extent to which an action plan was developed was significantly correlated with the intervention length (r_s = 0.4, p = < 0.001).

Overall, explorations of the texter's situation was “adequate” or “exhaustive” in 95.88% of interventions, and the exploration of resources and solutions was “adequate” or “exhaustive” in 91.08%. Counselor initiated the end of the exchange in 55.56% of the interventions where a suicidal risk assessment was “absent” or “incomplete”, and in 53.13% of the interventions where the development of an action plan was “absent or incomplete”.

3.3. Texters' Perceptions of the Impact of the Interventions

Most texters reported that they were less upset after the intervention than before (114/146, 78.08%) (see Table 4). A paired‐sample t‐test using pre‐test (M = 3.79; SD = 0.99) and post‐test (M = 2.32; SD = 1.16) distress scores indicates a significant decrease in distress between the intervention's beginning and the end (t(145) = 12.89, p = < 0.001, Cohen's d = 1.38).

TABLE 4.

Self‐reported impact and satisfaction.

Distress relief and perceived ability to cope (changes from pretest to post‐test)	Worse	No change	Better
How upset are you now?	11 (7.53%)	21 (14.38%)	114 (78.08%)
Do you feel able to cope with your situation?	24 (16.43%)	43 (29.45%)	79 (54.11%)
Satisfaction (In your text session…)	Not at all true or Somewhat true (1–2)	Moderately true (3)	Very true or Completely true (4–5)
My responder seemed warm, supportive, and concerned.	8 (5.47%)	14 (9.58%)	124 (84.93%)
I felt heard, understood, and respected.	12 (8.21%)	11 (7.53%)	123 (84.24%)
I was able to express my feelings.	10 (6.84%)	12 (8.21%)	124 (84.93%)
We talked about what I wanted to talk about.	9 (6.16%)	20 (13.69%)	117 (80.13%)
I figured out possible ways of coping with a particular situation or problem.	22 (15.07%)	29 (19.86%)	95 (65.68%)
The session was helpful to me.	17 (11.64%)	16 (10.96%)	113 (77.39%)
Overall, I was satisfied with the session.	14 (9.59%)	17 (11.64%)	115 (78.76%)
Impact on feelings (As a result of the text session, I feel more…)	Negative (1–2‐3)	Neutral (4–5–6‐7)	Positive (8–9‐10)
Sad – Happy	9 (6.16%)	71 (48.63%)	66 (45.15%)
Tense – Relaxed	13 (8.9%)	70 (47.94%)	63 (43.15%)
Angry – Calm	6 (4.1%)	56 (38.35%)	84 (57.53%)
Afraid – Confident	6 (4.1%)	94 (64.38%)	46 (31.51%)
Uncertain – Definite	14 (9.59%)	95 (65.06%)	37 (25.34%)
Helpless – Capable	15 (10.27%)	78 (53.42%)	53 (36.3%)
Hopeless – Hopeful	16 (10.96%)	76 (52.05%)	54 (36.98%)

About half of texters reported that their abilities to cope with the situation improved (79/146, 54.41%). A paired‐sample t‐test using pre‐test (M = 2.49; SD = 1.04) and post‐test (M = 3.09; SD = 1.17) of coping ability scores indicates a significant increase in texters' ability to cope with their situation (t(145) = −5.63, p = < 0.001, Cohen's d = 1.29).

A large proportion of texters were completely satisfied with the session (78.76%) and felt the session was helpful (77.39%). Their ratings of the counselor were very positive (see Table 4). Two‐thirds (65.68%) said that it is very true or completely true that “I figured out possible ways of coping with a particular situation or problem.”

About half of texters experienced no or few positive changes in feelings during the session, but when there were changes, the changes were most often positive (see Table 4).

