Development and implementation of a self-directed violence prevention training program for correctional behavioral health providers: a clinical trial study protocol

Abstract

Background

Self-directed violence (SDV) comprises both suicide and self-injury and represents a pressing problem among incarcerated persons. Negative impacts of SDV in correctional settings also extend to behavioral health clinicians (BHCs) (e.g., job turnover). Correctional SDV risk assessment and management standards include staff training as part of the comprehensive approach. The Core Competency Model for Corrections (CCM-C) is a novel, evidence-informed training program for BHCs covering both clinician self-management and clinical care skills.

Methods

This pilot trial is a type 3 hybrid implementation–effectiveness approach. It will employ a wait-list control sequential cross-over design. Participants (N = 50–100) will be BHCs employed by the North Carolina Department of Adult Corrections. Following stratification for years of clinical experience, BHCs will be randomly assigned to (1) a training group that receives CCM-C immediately and (2) a wait-list control receiving CCM-C approximately 6 weeks later. Electronically administrated survey evaluation will occur across baseline and two follow-up (i.e., 2 weeks after each training session) time points.

Discussion

The primary outcome is feasibility assessed through collaboration with a Corrections Advisory Panel and feedback from BHCs. Secondary effectiveness outcomes that will be evaluated over time include SDV-related knowledge, attitudes, stigma, and intent to use training content. We will examine a tertiary outcome, namely compassion fatigue. Clinical trial limitations and impacts are discussed.

Trial registration

Clinicaltrials.gov, NCT06359574. This study was registered on 04/05/2024.

Background

Self-directed violence (SDV) comprises both suicide and self-injury [1]. Incarcerated persons are at an elevated risk for SDV compared to the general population [2]. SDV in correctional settings represents a public health problem, as incarcerated persons may experience a range of negative health outcomes, and handling SDV occupies a considerable amount of correctional officer, correctional behavioral health clinician (BHC), and administrative staff time and resources. Furthermore, a need for better training, implementation, and use of evidence-based clinical practices is necessary [3]. Despite the pressing challenges associated with SDV in correctional settings, there is a lack of empirically evaluated evidence-based correctional SDV prevention training programs for BHCs.

SDV in correctional settings: scope and impacts

Suicide remains a leading cause of death among state prisoners as in 2019; 8.1% of correctional deaths were attributed to suicide [4]. The rate of suicide is higher among incarcerated individuals compared to U.S. residents [4]. Of all state and federal prisons, 19% had at least one suicide in 2019, with 9% of all jails reporting at least one suicide [4]. Self-injury was more common, with a national survey indicating that self-injury was a near-universal behavior across prisons (i.e., 98% of prisons had persons engaging in self-injurious acts), but with a prevalence rate of 2.4% of the surveyed prison population [5].

Negative impacts of correctional SDV exist on the system and provider levels. For example, estimates of economic impact show that correctional mental health and related services cost more than $1 billion annually [6]. In terms of providers, there is a significantly higher turnover in correctional staff than in the general workforce, with correctional nurse turnover in a 3-year period being 20% compared to 13% of the national nurse workforce [7]. Research has also indicated a higher prevalence of mental health disorders such as post-traumatic stress disorder (PTSD) and major depressive disorder among correctional workers relative to the general public [8]. These problems may be rooted in part by traumatic incidents (e.g., witnessing the death of an incarcerated individual) [7, 8].

Correctional SDV risk assessment and management practices and training

The American Correctional Association (ACA) [9] and the National Commission on Correctional Health Care [10] both establish standards for health services in correctional facilities, including guidelines for SDV prevention. These standards comprise screening and assessment of individuals at risk of suicide during intake or other mental health assessments, assessment of correctional staff training in identifying and responding to signs of suicidal behavior, assessment of timely and effective access to mental health services, and guidelines for minimizing the risk of SDV in correctional environments [10]. However, due to the flexibility of the ACA and NCCHC standards, the policies and implementation of guidelines may vary among facilities [11]. A recent example of best practice following these guidelines is the development and initial validation of the Self-Injury Risk Assessment Protocol for Corrections (SIRAP-C). Resulting from a corrections–academic partnership to meet the correctional standards for managing and preventing SDV, the SIRAP-C was developed to provide an evidence-based structured approach to risk determination and treatment of SDV in correctional settings [12].

The NCCHC’s current training standards for SDV assessment and management recommend that BHCs receive an initial in-depth mental health orientation that is reviewed once every 2 years or as needed and a minimum of 12 h of continuing professional education per year [13]. Speaking to the gap in existing trainings, the suicide-specific trainings that exist for professionals in correctional settings, such as the Question Persuade Refer [14, 15] Model for corrections, their primary emphasis is on correctional officers; the central training focus concerns basic skills such as merely identifying basic warning signs for suicidal behavior [3]. Adequate training tailored to BHCs to address SDV is the next needed step in addressing correctional SDV. The Core Competency Model (CCM) [16, 17] and its adapted correction version [2] provide a framework to address SDV risk assessment and management in correctional settings.

