Suicide Intervention Gatekeeper Training

Abstract

Objective:

Suicide is a significant public and mental health crisis in the United States. Training providers in suicide assessment and response is designated as one of the primary strategies for reducing deaths by suicide. Research has established that suicide intervention training is effective, but little work has been published on potential mediators of skill development and use.

Method:

Secondary data analysis of a randomized trial of the Question, Persuade, and Refer gatekeeper training with master of social work students. Path analysis was used to estimate mediated effects of knowledge, attitudes, reluctance, and self-efficacy on behavior outcomes.

Results:

Results suggest improvements in posttraining measures for knowledge, attitudes, self-efficacy, reluctance, and the use of gatekeeper behaviors, but there was no supporting evidence for the presence of mediated effects on behavior. Only self-efficacy demonstrated a strong direct relationship with gatekeeper behaviors.

Conclusions:

Ongoing evaluation is needed with an added interest in self-efficacy and how it can be enhanced through training.

Keywords

suicide suicide prevention gatekeeper training self-efficacy

Suicide is a significant public health and mental health crisis in the United States. The number of deaths by suicide in the United States continues to grow each year with no significant reduction in the rate of suicide deaths occurring in the United States in the past 50 years (National Action Alliance for Suicide Prevention [Action Alliance] Research Prioritization Task Force [RPTF], 2014). According to the Centers for Disease Control (CDC, 2015), there were more than 41,000 suicides in 2013; this equates to a rate of 12.6 suicides per 100,000 people or 1 suicide every 13 min or 117 suicides every day. Suicide is the 10th leading cause of death among all age-groups; the highest rate of suicide (19.3 per 100,000) is among people 85 years or older followed closely by middle-age adults aged 45–64 (19.2 per 100,000). More than 9 million adults reported having suicidal thoughts in the past year, and 1.3 million adults reported nonfatal suicide attempts (CDC, 2015). In 2014, nearly 500,000 people were treated in hospitals in the United States for injuries due to self-harm.

Suicide Intervention Training

In February 2014, the Action Alliance RPTF released its report outlining multiple pathways for reducing deaths by suicide using rigorous research and innovative ideas. The Action Alliance identified training health and behavioral health providers in suicide assessment and response as one of the top research goals for the country (p. 38). The Zero Suicide project is a highly promoted program that targets intervention efforts in health and behavioral health systems. Two essential elements outlined in the program are training to develop a “competent, confident, and caring workforce” and “systematically identify and assess suicide risk among people receiving care” (Suicide Prevention Resource Center, “Zero Suicide”, n.d.).

Developing and testing suicide intervention training curricula has been a challenge for the field because of the wide variation in the populations of individuals who need training, the populations of individuals at risk, and the settings in which suicide intervention may occur (Osteen, Frey, & Ko, 2014). Furthermore, the task and roles individuals play in suicide intervention are diverse and directly impact the content of a given training, making some interventions appropriate for a target audiences but without widespread appeal. The lack of standardized measures and limited access to client outcome data has also severely limited efforts to test if training leads to fewer suicide attempts and deaths (Osteen et al., 2014; Pisani, Cross, & Gould, 2011). Based on a review of the training literature, Osteen et al. (2014) identified and outlined core curriculum components for training a variety of potential interventionists including community gatekeepers, health providers, and behavioral providers to effectively identify, assess, and manage suicide risk.

Core Training Constructs

Despite the heterogeneity of training issues identified previously, core training needs remain consistent (Beidas & Kendall, 2010; Osteen et al., 2014). Core components of suicide intervention training are knowledge about suicide, attitudes toward suicide and suicide intervention, self-efficacy for engaging in intervention behaviors, and intervention behaviors (Osteen et al., 2014). The field has grown significantly over the past 20 years, and there are many different training programs available with published research literature on training effectiveness. However, one limitation is that the majority of repeated measures research on the effectiveness of training has evaluated training outcomes on the four core components as if they were unrelated to one another. For example, researchers may report the impact of training on practice behaviors, attitudes, self-efficacy, and knowledge for each individual construct without regard to the interaction of these multiple factors. Research is available that explores the potential moderating roles of baseline characteristics such as professional role and experience on the intervention behaviors with evidence supporting the notion that the development of intervention behaviors is impacted by more than just the training in and of itself (Osteen, Frey, Woods, Ko, & Shipe, 2016; Tompkins, Witt, & Abraibesh, 2010). This approach has allowed the field to make statements about whether or not training is effective in improving these outcomes, but it has done little to explore the “how/why” question or to develop strong causal models linking these components.

Social Cognitive Theory (SCT)

SCT provides a rationale for the interrelatedness of attitudes and self-efficacy and how they may mediate and moderate the development and use of intervention behaviors as a training outcome. Several researchers have argued that attitudes about suicide and intervention may predict changes in intervention behaviors (Gibb, Beautrais, & Surgenor, 2010; Valente & Saunders, 2004). Self-efficacy, believing that you can successfully engage in a behavior, may play a significant role in whether or not you actually attempt the behavior (Bandura, 1997; Ozer & Bandura, 1990) and function as a mediator of the application of behaviors (Maibach & Murphy, 1995). Results from a variety of suicide intervention training studies support associations between attitudes, self-efficacy, knowledge, and behavior (e.g., Conner, Pisani, & Kemp, 2013; Cross, Seaburn, Schmeelk-Cone, White, & Caine, 2011; Jacobson, Osteen, Jones, & Berman, 2012; Jacobson, Osteen, Sharpe, & Pastoor, 2012), but there is little research exploring interactions between constructs and the potential impact on behavioral outcomes (Osteen et al., 2014). Lee, Osteen, and Frey (2016) applied the SCT framework to a hierarchical analysis of change in clinical behaviors and found mixed support for the model; the results supported the role of self-efficacy in behavioral change but not for attitudes after controlling for individual and work-related characteristics. Osteen, Frey, Woods, Ko, and Shipe (2016) also used an SCT approach to analyze the direct and indirect effects of knowledge, attitudes, efficacy, and motivation of posttraining intervention behaviors and found strong evidence for the indirect effects of attitudes on self-efficacy and motivation and the indirect relationship of self-efficacy to posttraining intervention behaviors.