3.4. Variables Associated With Outcome Measures

In line with multiple studies examining the impact of intervention practices in suicide prevention helplines (Côté and Mishara 2022; Gould et al. 2013; Mishara et al. 2007; Mokkenstorm et al. 2017; Sindahl et al. 2019), we employed a linear model to investigate the relationships between counselors' behaviors, session characteristics, and outcome measures. Since this research is exploratory, we developed multiple regression models.

To examine the impact of the completeness and duration of the interventions, we developed an initial set of predictors: intervention duration, along scores from the completeness scales for each intervention stage (risk assessment, situation exploration, resource exploration, discussion of possible solutions, and agreement on plan of action) (see Table 3). To examine the impact of using different intervention techniques, we used the frequencies of use of validation, moral support, reinforcing strengths or positive actions, reformulation, questions about feelings, psychoeducation, setting boundaries, questions about past and present resources, empowering texters to identify their resources and solutions, and suggesting resources and solutions. The outcome variables were: (1) distress relief, measured by post‐intervention responses to “How upset are you now?”; (2) perceived ability to cope, measured by post‐intervention responses to “Do you feel able to cope with your situation?”; and (3) improvements in feelings scores, from the aggregation of the emotions items.

We tested our two sets of predictors for each outcome variable, which resulted in six regression models. Since the outcomes “distress relief” and “perceived ability to cope” are post‐intervention measures, we conducted hierarchical regression models to control for the corresponding pre‐intervention measures for each item (“How upset are you now?” and “Do you feel able to cope with your situation?” responses at pre‐intervention). Only the models that respected statistical assumptions and yielded statistically significant results are presented here (see Supporting Information).

3.5. Preliminary Analysis

Most intervention techniques had one or two z‐scores greater than 3 standard deviations from the mean, so their frequencies were winzorized. The variables Exploration of situation and Exploration of resources and solution also had multiple z‐score outliers and were winzorized. We extracted raw residuals, studentized residuals, and Cook's Distances to assess the quality of our regression models using diagnostic tests. We then calculated Mahalanobis distances for each model to check for multivariate outliers, ran Shapiro–wilk tests for assessing the normality of the distribution of the residuals, and extracted Variance Inflation Factor (VIF) scores to check for multicollinearity.

3.6. Model 1. Impact of Completeness of the Interventions on Improvements in Feelings

The dependent variables for Model 1 were the sum of positive/negative emotions scores. The predictors were the session duration and the completeness of each step of the intervention. In the preliminary analysis, 12 participants had scores above our threshold for Cook distances. However, visual inspection indicated that these participants' scores did not significantly stand out from the rest of the sample. They were therefore retained for subsequent analyses. No multivariate outliers were detected, nor were there any problems with the distribution of residuals or multicollinearity. The model as a whole is statistically significant (F(5, 140) = 5.902, p ≤ 0.001) (Table 5), which explains 15% of the variance in the sum of positive/negative emotions (r ² = 0.18; adjusted r ² = 0.15). Only the extent of the Exploration of resources and solutions was associated with improvement in feelings. Excluding the 12 participants identified during the preliminary analyses did not yield different results.

TABLE 5.

Multiple linear regression model of improvement in feelings.

	B	p	R 2	Adj. R 2
Model			0.18	0.15
Suicide risk assessment	−0.04	n. s.
Exploration of situation	0.1	n. s.
Exploration of resources and solutions	0.39	p ≤ 0.001
Establishment of an action plan	−0.04	n. s.
Length	−0.14	n. s.

3.7. Model 2. Impact of Intervention Techniques Frequencies on Distress

Dependent variables for Model 2 were responses to the post‐test question How upset are you? The predictors were the frequencies of the intervention techniques used by counselors. Since the pre‐session responses to the questions of “How upset are you now?” were potentially associated with the post‐session responses, they were included in the analyses in order to adjust for baseline covariates (Dalecki and Willits 1991). This approach allowed to control for baseline variance and to examine the effect on the outcome variable. Four participants obtained Cook's distance scores above our threshold of 0.027. However, visual inspection indicated that these participants' scores did not significantly stand out from the rest of the sample. They were therefore retained in subsequent analyses. No multivariate outliers were detected, Shapiro–wilk test indicated no problem with the distribution of the residuals. VIF scores indicated no multicollinearity problems. The F test F(11, 134) = 2.71, p = 0.003) indicates that the model as a whole is statistically significant (Table 6). The model explains 11.5% of the variance in the post‐test How upset are you now? scores (r ² = 0.18; adjusted r ² = 0.12). Only the technique Reinforcing a strength or a positive action was associated with reduction of distress at post‐test. Excluding the three participants identified during the preliminary analyses did not yield different results.