The core competency model: origins and evidence

The CCM [16, 17] is an evidence-based training approach for BHCs. The CCM is grounded in the social cognitive model of suicide prevention training [18]. This model posits that training, in this instance in ten core clinical care and clinician self-management skills, will result in improved BHC suicide-related knowledge, attitudes, and perceived skill mastery. The ten core competencies are (1) manage personal attitudes and reactions to suicide, (2) maintain a collaborative, empathic stance toward the client, (3) elicit evidence-based risk and protective factors, (4) focus on current suicide plan and intent of suicidal ideation, (5) determine the risk level, (6) enact a collaborative evidence-based treatment plan, (7) notify and involve other persons, (8) document risk, plan, and reasoning for clinical decisions, (9) know the law concerning suicide, and (10) engage in debriefing and self-care. Two important CCM principles inform the present trial. First, CCM training approaches involve a combination of psycho-educational content, self-assessment tools, resource provision, and interactive exercises. Second, the CCM is intentionally flexible, in that the training content and delivery can be tailored to the needs of a treatment setting or population. A fundamental part of the CCM training approach is the development and validation of the Suicide Competency Assessment Form (SCAF) [16, 19]. The SCAF assesses mastery or self-efficacy regarding the ten core competencies and can be used as a total score or individual item ratings. The SCAF is flexible in its use in terms of self-report or observer (e.g., a clinical supervisor) versions.

The CCM has been applied primarily in community mental health, general mental health provider, and university counseling center contexts to date [17, 20]. Numerous studies to date show positive, meaningfully sized impacts on BHC suicide risk assessment and management-related knowledge, attitudes, and perceived skill mastery [3, 17, 20,21,22]. For instance, the evaluation of a virtual self-paced CCM training yielded reductions in negative beliefs about suicidal patients, as well as positive gains in provider suicide-focused knowledge and self-efficacy in implementing the ten competencies [17]. Training has also demonstrated objective improvements in appropriate clinical responses to persons experiencing suicidal thoughts or behaviors [3], suicide risk recognition in response to standardized clinical vignettes [17], and use of screening tools in clinical practice [20]. Overall, the CCM provides a foundation for the development of an SDV risk assessment and management training program for correctional settings.

Core Competency Model for Corrections (CCM-C)

An adapted version of the CCM was developed and named the Core Competency Model for Corrections (CCM-C) [2]. Specifically, the research outlined the rationale, need for adaptation, and a sample training curriculum for mental health providers working in correctional settings. Owing to the high rates and negative impacts of SDV in correctional settings, the CCM-C adapts the social cognitive model of suicide prevention training [18] and CCM approach in several ways. First, the sole focus on suicide is expanded to include self-injury throughout the model and training curriculum (see Fig. 1). For example, mediating mechanisms of action explaining training impacts on intent and actual use of skills all focus on SDV (e.g., SDV risk assessment and management knowledge, attitudes, skills). Second, contextual factors such as the type of correctional facility need to be factored into training content and trainee selection. Third, the phrasing of competencies and associated SCAF items are reframed to address SDV. Finally, although some general CCM content remains the same, much of the psycho-educational information, clinical resource materials, and interactive training are largely adapted to focus on evidence-based content in the correctional SDV literature [23,24,25]. Moreover, case law and standards [26,27,28] specific to SDV management in correctional settings are included in that respective competency. More details regarding the CCM-C training are provided in the study protocol methods below.

Fig. 1

CCM-C social cognitive training model. SDV self-directed violence, CCM-C Core Competency Model for Corrections

Core competency model for corrections project aims and hypotheses

The overall purpose of this study is to conduct a pilot trial focused on the feasibility and preliminary effectiveness evaluation of the CCM-C [2] among a sample of state correctional BHCs in North Carolina. The CCM-C is an evidence-based training program aimed at training BHCs in SDV prevention skills. Using both quantitative and qualitative assessment approaches, we propose a protocol featuring a wait-list control sequential cross-over design. In line with program implementation models [29, 30], we will assess CCM-C training feasibility, acceptability, appropriateness, and usability from the perspectives of BHCs with input from an expert corrections advisory panel (CAP). Based on the social cognitive model of suicide/SDV prevention training [2, 18], preliminary training effectiveness will be judged by trainee SDV risk assessment and management knowledge, attitudes/stigma, perceived skill mastery, and intention to use training content. Because BHC compassion fatigue is associated with interactions with high-risk (e.g., trauma, suicide, violence) patients [31, 32], we selected compassion fatigue as an exploratory outcome. Therefore, this pilot feasibility and preliminary effectiveness evaluation holds three aims, with the following respective research questions (RQ) and hypotheses (H).

• Aim 1: To explore CCM-C implementation feasibility, acceptability, appropriateness, and usability.

  • RQ1: What are the barriers and facilitators of CCM-C implementation?

  • RQ2: What CCM-C improvements can be made to the training program?

• Aim 2: To conduct a preliminary CCM-C effectiveness evaluation.

  • H1: Participation in CCM-C training will result in improved SDV prevention knowledge.

  • H2: Participation in CCM-C training will result in improved SDV prevention attitudes (reduced stigmatizing beliefs).

  • H3: Participation in CCM-C training will result in improved perceived SDV prevention skill mastery.

  • H4: Participation in CCM-C training will result in improved intent/actual use of CCM-C training content.

• Aim 3: To explore the impact of CCM-C training participation on BHC compassion fatigue.

  • RQ3: Does compassion fatigue change after CCM-C participation?

Methods

Trial design

This pilot trial is a type 3 hybrid implementation–effectiveness trial [33]. A type 3 design has a primary purpose of determining the impact of an implementation strategy and secondary aim of assessing clinical outcomes (i.e., effectiveness) [33]. Previous research clearly states:“A hybrid type 3 is essentially an implementation trial plus an evaluation of outcomes. The primary outcomes are implementation outcomes” [33]. Additionally, prior literature specifies implementation outcomes, which include feasibility, acceptability, and appropriateness [30]. The primary aim of the CCM-C trial is to test the feasibility, appropriateness, intervention revision, etc. of the training intervention (i.e., implementation). The secondary aim is a pilot effectiveness evaluation focused on social learning training-focused outcomes (e.g., knowledge, attitudes).

