Factor Structure and Measurement Invariance of the Suicide Status Form-IV

Abstract

The Suicide Status Form-IV (SSF-IV) is the measure used in the Collaborative Assessment and Management of Suicidality (CAMS). The SSF-IV Core Assessment measures various domains of suicide risk. Previous studies established a two-factor solution in small, homogeneous samples; no investigations have assessed measurement invariance. The current investigation sought to replicate previous factor analyses and used measurement invariance to identify differences in the Core Assessment by race and gender. Adults (N = 731) were referred for a CAMS consultation after exhibiting risk for suicide. Confirmatory factor analyses indicated good fit for both one- and two-factor solutions while the two-factor solution is potentially redundant. Configural, metric, and scalar invariance held across race and gender. Ordinal logistic regression models indicated that neither race nor gender significantly moderated the relationship between the Core Assessment total score and clinical outcomes. Findings support a measurement invariant, one-factor solution for the SSF-IV Core Assessment.

Keywords

CAMS Suicide Status Form measurement invariance

In 2020, over 12 million adults reported experiencing serious thoughts of suicide in the United States and inpatient psychiatric care is often the primary point of contact for these individuals (Drapeau & McIntosh, 2021; Substance Abuse and Mental Health Services Administration, 2021; Ward-Ciesielski & Rizvi, 2021). During inpatient psychiatric care, health care providers may employ several different self-report or clinician-administered risk assessment instruments designed to determine risk for outpatient suicidal behavior (Myles et al., 2021).

These self-report and semi-structured instruments typically direct clinical focus toward recent suicidal ideation (SI), a history of suicidal behavior, suicidal planning, and suicidal preparation as the primary determinants of individual risk for death by suicide (Brown et al., 2020). While the Joint Commission recommends the use of these tools as part and parcel of effective suicide prevention (The Joint Commission, 2019), prevailing evidence suggests that these variables (e.g., SI, history of suicide attempts) insufficiently characterize and/or predict greater risk for death by suicide in both univariate and multivariate analyses (Franklin et al., 2017).

These risk assessment instruments implicate a “complicated” model of risk conceptualization; that is, they prescribe “a specific number of factors” that are “sufficient for accurate distinction” between those who think about suicide, those who attempt suicide, or both (Huang et al., 2020, p. 556). Paradoxically, some evidence suggests that these risk assessment instruments attempt to capture such “complicated” risk, but the majority of explained variance captured by such instruments may be attributed to individual suicide attempt history (Brown et al., 2020), far afield of their intended goal.

Instead, Huang and colleagues (2020) contend suicide risk better conceptualized as a “complex” phenomenon; that is, the factors which determine risk for the spectrum of suicidal thoughts and behavior (STB; for example, SI, suicidal planning, suicide attempts) are not always mutually exclusive, are likely infinite, and may be highly idiographic in nature (Huang et al., 2020). This does not mean that personal history of STB is not important in conceptualizing suicide risk. However, self-report and clinician-administered risk assessment instruments consistently struggle to accurately quantify current and future risk for STB, demonstrating questionable sensitivity, specificity, positive predictive value, and do not consider protective factors in their conceptualization of suicide risk. For example, Gutierrez and colleagues (2021) determined that four commonly used assessments of suicide risk evince sensitivity and specificity estimates for STB little better than a coin flip.

Such findings suggest that (a) many commonly used risk assessment instruments do not accurately identify the phenomena of concern and (b) identifying “risk” for individual outpatient suicide risk is extremely difficult when using a risk centric approach. Moreover, in the context of inpatient psychiatry, suicide risk assessment is even more difficult when considering measurable phenomena like non-disclosure or concealment of STB (Podlogar & Joiner, 2020) and patient coercion (Ward-Ciesielski & Rizvi, 2021) within these service delivery systems. Thus, because self-report and clinician-administered risk assessment instruments consistently struggle to accurately quantify current and future risk for STB, risk assessment should likely be conceptualized as more than a simple data gathering tool.

Accordingly, suicide attempt survivors advise that health care providers take time to consider “the various factors contributing to their current problems” and consider “relevant social, physical, and/or environmental health factors” rather than conceptualizing suicide risk only in terms of STB and/or psychiatric diagnoses (Hom et al., 2020, p. 5). Thus, clinicians must adopt a patient-centered, therapeutic approach to effectively assess risk and mitigate future STB (Michel, 2021). In fact, meta-analytic evidence suggests that when assessment procedures are personalized and collaborative, assessment, testing, and feedback may represent brief “therapeutic assessment” wherein the assessment by itself is helpful, leading to improved treatment outcomes in the long term (Duronsini & Aschieri, 2021; Hanson & Poston, 2011; Poston & Hanson, 2010). These adaptive effects of therapeutic assessment may be attributed to collaboration, compassion, and openness and may lead to a measurable impact on client symptoms and client self-enhancement (Duronsini & Aschieri, 2021). The Collaborative Assessment and Management of Suicidality (CAMS; Jobes, 2016) is a suicide-focused therapeutic framework that provides a personalized, collaborative structure for the assessment, management, and treatment of what makes an individual patient vulnerable to suicide.

The flexibility of CAMS has facilitated adaptations to a variety of clinical contexts, including inpatient psychiatric care (Ellis et al., 2015; Jobes et al., 2018; Swift et al., 2021). Using the Suicide Status Form-IV (SSF-IV; Jobes, 2016), CAMS facilitates a personalized, functional assessment of risk for STB. Section A of the SSF-IV includes the SSF “Core Assessment,” which assesses six variables: Psychological Pain, Stress, Agitation, Hopelessness, Self-hate, and overall lifetime risk of suicide. These variables were derived from extant theoretical models of suicide (Baumeister, 1990; Beck & Rush, 1978; Jobes, 2016; Shneidman, 1987) and are rated quantitatively on a Likert-type scale from 1 to 5 which integrate qualitative responses where patients are encouraged to provide rationale for their chosen response category.

A handful of psychometric studies have examined SSF Core Assessment, generally yielding a superior two-factor solution. In the initial psychometric investigation of 106 treatment-seeking undergraduates (mostly White men) deemed at risk for suicide, Jobes and colleagues (1997) found that a two-factor solution accounted for 36% of the variance in responding. The items of the original SSF were labeled “pain,”“external pressure,”“agitation,”“hopelessness,” and “self-regard.” This model produced highly variable communalities and suggested that it worked best for the agitation (.72) and hopelessness (.55) Core Assessment variables (all others between .16 and .30). Despite the high communalities, the item level evidence supported the “quasi-independent nature” (p. 370) of the items. Interestingly, the authors also conducted a model with a forced single-factor solution and observed that reliability estimates increased to the acceptable range (.70; Jobes et al., 1997). Although Jobes and colleagues (1997) ultimately opted for a two-factor solution, the aforementioned findings also lend support for a possible one-factor solution.

Conrad and colleagues (2009) sought to replicate the foregoing study in high-risk inpatient sample (n = 149) using the first revision of the SSF (SSF-II) which excluded the overall risk of suicide item. They found support for two-factor solution in which common variance (i.e., 74%) and communality scores (i.e., ranging from .63 to .80) were much higher when compared to the original psychometric study (Conrad et al., 2009; Jobes et al., 1997). Accordingly, Conrad and colleagues (2009) determined that the Core Assessment followed a two-factor model, characterizing Psychological Pain, Hopelessness, and Self-Hatred (originally “self-regard”) as a “chronic” set of variables typical of patients with persistent STB. Conversely, the Stress (originally, “external pressures”) and Agitation items were characterized as an “acute” set of variables typical of patients experiencing unusual or uncomfortable episodes of suicidal thinking (Conrad et al., 2009).