TABLE 6.

Hierarchical linear regression of the impact of intervention techniques frequencies on distress.

	B	p	R2	Adj. R 2
Model 1 (baseline covariate)			0.04	0.03
How upset are you now? (pretest)	0.19	0.02
Model 2			0.18	0.12
Validation	0.07	n. s.
Moral support	0.03	n. s.
Reinforcing a strength or a positive action	−0.26	p ≤ 001
Reformulation	0.12	n. s.
Questions about feelings	−0.08	n. s.
Psychoeducation	−0.1	n. s.
Setting boundaries	−0.14	n. s.
Questions about past and present resources	−0.15	n. s.
Empowerment: resources & solutions	0.06	n. s.
Suggestions: resources & solutions	0.21	n. s.

4. Discussion

Consistent with what has been described in the literature on chat and text‐based suicide prevention services, the majority of CSPS users were women aged 19–44, and one in four was an adolescent (Mokkenstorm et al. 2017; Gould et al. 2021; Gould et al. 2022). The problems discussed were mainly mental health and interpersonal issues, which are comparable to what is observed in telephone interventions.

Unfortunately, the CSPS was not able to provide us with information about response rates to the questionnaire, due to shortcomings in the computer program they used for data collection. Given the small sample size in our study and the lack of information regarding the questionnaire response rate, we cannot generalize our results to all CSPS texters. Nevertheless, we found that 78.76% of texters had a SIDAS score indicating a high severity of suicidal ideation. This is consistent with observations from the Quebec‐based chat intervention service, suicide.ca, where 75.56% of all texters had SIDAS scores indicating severe suicidal ideation (Côté and Lane 2024). Moreover, 26.71% of texters in the present study indicated that they had a specific plan to attempt suicide. These results support the notion that text‐based intervention services effectively engage young individuals experiencing suicidal ideation who have significant suicide risk, which is particularly significant given this population's frequent reluctance to contact telephone helplines (Gould et al. 2006). Although some quantitative descriptive studies have provided insights into the characteristics of textline users (Côté and Mishara 2022; Gould et al. 2021; Gould et al. 2022; Mokkenstorm et al. 2017; Sindahl et al. 2019; Williams et al. 2020), further investigation is essential to uncover the motivations driving suicidal youth and young adults to utilize textlines and the specific benefits they derive from this type of support. Additionally, future studies should use standardized measures, such as the SIDAS, to facilitate comparisons across different textlines.

Our results indicate that interventions were mainly focused on exploring the texters' situation and identifying possible solutions to their problems. However, these discussions rarely resulted in the development of an action plan to help the texter implement the solutions identified. Counselors rarely asked questions about all three components of suicide planning (how, where, when) and other risk factors were rarely explored. Texters almost always responded to the risk assessment questions they were asked, which is contrary to what has sometimes been reported (Drexler 2013). Counselors should not assume that suicide risk assessment will be difficult to conduct. However, counselors may not conduct complete risk assessments because of the amount of time required, which is supported by the finding that the completeness of intervention stages was correlated with the intervention duration. It appears to be possible to carry out complete texting interventions, but the counselors must take enough time to do so.