We will employ a wait-list control sequential cross-over design for the present study (see Fig. 2). Based on similar BHC suicide prevention training pilot evaluations in the literature [34, 35], we anticipate that between 50 and 100 BHCs will take part in the CCM-C pilot. We will conduct simple random sampling of eligible BHCs to generate two training intervention groups: training group one versus wait-list control. We will use the population stratification of number of years’ clinical experience in adult corrections to randomly assign 50% of each stratum into the two treatment groups. An overview of the study data collection and training implementation timeline is provided in Table 1. In summary, all BHCs will complete a baseline assessment. Those assigned to training group one will complete the CCM-C training program first (week 3). After the first training session has been completed, all BHCs will complete a second assessment (week 7). Then, training group two (wait-list control group) will receive the CCM-C training (week 9). Following training completion, all BHCs will complete a third assessment 1 month from the last training session (week 13).

Fig. 2

CCM-C pilot study design

Table 1 Core Competency Model for Corrections (CCM-C) pilot study timeline

Study setting

This study is part of an ongoing community–academic partnership between the University of North Carolian (UNC) at Charlotte and North Carolina Department of Adult Corrections (NC DAC). The partners are as follows.

NC DAC

The state of North Carolina prison system has an average daily population of approximately 31,000 incarcerated persons [36]. The pilot study setting, the NC DAC, is a statewide agency comprising 55 prisons. With regard to security level, 21 are minimum, 14 are medium, and the remaining are a mixture of custody levels. Across all prison facilities statewide, the custody level assignment for those incarcerated reflect 19% close custody, 49% medium custody, and 32% minimum custody assignment. The prison population is 92% men overall, and of the 55 total facilities, 4 house women [36]. There are two prisons that include hospital facilities within the complex. NC DAC facilities are organized by four regions: region 1 (Western, 15 facilities), region 2 (South Central, 12 facilities), region 3 (Central, 15 facilities), and region 4 (Eastern, 13 facilities). The Behavioral Health Services section within NC DAC comprises approximately 160 BHCs across the state including psychologists, clinical social workers, and clinical mental health counselors. While available behavioral health services vary by prison site, the following services are available within the NC DAC: screenings on intake, outpatient services, crisis intervention, management of acute psychiatric episodes, crisis stabilization, prevention of psychiatric deterioration, elective therapy services, preventive treatment, provision for referral and admission to mental health facilities for higher levels of care, procedures for obtaining and documenting informed consent, follow-up services upon discharge from an inpatient psychiatric facility, substance-use-specific treatment services, and aftercare planning for community re-entry. To address SDV, BHCs also conduct mental health screenings of incarcerated persons and self-injury risk assessments (SIRAs) at any indication of risk for self-injury or suicide. To illustrate the scope of the need statewide, in 2023, NC DAC BHCs conducted a total of 3028 SIRAs.

UNC Charlotte

The academic team comprises faculty, staff, and graduate students in the Department of Public Health Sciences within the College of Health and Human Services. At any time, the team is composed of between four and eight team members with disciplinary backgrounds such as clinical psychology, community health, experimental psychology, social work, informatics/decision sciences, and epidemiology. As such, the team contributes expertise in SDV prevention, clinical trial design, and quantitative and qualitative methods and analysis to the partnership. The academic team also contributes project infrastructure in terms of institutional review board application and review, survey data collection and analytic software, and grants management support.

Training sites

The NC DAC will utilize a local community college training venue to implement CCM-C sessions. NC DAC maintains existing agreements with regional community colleges, as well as other training locations made available by counties and municipalities that offer governmental waivers for the locations. Training locations are outside of a prison facility, designed for trainings, and include necessary training facility resources. For this protocol, training will be delivered at the Wake Tech Community College campus facility.

Correctional advisory panel (CAP)

In line with prior correctional healthcare literature [37], we will employ a corrections advisory panel (CAP) to provide expert input for CCM-C training co-design. The CAP will be comprised of six internal BHCs (to the NC DAC) and three to six external BHCs (e.g., from other state departments of corrections) with experience providing correctional behavioral healthcare services and in the administration of correctional behavioral healthcare systems. Each CAP member will be a master’s or doctoral-level BHC in psychology, clinical social work, or clinical mental health counseling.

The CAP members will provide an expert review of the training content, approaches, and evaluation plan at two points. During the preparation phase, CAP members will provide a review of the initial training program with respect to usability, feasibility, applicability, and appropriateness. Considerations will include how well content matches the realities of conducting SDV risk assessment and management work in correctional systems and how correctional behavioral healthcare providers may respond to the training. We will collect this formal input from the CAP via a structured feasibility form (see description in measures section). Once the initial pilot evaluation is completed, the evaluation team (Cramer, Cacace, Moxie, Prowten) will present findings to various NC DAC stakeholders including the CAP. At this time, CAP members will provide additional feedback for training revision focusing on content, training techniques, and implementation guidance for the completion of a CCM-C toolkit. The CAP will further advise on the potential to generalize the training to new prison sites of varying sizes and in different states.

Participants and eligibility criteria

Participants will be 50–100 NC DAC BHCs. The study eligibility criteria include (1) being a BHC, (2) being 18 years of age or older, (3) living in the U.S., and (4) being currently employed by the NC DAC. Study exclusionary criteria include (1) decisional or cognitive impairments that preclude being able to consent to study participation and (2) being a member of the study correctional advisory panel (CAP; see details below).