The factor structure of the Core Assessment was reevaluated again by Brausch and colleagues (2019) in a psychiatric adolescent sample using the latest revision of the SSF (SSF-IV). Compared with previous work, they observed support for a two-factor structure characterizing Hopelessness and Self-hatred as indicative of “chronic” suicidality. Contrary to Conrad and colleagues (2009), this investigation found that the “acute” risk subscale comprised Stress, and Agitation and Psychological Pain variables (not simply Stress and Agitation). The authors compared the fit indices of the two-factor and one-factor solutions and found that the two-factor model indeed demonstrated better fit, providing additional evidence that the SSF may distinguish between elements of STB that could be acute or chronic in nature (Brausch et al., 2019). The discrepant factor structures observed in the adult patients versus adolescent patients may be due to adults registering the longevity of their STB as more psychologically painful, while the adolescent sample may have attributed psychological pain to what they were experiencing in the present moment (Brausch et al., 2019). Given the potential of the SSF-IV to provide a wholistic, collaborative approach to suicide risk assessment, ensuring the strength of the psychometric properties of its core components would prove invaluable to practitioners and researchers alike. Additional replication studies are needed given these discrepancies and divergent samples.

While the psychometric results of these studies were encouraging, several limitations remain. The sample sizes were small (ranging from 100 to 149), and the demographics of said studies were predominantly White (ranging from 79% to 90%) women (ranging from 67.5% to 79%; Brausch et al., 2019; Conrad et al., 2009; Jobes et al., 1997). This absence of equitable representation from racial and ethnic minority groups and gender is problematic, limiting both the generalizability of findings and the evidence base for translating said findings into clinical practice (Landry et al., 2018). Unfortunately, such disparity applies to suicide prevention research as well; Cha and colleagues (2018) observed that more than 20% of longitudinal suicide risk factor studies since 1985 did not report race and over 50% did not report ethnicity.

Clinically, significant differences have been observed on variables assessed by the SSF-IV Core Assessment between gender identity (Bryan et al., 2014; Romanowicz et al., 2013) and racial identity (Brooks et al., 2020), but it is telling that the reasons why hopelessness, for example, may emerge among racial and ethnic minorities may be very distinct. For example, acculturative stress and perceived discrimination demonstrate a significant relationship with hopelessness among racial and ethnic minorities, which in turn explains the relationship between these culturally related stressors and greater SI intensity (Polanco-Roman & Miranda, 2013). The aforementioned self-report and clinician-administered measures of suicide risk do not facilitate exploration of these issues; however, the SSF-IV facilitates collaborative exploration of individual identity and group-level differences that often arise in response to items assessed by the Core Assessment. Thus, reevaluation of the factor structure in a larger and more representative sample is needed to more appropriately characterize the factor structure of the Core Assessment of the SSF-IV.

Relatedly, endorsement of suicide-related experiences may differ by racial identification and gender identity (e.g., Baiden et al., 2020; Coley et al., 2021; Stephenson et al., 2006). For example, research demonstrates that men and women tend to differ on frequency of and risk factors associated with suicidal thinking (e.g., alcohol consumption related to suicidal thinking in women; Stephenson et al., 2006). Differences have also been found in relation to racial/ethnic identity; research in adolescent samples has found that Black men demonstrate a higher prevalence of SI, planning and attempts in the past year compared with Black women and White men and women (Valois et al., 2013). Recent investigations also reveal that predictive measurements of suicide risk demonstrate poorer performance in Black individuals compared with all other racial categories (Coley et al., 2021). Although the Core Assessment is not necessarily intended to be used as a predictive tool, it follows that any tool intended to provide clinical data regarding a patient’s prognosis should perform adequately across populations. However, some of the variables assessed in the SSF-IV Core Assessment (e.g., Psychological Pain, Hopelessness) demonstrate equivalent predictive utility when compared to a personal history of STB (Franklin et al., 2017), and evaluating the relationship of these core variables with other indicators of current and future suicide risk as moderated by demographic identities fits a “complex” model of suicide risk conceptualization (Huang et al., 2020).

Accordingly, measurement invariance techniques are warranted. This strategy assesses the extent to which a specific instrument is psychometrically equivalent across groups (e.g., race, gender; Cheung & Rensvold, 2009; Putnick & Bornstein, 2016; Raykov et al., 2012). If measurement invariance is observed, it can be inferred that patients across an array of identities interpret each item and the underlying construct in the same way; on the other hand, if a measure is non-invariant across groups, these groups likely interpret the meaning of instruments in distinct ways and sum, or mean scores cannot be compared meaningfully across these groups.

No investigations of the SSF Core Assessment have tested measurement invariance (e.g., metric invariance, configural invariance) and the purpose of the current investigation was to conduct a robust psychometric evaluation of the Core Assessment of the SSF-IV in a large, diverse sample of patients admitted for psychiatric care. We sought to (a) replicate previous factor analytic work to determine ideal factor structure, (b) conduct measurement invariance to determine if measurements of the Core Assessment differ by race, (c) conduct measurement invariance to determine if measurements of the Core Assessment differ by gender, and (d) determine if the SSF-IV Core Assessment is related to other important clinical indicators of suicide risk (i.e., self-reported risk of suicide, suicide ambivalence). Meeting these aims will help further theory building initiatives aimed at understanding the Core Assessment. Furthermore, it will increase clinician and researcher confidence in using the Core Assessment across populations.

Method

Participants

The sample consists of N = 731 adults referred for a CAMS consultation at a large academic medical center in the southeastern United States. Patients were referred for a CAMS Consultation at the behest of six psychiatrists across three psychiatric inpatient units and the Medical/Surgical floor(s) at said hospital between 2017 and 2021 if they were 18 years of age or older and exhibited some risk for an outpatient suicide attempt (e.g., expressing passive to active SI, admitted for a suicide attempt, history of a suicide attempt, collateral report of death or suicide-related statements, comorbid psychiatric diagnosis). Such risk was determined by the aforementioned attending psychiatrists or psychiatric residents based largely on the clinical judgment of each attending psychiatrist after a comprehensive History and Physical (H&P) Examination was conducted and admitting concerns were reviewed. No formal exclusion criteria are included.

Procedures

The current study is an archival investigation of patient charts from the aforementioned clinical consultations. Graduate-level clinicians (i.e., Clinical Psychology graduate students) completed integrated CAMS training and preliminary fidelity “check-outs” with a nationally accredited CAMS Trainer (author R.P.T.) prior to initiating services. Initial CAMS consultation included the collaborative completion of an adapted version of the SSF-IV (Jobes, 2016) and a suicide safety plan.¹ After the session was complete, the SSF-IV and the safety plan were photocopied, and the master copies were integrated into the patient’s Electronic Health Record (EHR) while the photocopy was stored in secure location on medical center premises. Demographic and diagnostic information were garnered from the patient’s EHR by an undergraduate Research Associate. Hard copies of patient protected health information (PHI) and personal identifying information (PII), including the SSF-IV and the safety plan, were de-identified by the Research Associate and stored in a secure location on medical center premises. Digitized PHI and PII were stored on a password-protected, encrypted hard drive which was also stored in a locked filing cabinet. The foregoing study procedures were approved by the associated Institutional Review Board of the hospital.

