Abstract
Mental disorder is widespread in correctional settings, as is the potential for suicide and self-harm among detained individuals. The ability of the Personality Assessment Screener (PAS) to identify potential mental health problems has been the focus of recent research in corrections, but few studies have examined the prediction of crisis events related to self-injuring thoughts and behaviors. We explored the utility of repeated PAS administrations among men incarcerated in prison (N = 270) to identify those at risk for a subsequent crisis event involving potential self-harm. Baseline PAS total scores predicted the odds of experiencing a crisis event (odds ratio = 3.76–4.46) and, importantly, changes in PAS total scores over time incrementally predicted such events beyond scores assessed at a single time-point, t(49) = –2.14, p = .037. These preliminary findings demonstrate promise for the PAS as an indicator of potential self-injuring crisis events when administered at intake and regular intervals throughout incarceration.
Keywords
Current estimates suggest that the United States detains over 2.2 million individuals in state and federal prisons and local jails (Carson, 2018; Zeng, 2018). The percentage of persons with diagnosed or undiagnosed mental disorders in correctional environments appears to be well beyond that of the general population in the United States. A national survey conducted between 2011 and 2012 reported that more than 26% of individuals in jails and 14% in federal prisons met the threshold for serious psychological distress within a 30-day period (Bronson & Berzofsky, 2017). Epidemiological studies indicate that major depression and psychotic illnesses have the highest prevalence in prisons, with approximately one in seven individuals recently experiencing a serious mental disorder (Fazel & Seewald, 2012). However, despite this high rate of mental health problems, they are frequently underdiagnosed and undertreated in such settings (Gould et al., 2018; Martin et al., 2013).
One important implication of undetected mental disorder involves the potential risk for suicide and self-harm, given that they are an important predictor of subsequent suicidal and self-injurious behaviors in both community (Turecki & Brent, 2016) and correctional settings (Fazel et al., 2008; Fazel & Seewald, 2012). Mental health risk factors for potential self-harm may be particularly salient within correctional settings, as between 2012 and 2013, the rates of suicide in North America increased 9% and accounted for more than a third of deaths in correctional institutions (Noonan & Ginder, 2013). Similarly, a recent study reported that the United States had an annual suicide rate of 23 per 100,000 incarcerated individuals (Fazel et al., 2017). As such, early identification of relevant risk factors could be critical in preventing subsequent suicidal behaviors in such environments. Even when such behaviors do not result in death or serious injury, suicidal and self-harm behaviors may be extremely disruptive in correctional settings for the individual, others incarcerated in such facilities (e.g., cell mates), and the mental health and security staff who are responsible for maintaining institutional order and the health and welfare of individuals under their supervision (Kapoor, 2015; Metzner & Hayes, 2006; Thienhaus, 2007).
Unfortunately, efforts to identify serious mental disorders and other issues of clinical concern within corrections historically have been limited by an overabundance of individuals with such problems in conjunction with a scarcity of appropriate assessment resources to detect them (Grisso, 2005). Reflecting this poor state of affairs, research in the early 1990s indicated that only 32.5% of those in jail settings who had a severe mental illness were positively identified at intake, with identification rates of only 45% for psychotic disorders and 7% for major depression (Teplin, 1990). In a more recent study, Brooker et al. (2009) similarly found that only 25% to 33% of individuals with serious mental illness in prison were accurately identified by prison staff using current screening practices.
In part, because of the high rates of undiagnosed psychiatric disorders, in recent years, more attention has been given to the potential application of mental health screening instruments in correctional settings. These brief measures may offer a time- and cost-effective method of identifying those who are in need of further and more extensive evaluation procedures (Grisso et al., 2005; Lurigio & Swartz, 2006). Indeed, utilization of screening instruments has been shown to significantly increase early detection of mental illness in these environments (Martin et al., 2013). However, many of the instruments being used have little validation evidence, and it remains unclear the extent to which these measures in correctional settings demonstrate adequate psychometric properties, such as acceptable diagnostic accuracy or predictive validity in relation to adverse outcomes or events, such as suicidal or self-injurious acts or other behaviors requiring some type of institutional crisis management (for review, see Gould et al., 2018; Martin et al., 2013).