Intervention: CCM-C training program

The CCM-C training will be delivered via an in-person format to 25–50 trainees per session. The full-day training will contain pre-scheduled breaks and opportunities to debrief about content in smaller groups. The initial CCM-C training program follows the general structure of the CCM competencies, with the exception of previously noted wording changes to fit correctional setting needs (e.g., replacing suicide with SDV throughout the competencies). Table 2 summarizes each CCM-C training module. A total of 10 modules is preceded with a brief introduction to the CCM training framework. Each module contains a combination of psycho-educational material, handouts, case studies, and discussion. Following the social cognitive model of CCM-C training [2], each module intentionally targets mechanisms (e.g., improving knowledge, reducing stigmatizing attitudes) hypothesized to increase the likelihood of the use of evidence-based SDV risk assessment and management practices.

Table 2 Core Competency Model for Corrections training program module structure and program targets

Trainers

In line with the principles in community–academic partnership (e.g., co-ownership of knowledge generation) [86], a member of both NC DAC and UNC Charlotte will co-deliver the training. Specifically, study investigators RC and LJP will coordinate training through two sessions. Combined, the trainers possess experience and expertise in competency-based training delivery for legal, clinical, and correctional audiences; behavioral health services; suicide risk assessment and management practice, training, and research; and correctional psychology. One trainer is a licensed clinical psychologist with correctional practice experience, and the other is a trained clinical psychologist with an established research record in suicide prevention and forensic-correctional behavioral health.

Outcomes

Implementation (e.g., feasibility, usability) is the primary outcome given the novel nature of the CCM-C program and the focus on iterative intervention design [87]. We selected a multi-dimensional model of assessment [88]. We did so for two reasons. First, prior hybrid implementation–effectiveness trials [89, 90] employed more than one facet of implementation, yielding valuable information compared to more restricted sets of outcomes. Second, models of implementation science [91] frame implementation or feasibility as a multi-dimensional outcome set. The secondary aims are SDV risk assessment and management knowledge, attitudes, perceived skill mastery, and intent to employ training content. We identified these outcomes in line with prior general CCM training evaluation studies [3, 17, 20,21,22], as well as the social cognitive model of training applied to corrections [2, 18]. The tertiary exploratory outcome of this trial is compassion fatigue. We elected to explore training impacts on compassion fatigue given it is a common challenge for BHCs (especially those working in difficult treatment environments) [92], and CCM-C training addresses matters of self-care, critical incident debriefing, and consultation that may mitigate compassion fatigue among BHCs.

Sample size

We opted for a sample size range of 50–100 BHCs for several reasons. First, the total population of NC DAC mental health providers is no more than 160. We therefore wanted to draw a large proportion of the overall group, while leaving room for NC DAC to implement the eventual CCM-C toolkit in a sustainable way (e.g., train-the-trainer materials). Second, for pragmatic reasons (e.g., room capacity), we wanted to maintain training groups small enough to leverage the interactive nature of training with two trainers. Finally, prior CCM [3, 17, 22] and mental health provider [93, 94] training pilot evaluations have demonstrated success with training groups of this size.

Recruitment

Eligible BHCs will be recruited from the South, Central, Eastern, and Western regional NC DAC facilities. Two recruitment methods will be used in this study: in-person recruitment and email recruitment.

In-person

The NC DAC lead investigator will introduce the study at an all-staff meeting in-person. A study introductory script will be used to explain that this training is mandatory per their position requirements at the NC DAC; this summary will include repetition that participation in the survey evaluation research study is voluntary.

Email

The NC DAC lead investigator will summarize the training and optional evaluation survey research to the full roster of NC DAC BHCs by email after the in-person meeting. He will use the same verbal introductory script, but this time written in an email in text. After he provides the UNC Charlotte research team with the full set of NC DAC BHC email addresses, the UNC Charlotte team will randomly select (see random sampling details section) up to 100 participants to contact for possible participation. Each person will receive a study summary and invitation email. It summarizes how they were randomly selected, why study participation is important, and what to expect next (i.e., a follow-up email with the study link and e-consent form in the following weeks).

BHCs working at the NC DAC who are randomly selected for possible participation will be provided an e-consent document by email inviting them to participate in the CCM-C training program through Qualtrics. This contact will occur several weeks after the initial study advertisements (i.e., in-person and email by the NC DAC lead investigator) in order to allow time for BHCs to carefully consider participation. The consent form will be accessed via a unique Qualtrics link generated using the Qualtrics distribution function. Details about the consent form are contained in the consent section below.

Randomization/allocation

We will use the full pool of eligible BHCs to randomize participants into two training groups. Then, we will use the population stratification of years of experience in adult corrections to randomly assign 50% of each stratum into the two treatment groups: training group one versus training group two (wait-list control). Population strata of years of experience will be calculated using the 10th, 25th, 75th, and 90th percentiles of the NC DAC clinical staff in adult corrections. The sampling procedures will be conducted by co-author SC (statistician) using Stata standard edition (SE) v. 18.0 [95] and stratified random assignment conducted using the randomizr package [96].

Blinding

No study blinding procedures will be used for study personnel. Training groups will be aware which sessions (first or second) they are assigned to but are not privy to the wait-list control design.

Data collection methods

Schedule of assessments

The planned assessment schedule can be seen in Table 3. All surveys will be completed via the Qualtrics survey distribution feature. In addition to being completed by the CAP, the primary outcome (feasibility assessment) will be completed by each training group at the respective follow-up data collection time point closest in time to complete the CCM-C training. Secondary SDV-focused and tertiary compassion fatigue outcomes will be completed at each time point. Characteristics to describe the sample (i.e., demographics and previous SDV exposure) will be collected at baseline only.