Materials

Demographics

The following information was gathered from each patient’s EHR: Age, Gender, Race, Ethnicity, Marital Status, and Insurance Type. Unfortunately, it is unclear as to whether these identity categories are typically self, or provider identified as no standard assessment procedure is used across the different settings (e.g., inpatient psychiatry, medical/surgical floor) from which data were extracted. A detailed description of the sample characteristics may be found in Table 1.

Table 1

Sample Characteristics.

Variable	M	SD	Range
Age	40.71	9.77	18–90
Days in hospital	7.9	9.7	1–145
Ethnicity
Hispanic	2.4%
Non-Hispanic	97.1%
Race
White	61.5%
Black/African American	33.6%
First Nations	0.7%
Latino/a	2.3%
Biracial	0.01%
Gender
Woman	43.5%
Man	56.5%
Gender diverse	0%
Marital status
Single	58.5%
Married	15.3%
Divorced	16.9%
Widowed	3.1%
Insurance
No insurance	1.2%
Medicare	22.9%
Medicaid	56.3%
Private	19.5%
Lifetime SI
Yes	69.8%
No	27.3%
N/A	2.8%
Lifetime suicidal planning
Yes	27.4%
No	68.8%
N/A	3.7%
Lifetime suicidal preparation
Yes	19.7%
No	76.1%
N/A	4.1%
Lifetime suicidal rehearsal
Yes	14.0%
No	81.3%
N/A	4.6%
Lifetime suicide attempt
Yes	51.0%
No	46.6%
N/A	2.2%

Note. SI = suicidal ideation.

Suicide Status Form-IV

The SSF-IV (Jobes, 2017) is characterized by three components. First, Section A includes a Core Assessment of theoretically derived items commonly used to capture psychological vulnerability for suicide (i.e., Psychological Pain, Stress, Agitation, Hopelessness, and Self-Hate) which patients are asked to rate on a scale from 1 (low) to 5 (high). This Core Assessment component also asks patients to rate their Overall Risk of death by suicide in the future on a scale from 1 (extremely low risk) to 5 (extremely high risk) and rank the five Core Assessment items in terms of their salience for contributing to their individual experience of STB. Section A continues, asking patients to rate the extent to which their being suicidal is related to thoughts about themselves and thoughts about others on a scale from 1 (not at all) to 5 (completely). Finally, Section A asks patients about their reasons for dying (RFD), reasons for living (RFL), and wish to live, wish to die (each rated on a scale from 0 [not at all] to 8 [very much]). Section B of the SSF-IV contains and assesses several empirically derived risk factors and warning signs used to further characterize the patient’s risk for engaging in future suicidal behavior (e.g., SI frequency and intensity, suicidal planning history, suicide attempts, perceived burdensomeness) while Section C of the SSF-IV includes the collaborative process of treatment planning, driver(s) identification, mutual agreement to treatment conceptualization, and mutual agreement for whether hospitalization is indicated (omitted in this case due to patients already being hospitalized).

Analytical Plan

Demographic information was calculated across patients within the sample, data were examined for missingness, the distributions of the five Core Assessment questions examined, and a correlation matrix among variables of interest was calculated. As is observed by Conrad and colleagues (2009), the Overall Risk item was excluded from the factor analyses as it is a summative variable, it is not theoretically derived, and it is the main outcome of interest within the CAMS framework (Jobes, 2017). Of note, CAMS is collaborative by design and, if a patient elected to terminate a session for varying reasons, the session was terminated and data were not used in this investigation. In addition, due to the highly fluid nature of the medical/surgical floor in particular, many participants may be missing portions of data due to session interruptions from other providers or family members, medical emergencies, and so on. Nevertheless, missing data analyses revealed that approximately 2.47% of data were missing for variables of interest and were handled via listwise deletion.

Confirmatory Factor Analysis

Three confirmatory factor analyses (CFAs) using diagonally weighted least squares (WLSMV) were conducted on the SSF-IV Core Assessment variables in accordance with those outlined in the CAMS literature to date. The WLSMV estimator was used in lieu of maximum likelihood considering Likert-type scale items of five or fewer choices and asymmetric category thresholds may be best considered ordinal rather than continuous data (Rhemtulla et al., 2012). Of note, the collaborative nature of CAMS and the SSF-IV permits recording verbatim ranks (i.e., zero and “in between”) of experience on the Core Assessment. Thus, Core Assessment data for analysis rounded patient self-rated “zero” scores up to one; “in between” scores (e.g., 1.5, 2.5) were also rounded up to closest rank category (e.g., 1.5 rounded to 2). The extant SSF-IV Core Assessment factor solutions are as follows: (a) an inferred one-factor solution (Jobes et al., 1997) and (b) an inferred two-factor solution with Hopelessness and Self-Hate indicating chronic and enduring risk states in factor 1 and Psychological Pain, Stress, and Agitation indicating acute risk states in factor 2 (Brausch et al., 2019). The item loadings for factor two were held constant when specifying the model. Finally, an inferred two-factor solution with Psychological Pain, Hopelessness, and Self-Hate fitting onto Factor 1 (i.e., the “chronic” risk factor) and Stress and Agitation fitting on to Factor 2 (i.e., the “acute” risk factor; Conrad et al., 2009). Again, item loadings of the “acute” risk factor were held constant in this inferred solution. Model fit was determined using the comparative fit index (CFI; ≥.95), Tucker–Lewis Index (TLI; ≥.95), root mean square error of approximation (RMSEA; ≥.06), and standardized root mean square residuals (SRMR; ≥.05) with the associated cut points indicating good to excellent fit for the inferred factor solutions (Hu & Bentler, 1999). In addition, item level loadings (i.e., <.35) and standardized residuals were examined for areas of potential model misfit, as was the relationship between the factors within the foregoing two-factor models (Bagby & Cox, 1991). All CFA models were tested in R (R Core Team, 2019) using the lavaan package (Rosseel, 2012) and semTools package (Jorgensen et al., 2021) to compare the fit for the foregoing solutions and following measurement invariance analyses.

Measurement Invariance

Multi-group CFAs were conducted by race (i.e., White, Black) and gender (i.e., cisgender man, cisgender woman); unfortunately, low sample representation from other marginalized groups within the United States (e.g., American Indian, gender-diverse individuals, Hispanic ethnicity) prohibited measurement invariance analyses of these groups. The following multigroup CFAs in White and Black and men and women subgroups evaluated configural invariance (i.e., identical factor structure), metric invariance (i.e., equivalent factor loadings), and scalar invariance (i.e., equivalent indicator thresholds [item intercepts]). Strict invariance (i.e., the equivalence of factor and error variances) was not tested as invariance of the item residuals is inconsequential to interpretation of latent mean differences (Vandenberg & Lance, 2000) and can be omitted per Putnick and Bornstein (2016).

Latent group means were examined by fixing the mean of the reference group (i.e., White, male) to zero while allowing the alternative group mean to vary freely within the respective scalar invariance models (i.e., race, gender). The critical ratio (CR) was used to assess the extent of the latent mean differences; the CR is calculated by dividing the parameter estimate (i.e., factor intercept) by its standard error, with a CR value greater than or equal to 1.96 indicating a statistically significant difference between latent means based on race and gender categories (Byrne, 2013; Chen et al., 2019).