Mediation of Gatekeeper Intervention Behaviors

A path model of direct and indirect relationships linking core training constructs to the primary outcome of use of gatekeeper suicide intervention behaviors is presented in Figure 1. Hypothesized relationships are based on theory and empirical data from previous studies. Pathways move left to right indicating the passage of time from pretraining to posttraining. The left side of the model represents a typical multiple regression model in which pretraining scores for participants’ knowledge about suicide, attitudes about suicide and suicide intervention, self-efficacy for engaging in gatekeeper behaviors, reluctance to engage in those behaviors, and hours of previous suicide training are tested as predictors of pretraining use of gatekeeper intervention behaviors. The right side of the model represents a similar regression model in which posttraining scores are tested as predictors of the use of posttraining gatekeeper behaviors.

Figure 1.
Standardized point estimates (STDXY values) and 95% confidence intervals for direct and indirect effects. *p < .05. **p < .01.

It is the center of the model that represents the primary interest of this study. As illustrated by their position between pretraining and posttraining behaviors, it is hypothesized that posttraining scores on study outcomes mediate how much participants’ use of intervention behaviors actually changes, meaning that the degree of change seen over the course of the intervention is influenced by the degree of change in participants’ knowledge about suicide, attitudes about suicide and suicide intervention, self-efficacy for engaging in gatekeeper behaviors, and reluctance to engage in those behaviors.

Research Questions

Using SCT as a framework, this study was designed to answer the following questions to guide the development and testing of existing and new interventions:
Research Question #1: “Are pretraining scores on use of gatekeeper behaviors associated with pretraining scores on training constructs?” Specific training constructs are:
participants’ knowledge of suicide warning signs and risk factors,

participants’ negative attitudes toward suicide prevention (ASP),

participants’ self-efficacy to engage in gatekeeper behaviors,

participants’ reluctance to engage in gatekeeper behaviors, and

hours of previous suicide intervention training.

Researcher Question #1 is: “Are posttraining scores on training outcomes better than pretraining scores?” Specific training outcomes are:
participants’ knowledge of suicide warning signs and risk factors,

participants’ negative ASP,

participants’ self-efficacy to engage in gatekeeper behaviors,

participants’ reluctance to engage in gatekeeper behaviors, and

use of gatekeeper behaviors.

Research Question #2 is: “Is there evidence that training outcomes may mediate the use of gatekeeper intervention behaviors? Specific training outcomes are:
participants’ knowledge of suicide warning signs and risk factors,

participants’ negative ASP,

participants’ self-efficacy to engage in gatekeeper behaviors, and

participants’ reluctance to engage in gatekeeper behaviors.

Method

The current study is a repeated measures path analysis of secondary data originally collected during a randomized trial of the “Question, Persuade, and Refer” (QPR) suicide intervention gatekeeper training with advanced master of social work (MSW) students (Jacobson et al., 2012). The introductory-level QPR training covers the following topics: suicide rates and statistics across the life span; suicide warning signs, risk factors, and protective factors; procedures regarding how to ask about suicide risk; instructions for persuading clients at risk to seek additional help; and local and national referral resources for support and response to suicide risk (Quinnett, 1995). Data were collected at three time points for the intervention group (pretest, posttest, and follow-up) and at two time points for the control group (pretest and follow-up). An online pretest survey was administered prior to randomization, and participants in the intervention group were assigned by the researchers to one of the two trainings based on the student’s schedule. Participants in the intervention group received a paper version of the posttest immediately following the training. All study participants received an online follow-up survey at 4 months. The trainings were 90 min in length and delivered by the same trainer who was certified by the QPR institute. Additional information on the QPR training, method used in the original study, including sample, measures, and procedures, is available in Jacobson et al. (2012).

Sample

In the original study, the researchers randomly selected 112 MSW students from a possible population of 417 MSW students who were enrolled in a large school of social work in a mid-Atlantic university. Inclusion criteria were that students had to be enrolled in their concentration (or advanced) year of their MSW program at the time of the study and entering their advanced or second-year field placement. Seventy-two students consented to be in the study with 38 (52%) assigned to the control group and 34 (48%) assigned to the intervention group. All participants completed a pretest survey and a 4-month follow-up survey, but only participants in the intervention group completed posttest surveys immediately following the training. Only intervention arm participants are included in the current analyses because this study uses posttest scores as mediators and participants in the control group did not complete posttests. Participants were primarily female (n = 31, 91%) and Caucasian (n = 22, 64%). Students identified as being in either the clinical concentration (n = 27, 80%) or macro concentration (n = 7, 20%). Most participants were providing direct client services (yes [n = 27, 80%] or no [n = 7, 20%]). The majority of students were second year (n = 25, 76%) compared to advanced standing (n = 9, 24%). Previous hours of training in suicide intervention were also measured (mean = 8.62, SD = 6.17, median = 2.00).