The development of an action plan was the stage least likely to be completed. It remains unclear whether this is due to counselors neglecting this step or texters being reluctant to engage in this. Furthermore, we found that texters ended approximately half of the incomplete interventions. Further research is needed to understand what keeps texters engaged in the exchanges and what motivates counselors to conclude interactions prematurely. For example, texters might be multitasking, on the move, or may end the conversation because they feel better or have received the support they sought. Counselors might shorten a conversation because they are at the end of their shift, view certain steps as being nonessential, or end interactions they deem unproductive. Qualitative research could shed light on these dynamics. Future studies might explore strategies counselors can use to foster greater texter engagement.

Several indicators suggest that the service is more effective in reducing texters' distress than in improving texters' confidence in their ability to cope with their situation. Most texters (78.08%) reported that their distress decreased, and 54.11% felt more able to cope with their situation. The impact of the interventions on emotions was better for affects associated with emotional regulation (Sad—Happy, Tense—Relaxed, Angry—Calm) than for feelings associated with problem‐solving (Afraid—Confident, Uncertain—Definite, Helpless—Capable, Hopeless—Hopeful). These results align with previous studies on chat and text services, which reported that services are effective in reducing distress, hopelessness, and helplessness in suicidal individuals (Côté and Mishara 2022; Gould et al. 2021; Gould et al. 2022; Mokkenstorm et al. 2017; Sindahl et al. 2019; Williams et al. 2020).

Comprehensive exploration of texters' resources and solutions was associated with improved affect. This result is consistent with research on best practices in telephone intervention, which indicates the importance of collaborative problem solving in crisis intervention (Mishara et al. 2007; Cox et al. 2021). The practice of highlighting texters' strengths and positive actions was associated with positive outcomes. This technique manifested in various ways, including counselors praising texters for acknowledging their need for help or for reaching out; commending their actions; highlighting qualities such as courage, strength, perseverance, and resilience; and validating the small steps they had already taken to deal with their problems. The association of this technique with positive outcomes might reflect counselors' ability to recognize texters' personal resources and empower them to navigate the adversities they face. However, it should be noted that the use of this technique was associated with a decrease in distress, but not with their feelings of being able to cope with their situation. Perhaps, in texting, helping the texter recall a memory of success is sufficient to evoke positive feelings that facilitate a return to calm. This is the third study that found that this practice is associated with positive outcomes (Côté and Mishara 2022; Salmi et al. 2024).

Suicide risk assessment was not associated with impacts in this study. Given the ongoing doubts about the reliability, validity, and clinical significance of suicide risk assessments, it is important to evaluate their value and manner of application in text‐based interventions. A crucial aspect of suicide risk assessment is exploring whether the texter has a plan. This element is essential for ensuring the safety of the texter, particularly for helplines that adopt the ethical stance that if life is in danger, a rescue must be undertaken under all circumstances (Mishara and Weisstub 2024). However, when it comes to other risk factors, it may be less important to explore all the elements listed in suicide risk assessment scales in the context of text interventions. Peart et al. (2023), in their study of 35 text exchanges with youths age 13 to 25 on Instagram, demonstrated how counselors used rating scale questions about suicidal intent to explore protective factors, which subsequently facilitated the identification of the person's strengths to help prevent suicide attempts. Future studies could explore how counselors use suicide risk assessments to identify texters' resources and strengths as an intervention strategy.

Finally, more research is needed to explore the impact of the intervention's duration and the counselor's approach on outcomes. The results of our study suggest that both the counselor and the texter can affect the intervention's length and comprehensiveness, raising important questions about whether the effectiveness of the intervention depends more on its duration, the counselor's approach, or the interaction dynamics between the texter and the counselor. Future studies with larger samples and more sophisticated statistical models could explore the relationships among intervention duration, texter characteristics (such as age, gender, presenting problems, distress level, and suicidality), the counselor's intervention style, and factors such as the quality of contact, the level of engagement in the interaction, and the effectiveness of collaborative problem‐solving efforts.