Per Chen (2007), the following cutoff criteria were used to evaluate measurement invariance, assuming relatively unequal sample size between groups: a −.005 change in CFI supplemented by .010 change in RMSEA for loading (i.e., metric) and intercept (i.e., scalar) invariance. Configural invariance is the least strict invariance criteria and is not accompanied by predetermined cutoffs within the literature to date and was thus evaluated via inspection of item level loadings for the best fitting solution from the foregoing pooled sample CFAs.

Finally, measurement invariance assesses the extent to which the summed outcome of an instrument is equivalently related to an outcome of interest across groups. Considering one of the primary resolution criteria within the CAMS framework is improvement in the Overall Risk score, this was used as a dependent variable within the ordinal regression models (Jobes, 2016). The Overall Risk score is a single item which asks each patient to rate their relative risk of death by suicide in the future on a scale from 1 (extremely low risk) to 5 (extremely high risk). While a single-item relative risk score has inherent limitations, extant literature suggests that relative risk indicators can effectively distinguish between life and death orientation during treatment and appropriately correspond to changes in these orientations over the course of treatment for STB (Corona et al., 2013; Jobes, 2016; O’Connor et al., 2012). Accordingly, the five variables of the Core Assessment were summed to yield a total score for each patient, which were then mean centered. The mean centered Core Assessment sum score was used to build two ordinal regression models: one model assessing the extent to which the mean centered total score is related to Overall Risk at the intersection of gender and the other at the intersection of race. Within these models, the mean centered Core Assessment total score and race/gender were entered into Step 1 while the moderation term(s) (i.e., Total Score × Race, Total Score × Gender) was entered into Step 2 of the model.

Wish to live (WTL), wish to die (WTD), and suicidal ambivalence (AMB) are additional categories which have evinced a meaningful relationship with suicide risk in the context of the CAMS framework (O’Connor et al., 2012) and are data points recommended to be discussed in the post-session evaluation (clinical documentation procedures) of CAMS. WTL and WTD scores derive from the internal debate hypothesis (Kovacs & Beck, 1977) and have been observed to reliably correlate at a coefficient of r = −.60 (Brown et al., 2005; Bryan et al., 2016). Research has found that those presenting to a psychiatric ER with ambivalence about living are at twice the risk for a suicide attempt (Naherniak et al., 2019). Patient WTL scores were subtracted from patient WTD scores to create an overall AMB score, which was then categorically coded. This is typical practice (Brown et al., 2005; Bryan et al., 2016; Goods et al., 2020; Kovacs & Beck, 1977; O’Connor et al., 2012) and WTL, WTD, and AMB have been linked to risk for suicidal desire and suicidal behavior cross-sectionally and prospectively (Brown et al., 2005; Bryan et al., 2016). Within the current study, life-oriented individuals were coded as (0) representing a “suicidal ambivalence score”≥1, ambivalent individuals were coded as (1) representing a “suicidal ambivalence score” of exactly zero, and death-oriented individuals were coded as (2) representing a “suicidal ambivalence score” of ≤−1. Thus, as death orientation increases, so does suicidal ambivalence category.

Next, using the aforementioned mean centered Core Assessment total score, two ordinal regression models were run: one model assessing the extent to which the mean centered Core Assessment total score is related to suicidal ambivalence categories at the intersection of gender and the other at the intersection of race. Within these models, the mean centered Core Assessment total score was entered into Step 1 while the moderation term(s) (i.e., Total Score × Race, Total Score × Gender) was entered into Step 2 of the model; only Black and White men and women were included in these analyses. All ordinal regression analyses were run in SPSS (IBM Corp., 2020).

Results

Factor Structure in the Full Sample

A detailed description of counts for each response category within each question may be found in Table 2. The initial CFA model was structured as an inferred one-factor solution (i.e., Model 1). As can be observed in Table 3, Model 1 demonstrated good fit for the data (χ² [5] = 25.673, p < .001; RMSEA = .099; CFI = .978; TLI = .956; SRMR =.031) with three of four fit indices falling above the predetermined criteria for good to excellent fit (Hu & Bentler, 1999). Table 3 shows that Model 2 also demonstrated good to excellent fit and better than that of Model 1 (χ² [5] = 24.929, p < .001; RMSEA = .081; CFI = .985; TLI = .970; SRMR = .030); however, the estimated relationship between the two inferred factors was large (r = .91), suggesting that they are more than likely redundant. As can be observed in Table 3, Model 3 demonstrated observably worse fit than the two aforementioned models (χ² [5] = 47.975, p < .001; RMSEA = .122; CFI = .967; TLI = .934; SRMR = .047). In addition, it was observed that the two factors within this inferred solution were highly correlated (r = .93), suggesting that they are more than likely redundant. Considering the observed results, Model 1 was selected for measurement invariance analyses. Factor loading for all inferred solutions may also be found in Table 3. Because the inferred one-factor solution does not appear to be adequately tau-equivalent, omega-categorical ( $ω$ _u-cat) was used to calculate the internal consistency of the Core Assessment total score. Omega coefficients do not assume tau-equivalence but rather a congeneric model where factor loadings are allowed to vary freely within the inferred solutions (Flora, 2020). The calculation of $ω$ _u-cat uses the polychoric correlation matrix as in the original CFA using WLSMV and would represent the reliability of a hypothetical latent total score, but it has been rescaled to represent the observed total score (Flora, 2020; Green & Yang, 2009). Accordingly, $ω$ _u-cat = 0.821 for the SSF-IV Core Assessment within the current sample and thus appears suitable for follow-up analysis.

Table 2

Category Counts and Percentages for SSF-IV Core Variables.

Response category	Psych. pain (n = 731)	Stress (n = 731)	Agitation (n = 726)	Hopeless. (n = 716)	Self-hate (n = 716)
1	36.25% (265)	32.01% (234)	45.86% (333)	49.16% (352)	55.73% (399)
2	16.41% (120)	17.0% (124)	16.3% (118)	13.4% (96)	10.8% (77)
3	21.9% (149)	16.6% (122)	13.2% (96)	12.0% (86)	10.5% (75)
4	11.1% (78)	13.5% (99)	8.5% (62)	8.5% (61)	9.5% (68)
5	16.7% (119)	20.7% (152)	15.9% (116)	16.9% (121)	13.5% (97)

Note. SSF = Suicide Status Form.

Table 3

Global Fit Indices and Factor Loadings for CFA of the SSF-IV Core Assessment.

Model #	χ²	df	RMSEA	SRMR	CFI	TLI	Psych. pain	Stress	Agitation	Hopeless.	Self-hate
Model 1	25.673**	5	.099	.031	.978	.956	0.84	0.77	0.62	0.79	0.71
Model 2	24.929**	5	.081	.030	.985	.970	0.85	0.78	0.63	0.79	0.79
Model 3	47.975**	5	.122	.047	.967	.934	0.85	0.74	0.74	0.80	0.71

Note. CFA = confirmatory factor analysis; SSF = Suicide Status Form; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CFI = comparative fit index; TLI = Tucker–Lewis Index.

Bolded parameters indicate the respective two item factors.

p < .01.

Measurement Invariance by Race

Configural invariance by race held for the SSF-IV Core Assessment; no significant differences were observed across factor loadings between Black and White patients, retaining good model fit in the combined sample (χ² [10] = 38.600, p < .001; RMSEA = .093; CFI = .988; TLI = .976; SRMR = .037). As can be observed in Table 4, test of metric invariance was observed to hold by race (ΔCFI = .004; ΔRMSEA = −.030). Test of scalar invariance was also observed to hold by race (ΔCFI = .003; ΔRMSEA = −.026; see Table 4). Tests of latent mean differences between White and Black patients demonstrated no significant difference, z = −0.554, p = .579.