One mental health screening instrument that increasingly has been the focus of validation research in forensic and correctional settings (Edens et al., 2018; Kelley et al., 2018) is the Personality Assessment Screener (PAS; Morey, 1997), a 22-item screening measure derived from the 344-item Personality Assessment Inventory (PAI; Morey, 1991). The primary purpose of the PAS is to efficiently identify individuals likely to experience clinically significant problems across 10 domains, including suicidal ideation. Despite the fact that the PAS has garnered relatively less empirical attention in corrections than its parent instrument, studies have demonstrated support for its utility as a screen for the presence of general psychopathology and other indicators of distress, dysfunction, and maladjustment (Creech et al., 2010; Edens et al., 2018; Harrison & Rogers, 2007; Kelley et al., 2018; Porcerelli et al., 2012, 2015). For instance, in a primary care sample, the PAS total score has been effective in identifying those with a history of childhood and adult abuse (Porcerelli et al., 2015), as well as mood disorders, personality disorders, and alcohol use disorder (Porcerelli et al., 2012).
Given its brevity and ease of administration, the PAS may have considerable utility in the context of screening evaluations performed in resource-limited correctional settings (Kelley et al., 2018). Several studies (Edens et al., 2018; Harrison & Rogers, 2007; Kelley et al., 2018) investigating the PAS in such environments found that, with some modification to recommended cut scores, the PAS total score can effectively identify justice-involved individuals with clinically significant elevations on the full PAI. The PAS manual recommends a cutoff score of 19 for identifying the immediate presence of a clinically significant problem (sensitivity = 84.7, specificity = 78.7 using clinical and community norms). However, Kelley et al. (2018) suggested a cutoff score of 29 for criminal justice–involved individuals, which demonstrated greater discrimination and greater utility as compared with the manual-derived cut score of 19 (sensitivity = 52.9, specificity = 99.2 using community norms; sensitivity = 72.7, specificity = 76.6, using corrections norms). PAS element scores also have demonstrated significant relationships with various clinically or legally relevant criterion measures (Edens et al., 2018; Harrison & Rogers, 2007; Kelley et al., 2018; McLeod et al., 2004). For instance, Negative Affect (NA) has been significantly associated with dissociative experiences (Kelley et al., 2018; McLeod et al., 2004), degree of offense recall (McLeod et al., 2004), and presence of major depression in correctional environments (Harrison & Rogers, 2007). Notably, however, only a few of the element scores thus far demonstrate significant associations with theoretically relevant external correlates beyond what is captured by the PAS total score (Edens et al., 2018; Kelley et al., 2018).
Although the focus of some research in correctional settings, examination of the ability of the PAS to identify individuals who are at heightened risk for suicidal and/or self-injurious behavior has received limited attention to date. Harrison and Rogers (2007) reported associations between measures of major depression and suicide risk and the NA and Suicidal Thinking (ST) element scores in a jail sample. NA is a measure of personal distress and indicates the possible presence of anxiety or depression symptoms, whereas ST screens for the presence of past or current suicidal ideation. NA scores outperformed other screening measures in the prediction of major depression with an overall hit rate of 80%. ST score also demonstrated some promise as a predictor of suicide risk (as operationalized by the Suicide Probability Scale [SPS]; Cull & Gill, 1988), with a negative predictive power of .96 and sensitivity of .85. Nevertheless, 15% of individuals who were classified as high risk on the SPS scored zero on ST, leading to concerns regarding potential false-negative classifications based on this element score in high-volume screenings in jail facilities. Furthermore, it should be noted that the SPS does not directly measure behavioral outcomes (e.g., counts of suicidal ideation and suicide attempts) and thus the generalizability of these findings to more concrete outcomes are unknown. A more recent study reported that ST was strongly associated with a history of self-reported suicide attempts (r = .52; area under the curve [AUC] = .80), even more so than the total score (r = .30; AUC = .69; Kelley et al., 2018). However, there has yet to be a prospective study examining the PAS’s ability to predict the occurrence of serious adverse outcomes, such as crisis events resulting from suicidal or other self-injurious thoughts or behaviors.