Model Variables

Five training outcomes were analyzed in the current study: knowledge, attitudes, self-efficacy, reluctance, and gatekeeper behaviors. Most of the measures used in this study were developed specifically for use with the QPR training, with the exception of the attitudes measure, and have been widely used in evaluations of the QPR suicide intervention training. Hours of previous suicide intervention training were included as a covariate.

Knowledge

To assess declarative or factual knowledge about suicide prevention, the researchers used a standardized 14-item self-report measure developed by Wyman et al. (2008) for use of suicide prevention within school settings; items were modified with the permission of the developer to align with the study population (e.g., “students” were changed to “clients”). The Knowledge of Suicide Warning Signs and Intervention Behaviors Scale is comprised of eight questions focused on knowledge related specifically to the QPR training and six questions focused on suicide risk factors. Sample items include: (1) Which of the following is not a warning sign of suicide…? and (2) the number one contributing cause of suicide it…? Responses were scored as correct or incorrect, and the total score is summed from the percentage of correctly answered questions. Items for this scale were reviewed by an expert panel for content validity (Wyman et al., 2008) and have been previously shown to differentiate between those who have taken the training and those who have not (Quinnett, 2013).

Attitudes

The researchers used the ASP Scale (Herron, Ticehurst, Appleby, Perry, & Cordingley, 2001) to assess stigma regarding suicide and suicide prevention. The ASP has 14 items that are answered using a 5-point rating scale ranging from 1 (strongly disagree) to 5 (strongly agree). Examples of items are (1) suicide is not my responsibility and (2) I fell defensive when people offer advice about suicide prevention. The researchers recoded individual responses and then summed the scale; higher scores indicate more negative ASP. The scale has demonstrated acceptable reliability (Cronbach’s α = .77 [original] and .75 [observed]) and evidence supporting validity.

Self-efficacy

Self-efficacy was measured using the Efficacy to Perform Gatekeeper Role Scale (Wyman et al., 2008). This scale is comprised of 7 items designed to assess perceived efficacy to perform suicide prevention activities. Participants are instructed to respond to items using a 7-point rating scale ranging from 1 (strongly disagree) to 7 (strongly agree). Scale items include (1) I am aware of the warning signs for suicide and (2) I can make appropriate referrals. A total score is computed as the mean of all items after recoding responses per the developers’ instructions. Higher scores indicate greater perceived efficacy; the scale has acceptable reliability (Cronbach’s α = .80 [original] and .72 [observed]).

Reluctance

The Reluctance to Engage With Suicidal Clients Scale is comprised of 8 items designed to address a participants’ reluctance to engage in suicide prevention activities (Wyman et al., 2008). Items are scored on a 7-point scale, ranging from 1 (strongly disagree) to 7 (strongly agree). One scale item is I am too busy to participate in suicide prevention activities. A total score is computed as the mean of all items after recoding responses per the developers’ instructions; higher scores indicate greater reluctance. Internal consistency meets the minimum requirement (Cronbach’s α = .68 [original] and .63 [observed]).

Gatekeeper behaviors

The researchers used the Asking Clients about Suicide in Response to Warning Signs to assess suicide risk assessment behaviors. The measure is a 6-item scale that measures participants’ behaviors regarding asking clients about suicide and responding to client suicide behavior when signs and symptoms of depression were present (Wyman et al., 2008). Responses were coded using a 5-point rating scale ranging from 1 (never) to 5 (always), with higher total scores suggesting more frequent use of suicide risk assessment behaviors. Internal consistency for the scale was high (Cronbach’s α = .94 [original] and .83 [observed]).

Data Analysis

Mplus statistical software (v.7; Muthén & Muthén, 1995–2011) was used for all analyses. Paired-samples t-tests were used to assess changes in knowledge, attitudes, self-efficacy, reluctance, and use of gatekeeper behaviors following QPR training and hypothesized direction of effects based on previous research and theory. A repeated measures path analysis was used to estimate direct and indirect effects of training constructs on posttraining use of gatekeeper behaviors. Path analysis is an extension of multiple regression that allows the researcher to test for the presence of hypothesized mediating variables. The path analysis approach is a subcategory of structural equation modeling that only uses observed variable and allows the researcher to test for both direct relationships and indirect (mediated) relationships over time. The analysis is based on the fit between the hypothesized relationships in the model and the observed relationships between the variables in the data.

As illustrated in Figure 1, hours of previous training, pretraining knowledge, attitudes, self-efficacy, and reluctance were tested for a direct effect on pretraining gatekeeper behaviors. Posttraining knowledge, attitudes, self-efficacy, and reluctance were tested as mediators of the development and demonstration of gatekeeper behaviors from pretraining to a 4-month follow-up. One-tailed tests were used in the analyses with directionality based on previous research. The focus of the analysis was magnitude of any observed effects in regard to practical significance. Results are provided for sample-based point estimates, with observed effect sizes and confidence intervals (Greenland et al., 2016; Wasserstein & Lazarm, 2016) and p values. Guidelines provided by Hu and Bentler (1999) and Kline (2005) were used as model fit criteria (root mean square error of approximation [RMSEA] < .05, comparative fit index [CFI]/Tucker–Lewis index [TLI] > .90, standardized root mean square residual [SRMR] < .08). Full information maximum likelihood estimations were used to handle missing data at the 4-month follow-up for four participants. A priori power analyses were not calculated, as this was a secondary data analysis; however, post hoc power analyses were conducted and are discussed in more detail below.