Table 4

Measurement Invariance Testing of the SSF-IV Core Assessment by Race and Gender.

	Race						Gender
Invarianc test	χ²	df	RMSEA	SRMR	CFI	TLI	χ²	df	RMSEA	SRMR	CFI	TLI
Configural	38.600**	10	.093	.037	.988	.976	29.512**	10	.077	.034	.992	.983
Metric	32.201**	14	.063	.040	.992	.989	34.480**	14	.067	.041	.991	.987
Scalar	40.383	28	.037	.037	.995	.996	49.906**	28	.049	.036	.991	.993

Note. SSF = Suicide Status Form; RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CFI = comparative fit index; TLI = Tucker–Lewis Index.

p < .01.

Measurement Invariance by Gender

Configural invariance by gender held for the SSF-IV Core Assessment; no significant differences were observed across factor loadings between cisgender men and women, retaining good model fit in the combined sample (χ² [10] = 29.512, p < .002; RMSEA = .077; CFI = .992; TLI = .983; SRMR = .034). As can be observed in Table 4, test of metric invariance held by gender (ΔCFI = −.001; ΔRMSEA = −.010). Test of scalar invariance also held by gender (ΔCFI = −.000; ΔRMSEA = −.018; see Table 4). Tests of latent mean differences between cisgender men and cisgender women patients demonstrated no significant difference, z = 1.370, p = .171.

SSF-IV Total Score Predicting Overall Risk

An ordinal logistic regression analysis was used to investigate the relationship between the Core Assessment total score, race, and individual patient Overall Risk score. The Core Assessment total score was found to contribute to the model, ordered log odds (β = .239, SE = 0.0533, Wald[1] = 20.073, p < .001) while race was not found to contribute to the model, ordered log odds (β = −.063, SE = 0.1806, Wald[1] = 0.122, p > .05). In addition, the moderation term of Core Assessment Total Score × Race was not found to contribute to the model, ordered log odds (β = −.002, SE = 0.0310, Wald[1] = 0.005, p > .05). The estimated odds ratio of the Core Assessment total score favored a positive relationship with Overall Risk (OR = 1.27, 95% CI = [1.144, 1.410]). Thus, as the Core Assessment score increases, the odds of a higher rating on Overall Risk increases by 27%, holding all other variables constant.

An ordinal logistic regression analysis was used to investigate the relationship between the Core Assessment total score, gender, and individual patient Overall Risk. The Core Assessment total score was found to contribute to the model, ordered log odds (β = .222, SE = 0.0499, Wald[1] = 19.723, p < .001) while gender was not found to contribute to the model, ordered log odds (β = .011, SE = 0.1686, Wald[1] = 0.004, p > .05). In addition, the moderation term of Core Assessment Total Score × Race was not found to contribute to the model, ordered log odds (β = −.007, SE = 0.0297, Wald[1] = 0.052, p > .05). The estimated odds ratio of the Core Assessment total score favored a positive relationship with Overall Risk (OR = 1.25, 95% CI = [1.132, 1.377]). Thus, as the Core Assessment score increases, the odds of a higher rating on Overall Risk increases by 25%, holding all other variables constant.

SSF-IV Total Score Predicting Suicidal Ambivalence

An ordinal logistic regression analysis was used to investigate the relationship between the Core Assessment total score, race, and suicidal ambivalence category. Neither the Core Assessment total score, ordered log odds (β = .102, SE = 0.0656, Wald[1] = 2.428, p > .05), race, ordered log odds (β = −.131, SE = 0.2631, Wald[1] = 0.246, p > .05), nor the moderation term of Core Assessment Total Score × Race ordered log odds (β = −.057, SE = 0.0393, Wald[1] = 2.115, p > .05) contributed to the model.

An ordinal logistic regression analysis was used to investigate the relationship between the Core Assessment total score, gender, and suicidal ambivalence category. The Core Assessment total score was found to contribute to the model, ordered log odds (β = .213, SE = 0.0668, Wald[1] = 10.124, p < .001) while gender was not found to contribute to the overall model, ordered log odds (β = .229, SE = 0.2611, Wald[1] = 0.768, p > .05). In addition, the moderation term of Core Assessment Total Score × Race was not found to contribute to the model, ordered log odds (β = −.010, SE = 0.0393, Wald[1] = 0.067, p > .05). The estimated odds ratio of the Core Assessment total score favored a positive relationship with suicidal ambivalence (OR = 1.24, 95% CI = [1.085, 1.410]). Thus, as the Core Assessment score increases, the odds of a higher rating on suicidal ambivalence increases by 24%, holding all other variables constant.

Discussion

STB are pervasive phenomena that affect individuals from a variety of backgrounds (Brooks et al., 2020; Polanco-Roman & Miranda, 2013). Accordingly, clinical tools that support the treatment thereof should adequately account for and accurately characterize individuals from diverse backgrounds. Measures used for tracking risk for STB throughout treatment are critical as these measures are used in important clinical decision making (e.g., hospitalization, treatment efficacy, categorizing risk post-discharge). The present study evaluated the psychometric properties of the SSF Core Assessment, measurement invariance of the SSF-IV Core Assessment across demographic factors (e.g., race and gender), and SSF-IV Core Assessment total score and important clinical outcomes (e.g., overall suicide risk, suicidal ambivalence).

In comparing the one-factor solution for the SSF Core Assessment (Jobes et al., 1997) to the two-factor solution found in a sample of adolescent psychiatric inpatients (Brausch et al., 2019) and adult psychiatric inpatients (Conrad et al., 2009), the current factor analysis supported a one-factor solution. While the present factor analysis also revealed a good to excellent fit for a two-factor solution, mirroring findings from Brausch and colleagues (2019), the two factors were highly correlated (r = .91) demonstrating redundancy. The present findings support the use of the total score of the SSF Core Assessment as a measure to track core psychological risk for suicide in patients (along with single item tracking if desired by the clinician). This assertion is made based on the inferred factor structure; however, subscales defined by previous work may delineate specific subsets of Core Assessment items that best relate to acute suicide risk or chronic and enduring risk for death by suicide (e.g., one subscale relates to higher baseline SI above and beyond the other). In previous research, individual items from the SSF-IV demonstrated convergent validity with clinically relevant scales within the literature (Brausch et al., 2019; Conrad et al., 2009); however, factor subscale scoring or total scoring was not used in past studies. Follow-up studies should include longitudinal designs to examine concurrent and prospective discriminant validity of the SSF-IV Core Assessment to determine clinical utility of considering subscales of the measure or if the total score suffices in terms of substantial predictive validity.

Measurement invariance of the SSF-IV Core Assessment across race (i.e., White vs. Black) and gender (i.e., cisgender man vs. woman) was examined. Relative to race, no significant differences in factor loadings were observed between White and Black patients. Furthermore, tests of metric and scalar invariance held by race. Relative to gender, no significant differences in factor loadings were observed between cisgender men and women patients. Again, tests of metric and scalar invariance held by gender. Our findings provide support for the use of the SSF-IV Core Assessment across the included demographic groupings. This is critical, as those from minoritized backgrounds may disproportionately experience suicidal thoughts and non-lethal suicidal behavior (Denneson et al., 2021; Talley et al., 2021). It is imperative that measures designed to assess psychological risk factors for suicide function equitably across those from different demographic backgrounds.

