Checking the Math: Do Restrictive Housing and Mental Health Need Add Up to Psychological Deterioration?

Abstract

The current study sought to determine whether restrictive housing leads to psychological deterioration. A growth mixture modeling analysis of clinician ratings on the Brief Psychiatric Rating Scale identified two classes of participants in a sample of 266 state prison inmates: A large group of prisoners whose psychological status gradually improved over the course of a year and a small group of inmates whose psychological status deteriorated over the course of a year. Inmates with a history of mental health need were significantly more likely to experience severe psychological reactions to administrative segregation (AS) than inmates with no history of mental health need, although inmates with a history of mental health need were just as likely to experience severe psychological deterioration in general population as in AS. These results indicate that psychological deterioration in mentally ill inmates may have less to do with AS than with incarceration in general.

Keywords

restrictive housing administrative segregation psychological deterioration mentally ill inmates

Virtually every jurisdiction in the United States uses some form of restrictive housing to separate hard-to-manage prisoners from the general inmate population (Association of State Correctional Administrators-Liman, 2015). Although there is growing concern among administrators in nearly every jurisdiction that long-term placement in restrictive housing can have a deleterious effect on the mental health status and psychological adjustment of inmates so housed, there has not been a corresponding decrease in the use of restrictive housing for hard-to-manage inmates, except in a small number of jurisdictions (U.S. Department of Justice, 2017). The purpose of the current investigation was to explore the possibility that placement in restrictive housing has a significant negative impact on the psychological well-being and mental health status of inmates with and without a history of mental health problems. If it can be shown that restrictive housing has a demonstrably negative effect on the psychological well-being and adjustment of inmates then it would add urgency to current efforts to find alternatives to restrictive housing.

Literature Review

With most prison systems continuing to rely on long-term segregation to manage disruptive inmates and security threat groups, attention has turned to the effect such confinement has on the short- and long-term psychological adjustment and well-being of inmates housed in these institutions, facilities, and units. Larger jurisdictions often dedicate an entire institution to housing disruptive inmates and gang members in long-term segregation; smaller jurisdictions will more often set aside a unit within a larger institution to house individuals who create problems within the system. In everyday parlance, these facilities and units are referred to as supermaximum security (supermax) prisons, institutional control units, administrative segregation, or the segregation housing unit (SHU). The stated purpose of supermax prisons is to reduce institutional violence toward staff and other inmates. Research on this issue is mixed, however. One study, for instance, found that supermax prisons had no effect on inmate-on-inmate violence, although they did reduce violence toward staff under some circumstances (Briggs, Sundt, & Catellano, 2003). Another possible consequence of supermax confinement is one that many correctional administrators would rather not discuss, that being a rise in psychological deterioration in what is sometimes referred to as SHU syndrome.

SHU syndrome was originally defined by Grassian (1983, 2006) who classified it as a “specific psychiatric syndrome” (2006, p. 335) characterized by a unique set of symptom patterns: stupor and delirium, perceptual anomalies and hallucinations, affective difficulties like anxiety and depression, disturbances in memory and concentration, abnormalities in thought content and process, and problems with impulse control. This syndrome was attributed to long-term solitary confinement in administrative segregation or segregation housing units (i.e., SHU). Although inmates with a history of mental illness are at greatest risk for SHU syndrome, an inmate need not have a prior history of psychiatric disorder or treatment to suffer from its effects (Grassian & Friedman, 1986; Kupers, 1999). Early support for the assertion that placement in restrictive houses places an inmate at increased risk for psychological deterioration was based almost exclusively on anecdotal accounts (Jackson, 2001), case histories (Kupers, 2016), and uncontrolled studies on symptom patterns in offenders housed in lockdown units and facilities (Blanchette, 2001; Brodsky & Scogin, 1988; Haney, 2003). The general assertion that restrictive housing causes psychological deterioration in general and SHU syndrome in particular remains unverified and the symptom patterns proposed for SHU syndrome have yet to be fully replicated (Morgan et al., 2016). More research is obviously required.