Results

Univariate (mean, standard deviation, and range) statistics and bivariate correlations are provided for all study variables in Table 1. The assumption of normality was met for all model variables except for gatekeeper behaviors at both time points. Skewness (S) and kurtosis (K) were minimal for both distributions (S = .52/K = −1.45 and S = −.13/K = −1.54, respectively). There was no evidence that the assumption of independence of pairs was violated. The assumption of nonmulticollinearity was upheld based on all bivariate correlations <.62. The repeated measures path model included direct and indirect effects (see Figure 1), and the results of the analysis suggest acceptable fit of the model to the data based on standard cutoff values (RMSEA = .03, CFI = .98, TLI = .96, SRMR = .08). Means, standard deviations, 95% confidence interval for difference in pre- and posttraining means, and Cohen’s d effect sizes are reported below for paired-samples t-tests. Standardized point estimates for all variables (STDXY values) and 95% confidence intervals for direct and indirect effects are reported below and in Figure 1. The model is presented in its original design; no modifications were made to the model. In order to make the figure easier to read, the pathways between pre- and posttraining scores for knowledge, attitudes, self-efficacy, and reluctance are not included in the figure but are reported below in text.

Table 1.
Correlations, Means, and Standard Deviations.

Hours GK Behaviors (Pre) GK Behaviors (FU) Knowledge (Pre) Knowledge (Post) Attitudes (Pre) Attitudes (Post) Self-Efficacy (Pre) Self-Efficacy (Post) Reluctance (Pre) Reluctance (Post)

Hours 1.00

GK Behaviors (pre) .39* 1.00

GK Behaviors (FU) .22 .39* 1.00

Knowledge (pre) −.10 .16 .04 1.00

Knowledge (post) .11 .29 .32* .12 1.00

Attitudes (pre) .11 −.23 −.24 −.21 .02 1.00

Attitudes (post) −.17 −.24 −.12 −.10 −.08 .59** 1.00

Self-Efficacy (pre) .38* .50 .45 .21 .14 −.35* −.22 1.00

Self-Efficacy (post) .16 .37* .36* .32* .04 −.32* −.42 .64 1.00

Reluctance (pre) −.08 −.31* −.06 −.43 .12 .55 .28 −.46** −.37* 1.00

Reluctance (post) .04 −.32* −.09 −.51 .03 .27 .26 −.22 −.29 .62 1.00

Mean (SD) 8.62 (6.17) 2.38 (1.59) 2.90 (1.52) 0.75 (0.13) 0.86 (0.10) 1.92 (0.25) 1.64 (0.28) 3.98 (0.84) 4.75 (0.70) 2.42 (0.64) 1.95 (0.66)

Range 0–24 1–5 1–5 0–1 0–1 1–5 1–5 1–7 1–7 1–7 1–7

Statistical significance indicated as p < .05. p < .01. p < .001.

FU: follow-up.

Pre/Posttraining Changes

Changes were observed in every pair of constructs. Knowledge scores were 75% correct (SD = 0.13) at pretest compared to 85% correct (SD = 0.10) at posttest (95% CI for difference: [−0.15, −0.04]; Cohen’s d = .86, p < .001). Notably, correct responses for 4 items increased by 25% or more; 1 item addressed skills needed to be effective in suicide prevention and the other 3 questions were related to warning signs. Mean ASP scores were 1.93 (SD = 0.25) at pretest and 1.66 (SD = 0.27) at posttest (95% CI for difference: [0.18, 0.37]; Cohen’s d = 1.04, p < .001). All items on the scale showed improvement, and the largest change (.91 points on a 5-point scale) occurred for the item asking respondents if they thought someone serious about suicide would actually tell anybody.

Scores for self-efficacy changed from 3.93 (SD = 0.82) to 4.71 (SD = 0.68; 95% CI for difference: [−1.01, −.53]; Cohen’s d = .94, p < .001). Changes on individual items ranged from 1.3 to 2.1 points (on a 7-point scale), with the largest increases seen for the items on responding appropriately to disclosures of suicide ideation and identifying risk indicators based on client history. Means scores for reluctance were 2.49 (SD = 0.64) at pretest and 1.94 (SD = 0.67) at posttest (95% CI for difference: [0.34, 0.74]; Cohen’s d = .84, p < .001). The largest change was for being responsible for discussing suicide with clients (1-point drop in reluctance on 7-point scale). The mean pretest score for gatekeeper behaviors was 2.31 (SD = 1.53) compared to the mean 4-month follow-up score of 2.96 (SD = 1.58; 95% CI for difference: [−1.25, −0.04]; Cohen’s d = .42, p = .045). Three items increase by one or more points (on a 5-point scale); all 3 items involved asking a client about suicide when the client had a history of depression or trauma or if the student had a feeling “something was wrong.” Based on Cohen’s d, approximately 80% of scores on the posttest were improved over the mean score of the pretest for knowledge and reluctance. Effects for reluctance and self-efficacy suggest 84% of posttest scores were higher than the pretest means. Cohen’s d for gatekeeper behaviors translates to roughly 66% of participants scoring higher at 4 months than the mean of pretraining scores.