Similarly, this research provides confidence that the five variables theorized to create psychological vulnerability for suicide may function similarly across these demographic groupings and can be used in future research to compare suicide risk across these populations. Although adult psychiatric inpatients from minoritized versus majority backgrounds experience different and disproportionate risk for suicide (e.g., Tucker, 2019), the current research supports core psychological risk, as measured by the SSF-IV only, may be more similar than different.

This assertion of similarity is further supported as race and gender did not contribute to the concurrent prediction of SSF Overall Risk rating and suicidal ambivalence and did not moderate the relationship between the SSF Core Assessment total score and these important clinical indicators of suicide risk. In considering the utility of using the SSF-IV to assess psychological risk for suicide in both clinical practice and research, similarities/differences in factor structure and predictive validity of important outcomes should be considered. For example, research has demonstrated that suicide risk predictive models may under perform in the prediction of suicide in racial minority versus majority populations (Coley et al., 2021). The current results provide very preliminary support that the SSF-IV may concurrently related to important indicators of risk for suicide across Black and White individuals as well as across cisgender men and women. However, the current investigation is limited in testing this important aspect of equitability of the SSF-IV so interpretations regarding predictive similarity should be interpreted cautiously.

Limitations

The findings should be viewed within the context of several limitations. An overarching methodological weakness of this study is measurement contamination. All variables were derived from the SSF-IV with no other indicators of validity. Ideally, other measures of psychological risk for suicide (e.g., Suicide Cognitions Scale-Revised; Bryan et al., 2020) and STBs (Self-Injurious Thoughts and Behaviors Interview; Nock et al., 2007) would have been included in the study to test potential differences in concurrent predictive validity of the SSF-IV. Similarly, although concurrent predictive validity should be established, the current study would ideally have included a longitudinal design to determine potential differences in how the SSF-IV Core Assessment relates to future risk for suicide. Another limitation includes the skewed distribution toward lower scores on the SSF-IV for overall risk of death and few participants that were death oriented at the time of assessment (i.e., their wish to die rating was higher than their wish to live). Although data were collected from adult psychiatric inpatients presenting with suicidal thoughts and/or behaviors, replication and extension in populations experiencing more intense suicide risk is warranted.

Unfortunately, limited representation from diverse racial, ethnic, and gender backgrounds prevented important tests of measurement invariance (e.g., across Hispanic and non-Hispanic adults or transgender/gender diverse and cisgender individuals). Similarly, analyses at the intersection of race and gender were unable to be completed due to low cell size once the sample was stratified. Scholars argue for further integration of an intersectional lens to understanding suicide risk (Standley, 2020), and this work should extend to testing whether clinical and research tools such as the SSF-IV perform equitably across people from multiple underrepresented populations. The final concern regarding demographics is how these groupings were established; the collection of demographic data was not standardized across the sample. Data were pulled from each patient’s EHR, with no record of whether the information is based on self-report or determined by the provider which limits the accuracy of these data points.

Similarly, categorization of race and gender is not the ideal way to capture felt identity and factors that could contribute to differences in psychological risk for suicide in marginalized populations (Tucker, 2019). Future research should include not only standardized assessment of patient demographics but also the measures of identity, minority stress, and other potential explanatory factors that could explain why (or why not) a measure of psychological risk for suicide performs differently across demographic groupings.

Conclusion

Notwithstanding the noted limitations, the present study includes several promising results with important clinical applications. Findings suggest that the SSF Core Assessment (i.e., psychological vulnerability for suicide) is a quick-to-administer measure of tracking psychological vulnerability for STB, which equivalently assesses individuals from different backgrounds and relates to a number of important clinical outcomes. What is more, the Core Assessment and SSF-IV more broadly are particularly well-suited to culturally competent suicide risk assessment; as previously noted, hopelessness, stress, and psychological pain may be invariant by race and gender, but the source of said hopelessness, stress, and psychological pain may differ by culture (e.g., racial discrimination, acculturation; Brooks et al., 2020; Polanco-Roman & Miranda, 2013). The Core Assessment permits exploration of these important cultural differences via qualitative response to items and facilitates a connection of these variables as contributing to the phenomenology of STB and may thus appropriately model STB as truly “complex” (Huang et al., 2020). However, that without adequate training in the administration of the SSF-IV and relevant background information in cultural considerations for suicide (e.g., Chu et al., 2017), important culturally relevant information may not be assessed. Adequate training is needed to ensure the tool can be used appropriately.

More and more, psychometric research in suicide is demonstrating that the psychosocial variables used to measure risk may be amalgamated into a suicidal belief system or “two sides of the same medal” (Bryan et al., 2020; Forkmann et al., 2018; Millner et al., 2020). However, ecological momentary assessment (EMA) using the SSF Core Assessment items may prove enlightening; such research would provide insight into the individual function of each item and the amalgamated psychological vulnerability for suicide construct. Future research should also examine how psychosocial vulnerability for suicide varies over the course of treatment compared with other measures of risk and predicts positive therapeutic change in patients treated for STB. Finally, future research would benefit from continued examination of measurement invariance in this and other widely used clinical tools across individuals from a variety of demographic backgrounds (e.g., Hispanic and Latino/a individuals, sexual and gender minority individuals).

Footnotes

Acknowledgements

Many thanks to Dr. Thomas Olino for his thoughtful consultation with the analyses.

Declaration of Conflicting Interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: David A. Jobes has research funding from NIMH and NIAAA and receives book royalties from American Psychological Association Press and Guilford Press. He is the founder and co-owner of CAMS-care, LLC (a professional training and consultation company). Raymond P. Tucker receives financial compensation for training providers in the Collaborative Assessment and Management of Suicidality (CAMS) and the related Suicide Status Form (SSF) that is used in this investigation.

Funding

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

Ethical Approval

The data collection for this study was approved by the Franciscan Missionaries of Our Lady University Institutional Review Board (IRB) and the research protocol was deemed exempt from informed consent procedures.

ORCID iDs

Nicolas Oakey-Frost

Emma H. Moscardini

Anthony Robinson

Matthew Calamia

Data Accessibility Statement

Research data are not shared.

Notes

References

Bagby

R. M.

Cox

B. J.

(1991). Multidimensionality of state and trait anxiety: Factor structure of the endler multidimensional anxiety scales. Journal of Personality and Social Psychology, 60(6), 919–926.

Baiden

LaBrenz

C. A.

Asiedua-Baiden

Muehlenkamp

J. J.

(2020). Examining the intersection of race/ethnicity and sexual orientation on suicidal ideation and suicide attempt among adolescents: Findings from the 2017 Youth Risk Behavior Survey. Journal of Psychiatric Research, 125, 13–20.

Baumeister

R. F.

(1990). Suicide as escape from self. Psychological Review, 97(1), 90–113.

Beck

A. T.

Rush

A. J.

(1978). Cognitive approaches to depression and suicide. In Serban

(Ed.), Cognitive defects in the development of mental illness (pp. 235–257). Brunner/Mazel.

Brausch

A. M.

O’Connor

S. S.

Powers

J. T.

McClay

M. M.

Gregory

J. A.

Jobes

D. A.

(2019). Validating the suicide status form for the collaborative assessment and management of suicidality in a psychiatric adolescent sample. Suicide and Life-Threatening Behavior, 50(1), 263–276.

Brooks

J. R.

Hong

J. H.

Cheref

Walker

R. L.

(2020). Capability for suicide: Discrimination as a painful and provocative event. Suicide and Life-Threatening Behavior, 50, 1173–1180.