Moreover, research investigating the PAS generally utilizes PAS scores at intake (i.e., baseline) to predict long-term outcomes. There is currently no research investigating the dynamic predictive validity of the PAS over time in relation to institutional crisis events concerning individuals who may engage in self-harm or suicide. Specifically, the question of the extent to which repeated PAS screenings can predict the occurrence of such events above and beyond a one-time intake screening has never been examined. Research to date has only examined the validity of PAS scores at a singular time point, and there is little understanding of the potential meaningfulness of changes in PAS scores over time.
The present study explores the utility of the PAS as an indicator of overall and proximal risk of an institutional crisis event occurring as a result of ideation or behaviors related to potential suicide or self-harm, assessed in an incarcerated sample across five time points spanning 1 year. Within this longitudinal dataset, three questions are of specific interest:
These research questions are intended to explore the extent to which the PAS is associated with long-term and proximal risk for such an event, with an additional consideration of whether changes in PAS scores over time are more indicative of risk than PAS scores assessed at a single time point.
Method
Data were obtained from a public dataset of a research project sponsored by the U.S. Department of Justice examining persons incarcerated in the Colorado Department of Corrections prison system over approximately 1 year. A full description of the sample, setting, and data collection procedures has been published previously (O’Keefe et al., 2013).
Participants
Participants included 270 men with ages ranging from 17 to 59 years (M = 31.8, SD = 9.1). The racial/ethnic composition of the sample was 40% White/Caucasian, 36% Hispanic, 18% Black/African American, 4% Native American, and 1% Asian. In accordance with O’Keefe et al.’s (2013) study examining psychological effects of administrative segregation, participants were sampled from a variety of correctional placements: administrative segregation (47%), general population (28%), and psychiatric prison (25%). Participants were given the option to voluntarily withdraw consent and discontinue testing at any time; a total of 23 participants chose to voluntarily withdraw during the 1-year study period and were not included in the final sample of 270 participants. During the data collection period, a total of 30 participants (11.1%) were reported by mental health staff as experiencing a mental health crisis related to self-injuring thoughts and behavior (SITB).
Assessments and Measures
PAS
The PAS (Morey, 1997) is a 22-item self-report questionnaire designed as a general psychopathology screening measure derived from its parent instrument, the PAI (Morey, 1991). In addition to the total score, the PAS also offers an evaluation of clinical functioning across 10 major domains or “elements”: NA, Acting Out, Health Problems, Psychotic Features, Social Withdrawal, Hostile Control, ST, Alienation, Alcohol Problem, and Anger Control. Each of these elements consists of two or three items selected from existing PAI items found to optimally differentiate respondents with significant pathology from those with minimal clinical difficulties. Responses are provided on a 4-point rating scale ranging from false, not at all true to very true with a possible total raw score of 0 to 66.
Reliability studies in the PAS manual demonstrated internal consistencies of .75 for the PAS total score, .73 for NA, and .84 for ST in a clinical normative sample (Morey, 1997). In the current sample, internal consistency ranged from a coefficient alpha of .76 to .79 (mean α = .78 across the five time points) for the total score, .61 to .70 for NA (mean α = .66), and .86 to .95 to ST (mean α = .91). Morey (1997) reported test–retest reliability coefficients of .79 for the total score, .73 for NA, and .84 for ST. The current sample demonstrated appreciable stability in PAS total score over time with test–retest scores ranging from .76 to .84 between administration periods (McCredie et al., 2020).