Direct Effects on Pretraining Gatekeeper Behaviors

Pretraining self-efficacy had the strongest relationship with pretraining gatekeeper behaviors (β = .32, 95% CI [0.03, 0.60], p = .02), followed by the number of hours of previous suicide intervention training (β = .29, 95% CI [0.01, 0.57], p = .02), both indicating moderate effect sizes in the sample for the potential impact of self-efficacy and training on gatekeeper behaviors prior to completing the QPR training intervention. Very weak direct effects were observed for knowledge (β = .07, 95% CI [−0.32, 0.40], p = .35), attitudes (β = −.09, 95% CI [−0.44, 0.24], p = .28), and reluctance (β = −.06, 95% CI [−0.45, 0.34], p = .39), suggesting no relationships with pretraining gatekeeper behaviors in the current sample. A large effect was observed for this portion of the model (R ² = .33, p = .02) with one third of the variance in pretraining gatekeeper behaviors associated with variance in the combined set of pretraining predictors.

Direct Effects on Posttraining Constructs

Strong relationships were observed between pretraining and posttraining scores for attitudes (β = .57, 95% CI [0.36, 0.78], p < .001), self-efficacy (β = .62, 95% CI [0.37, 0.87], p < .001), and reluctance (β = .57, 95% CI [0.39, 0.75], p < .01), but not for knowledge (β = .08, 95% CI [−0.27, 0.43], p = .32) in the sample. With the exception of knowledge, these results indicate that higher pretraining scores were associated with higher posttraining scores on these constructs; for knowledge, there appears to be no relationship between pre- and posttraining scores. The analyses yielded small-to-moderate estimates between pretraining gatekeeper behaviors and posttraining scores on knowledge (β = .27, 95% CI [−0.05, 0.61], p = .05), attitudes (β = −.18, 95% CI [−0.41, 0.34], p = .26), self-efficacy (β = .06, 95% CI [−0.22, 0.33], p = .34), and reluctance (β = −.15, 95% CI [−0.41, 0.12], p = 14) in the sample, but large p values and wide confidence intervals around β = 0 for attitudes, efficacy, and reluctance minimize any practical interpretation.

Direct and Mediated Effects on Posttraining Gatekeeper Behaviors

The results of the analyses yielded no support for the mediation of changes in gatekeeper behaviors by knowledge (β = .07, 95% CI [−0.06, 0.19], p = .16), attitudes (β = −.01, 95% CI [−0.06, 0.05], p = .39), self-efficacy (β = .06, 95% CI [−0.07, 0.10], p = .35), or reluctance (β = −.01, 95% CI [−0.05, 0.03], p = .37). There was, however, a moderate direct effect of posttraining self-efficacy scores on the use of gatekeeper behaviors at the 4-month follow-up observation (β = .30, 95% CI [−0.04, 0.75], p = .04), suggesting that posttraining self-efficacy is associated with engagement in gatekeeper intervention skills. Similar point estimates were observed for knowledge (β = .24, 95% CI [−0.08, 0.55], p = .07) and pretraining gatekeeper behaviors (β = .24, 95% CI [−0.17, 0.65], p = .12), but the confidence intervals for these two estimates provide only minimal support for a true effect (β ≠ 0). Small effects were observed for reluctance (β = .04, 95% CI [−0.21, 0.13], p = .36) and attitudes (β = .07, 95% CI [−0.28, 0.36], p = .35), again with no evidence of an effect based on large p values and wide confidence intervals. A large effect size for the overall model was found (R ² = .28, p = .03), suggesting that about one third of the variance in use of posttraining gatekeeper suicide intervention skills is associated with variances in scores in the full model.

Discussion and Applications to Practice

Certain characteristics of the original study and current analysis should be considered when interpreting the results of this study. The original study used a strong design including random sampling and random assignment, thereby increasing both internal and external validity and supporting hypothesis testing within the population of MSW students in the program. Generalizations beyond that particular program should be tempered. The researcher sought to examine the evidence on the direction and magnitude of effects associated with the use of a specific suicide intervention training curriculum (QPR), with a specific sample of advanced MSW students in order to inform a broader causal model of training that is being developed from multiple samples using different training curriculum and different populations.

One major limitation in the study is the small sample size, and it impacts several aspects of the model. The analysis was limited to the training outcomes and the inclusion of hours of previous training as a covariate. Arguably, there are a number of additional covariates that may have bearing on the model, but additional covariates were included in the model based on three considerations. (1) The statistical power of the analysis was already limited, given the small sample as noted below, and the addition of more parameters to be estimated would further reduce it. (2) There is no empirical evidence to suggest that relationships exist between the study variables and personal demographics (i.e., gender, race). (3) Potentially important variables such as being in a clinical placement or providing direct clinical practice had highly disproportionate subgroups.

Post hoc power analyses were conducted for direct effects (power = .47–.69), mediated effects (power = .15–.30), and model fit (RMSEA, power = .53) and indicated that the study was underpowered for testing all parameters. Additionally, the small sample size is likely to impact the stability of the parameter estimates in replication analyses. It should be noted however that observed estimates for mediated paths are consistent with those reported in Osteen et al.’s (2016) mediation analysis of the Recognizing and Responding to Suicide Risk (RRSR) training. These consistencies in magnitude of estimates may offer useful starting values in future research.

Moderate to large improvements in the five training outcomes were observed from pretraining to posttraining. Arguably, the most important outcome of any training is the acquisition and use of learned behaviors, and improvements were observed in the study sample for all gatekeeper behaviors. The largest increases were for items pertaining to asking clients about suicide thoughts when certain risk factors were present; this is a positive result, given that a primary behavioral goal of the training is to ask questions about suicide (Quinnett, 1995).