In answering the call for more research, several studies compared prisoners in restrictive housing or administrative segregation with general population inmates in an effort to gauge the effect of segregation on mental health symptoms, although the results have been mixed and inconclusive. Suedfeld, Ramirez, Deaton, and Baker-Brown (1982), for instance, observed no differences between segregated and general population inmates on a range of psychosocial measures. Hodgins and Côté (1991), on the contrary, discovered that schizophrenia and bipolar disorder were more commonly observed in segregation than in general population, but that depression was more common in general population. In a cross-sectional study comparing preexisting characteristics of inmates housed in segregation and inmates housed in general population, Motiuk and Blanchette (2001) determined that inmates assigned to restrictive housing were more likely to have prior criminal justice involvement, greater family disruption, more antisocial attitudes, weaker educational skills, and stronger antisocial associations than general population inmates. To adequately address the issue of whether long-term confinement in a segregation unit has a deleterious effect on an inmate’s mental health status, however, a longitudinal investigation is required.

The first study to address the longitudinal effects of restrictive housing or administrative segregation on psychological status was conducted by Zinger, Wichmann, and Andrews (2001) using a group of 60 Canadian prison inmates. The inmates were interviewed by a trained research assistant and completed a battery of tests at three different points in time: baseline, 30 days postbaseline, and 60-days postbaseline. Although prisoners in administrative segregation displayed higher risk levels and poorer mental health functioning and psychological adjustment compared with general population inmates, there was no evidence that inmates in restrictive housing suffered psychological deterioration over the three assessments. On the basis of these results, Zinger et al. (2001) concluded that administrative segregation may not have as detrimental an effect on an inmate’s current and future psychological adjustment as had originally been thought, although they did acknowledge that it may take more than 2 months in solitary confinement to achieve a noticeable psychological effect. The authors therefore encouraged professional staff to continue monitoring inmates in administrative segregation as a means of identifying potential problems and providing programming and support services where necessary.

Two years after Zinger et al. (2001) published their results, a second longitudinal study was performed on psychological change in inmates placed in administrative segregation (Andersen, Sestoft, Lillebæk, Gabrielsen, & Hemmingsen, 2003). In this study, 113 Danish prisoners housed in administrative segregation and 95 Danish general population inmates were evaluated using self-report surveys and interviews conducted by trained staff. Participants were interviewed and surveyed on four separate occasions: at baseline and then again after 3 weeks, 3 months, and 4 months. Cross-sectional comparisons revealed a higher rate of adjustment disorder in inmates housed in administrative segregation, whereas longitudinal analysis showed reduced symptomatology over time in the general population group and no change in symptomatology over the course of the study in the administrative segregation group. Andersen et al. (2003) further reported that transfer from administrative segregation to general population resulted in a reduction in psychological distress and symptomatology. Unlike Zinger et al. (2001), these results provide modest, but consistent support for the possibility that placement in restrictive housing can lead to psychological deterioration. One way to explain the differences in outcome between the two studies is that while inmates in the Anderson et al. study were in the community just prior to being placed in restrictive housing, inmates in the Zinger et al. study had been in prison just prior to their placement in restrictive housing.

O’Keefe and colleagues (2013) conducted what is considered the largest longitudinal study on restrictive housing and psychological functioning in inmates to date. In this study, the authors followed several groups of prisoners housed in Colorado high security facilities for a period of 1 year. Two groups of inmates from this study were housed in long-term segregation in Colorado’s supermax facility, one with and the other without a history of mental illness. This determination was based on clinical diagnosis (bipolar disorder, major depression, dysthymia, schizophrenia, and other psychotic disorders), acuity of symptoms, and consumption of mental health resources. There were also two groups of general population inmates, one of which presented with mental health needs and the other of which had no documented history of mental problems. A fifth group of inmates was sampled from a specialized facility for seriously mentally ill offenders. A self-report measure designed to assess psychological functioning was administered 5 times over a period of 1 year, with results showing that participants in all five groups displayed improved psychological functioning. More importantly, there was no difference between participants who were and were not in restrictive housing. The Colorado study was not without limitations. Questions, for instance, have been raised as to the reliability and validity of the self-report data on which O’Keefe et al. (2013) based their conclusions (Grassian & Kupers, 2011). This was one reason why staff ratings were used instead of inmate ratings in determining psychological deterioration for the current study.