Self-Injuring Thoughts and Behavior Crisis Events
Self-injuring thoughts and behavior crisis events (henceforth referred to as “SITB crisis events” or simply “crisis events”) were defined and recorded as a situation in which participants had an unscheduled mental health visit and immediate psychological intervention was necessary. These events were originally coded by clinicians for both suicidal and self-harm ideation and behavior on a 3-point scale (1—suicidal/self-harm ideation, 2—self-harming behavior, 3—attempted suicide); Categories 2 and 3 were combined by the original coders into a fourth category, suicide attempt/self-harming behavior. Due to the nature of the data available and for the purpose of this study, crisis events were collapsed such that the crises were dichotomized as either the absence (0) or the presence (1) of a suicidal or self-harming event. This was done given the low base rate of SITB crisis events (11.1% of total sample; 8.7% of administrative segregation; 2.6% of general population; 25.4% of psychiatric prison), which would have precluded analysis at the crisis event–level due to attenuated power. In addition, the number of days that the individual had participated in the study until the crisis event was recorded.
Procedure
The study protocol spanned 1 year with five testing sessions administered at 3-month intervals, beginning with the initial administration at the time of study consent. Measures were completed at baseline, 3 months, 6 months, 9 months, and 12 months (i.e., Time 1 to Time 5). In addition, individuals in the administrative segregation group were on a waitlist for an average of 3 months and thus some were administered a “pre-baseline” measure during this period. All study materials were completed on pencil-and-paper with as-needed assistance from the field researchers. Participants were compensated US$10 per testing session. The study was approved and conducted under the institutional review board of the University of Colorado at Colorado Springs.
Analyses
A total of three sets of analyses were conducted to examine group mean differences in PAS scores between individuals with and without an SITB crisis event. The first set of analyses examined mean differences of baseline PAS scores between those with a crisis event (n = 30) versus no crisis event (n = 240) using a series of three independent-samples t tests. Occurrence of a crisis event was entered as the independent grouping variable, and baseline scores on the PAS total score, NA, and ST indicators were evaluated as dependent variables. In the remaining analyses, a subset of the sample was derived, which included those who experienced a crisis event during the study duration and matched individuals without a crisis event (n = 60 participants). Participants in the subsample with a crisis (n = 30) were matched with participants without a crisis (n = 30) on the basis of obtaining similar baseline PAS total scores, to control for baseline symptomatology. No other matching criteria were used.
The second set of analyses examined the extent to which there were group differences between individuals with and without an SITB crisis event in PAS scores as assessed in the time period just prior to the crisis event, controlling for any differences observed at baseline. To do so, we calculated “proximal” scores by identifying PAS total, NA, and ST scores at the time period immediately prior to the individuals’ recorded date of a crisis event, with similar steps taken for the matched subsample. Four date ranges were specified: An incident occurring 0 to 90 days would be in the “Time 1” range, 91 to 181 days would be in the “Time 2” range, and so on. For instance, an individual with a recorded crisis event on day 161 would fall in the “Time 2” range and the PAS scores from this period would be used as the proximal score, because it is the nearest data collection point prior to the crisis event. Group differences were evaluated using a series of three independent-samples t tests, with the occurrence of a crisis event entered as the independent grouping variable, and proximal scores on the PAS total score, NA, and ST indicators evaluated as dependent variables.
The third set of analyses examined the extent to which changes in PAS scores over the time period just prior to the recorded crisis event are associated with SITB risk above and beyond proximal PAS scores, to reflect dynamic indications of risk rather than static predictors. To do so, the researchers calculated “pre-proximal” PAS scores, or PAS scores at the time period just prior to the proximal time period. For instance, a participant with a recorded crisis event between Times 3 and 4 (e.g., about 190 days into the start of the study) would have a proximal score collected at Time 3 and a pre-proximal score at Time 2. A series of three linear regressions were conducted using pre-proximal PAS total, NA, and ST scores as a predictor of proximal PAS scores. The unstandardized residuals from these analyses were saved as residualized change scores, indicating the difference between the observed and predicted proximal scores by the regression model. Three independent-samples t tests (PAS total, NA, and ST) were conducted to analyze differences in these residualized change scores between the dichotomized present/absent crisis event groups. Occurrence of a crisis event was entered as the independent grouping variable, and residualized change scores for the PAS total score, NA, and ST indicators were evaluated as dependent variables. Results of this analysis indicate whether changes in PAS scores from the pre-proximal to proximal period were significantly different between groups.