Brown

G. K.

Steer

R. A.

Henriques

G. R.

Beck

A. T.

(2005). The internal struggle between the wish to die and the wish to live: A risk factor for suicide. The American Journal of Psychiatry, 162, 1977–1979.

Brown

L. A.

Boudreaux

E. D.

Arias

S. A.

Miller

I. W.

May

A. M.

Camargo

C. A.

Bryan

C. J.

Armey

M. F.

(2020). C-SSRS performance in emergency department patients at high risk for suicide. Suicide and Life-Threatening Behavior, 50, 1097–1104.

Bryan

C. J.

Hitschfield

M. J.

Palmer

B. A.

Schak

K. M.

Roberge

E. M.

Lineberry

T. W.

(2014). Gender differences in the association of agitation and suicide attempts among psychiatric inpatients. General Hospital Psychiatry, 36, 726–731.

10.

Bryan

C. J.

May

A. M.

Thomsen

C. J.

Allen

M. H.

Cunningham

C. A.

Wine

M. D.

Taylor

K. B.

Baker

J. C.

Bryan

A. O.

Harris

J. A.

Russell

W. A.

(2020). Psychometric evaluation of the Suicide Cognitions Scale-Revised (SCS-R). Military Psychology, 34(3), 269–279.

11.

Bryan

C. J.

Rudd

M. D.

Peterson

A. L.

Young-McCaughan

Wertenberger

E. G.

(2016). The ebb and flow of the wish to live and the wish to die among suicidal military personnel. Journal of Affective Disorders, 202, 58–66.

12.

Byrne

B. M.

(2013). Structural equation modeling with AMOS: Basic concepts, applications, and programming. Routledge.

13.

Cha

C. B.

Tezanos

K. M.

Peros

O. M.

M. Y.

Ribeiro

J. D.

Nock

M. K.

Franklin

J. C.

(2018). Accounting for diversity in suicide research: Sampling and sample reporting practices in the United States. Suicide and Life-Threatening Behavior, 48(2), 131–139.

14.

Chen

F. F.

(2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modelling, 14(3), 464–504.

15.

Chen

Dai

Gao

(2019). Measurement invariance and latent mean differences of the Chinese version physical activity self-efficacy scale across gender and education levels. Journal of Sport and Health Science, 8(1), 46–54.

16.

Cheung

G. W.

Rensvold

R. B.

(2009). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modelling: A Multidisciplinary Journal, 9(2), 233–255.

17.

Chu

Buchman-Schmitt

J. M.

Stanley

I. H.

Hom

M. A.

Tucker

R. P.

Hagan

C. R.

Rogers

M. L.

Podlogar

M. C.

Chiurliza

Ringer

F. B.

Michaels

M. S.

Patros

C. H. G.

Joiner

T. E.

Jr (2017). The interpersonal theory of suicide: A systematic review and meta-analysis of a decade of cross-national research. Psychological Bulletin, 143(12), 1313–1345.

18.

Coley

R. Y.

Johnson

Simon

G. E.

Cruz

Shortreed

S. M.

(2021). Racial/ethnic disparities in the performance of prediction models for death by suicide after mental health visits. JAMA Psychiatry, 78(7), 726–734.

19.

Conrad

A. K.

Jacoby

A. M.

Jobes

D. A.

Lineberry

T. W.

Shea

Arnold Ewing

T. D.

Schmid

P. J.

Ellenbecker

S. M.

Lee

J. L.

Fritsche

Grenell

J. A.

Gehin

J. M.

Kung

(2009). A psychometric investigation of the Suicide Status Form II with a psychiatric inpatient sample. Suicide and Life-Threatening Behavior, 39(3), 307–320.

20.

Corona

C. D.

Jobes

D. A.

Nielsen

A. C.

Pedersen

C. M.

Jennings

K. W.

Lento

R. M.

Brazaitis

K. A.

(2013). Assessing and treating different suicidal states in a Danish outpatient sample. Archives of Suicide Research, 17(3), 302–312.

21.

Denneson

L. M.

Hoffmire

C. A.

Blosnich

J. R.

Dichter

M. E.

Fitelson

Holliday

Monteith

L. L.

Smolenski

D. J.

Yano

E. M.

(2021). Advancing knowledge of suicide risk and prevention among women: Introduction to the Supplement. Medical Care, 59, S1–S3.

22.

Drapeau

C. W.

McIntosh

J. L.

(2021). U.S.A suicide: 2020 Official final data. Suicide Awareness Voices of Education (SAVE).

23.

Duronsini

Aschieri

(2021). Therapeutic assessment efficacy: A meta-analysis. Psychological Assessment, 33(10), 962–972.

24.

Ellis

T. E.

Rufino

K. A.

Allen

J. G.

Fowler

J. C.

Jobes

D. A.

(2015). Impact of a suicide-specific intervention within inpatient psychiatric care: The collaborative assessment and management of suicidality. Suicide and Life-Threatening Behavior, 45(5), 556–566.

25.

Flora

D. B.

(2020). Your coefficient alpha is probably wrong, but which coefficient omega is right? A tutorial on using R to obtain better reliability estimates. Advances in Methods and Practices in Psychological Science, 3(4), 484–501.

26.

Forkmann

Teismann

Stenzel

J. S.

Glaesmer

de Beurs

(2018). Defeat and entrapment: More than meets the eye? Applying a network analysis to estimate dimensions of highly correlated constructs. BMC Medical Research and Methodology, 18, 16.

27.

Franklin

J. C.

Ribeiro

J. D.

Fox

K. R.

Bentley

K. H.

Kleiman

E. M.

Huang

Musacchio

K. M.

Jaroszewski

A. C.

Chang

B. P.

Nock

M. K.

(2017). Risk factors for suicidal thoughts and behaviors: A meta-analysis of 50 years of research. Psychological Bulletin, 143, 187–232.

28.

Goods

N. A. R.

Page

A. C.

Stritzke

W. G. K.

Kyron

M. J.

Hooke

G. R.

(2020). Daily monitoring of the wish to live and the wish to die with suicidal inpatients. In Page

A. C.

Stritzke

W. G. K.

(Eds.), Alternatives to suicide: Beyond risk and toward a life worth living (pp. 106–127). Elsevier.

29.

Green

S. B.

Yang

(2009). Reliability of summed items scores using structural equation modeling: An alternative to coefficient alpha. Psychometrika, 74, 155–167.

30.

Gutierrez

P. M.

Joiner

T. E.

Hanson

Avery

Fender

Harrison

Herns

McGowan

Stanley

I. H.

Silva

Rogers

M. L.

(2021). Clinical utility of suicide behavior and ideation measures: Implications for suicide risk assessment. Psychological Assessment, 33(1), 1–13.

31.

Hanson

W. E.

Poston

J. M.

(2011). Building confidence in psychological assessment as a therapeutic intervention: An empirically based reply to lilienfeld, garb, and wood (2011). Psychological Assessment, 23(4), 1056–1062.

32.

Hom

M. A.

Bauer

B. W.

Stanley

I. H.

Boffa

J. W.

Stage

D. L.

Capron

D. W.

Schmidt

N. B.

Joiner

T. E.

(2020). Suicide attempt survivors’ recommendations for improving mental health treatment for attempt survivors. Psychological Services, 18, 365–376.

33.

Bentler

P. M.

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

34.