Table 3 Assessment schedule

Participant retention

The following strategies will be used to ensure maximal survey completion. First, consistent with empirically supported practices [97, 98], we plan to employ pre-study advertisement and a strategically planned survey distribution reminder. We will be transparent with BHCs in study advertisements that their feedback will have direct impacts on improving the training. In doing so, study advertisements will highlight the value and contributions. Second, carefully selected instruments and pilot-tested survey length to minimize the likelihood of survey fatigue. We plan a survey reminder schedule to optimize timing during the week in which to amplify attention to surveys. Finally, the lead NC DAC study investigator will communicate to BHCs that they will be afforded time while at work to complete training evaluation surveys.

Primary outcome: training implementation (e.g., feasibility)

In line with program implementation models [29, 30], we will assess CCM-C training feasibility, acceptability, appropriateness, and usability from the perspectives of BHCs using a combined strategy of established and newly adapted measures and questions. We will use the Feasibility of Intervention Measure (FIM), Acceptability of Intervention Measure (AIM), and Intervention Appropriateness Measure (IAM) [88]. Each of these measures contain four theoretically supported [30] items measured on a 5-point Likert scale (1 = completely disagree; 5 = completely agree). Each item requires the study team to insert the name of the intervention being tested. For instance, sample items are as follows: FIM (“CCM-C seems possible”), AIM (e.g., “CCM-C is appealing to me”), and IAM (e.g., “CCM-C seems applicable”). The FIM, AIM, and IAM have been validated in past research, demonstrating strong subscale internal consistency (Cronbach’s alpha range 0.85–0.91) and test–retest reliability (r = 0.73–0.88) [88]. For the present study, we added four items assessing usability because the assessment of usability is consistent with models of use of new programs [99] and has shown utility in the evaluation of new training programs [100]. We title this set of items Usability of Intervention Measure (UIM); a sample item is “CCM-C seems usable.” In line with recent recommendations for mixed methods assessment of program feasibility [101], we incorporated qualitative assessment in our strategy. Namely, BHCs will have the opportunity to respond to a series of eight open-ended questions (two per domain of feasibility, acceptability, appropriateness, and usability).

Secondary outcomes: SDV-focused knowledge, attitudes, stigma, perceived skills, and intent to use training content

Suicide Competency Assessment Form (SCAF)

The Suicide Competency Assessment Form (SCAF) [16, 19] will be used to measure participants’ perceived SDV prevention skill mastery. The SCAF contains ten items capturing core competencies of the CCM-C training; these items are measured on a 4-point scale of perceived competency (1 = incapable; 4 = advanced). For the present study, these ten items were adapted to capture SDV-focused skills. Specifically, the term “suicide” was replaced with “self-directed violence” within SCAF items. Studies have indicated good internal consistency in SCAF items (Cronbach’s alpha range = 0.88–0.93). Validity was demonstrated in that the SCAF total score was associated with positive attitudes toward suicide prevention [19, 102].

Attitudes about Intervening with a Suicidal Person (AIBS)

The Attitudes about Intervening with a Suicidal Person (AIBS) [103] questionnaire will be used to measure SDV prevention-focused attitudes and the willingness to intervene in an event where a person is experiencing SDV. The AIBS is a subscale of the larger, recently revised Willingness to Intervene against a Suicidal Person Enhanced Questionnaire [104]. The AIBS contains 14 items measured on a 5-point Likert scale (1 = strongly disapprove; 5 = strongly approve). After reverse-coding seven items, these items will be summed to derive a total score. Previous research has suggested that the AIBS is internally consistent with strong reliability (Cronbach’s alpha range = 0.78–0.89), and the AIBS total score significantly positively predicted the intention to intervene with a suicidal person [103, 104].

Attitudes toward Prisoners who Self-Harm Scale (APSH)

The Attitudes toward Prisoners who Self-Harm (APSH) [24, 45] scale will be used to measure SDV prevention-focused attitudes and stigma. The APSH consists of 25 items scored on a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree). After 14 items are reverse-coded, these items are summed to derive a total score where a higher score indicates more positive attitudes toward incarcerated persons who self-harm. The APSH also allows for four subscales; however, item scoring differs for subscale tabulation. The four subscales are (1) representing attitudes endorsing harsh treatment (four items; α range = 0.75–0.76); (2) attitudes reflecting understanding (six items; α range = 0.40–0.62); (3) attitudes endorsing negative myths (i.e., four items; α range = 0.71 for both studies), and; (4) relating self-harm to suicide (two items; α range = 0.28–0.58) [24]. Because of poor internal consistency across subscales, we will use the total score in the present study. The APSH total score has demonstrated good overall internal consistency (Cronbach’s alpha range = 0.72–0.74) [24, 45] and test–retest reliability (r = 0.79, p < 0.001) [24]. The APSH total score was positively related to a general measure of pro-social attitudes toward incarcerated persons [45].

Core Competency Model for Corrections (CCM-C) knowledge quiz

CCM trainings typically use a set of 10 to 15 multiple-choice questions based on training content [16, 17, 22]. Items are designed to reflect a range of difficulties from basic factual recall to detection and application of content in applied clinical situations. Accurate response ranges to CCM quiz questions suggest that (a) there is a range of difficulty and (b) the range of accurate responding increases post-training [22]. For this pilot evaluation, we developed a ten-question quiz (one question per competency). Correct answers are summed for a total score where higher scores indicate greater understanding of the CCM-C.