Results
Group Mean Differences in PAS Scores
Table 1 presents descriptive statistics of baseline PAS scores, residualized change scores of baseline PAS scores predicting proximal PAS scores, residualized change scores of pre-proximal PAS scores predicting Proximal PAS scores, and Cohen’s d effect sizes for the respective tests. The first series of independent-samples t tests examined the extent to which mean differences between those with and without a crisis event on the PAS total score, NA, and ST indicators were observed at baseline in the full study sample (N = 270). Mean differences between the two groups were statistically significant at baseline for PAS total, t(268) = –3.98, p < .001, NA, t(268) = –3.97, p < .001, and ST, t(33.22) = –3.52, p = .001, such that large effect size differences between participants with and without a crisis event were observed in the positive direction on all three PAS variables assessed at baseline.
Descriptive Statistics and Effect Sizes for SITB and No SITB Groups’ PAS Scores
Note. “Residual” scores are residualized change scores computed using regression analyses. SITB = self-injuring thoughts and behaviors. PAS = Personality Assessment Screener (Morey, 1997); NA = Negative Affect; ST = Suicidal Thinking.
n = 270. bn = 267. cn = 60. dn = 57. en = 51. fn = 50.
A second series of independent-samples t tests were conducted investigating group differences in proximal PAS scores between participants with and without a crisis event in the matched sample (n = 60). Results indicated that there were no significant group differences between those with and without a crisis on the proximal PAS total score, t(58) = –1.69, p = .096; NA, t(58) = –1.51, p = .137; and ST scores, t(58) = –1.43, p = .159. However, to remove the contribution of preexisting group differences at baseline as noted above, additional analyses were conducted utilizing unstandardized residualized scores from regression analyses of baseline PAS score predicting the event-proximal PAS score. There were significant group differences of a moderate effect size for PAS total scores, t(58) = –2.12, p = .038, indicating that individuals with a crisis demonstrated higher PAS total scores at the proximal time period than those without a crisis, after controlling for baseline PAS scores. However, participants’ proximal scores on PAS NA, t(58) = –1.36, p = .179, and ST, t(55) = –1.19, p = .239, did not demonstrate significant group differences above and beyond those indicated by baseline scores, although small effect size differences in the predicted direction were evident.
The final series of analyses examined group differences in changes of PAS scores just prior to the recorded suicidal incident between individuals with and without a crisis in the matched sample (n = 60). Residualized change scores from regression analyses of pre-proximal PAS total scores predicting proximal PAS total scores demonstrated significant group differences of a moderate effect size in the positive direction, t(49) = –2.14, p = .037, such that those with a crisis demonstrated larger changes in the PAS total score from the pre-proximal to proximal time period than their counterparts without a crisis. However, there were no significant group differences on PAS NA, t(49) = –1.60, p = .116, and ST, t(48) = –1.42, p = .163, change scores from the pre-proximal to proximal period, although small effect size differences in the positive direction were again observed for these indicators.
Group Classification Statistics
Although mean scores are useful in evaluating the magnitude of differences across groups, it is also informative to consider the performance of the PAS in relation to classification accuracy. First, we conducted a receiver operating characteristic (ROC) analysis, which produces an AUC value that represents the likelihood that a randomly selected individual positive for the criterion measure would obtain a higher PAS score than a randomly selected individual who was negative for the criterion measure. A value of .50 represents chance classification accuracy and 1.00 represents perfect accuracy. The AUC for PAS scores in identifying individuals who experienced a crisis event was .70 (SE = .049) for total scores, .72 for NA (SE = .047), and .71 (SE = .054) for ST.