SCT suggests that attitudes and self-efficacy impact changes in use of learned behaviors (Bandura, 1997; Gibb et al., 2010; Ozer & Bandura, 1990; Valente & Saunders, 2004), and the author (Osteen et al., 2016) found empirical support for this model. Measures of knowledge and reluctance were not available in Osteen et al.’s (2016) and Lee et al.’s (2016) previous repeated measures modeling studies, and the current data provided the opportunity to include them and explore different pathways to skill development and use.

It is a logical assumption that knowledge of suicide prevention and warning signs is required for use of gatekeeper behaviors involving assessment of risk based on observed warning signs, and yet level of knowledge was not predictive of engagement in intervention behaviors before or after training in this sample even though previous studies have shown correlations between knowledge and behaviors (Osteen, Jacobson, & Sharpe, 2014). Results from this study also suggest a large effect for the increase in knowledge scores after training, a finding consistent with the majority of suicide intervention training studies (e.g., Pisani et al., 2011; Smith, Silva, Covington, & Joiner, 2014; Tompkins et al., 2010), but there was no observed correlation between pretraining and posttraining knowledge scores. One interpretation of these results is that even though knowledge of a skill is requisite for engagement in that skill, knowledge does not equate to use, meaning that just because an individual knows how to do something does not mean that they will actually do it. A second interpretation is that although the lack of correlation between pre- and posttraining knowledge scores is counterintuitive, it may be that the training results in a similar knowledge base for most participants regardless of their knowledge before the training. The absence of a relationship between knowledge and behaviors is further emphasized by the lack of a mediated or indirect effect.

There is both clinical and theoretical reasoning that individuals with more positive attitudes toward suicide and suicide intervention would use intervention behaviors more frequently, but the results of the current study found little support for this idea. Although there was an increase in posttraining attitudes, attitudes did not predict gatekeeper behaviors at either time point nor was there evidence of a mediated or indirect effect. The literature is full of inconsistent findings regarding attitudes toward suicide and suicide intervention with some studies demonstrating improved attitudes (e.g., Jacobson et al., 2012; Herron et al., 2001; Matthieu & Swensen, 2014) and others finding no immediate and/or sustained improvements after training (e.g., Botega et al., 2005; Pisani et al., 2011; Tomkins, et al., 2010); in addition, the current findings are opposite to those found by Osteen et al. (2016) in similar models with other training curriculum. One explanation posited in previous research is that attitudes tend to be positive in samples of mental health professionals even before training; in such circumstances, the magnitude of any improvement will be attenuated thereby minimizing any shared variation with posttraining behaviors. Another possibility is limitations, such as low internal consistency, with the measure itself. This should not be interpreted to say that attitudes are not important but instead that they are not well measured or that their role is not well understood.

Self-efficacy was the only variable in the model to be associated with gatekeeper behaviors both pretraining and posttraining. Direct effects of self-efficacy on behaviors support the idea that as confidence in successfully engaging in a behavior increases, so does actual engagement in that behavior. The observed effects in the current study are consistent with findings in other studies relating self-efficacy to behavior (Coleman & Del Quest, 2015; Conner et al., 2013; Cross et al., 2011; Lee, Osteen, & Frey, 2016; Osteen et al., 2016). The author found very similar results using a repeated measures model such as that presented in the current study, but with data from a training curriculum different than the QPR program and with a much more diverse sample of mental health professionals (Osteen et al., 2016). The author also found evidence of self-efficacy as a mediator of engagement in clinical intervention behaviors (Osteen et al., 2016) but that finding was not replicated in the current study.

Studies demonstrating improvement in self-efficacy often use training curriculum that includes an experiential component in which participants engage in an activity requiring them to use the intervention behaviors they are learning (e.g., QPR, RRSR, and Applied Suicide Intervention Skills Training). It is a reasonable interpretation to think that “practicing” a behavior increases self-efficacy that in turn increases actual use of the behavior when appropriate. Shockley and Albright (2014) and Pasco, Wallack, Sartin, and Dayton (2012) examined the relationship between experiential activity and gatekeeper behaviors and found evidence that these activities impacted preparedness, efficacy, likelihood, and use of intervention behaviors. It would be interesting to see if training that incorporates experiential activities results in higher levels of posttraining self-efficacy as compared to training that does not include these activities.

This study yielded no evidence that participants’ reluctance to engage in gatekeeper behaviors had either a direct or a mediated effect on use of intervention behaviors. One could interpret this positively as meaning that students used gatekeeper behaviors even when they were reluctant to do so. Ajzen’s (1991) “theory of planned behavior” has been used extensively in the training literature and provides one possible explanation in that intervening with a suicidal individual is viewed as a positive action and constructive behavior that overall benefits the individual who is suicidal. It is an expected norm in practice to intervene despite reluctance (or negative attitudes) toward doing so.

It is important to train social workers to identify warning signs and risk factors, assess risk for suicidal behavior, and respond appropriately when encountering a potentially suicidal individual. According to the National Association of Social Workers (n.d.), social workers are the nation’s largest group of mental health providers, and there are more clinically trained social workers than psychologists, psychiatrists, and psychiatric nurses combined. Within this context, it is highly probable that clinical social workers will encounter clients with suicidal thoughts and behaviors. Previous studies estimate that one third of social workers will experience the death of a client by suicide and more than 50% of social workers will experience clients with nonfatal self-injurious behavior (Feldman & Freedenthal, 2006; Ting, Jacobson, & Sanders, 2011).