Importance and intent to use training

We will use the intention to use training content scale [17] to capture BHC’s perceived importance in using (pre-training) and intent to use (post-training) CCM-C training content. In its original form, this instrument is a seven-item questionnaire where each item queries intent to employ SDV prevention content or skills (e.g., self-reflection tools, clinical interview skills) in the CCM-C training. Items are rated on a 5-point Likert scale (1 = strongly disagree; 5 = strongly agree) and summed to derive a total score where higher scores indicate greater intent to use the training. The intent to use training tool will be measured at baseline and at time points one and two of this study. At baseline, BHCs will be asked about the perceived importance of using the SDV prevention skills. At time points one and two, they will be asked about their intent to use these SDV prevention skills in their future practice. The strategy of asking about perceived importance followed by actual intent to use training is informed by recommendations in the literature in the design of scales measuring intention to engage in behaviors [105] and aligns well with the social cognitive model underpinning CCM-C training [2, 18]. This approach to assessing perceived clinician importance and intent to use training content has been used in prior literature [106].

Tertiary outcome: Compassion Fatigue-Short Scale (CF-SS)

The Compassion Fatigue-Short Scale (CF-SS) [107] will be used to assess feelings of compassion fatigue, conceptualized as secondary traumatic stress and job burnout. This measure contains 13 items; each item is measured on a 10-point Likert scale (1 = rarely; 10 = very often). Items are summed to derive a total score where higher scores suggest greater compassion fatigue. The CF-SS has been validated in prior research demonstrating strong internal consistency (Cronbach’s alpha = 0.80–0.91) [107, 108]. This research has also found that the CF-SS was predictive of secondary traumatic stress and job burnout [108].

Covariates: demographics and self-directed violence exposure

We will collect information regarding participant demographics, work history, and history of SDV and suicide exposure. This assessment will query participant age, gender, race, ethnicity, education level, primary discipline, and years of experience as a BHC in adult corrections. Based on suicide exposure questions in prior CCM training studies [17, 22], we adapted a set of questions to assess SDV exposure via a series of “yes”/ “no” questions. Participants will be asked, “Have you ever known someone in your personal life who…” (1) engaged in non-suicidal self-injury, (2) attempted suicide, and (3) died by suicide. We will also query “Have you ever known anyone in your personal life who has…” (1) non-suicidal self-injury, (2) attempted suicide, and (3) died by suicide. For each SDV exposure item a BHC responds yes to, they will be provided further free response text to indicate who it was (e.g., family member, co-worker). Participants will also be asked to report the total number of hours of suicide prevention training received to date and the total number of patients lost to suicide.

Data management plan

The NC DAC lead investigator provided the university study team with NC DAC email addresses and years of experience via a password-protected excel file. NC DAC email addresses were used to recruit and randomly assign BHCs to participate in this study. These email addresses include BHC names within the address text. UNC Charlotte study members will link email addresses with a random, anonymous study-specific ID number via a password-protected excel document. NC DAC investigators will not have access to this linked summary, nor the dataset, in order to maintain the privacy and confidentiality of NC DAC behavioral health staff participating in the study.

The following infrastructure and training exists to ensure data security and privacy. The University holds a license for DropBox for Business, which utilizes the 256-bit Advanced Encryption Standard (AES) and Secure Sockets Layer (SSL)/Transport Layer Security (TLS) security tunnel, along with a two-step verification for all University faculty and staff, ensuring protection against unauthorized data access for participant responses. All data will be stored in a password-protected DropBox folder to which only the research team will have access. Study participant email addresses will be maintained in the secure study DropBox folder via a password-protected Excel sheet (only study team members at UNC Charlotte will have a password and file access). Third, datasets will be stripped of identifiers (e.g., IP addresses) upon being downloaded. Finally, all members of the research team have completed data security protocol training, per UNC Charlotte, to mitigate user error in sharing confidential research data.

We will take the following steps to maintain participant privacy and confidentiality. First, we will employ personalized Qualtrics links. This strategy will allow participants to complete surveys in a location of their own choosing, thereby enhancing the sense of privacy in responding to surveys. Second, all data will be stored in a password-protected DropBox folder that only the research team will have access. Study participant email addresses will be maintained in the secure study DropBox folder via a password-protected Excel sheet (only study team members at UNC Charlotte will have the password and file access). Third, datasets will be stripped of identifiers (e.g., IP addresses) upon being downloaded. Also, once all data analysis is completed, email addresses linked with study ID numbers will be deleted in all formats (i.e., from Qualtrics, the Excel sheet). This procedure will take place no less than 1 month from the study completion.

Statistical methods

CAP feasibility feedback

Initial CAP feasibility data will be quantified via descriptive statistics of individual and total scores in each domain (e.g., appropriateness). As reported in other intervention development work [109], we will examine open-ended responses using deductive thematic analyses in Dedoose [110] qualitative software. Relying on a previously used approach [111] of familiarizing ourselves with the data, we will create initial codes (based on our question domains of usability, feasibility, applicability, and appropriateness), identify patterns among codes for themes, and review, define, and name themes. A team of two coders will review the data and discuss themes in team meetings. The lead NC DAC investigator will hold at least two virtual meetings with CAP members in order to discuss CAP feedback prior to the revision of training materials for the primary pilot evaluation.

Primary evaluation data cleaning and handling missing data

Immediately following each data collection period, response data will be exported and analyzed for data quality by examining: (1) completion time extremes and linear response patterns [112] and (2) low-quality entries flagged for sensitivity analysis (i.e., comparisons including and excluding low-quality responses). Low-quality responses will be excluded if they significantly alter the results. Missing responses at follow-up time points will be examined for percent completion and missing at-random data patterns using Little’s MCAR test [113]. Missing data will be replaced using factored regression analysis, utilizing appropriate auxiliary predictors [114]. Auxiliary predictors for each time period are shown in Table 4.