In addition, odds ratios (ORs) were calculated to determine the odds that individuals scoring above the PAS total cut scores proposed by the PAS manual (i.e., 19) and Kelley et al. (2018) (i.e., 29) would experience an SITB crisis event during the duration of the study. We should note that these cut scores were not derived to serve specifically as indicators of suicidal or self-harm risk, but rather were selected to identify individuals who were likely to demonstrate a wide variety of mental health problems. Nevertheless, these scores were strongly related to the risk of a crisis event occurring: At the cut score of 19, the OR was 4.46 (95% confidence interval: [0.59, 33.9]), indicating that individuals scoring at or above a total score of 19 on the PAS had almost four and a half times higher odds of experiencing an SITB crisis event than those scoring below 19. At the cut score of 19, sensitivity to detect a crisis event was .97 and specificity was .13. At the cut score of 29, the OR was 3.76 (95% confidence interval: [1.39, 10.15]), indicating that individuals scoring at or above a total score of 29 had almost 4 times higher odds of experiencing an SITB crisis event than those scoring below 29. At a cut score of 29, sensitivity to detect a crisis event was .83 and specificity was .43.
Discussion
The current study investigated the utility of the PAS in predicting future crisis events related to suicidal and self-harming thoughts and behaviors in a prison environment during a year-long study with five data collection periods. PAS total, NA, and ST scores were compared at baseline administration as indicators of subsequent crisis events throughout the duration of the study. In addition, PAS scores at the nearest prior data collection period to the crisis were examined to determine the extent to which repeated PAS administrations increment baseline scores in identifying differences between individuals who did and did not experience a crisis event. Finally, the present study investigated changes in PAS scores over time as possible indicators of crisis events to determine if a more dynamic approach would increment baseline PAS scores as predictors of such events in the future.
Consistent with previous findings on the relationship between PAS scores and potential self-harm (Harrison & Rogers, 2007; Kelley et al., 2018), there were significant, large effect size differences in PAS total, NA, and ST baseline scores between study participants who experienced a crisis event and those who did not. Those who experienced a crisis during the study demonstrated higher PAS scores across all three indicators as compared with individuals who did not experience such a crisis. Furthermore, baseline scores were incremented by the inclusion of the PAS total score assessed during the data collection period just prior to the crisis, suggesting that there may be some benefit to repeated administrations of the PAS to augment baseline suicide risk predictions. However, this finding was only statistically significant for the PAS total score, not for the specific ST and NA element scores. Finally, short-term changes in PAS total scores additionally incremented the ability of static scores to identify risk for suicidal crises with medium effect sizes, suggesting that change in the PAS total score over time offers additional meaningful information beyond scores assessed at a single time point. However, this finding was again only demonstrated for the PAS total score and did not achieve significance with the ST or NA elements, although small to moderate effects were observed in the predicted direction with these elements.
The lack of significant findings in the analyses involving NA and ST, beyond baseline scores, may be attributed to the structure of the PAS. Specifically, the PAS was constructed with the intention of being primarily interpreted at the full-scale level rather than by its element scores. The PAS total score utilizes all 22 items and can range from zero to 66 raw score points, whereas elements’ scores are comprised of two or three items and can range from zero to six or nine raw score points. Due perhaps to the greater reliability or greater coverage of the total score, small changes that are not captured through element scores appear more likely to be captured in the total score, as demonstrated in previous studies (Creech et al., 2010; Edens et al., 2018; Kelley et al., 2018; Porcerelli et al., 2012, 2015). Furthermore, the ST element score consists of two items assessing the presence of historical suicidal thoughts as opposed to current suicidal ideation, thus potentially limiting the sensitivity of ST to changes over time. As such, changes in risk for an SITB crisis event appear to be best represented by the PAS total score rather than this particular element score.
The present study provides an important first step in conducting research regarding the prediction of future suicidal and self-harming behavior using the PAS. Previous research investigating the PAS and suicidal crises has used past suicide attempts (Kelley et al., 2018) or criterion measures such as the PAI or SPS (Harrison & Rogers, 2007; Kelley et al., 2018) to compare with PAS scores, and as such, the present study is the first to examine prospective self-harm crises and their relation to prior PAS administrations. Results of the current study suggest that the PAS total cut score of 19 indicated in the manual (Morey, 1997), derived to identify individuals in need of further psychological evaluation, may also serve as a useful screening threshold for individuals at risk of a suicidal or self-harm crisis event. Specifically, participants with scores exceeding 19 at baseline were significantly more likely to experience an SITB crisis event during the duration of the study than those scoring below 19. However, it should be reiterated that this cutoff score was not intended to serve as a screening measure for suicide or self-harm risk specifically and should not be relied upon as a sole indicator of risk in correctional or other samples.