In what is probably the most recently published study on restrictive housing and psychological deterioration, Chadick, Batastini, Levulis, and Morgan (2018) examined a sample of 48 adult male offenders remanded to the custody of the Kansas Department of Corrections (KDOC). The 24 inmates sampled from restrictive housing, where they had been living for past 1 to 4 years, were matched on age, ethnicity, primary psychiatric diagnosis, and most severe crime with 24 general population inmates. All participants received pretest and posttest administrations of the Millon Clinical Multiaxial Inventory–III (MCMI-III) in an effort to determine whether placement in restrictive housing was associated with higher levels of psychological symptomatology. The results indicated that those in restrictive housing recorded significantly higher posttest scores on the MCMI-III anxiety, somatoform, dysthymia, posttraumatic stress disorder, and major depression scales compared to general population inmates, controlling for pretest scores on these same measures. There was only one change in scores over time, however: Time spent in segregation correlated significantly with alcohol dependence. Comparing the two groups, Chadick et al. (2018) discovered that while the psychological status of general population inmate gradually improved over time, the psychological status of inmates in administrative segregation remained stable. The overall conclusion from this study was that placement in administrative segregation had no apparent deleterious effect on an inmate’s psychological functioning over time.

The results of the O’Keefe et al. (2013) and Zinger et al. (2001) studies failed to show evidence of increased psychological distress and mental health symptomatology in AS inmates compared with general population inmates, whereas results from the Chadick et al. (2018) study offered no support for the assertion that placement in AS inevitably leads to psychological deterioration in those who receive such treatment. These findings are inconsistent with the founding tenets of SHU syndrome and the notion that long-term confinement in restrictive housing aggravates mental health symptomatology (Blanchette, 2001; Brodsky & Scogin, 1988; Haney, 2003; Jackson, 2001). In other words, the numbers do not add up, but is this because of conceptual problems with SHU syndrome and restrictive housing effects or does this reflect a serious limitation in the research itself? More rigorous testing is required.

In determining whether SHU syndrome and belief in the aversive psychological effects of restrictive housing are valid, we must examine the two assumptions upon which these beliefs are based. The first assumption is that SHU syndrome can occur in those with no prior history of serious psychiatric disorder and may be just as prevalent in inmates with no history of mental illness as it is in inmates with a history of mental illness. This assumption is implied in the original Grassian (1983) statement on SHU syndrome, “[the study] population . . . was not preselected by psychiatric status” (p. 1453), and in a recent article by Kupers (2016), “SHU Post-Release Syndrome is reported to me by the vast majority of prisoners and ex-prisoners I meet who have been released from solitary confinement after a long stint” (p. 92). The second assumption is that SHU syndrome and the adverse effect of restrictive housing are not simply the result of incarceration. As such, they should be significantly more prevalent in inmates exposed to long-term administrative segregation, such as would be found in a supermax facility, than in a regular maximum security facility (Arrigo & Bullock, 2008).

The Present Study

The purpose of the current investigation was to determine whether placement in administrative segregation (AS) has a deleterious psychological effect on individuals with no history of mental health needs, and, if not, whether the effect is stronger for inmates with mental health needs housed in AS than it is for inmates with a mental health history not housed in AS. Data for this study came from the previously mentioned Colorado Longitudinal Study of Psychological Effects of Administrative Segregation, but the present study differed from the original O’Keefe et al. (2013) investigation in several key respects. First, psychological deterioration was defined by a latent class identified in a growth mixture modeling (GMM) analysis of clinician ratings rather than in a comparative analysis of self-report scores. Second, psychological deterioration was assessed with a measure of clinician-rated psychological functioning instead of a measure of offender self-reported psychological functioning. Third, the current study directly tested the two assumptions upon which SHU syndrome and psychological deterioration in response to restrictive housing are based: (a) it should be as prevalent in inmates without mental health needs as it is in inmates with mental health needs; and (b) it should be significantly more prevalent in AS inmates with mental health needs than it is in general population non-AS inmates with mental health needs.

Three hypotheses were tested in this study.

Hypothesis 1: A GMM analysis of clinician ratings of psychological functioning will yield one or more construct valid latent class groupings representative of psychological deterioration.

Hypothesis 2: The psychological deterioration latent class or classes will contain similar proportions of AS inmates with and without a history of mental health need if the SHU syndrome argument is correct and significantly more AS inmate with mental health need than AS inmates without mental health need if the SHU syndrome argument is incorrect.

Hypothesis 3: The psychological deterioration latent class or classes will contain significantly more AS inmates with a history of mental health need than non-AS inmates with a history of mental health need if the SHU syndrome argument is correct and roughly equal proportions of AS and non-AS inmates with mental health need if the SHU syndrome argument is incorrect.