Table 4 Predictors for factored regression analysis

Preliminary analyses

Training groups will be compared on years of clinical experience in order to ensure successful randomization. Demographics (e.g., age, race, previous hours of suicide prevention training) and prior experience with SDV information will be described using graphics (i.e., sample percentage bar graphs) and tables (i.e., frequency and percentage). Descriptive statistics tables will be provided for outcome measures, including average and standard deviation aggregated for all participants, and by training group. Potential for covariation for prior experience with SDV and previous suicide prevention training will be assessed for each outcome using visual inspection (i.e., scatterplots) and Pearson’s r correlation matrices. Covariates with significant associations with the outcome measures of interest at any time point will be included as covariates in the final models. Internal consistency for all measures will be assessed using Cronbach’s alpha and will be interpreted using established conventions in the literature [115].

Aim 1. Descriptive CCM-C feasibility, acceptability, appropriateness, and usability analyses

To examine the feasibility of the CCM-C training (RQ1), we will provide descriptive statistics about the acceptability, appropriateness, feasibility, and usability of the training for all participants and by training group as graphics (e.g., bar graphs). We will also examine associations of acceptability, appropriateness, feasibility, and usability by demographic characteristics by providing descriptive statistics by demographic categories and visuals (e.g., scatterplots or bar graphs).

Aim 1. Qualitative CCM-C feasibility, acceptability, appropriateness, and usability analyses

We will use Dedoose online qualitative software [110] to conduct both thematic and content analyses with survey responses to the eight open-ended implementation outcome form questions. First, content analysis [116] will be employed to examine trends, patterns, and relationships of words [117] used in these brief responses. Second, we will use an inductive approach to develop themes based on participant responses rather than a priori theories [116]. Based on our questionnaire, we will examine the domains of acceptability in content and activities, appropriateness of topics, feasibility of participation in the training, and usability of the training. We planned supplemental data sources in the instance open-ended questions do not generate sufficient data for thematic analysis. We will employ supplemental qualitative data generation in the form of process or field notes [118]. These written observational notes will be taken by research team members and trainers during and immediately after the training sessions. These notes will emphasize aspects of the training that worked, BHC’s responses to training, and other observations that can supplement survey data for analysis. In the scenario this strategy does not work, we will default to reliance on content analysis alone.

Content analysis

Themes will describe larger patterns within the domains, while categories will label the specificities of the content within the theme (i.e., “the coding tree”) [117]. As is standard in qualitative content analysis to enhance coding reliability [119, 120], we will use multiple independent raters (n = 3), and each response will be coded twice. Inter-rater reliability will be calculated using Cohen’s kappa scores using Dedoose’s “test” function. Any Kappa scores of less than 0.6 (indicating less than “substantial” agreement) [121] will be discussed until consensus is reached and the codebook will subsequently be refined. We will then re-test to confirm the reliability of all codes. As is common in qualitative team analyses [122], the research team will meet regularly to discuss analytic challenges and establish consensus. During the coding process, coders will document their assumptions, biases, and developing patterns using analytic memos in Dedoose. Participants will not provide feedback on the themes. Themes will be presented with representative excerpts. Themes and excerpts will be integrated into a feedback presentation for the CAP and other relevant stakeholders.

Aim 2: CCM-C Program effectiveness analyses

To address aim 2, outcome analysis will be performed using multi-group interrupted time-series analysis using the user-generated ITSA package [123] in Stata v. 18 [95]. Interrupted time-series analysis uses ordinary least-squares regression adjusted for time point autocorrelation [124, 125], and the Newey–West method will be used for standard error autocorrelation adjustment [126]. A total of five interrupted time series models will be examined for each of the following outcomes: (a) CCM-C knowledge quiz, (b) attitudes about intervening with SDV (AIBS), (c) attitudes toward offenders/prisoners who self-harm (APSH), (d) SDV prevention competency (SCAF), and (e) intent to use or actual use of the CCM-C training content. Each interrupted time series model will be presented in a table including the coefficient of time, the training intervention, training group membership, interactions between time, training, and group membership, covariate associations with the outcome (if applicable), Newey–West standard errors, and significance level for each coefficient, a table showing comparison of post-training trends, and a graphic showing actual and predicted values of change over time for each training group. To address H1–H4, we anticipate a positive association between the training intervention and each outcome, resulting in a statistically significant (p < 0.05), positive coefficient for the training intervention in each model. In the event that training group membership has a significant association with the outcomes above, we will investigate potential demographic differences between training groups (e.g., age, gender, race) using analysis of variance (ANOVA) and include demographics with significant variation between groups as covariates in each interrupted time-series model.

Aim 3. Exploratory training impact on compassion fatigue

We will run an additional interrupted time series analysis to address RQ3, where we anticipate an association between the training intervention and compassion fatigue, resulting in a statistically significant coefficient (p < 0.05) in either the positive or negative direction.

Sensitivity analyses

As a part of the interrupted time series analysis, membership to each training group will be entered as a predictor of all outcomes of interest, along with years of clinical training experience, and an interaction term between both training group membership and years of clinical experience. Statistical significance of any of these parameters will signify confounding and will be retained for all models to control for the impact of training group assignment, variation in years of clinical experience, or their interaction.

Monitoring

Data monitoring

A research monitoring plan nor data monitoring committee is required for studies of minimal risk. We do not plan to conduct interim analyses. Should unforeseen circumstances occur, the study administrative team (Cramer, Peiper, Cacace, Moxie) in consultation with the UNC Charlotte IRB and NC DAC research committee would make the determination whether to end the trial early.