This study also appears to be the first to examine the PAS as a possible indicator of risk from a dynamic perspective. The current findings suggest that changes in the PAS total score over time offer meaningful information regarding the identification of self-harm risk above and beyond that provided by static scores assessed just prior to the crisis event. Although the PAS was not developed with the intention of assessing change, given both the brevity and breadth of the measure, these findings suggest that small changes on the PAS may indicate meaningful changes in psychological functioning. These findings are further supported by previous research which suggest that small temporal changes on the PAS correlate meaningfully with changes on related self-report instruments (McCredie et al., 2020). Thus, the present study contributes additional support to the notion that even relatively small changes on the PAS may indicate meaningful psychological change, while also offering initial empirical evidence for the utility of a dynamic approach to crisis risk assessment. As such, correctional institutions may benefit from repeated administration of risk screening measures at regular intervals, as intake screening alone is unable to alert correctional staff to meaningful changes in incarcerated individuals’ functioning. Future research should seek to build upon these findings through prospective examinations of the PAS as an indicator of crisis risk in a variety of settings and samples, including clinical populations.
Some limitations of the current research should be noted to inform future work in this area. First, this sample only included men from a variety of correctional placements, including administrative segregation and a psychiatric prison. As such, it remains to be seen whether the results will generalize to women in similar correctional settings, the general prison population, or individuals in other clinical and community settings. Several limitations in the operationalization of suicidal and self-harming behavior also warrant further research attention. The base rate of crisis events in the current sample was relatively low (11.1%), and, although this base rate is not unusual for a phenomenon such as suicidality, replication with a larger sample may provide results more specific to discrete behavioral outcomes (e.g., overt suicidal acts as opposed to verbal threats or self-reported ideation). Likewise, crisis events as defined in the current study encompassed pragmatically important but somewhat conceptually distinct manifestations of potential self-directed harm, including ideation, self-injury, and suicide attempts (Silverman, 2006), and it is possible that a focus on one specific behavior (e.g., suicide attempts) might offer somewhat different results. Unfortunately, the nature of the present data does not allow for the parsing out of such analyses, as the original coders collapsed self-harming behavior and suicide attempts into a single category, and the low frequency of some categories would result in effect size estimates of questionable generalizability due to attenuated power. It is also important to note that the reported crisis events only represent those individuals who informed staff about their current mental state and/or overt self-destructive acts and does not include those who experience suicidal/self-harm ideation or engage in such behaviors but go undetected. Furthermore, participants were matched only on PAS total score, given that the nature of the dataset did not allow demographic data to be tied to the examined variables. It is possible that there may be other relevant differences between the matched samples that cannot be identified with the current data. Thus, due to the relatively small matched sample size and lack of control over data collection and operationalization of SITB events, the current findings should be considered preliminary and, from which, future research should expand upon. That said, from a clinical (as well as institutional management) perspective, all of the categories encapsulated in our criterion measure are of considerable practical consequence, as such documented “SITB crisis events” represent situations with the potential for serious injury or even death, as well as considerable liability exposure for the institution and individual staff members responsible for the health and welfare of persons detained in these facilities (Daniel, 2009; Daou & Noroian, 2018).
The limitations of this particular research notwithstanding, the PAS does appear to show initial promise as an indicator of risk for suicidal and self-injurious behaviors and ideation when administered at intake and regular intervals throughout incarceration. In doing so, the PAS may serve as a time- and cost-efficient measure for staff members and evaluators in correctional settings. However, additional studies are needed to further explore the potential role of the PAS as a static and dynamic suicide risk assessment tool in correctional and other settings, particularly research using longitudinal designs and prospective outcome measures.
Footnotes
Authors’ Note:
Leslie C. Morey is the author of the Personality Assessment Screener and derives royalties from its sale.