Method

Participants

The sample for this study was taken from the Colorado Department of Corrections (CDOC) Longitudinal Study of Psychological Effects of Administrative Segregation (O’Keefe, Klebe, Stucker, Sturm, & Leggett, 2011). Participants for the current study were 266 of the 270 inmates from the original CDOC study who had complete data on at least one of the three Brief Psychiatric Rating Scale (BPRS) evaluations conducted in the longitudinal analysis. Participants came from four settings: 124 (64 with mental health need and 60 without mental health need) came from a 756-bed facility for AS-classified offenders where, by policy, they were allowed out of their cells 5 hr per week for recreation and permitted to shower 15 min 3 times a week; 67 came from a dedicated 255-bed special needs prison for inmates with serious psychiatric problems; and 75 (42 with mental health needs and 33 without mental health needs) came from general population prisons of comparable security level. All participants, including those in AS, were followed for a period of 1 year. Individual demographic data were masked in the dataset accessed by the author through the Inter-University Consortium for Political and Social Research (ICPSR), although the mean age of the total sample of participants was 31.8 years (SD = 9.1, range = 17-59), with a racial/ethnic breakdown of 40% White, 36% Hispanic, 18% African American, 4% Native American, and 1% Asian.

Measures

Psychological deterioration was assessed in the current study using ratings from clinical staff on the BPRS (Overall & Gorham, 1962; Ventura et al., 1993). The BPRS was completed on three occasions, with 6 months between occasions. This 24-item rating scale hones in on rapidly changing symptoms of serious psychiatric disorder to include “somatic concern,” “anxiety,” “depression,” “suicidality,” “guilt,” “hostility,” “elevated mood,” “grandiosity,” “suspiciousness,” “hallucinations,” “unusual thought content,” “bizarre behavior,” “self-neglect,” “disorientation,” “conceptual disorganization,” “blunted affect,” “emotional withdrawal,” “motor retardation,” “tension,” “uncooperativeness,” “excitement,” “distractibility,” “motor hyperactivity,” and “mannerisms and posturing.” All 24 items are rated on a 7-point scale (1 = not present, 7 = extremely severe) and the ratings summed to produce a score that can range from 24 to 168. The BPRS has been found to be sensitive to psychological status in incarcerated forensic samples (Greenwood & Burt, 2001; van Beek et al., 2015). Six-month test–retest reliability ranged from .43 to .55 and internal consistency was good when the BPRS was completed during Time Periods 1 (α = .80), 3 (α = .79), and 5 (α = .79) of the current study.

The Level of Service Inventory–Revised (LSI-R; Andrews & Bonta, 1995) was routinely administered to inmates at the CDOC diagnostic and reception center prior to an inmate being assigned to a regular prison and prior to completion of the BPRS. Because of their availability and fact that they assess an area of interest to correctional professionals, namely, criminogenic needs (van der Knaap, Alberda, Oosterveld, & Born, 2012), these variables were selected as predictors in this study. Given the importance of assessing needs, individual scores/ratings based on several of the LSI-R items (scored 1 = “yes” or 0 = “no”) were included as predictors in this study: mental health needs, self-destruction needs (i.e., suicidality), gang membership (member, associate, or suspect), substance abuse, medical needs, academic needs, vocational needs, intellectual deficit needs, and anger management needs. A dichotomous measure of having ever been in administrative or disciplinary segregation (AS/DS) prior to the start of the current study (1 = “yes,” 0 = “no”) was also included as a predictor variable in this study. Predictor variables were organized into two categories: convergent and discriminant. The convergent category contained indicators that were expected to correspond with mental health (BPRS) status (i.e., mental health needs and self-destruction needs), whereas the discriminant category contained indicators that were not expected to correspond specifically with mental health status as measured by the BPRS (i.e., gang membership, substance abuse, medical needs, academic needs, vocational needs, intellectual deficit needs, anger management needs, and prior AS/DS placement).

One final measure was included in this study as part of a supplemental analysis designed to investigate certain relationships noted in the main analysis. This measure was the Global Severity Index (GSI) from the Brief Symptom Inventory (BSI; Derogatis, 1993). The BSI is a 53-item self-report inventory composed of nine subscales (Somatization, Obsessive–Compulsive, Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation, and Psychoticism). The GSI is the mean score across the subscales. Because the BSI scores were positively skewed and there were a significant number of outliers, they were transformed by means of a square root transformation and centered at the mean of the baseline (Time 1) period. There were actually six time periods assessed, but the final period had over half of its data missing and was dropped from the analysis. In the current study, the GSI achieved good test–retest reliability over 3 months (r = .66-.82) and excellent internal consistency (α = .91-.92).