Harm, adverse events, and auditing

The study leadership team will track any unintended impacts of training intervention or study team interactions with participants. Training trial outcomes are not deemed serious; however, in line with Institutional Review Board (IRB) requirements, we will track and report any serious adverse events through the university IRB electronic reporting form. All adverse events will be reported in compliance with university requirements of 5 days. Corrective actions will be determined by the study leadership team in consultation with the UNC Charlotte IRB and NC DAC research committee. We will comply with all IRB auditing requests. We will perform an initial data quality check within 3 days of each survey administration time point to ensure coding of study variables occurs correctly.

Ethics and dissemination

Protocol approvals and amendments

We will complete the annual UNC Charlotte IRB reports and file amendments as needed. Appropriate (re)training of all research staff will occur to ensure compliance with any protocol modifications.

Consent

Eligible participants will read the informed consent form at every time point. The consent form will contain the study investigator contact information and encouragement to contact the study team should they have questions prior to study participation. It also contains clear statements concerning the study purpose, voluntary nature of the training evaluation, study procedures and risks, rights as a research participant, and collection and management of survey information. Participants will indicate their consent by selecting, “Yes, I choose to participate” radio button. Participants who select “No, I do not agree to participate” will not be able to continue with the study and will receive a termination debrief (i.e., “Thank you for considering being a part of our study.”). Consenting participants will be given the option to screenshot and/or print a copy of the consent form for their records. They may also contact study investigators for a copy at any time. A copy of the most recent consent form is available by email to the corresponding author, and online in the clinicaltrials.gov study record.

Debriefing

Participants will be debriefed after each data collection session via an e-debriefing form at the end of the Qualtrics survey. It will contain a message of thanks, behavioral health support contact information, and either a reminder of the next survey administration timeline (for baseline and follow-up 1 (FU1)) or conclusion statement (for follow-up 2 (FU2)). Participants will be reminded to print or screenshot a copy of the debriefing form. A copy of the most recent debriefing form is available via email to the corresponding author.

Confidentiality

We will use the following strategies to ensure participant confidentiality. First, data collection is being conducted via secure Qualtrics administration with individualized, anonymous survey links. Identifying information collected to link surveys will be stripped in favor of study random ID numbers. Also, participants will be allowed to complete surveys in their private work stations or other personal locations of their choice. Data will be stored in a secure, credential login-required mechanism. The data management plan section of this manuscript contains additional information regarding ensuring participant confidentiality.

Access to data

Only university study investigators will have access to data files. NC DAC investigators will not have data file access due to a conflict of interest in having access to colleagues’ and subordinates’ survey responses. We will not allow external data access for secondary analysis because of the possibility of identifying survey responses linked to a specific provider in the NC DAC system and because of the NC DAC standing expectations for data use.

Dissemination

In line with prior practice of the NC-DAC/UNC Charlotte corrections–academic partnership [2, 12, 38, 127], we will engage in a robust dissemination plan. Clinical trial results will be communicated via a conference abstract, peer-reviewed journal article, and report and presentation to NC DAC stakeholders. We will also compose the CCM-C toolkit, a publicly available resource for future CCM-C implementation and evaluation. The CCM-C toolkit will include at least the following items: (1) primary and supplemental slide decks, (2) training handouts (e.g., ethics checklist), (3) self-assessment (e.g., modified SCAF) and clinical (SIRAP-C) instruments, (4) CCM-C training implementation guide, and (5) foundational peer-reviewed readings. The CCM-C toolkit will be jointly owned, hosted, and distributed by NC DAC Behavioral Health Services, NC Department of Health and Human Services, and UNC Charlotte Violence Prevention Center (VPC). External agencies will be able to request CCM-C training and consultation via the UNC Charlotte VPC. Authorship for all dissemination activity will be determined using the CRedIT Taxonomy [127, 128]. Per the corrections–academic partnership agreement, to be credited authorship on any project-related output, an investigator must satisfy at least four domains of CRedIT Taxonomy contributions including editing/approving the final output. NC DAC retains the rights of final approval of all dissemination outputs.

Ancillary and post-trial care

There will be no care provided during or after this training trial.

Discussion

SDV in corrections settings has deleterious impacts on incarcerated persons, BHCs, and the larger system [7, 11]. Correctional SDV risk assessment and management standards and best practices exist [10, 11]. Training for BHCs working in correctional settings is one such best practice requiring a theory-driven, evidence-informed approach. The Core Competency Model (CCM) of Suicide Prevention Training [16, 17] provides a foundation for the development of such a program for correctional BHCs.

This clinical trial will pilot test the Core Competency Model for Corrections (CCM-C), an adaptation of the CCM accounting for prison-specific needs (e.g., self-injury assessment and management) and content (e.g., relevant case law) [2]. We will conduct a hybrid pilot implementation–effectiveness trial via a cross-over stratified random sampling approach. In line with corrections–academic partnership principles [127], this clinical trial meets the needs of an identified NC DAC problem in the form of SDV. Results will be translated into a variety of practice- and research-focused dissemination outputs toward sustainability and spread of CCM-C training.

To date, the study team set up study infrastructure (e.g., training materials), obtained UNC Charlotte IRB and NC DAC research committee approvals, and validated study data collection procedures. We acknowledge several limitations of this pilot implementation–effectiveness design that will temper conclusions. First, we are limited to BHCs in one state system, impacting both the total possible sample and ultimate generalizability of study conclusions. Additional limitations encompass lack of a formal comparison group and lack of behavioral outcome measures. Study design limitations are offset by a number of strengths (e.g., use of a CAP); moreover, we will carefully document observed limitations and their impacts throughout the trial for inclusion in dissemination output.