Research suggests that a variety of suicide training interventions are effective for achieving these goals (Osteen et al., 2014), and suicide intervention training should be included in MSW education and continuing education for clinical social workers. However, the field lacks empirically based causal models of how and why these interventions work. Understanding these models may help researchers and practitioners develop more effective training as well as maximize the use of limited resources. For example, there is a lot of variability in training time with commonly used interventions ranging from 1 hr to 2 days; similarly, some training is online whereas others are face-to-face or hybrid models. If self-efficacy is truly a driving force for development and use of intervention behaviors, then perhaps training should devote more time to experiential exercises and less to didactic techniques. Alternatively, if knowledge is a prerequisite for intervention behaviors, as opposed to a causal mechanism, then there may be options for building a knowledge base external to the training itself. Researchers are encouraged to build mediation and moderation analyses into their repeated measures studies of suicide interventions and to disseminate these finding, so that the field can continue to pursue this important step in reducing deaths by suicide.

	Hours	GK Behaviors (Pre)	GK Behaviors (FU)	Knowledge (Pre)	Knowledge (Post)	Attitudes (Pre)	Attitudes (Post)	Self-Efficacy (Pre)	Self-Efficacy (Post)	Reluctance (Pre)	Reluctance (Post)
Hours	1.00
GK Behaviors (pre)	.39*	1.00
GK Behaviors (FU)	.22	.39*	1.00
Knowledge (pre)	−.10	.16	.04	1.00
Knowledge (post)	.11	.29	.32*	.12	1.00
Attitudes (pre)	.11	−.23	−.24	−.21	.02	1.00
Attitudes (post)	−.17	−.24	−.12	−.10	−.08	.59**	1.00
Self-Efficacy (pre)	.38*	.50**	.45**	.21	.14	−.35*	−.22	1.00
Self-Efficacy (post)	.16	.37*	.36*	.32*	.04	−.32*	−.42**	.64**	1.00
Reluctance (pre)	−.08	−.31*	−.06	−.43**	.12	.55**	.28	−.46**	−.37*	1.00
Reluctance (post)	.04	−.32*	−.09	−.51**	.03	.27	.26	−.22	−.29	.62**	1.00
Mean (SD)	8.62 (6.17)	2.38 (1.59)	2.90 (1.52)	0.75 (0.13)	0.86 (0.10)	1.92 (0.25)	1.64 (0.28)	3.98 (0.84)	4.75 (0.70)	2.42 (0.64)	1.95 (0.66)
Range	0–24	1–5	1–5	0–1	0–1	1–5	1–5	1–7	1–7	1–7	1–7

Footnotes

Acknowledgments

The author would like to acknowledge and thank Dr. Jodi Frey at the University of Maryland School of Social Work as the principle investigator of the original study.

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

References

Ajzen

(1991). The theory of planned behavior. Organizational Behavior and Human Design Process, 50, 179–211.

Bandura

(1997). Self-efficacy: The exercise of control. New York, NY: Freeman.

Beidas

R. S.

Kendall

P. C.

(2010). Training therapists in evidence-based practice: A critical review of studies from a systems-contextual perspective. Clinical Psychology, 17, 1–30. doi:10.1111/J.1468-2850.2009.01187.X

Botega

N. J.

Reginato

D. G.

de Silva

S. V.

Cais

C. F.

Rapeli

C. B.

Mauro

M. L.

… Steffanello

(2005). Nursing personnel attitudes towards suicide: The development of a measure scale. Revists Brasileira de Psiquiatria, 27, 315–318. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.1590/S1516-44462005000400011

Centers for Disease Control and Prevention. (2015). Web-based Injury Statistics Query and Reporting System (WISQARS) [Online]. (2013, 2011) National Center for Injury Prevention and Control, CDC (producer). Retrieved July 12, 2016, from http://www.cdc.gov/injury/wisqars/index.html

Coleman

Del Quest

(2015). Science from evaluation: Testing hypotheses about differential effects of three youth-focused suicide prevention trainings. Social Work in Public Health, 30, 117–128. doi:10.1080/19371918.2014.938397

Conner

K. R.

Wood

Pisani

A. R.

Kemp

(2013). Evaluation of a suicide prevention training curriculum for substance abuse treatment providers based on Treatment Improvement Protocol Number 50. Journal of Substance Abuse Treatment, 44, 15–16. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.1016/j.jsat.2012.01.008

Cross

W. F.

Seaburn

Gibbs

Schmeelk-Cone

White

A. M.

Caine

E. D.

(2011). Does practice make perfect? A randomized control trial of behavioral rehearsal on suicide prevention gatekeeper skills. Journal of Primary Prevention, 32, 195–211.

Feldman

B. N.

Freedenthal

(2006). Social work education in suicide intervention and prevention: An unmet need. Suicide and Life-Threatening Behavior, 36, 467–480. doi:10.1521/suli.2006.36.4.467

10.

Gibb

S. J.

Beautrais

A. L.

Surgenor

L. J.

(2010). Health-care staff attitudes towards self-harm patients. Australian and New Zealand Journal of Psychiatry, 44, 713–720.

11.

Greenland

Senn

S. J.

Rothman

K. J.

Carlin

J. B.

Poole

Goodman

S. N.

Altman

D. G.

(2016). Statistical tests, P values, confidence intervals, and power: A guide to misinterpretations. European Journal of Epidemiology, 31, 337–350. doi:10.1007/s10654-016-0149-3

12.

Herron

Ticehurst

Appleby

Perry

Cordingley

(2001). Attitudes toward suicide prevention in front-line health staff. Suicide and Life-Threatening Behavior, 31, 342–347. Retrieved from https://dx-doi-org.web.bisu.edu.cn/10.1521/suli.31.3.342.24252

13.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling, 6, 1–55.