5. Limitations

The exploratory nature of this study restricts the generalizability of our findings. The study essentially used a convenience sample. The CSPS provided us with 150 transcripts that included at least some content related to suicide among those who answered the pre‐and post‐questionnaires, and we retained 146 exchanges based on the inclusion criteria. Since CSPS handled the sample selection, there may have been selection bias that we are unaware of. Furthermore, they were unable to provide us with the overall response rate for the post‐intervention survey, nor did we have access to data to conduct analyses to compare our sample with texters who did not answer the post‐intervention questionnaire. Thus, it is possible that the participants are not representative of all texters using the CSPS text service, which further limits the generalizability of the findings.

The cross‐sectional design limits our ability to establish causal relationships, although the pre‐post comparisons provide an indication of changes perceived by texters as associated with the intervention. Moreover, since this study is primarily quantitative, we lack qualitative analyses that could possibly help identify the clinical dynamics explaining why certain intervention stages were more likely to be completed than others and why specific intervention practices are associated with better outcomes. Additionally, the absence of manipulation or random assignment complicates the determination of whether observed changes are directly caused by specific interventions or if other, unobserved factors influenced the outcomes. Although these limitations are shared by research on textlines to date, they imply that future research should include both quantitative and qualitative components, and more longitudinal research is needed.

The iCarol program questions and categories for counselor data collection where already determined by the CSPS before we began this project and we had to use the data gathered in their daily practice, without the ability to modify questions asked or categories of information. We do not know the specific criteria used by counselors to categorize texters presenting problems (e.g., how they differentiated between mental illness, depressed mood, and anxiety). Additionally, some measures in the post‐intervention questionnaire exhibited poor item construction (e.g., “my responder seemed warm, supportive, and concerned,” and “I felt heard, understood, and respected”). The different characteristics in the same question should have been separated to ensure measurement validity and reliability, allowing for a clearer understanding of which descriptor the participants were responding to.

Another limitation is our measure of the length of the intervention based on its duration in minutes. Time alone may not be a reliable indicator of the amount of communication taking place, and the pace of communication may influence the perceived quality of contact, with response delays potentially serving as indicators—or being interpreted as markers—of both counselors' and texters' engagement in the intervention. Future research should take this characteristic of text‐based communication into account.

6. Conclusions

The results of this study indicate that it is possible to deliver quality interventions by text, but that it takes time. Since organizations offering text services usually have a high demand, they often face a difficult choice: to deliver complete interventions, similar to those with telephone calls, that require much more time, so fewer texters obtain help, or to have shorter and less complete interventions with more texters. Our results suggest that a compromise is possible. Text exchanges were found to be beneficial when the interventions lasted long enough to enable the counselors to conduct an exhaustive exploration of the texters' resources and also helped texters identify possible solutions to their problems. Furthermore, because of pressure to limit the length of text exchanges in order to serve more texters, it is important to determine which factors are essential to explore to conduct an adequate suicide risk assessment.

In our sample, one out of four (26.71%) texters had a specific suicide plan. In these exchanges, the priority should be to ensure the texter's immediate safety. At the very least, helpline counselors should adequately explore the texter's suicide planning in order to focus on the need to ensure safety when the risk is imminent. One potential way to gather more useful information about the texters’ suicidal intentions could be to include additional questions in the pre‐intervention questionnaire. This approach may help streamline the risk assessment, allowing counselors to focus more on addressing texters' concerns, reinforcing their strengths and resources, discussing potential solutions, and exploring ways to implement them effectively.

More studies are needed to understand the clinical dynamics in interventions using written communications. We need to develop models for suicide prevention interventions using text that take into account the specificities of this means of communication, based on research that validates their usefulness. We also need to determine what adjustments to intervention techniques are needed to consider the specificities of different sub‐groups of texters. Once these models have been established and validated, we will be able to develop training programs for text interventions in suicide prevention that are evidence‐based.

Ethics Statement

Approved by the Université du Québec à Montréal ethics committee; certificate no. 2788_e_2020.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Data S1.

Data Availability Statement

The data that support the findings of this study are not publicly available due to privacy and ethical restrictions.