Huang

Ribeiro

J. D.

Franklin

J. C.

(2020). The differences between suicide ideators and suicide attempters: Simple, complicated, or complex? Journal of Consulting and Clinical Psychology, 88(6), 554–569.

35.

IBM Corp. (2020). IBM SPSS statistics for windows (Version 27.0). IBM Corp.

36.

Jobes

D. A.

(2016). Managing suicidal risk: A collaborative approach (2nd ed.). The Guilford Press.

37.

Jobes

D. A.

(2017). Clinical assessment and treatment of suicidal risk: A critique of contemporary care and CAMS as a possible remedy. Practice Innovations, 2(4), 207–220.

38.

Jobes

D. A.

Gregorian

M. J.

Colborn

V. A.

(2018). A stepped care approach to clinical suicide prevention. Psychological Services, 15, 243–250.

39.

Jobes

D. A.

Jacoby

A. M.

Cimbolic

Hustead

L. A. T.

(1997). Assessment and treatment of suicidal clients in a university counseling center. Journal of Counseling Psychology, 44, 368–377.

40.

The Joint Commission. (2019). R3 Report Issue 18: National patient safety goal for suicide prevention. https://www.jointcommission.org/standards/r3-report/r3-report-issue-18-national-patient-safety-goal-for-suicide-prevention/#.Y35_gOzMI4i

41.

Jorgensen

T. D.

Pornprasertmanit

Schoemann

A. M.

Rosseel

(2021). SemTools: Useful tools for structural equation modeling. R package version 0.5-5. https://CRAN.R-project.org/package=semTools

42.

Kovacs

Beck

A. T.

(1977). The wish to die and the wish to live in attempted suicides. The Journal of Clinical Psychology, 33(2), 361–365.

43.

Landry

Ali

Williams

Rehm

Bonham

(2018). Lack of diversity in genomic databases is a barrier to translating precision medicine research into practice. Health Affairs, 37(5), 780–785.

44.

Michel

(2021). Suicide models and treatment models are separate entities. What does it mean for clinical suicide prevention? Journal of Environmental Research and Public Health, 18(10), 5301.

45.

Millner

A. J.

Robinaugh

D. J.

Nock

M. K.

(2020). Advancing the understanding of suicide: The need for formal theory and rigorous descriptive research. Trends in Cognitive Science, 24(9), 704–716.

46.

Myles

Tan

S. Y.

Large

(2021). Zero Suicide requires a radical reimagining of inpatient care. Australian & New Zealand Journal of Psychiatry, 55(10), 935–936.

47.

Naherniak

Bhaskaran

Sareen

Wang

Bolton

J. M.

(2019). Ambivalence about living and the risk for future suicide attempts: A longitudinal analysis. The Primary Care Companion for CNS Disorders, 21(2), 18m02361.

48.

Nock

M. K.

Holmberg

E. B.

Photos

V. I.

Michel

B. D.

(2007). Self-Injurious Thoughts and Behaviors Interview: Development, reliability, and validity in an adolescent sample. Psychological Assessment, 19(3), 309–317.

49.

O’Connor

S. S.

Jobes

D. A.

Yeargin

M. K.

Fitzgerald

M. E.

Rodriguez

V. M.

Conrad

A. K.

Lineberry

T. W.

(2012). A cross-sectional investigation of the suicidal spectrum: Typologies of suicidality based on ambivalence about living and dying. Comprehensive Psychiatry, 53, 461–467.

50.

Podlogar

M. C.

Joiner

T. E.

(2020). Allowing for nondisclosure in high suicide risk groups. Assessment, 27(3), 547–559.

51.

Polanco-Roman

Miranda

(2013). Culturally related stress, hopelessness, and vulnerability to depressive symptoms and suicidal ideation in emerging adulthood. Behavior Therapy, 44(1), 75–87.

52.

Poston

J. M.

Hanson

W. E

. (2010). Meta-analysis of psychological assessment as a therapeutic intervention. Psychological Assessment, 22(2), 203–212.

53.

Putnick

D. L.

Bornstein

M. H.

(2016). Measurement invariance conventions and reporting: The state of the art and future directions for psychological research. Developmental Research, 41, 71–90.

54.

Raykov

Marcoulides

G. A.

C. H.

(2012). Measurement invariance for latent constructs in multiple populations: A critical view and refocus. Educational and Psychological Measurement, 72(6), 954–974.

55.

R Core Team. (2019). R: A language and environment for statistical computing. R Foundation for Statistical Computing.

56.

Rhemtulla

Brosseau-Liard

P. E.

Savalei

(2012). When can categorical variables be treated as continuous? A comparison of robust continuous and categorical SEM estimation methods under suboptimal conditions. Psychological Methods, 17(3), 354–373.

57.

Romanowicz

O’Connor

S. S.

Schak

K. M.

Swintak

C. C.

Lineberry

T. W.

(2013). Use of the Suicide Status Form-II to investigate correlates of suicide risk factors in psychiatrically hospitalized children. Journal of Affective Disorders, 151(2), 467–473.

58.

Rosseel

(2012). Lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1–36.

59.

Shneidman

E. S.

(1987). A psychological approach to suicide. In VandenBos

G. R.

Bryant

B. K.

(Eds.), Cataclysms, crises, and catastrophes: Psychology in action (pp. 147–183). American Psychological Association.

60.

Standley

(2020). Expanding our paradigms: Intersectional and socioecological approaches to suicide prevention. Death Studies, 46(1), 224–232.

61.

Stanley

Brown

G. K.

Brenner

L. A.

Galfalvy

H. C.

Currier

G. W.

Knox

K. L.

Chaudhary

S. R.

Bush

A. L.

Green

K. L.

(2018). Comparison of the safety planning intervention with follow-up vs usual care of suicidal patients treated in the emergency department. JAMA Psychiatry, 75(9), 894–900.

62.

Stephenson

Pena-Shaff

Quirk

(2006). Predictors of college student suicidal ideation: Gender differences. College Student Journal, 40(1), 109–117.

63.

Substance Abuse and Mental Health Services Administration. (2021). Key substance use and mental health indicators in the United States: Results from the 2020 National Survey on Drug Use and Health (HHS Publication No. PEP21-07-01-003, NSDUH Series H-56). Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. https://www.samhsa.gov/data/

64.

Swift

Trusty

Penix

(2021). The effectiveness of the Collaborative Assessment and Management of Suicidality (CAMS) compared to alternative treatment conditions: A meta-analysis. Suicide and Life-Threatening Behavior, 51(5), 882–896.

65.

Talley

Warner

Ş. L.

Perry

Brissette

Consiglio

F. L.

Capri

Violano

Coker

K. L.

(2021). Understanding situational factors and conditions contributing to suicide among Black youth and young adults. Aggression and Violent Behavior, 58, 101614.

66.

Tucker

R. P

. (2019). Suicide in transgender veterans: Prevalence, prevention, and implications of current policy. Perspectives on Psychological Science, 14(3), 452–468.

67.

Valois

R. F.

Zullig

K. J.

Kammermann

S. K.

Kershner

(2013). Relationships between adolescent sexual risk behaviors and emotional self-efficacy. American Journal of Sexuality Education, 8(1–2), 36–55.

68.

Vandenberg

R. J.

Lance

C. E.

(2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1), 4–70.

69.

Ward-Ciesielski

E. F.

Rizvi

S. L.

(2021). The potential iatrogenic effects of psychiatric hospitalization for suicidal behavior: A critical review and recommendations for research. Clinical Psychology: Science and Practice, 28(1), 60.