14.

Jacobson

J. M.

Osteen

P. J.

Jones

Berman

(2012). Evaluation of the recognizing and responding to suicide risk training. Suicide and Life-Threatening Behavior. doi:10.1111/j.1943-278x.2012.00105.x

15.

Jacobson

J. M.

Osteen

P. J.

Sharpe

T. L.

Pastoor

J. B.

(2012). Randomized trial of suicide gatekeeper training for social work students. Research on Social Work Practice, 22, 270–281. doi:10.1111/j.1943-278X.2012.00105.x

16.

Kline

R. B.

(2005). Principles and practices of structural equation modeling (2nd ed.). New York, NY: Guilford Press.

17.

Lee

S. J.

Osteen

P. J.

Frey

J. J.

(2016). Predicting changes in health professionals’ clinical practice behaviors for recognizing and responding to suicide risk. Journal of the Society for Social Work and Research, 7, 23–41. doi:10.1086/685037

18.

Maibach

Murphy

D. A.

(1995). Self-efficacy in health promotion research and practice: Conceptualization and measurement. Health Education Research, 10, 37–50.

19.

Matthieu

M. M.

Swensen

A. B.

(2014). Suicide prevention training program for gatekeepers working in community hospice care settings. Journal of Social Work End-of-Life & Palliative Care, 10, 95–105. doi:10.1080/15524256.2013.877865

20.

Muthén

L. K.

Muthén

B. O.

(1998–2011). Mplus user’s guide (6th ed.). Los Angeles, CA: Author.

21.

National Action Alliance for Suicide Prevention. (2014). Research Prioritization Task Force. A prioritized research agenda for suicide prevention and action plan to save lives. Rockville, MD: Author.

22.

National Association of Social Workers. (n.d.). Social work profession. Retrieved October 23, 2016, from http://www.socialworkers.org/pressroom/features/general/profession.asp

23.

Osteen

P. J.

Frey

J. J.

(2014). Advancing training to identify, intervene, and follow-up with individuals at risk for suicide through research. American Journal of Preventative Medicine, 47, S216–S221. doi:10.1016/j.amepre.2014.05.033

24.

Osteen

P. J.

Frey

J. J.

Woods

Shipe

(2016). Latent variable modeling of direct and indirect effects on clinical skills and behaviors following suicide intervention training. Suicide and Life-Threatening Behavior. doi:10.1111/sltb.12288

25.

Osteen

P. J.

Jacobson

J. M.

Sharpe

T. L.

(2014). Suicide prevention in social work education: How prepared are social workers to help clients at risk for suicide? Journal of Social Work Education, 50, 349–364.

26.

Ozer

E. M.

Bandura

(1990). Mechanisms governing empowerment effects: A self-efficacy analysis. Journal of Personality and Social Psychology, 58, 472–486. doi:10.1037/0022-3514.58.3.472

27.

Quinnett

(1995). QPR: Ask a question, save a life. Spokane, WA: QPR Institute and Suicide Awareness/Voices of Education.

28.

Quinnett

(2013). Research and theory. The QPR Institute. Retrieved October 23, 2016, from www.qprinstitute.com

29.

Pasco

Wallack

Sartin

R. M.

Dayton

(2012). The impact of experiential exercises on communication and relational skills in a suicide prevention gatekeeper-training program for college resident advisors. Journal of American College Health, 60, 134–140. doi:10.1080/07448481.2011.623489

30.

Pisani

A. R.

Cross

W. F.

Gould

M. S.

(2011). The assessment and management of suicide risk: State of workshop education. Suicide and Life-Threatening Skill, 41, 255–276. doi:10.1111/j.1943-278X.2011.00026.x

31.

Shockley

Albright

(2014, 4). Building suicide prevention skills using virtual Humans: A meta-analysis. Paper presented at the annual meeting of the American Association of Suicidology, Los Angeles, CA.

32.

Smith

A. R.

Silva

Covington

D. W.

Joiner

T. E.

(2014). Assessment of suicide-related knowledge and skills among health professionals. Health Psychology, 33, 110–119. doi:10.1037/a0031062

33.

Suicide Prevention Resource Center. Zero suicide in health and behavioral health care. Retrieved July 12, 2016, from http://zerosuicide.sprc.org

34.

Ting

Jacobson

J. M

Sanders

(2011). Current levels of perceived stress among mental health social workers who work with suicidal clients. Social Work, 56, 327–336.

35.

Tompkins

T. L.

Witt

Abraibesh

(2010). Does a gatekeeper suicide prevention program work in a school setting? Evaluating training outcomes and moderators of effectiveness. Suicide & Life-Threatening Behavior, 40, 506–515.

36.

Valente

Saunders

J. M.

(2004). Barriers to suicide risk management in clinical practice: A national survey of oncology nurses. Issues in Mental Health Nursing, 25, 629–648.

37.

Wasserstein

R. L.

Lazarm

N. A.

(2016). The ASA’s statement on p-values: Context, process, and purpose. The American Statistician, 70, 129–133. doi:10.1080/00031305.2016.1154108

38.

Wyman

P. A.

Inman

Guo

Brown

C. H.

Crosee

Schmeelk-Cone

… Pena

J. B.

(2008). Randomized trial of a gatekeeper program for suicide prevention: 1-year impact on secondary school staff. Journal of Consulting and Clinical Psychology, 76, 105–155. doi:10.1037/0022-006X.76.1.104