Data Analysis

There were three principal analyses performed as part of this study. First, a GMM analysis was conducted on the three longitudinally arranged BPRS scores using Mplus 5.2 (Muthén & Muthén, 1998-2007). This analysis employed an MLR estimator, a maximum of 100 iterations for the continuous outcome, a maximum of 500 iterations for the optimization specifications, 250 random starts, and 50 final stage optimizations. Based on prior simulation research (Nylund, Asparouhov, & Muthén, 2007), the stopping rule for this analysis was based on an increase in the Bayesian information criterion (BIC; Schwarz, 1978) and a nonsignificant p value on the bootstrap likelihood ratio test (BLRT; McLachlan & Peel, 2000), with any discrepancies resolved in favor of the BLRT. Whereas traditional growth modeling assumes that all individuals in a group come from the same population, GMM assumes the presence of two or more subsets of individuals with different growth or change patterns (Muthén & Asparouhov, 2008). GMM is therefore a potentially useful technique for identifying a temporal pattern of psychological deterioration within a larger group of prison inmates.

The second (construct validity) analysis involved contrasting the psychological deterioration class or classes identified in the GMM analysis with the nonpsychological deterioration class or classes on a series of two convergent (mental health needs, self-destruction needs) and eight discriminant (gang membership, substance abuse, medical needs, educational needs, vocational needs, intellectual deficit needs, anger management needs, and prior AS placement) predictors. It was anticipated that the convergent predictors would be significantly more prominent in the psychological deterioration group than in the nonpsychological deterioration group, but that there would be no difference between the groups on the discriminant predictors.

The third analysis was a supplemental investigation designed to determine whether a subjective measure of psychological distress (BSI) could explain why some individuals with mental health needs experience psychological deterioration in prison while others do not. In actuality, two supplemental analyses were performed: A full sample analysis of all participants organized into three groups (psychological deterioration cases, nonpsychological deterioration cases with mental health needs, and nonpsychological deterioration cases without mental health needs) and a partial sample analysis conducted just on AS participants.

Missing Data

Seven participants were missing data for BPRS-1, 32 were missing data for BPRS-3, and 57 were missing data for BPRS-5. No participants were missing data for BSI-1, eight were missing data for BSI-2, 15 were missing data for BSI-3, 23 were missing data for BSI-4, and 30 were missing data for BSI-5. In each case, missing BPRS and BSI data were handled with expectation maximization (25 iterations). Four of the predictor variables also had missing data: self-destruction needs (n = 13), substance abuse (n = 10), anger management needs (n = 4), and intellectual deficit needs (n = 5). Missing data on these four predictor variables were handled with listwise deletion based on the fact that all four variables were dichotomous and the proportion of missing cases never exceeded 5% for any one predictor variable (see Allison, 2001).

Results

Preliminary Analyses

Table 1 lists the descriptive statistics and intervariable correlations for the BPRS results at Time Periods 1, 3, and 5. The concurrent validity of the BPRS was tested by correlating scores on the BPRS with the self-reported BSI scores that covered the same time periods. BPRS–BSI correlations for Time Period 1, Time Period 3, and Time Period 5 were .42, .43, and .36, respectively. All three correlations fall into the moderately high range, and, as such, lend support to the construct validity of the three BPRS scores.

Table 1:

Descriptive Statistics and Correlations for the BPRS at Three Different Time Points (N = 266)

Variable	M	SD	Range	BPRS-3	BPRS-5
BPRS-1	33.51	8.13	24-69	.43	.32
BPRS-3	30.99	6.45	24-66		.55
BPRS-5	31.03	6.23	24-69

Note. All correlations significant at p < .001; Range = range of scores in current sample; BPRS-1 = Brief Psychiatric Rating Scale at Time 1; BPRS-3 = Brief Psychiatric Rating Scale at Time 3; BPRS-5 = Brief Psychiatric Rating Scale at Time 5.

GMM

The results of a GMM analysis of the BPRS are summarized in Table 2. The BLRT supported a two-class model, whereas the BIC indicated the presence of a six-class model. The problem with the six-class model was that three of the six classes had fewer than 10 members and the results could easily be incorporated into the two-class model. In addition, the Nylund et al. (2007) simulation study showed that the BLRT is superior to the BIC when GMM is performed with small samples.

Table 2:

Growth Mixture Modeling Results for the Brief Psychiatric Rating Scale (N = 266)

Model	LL	Par	BIC	BLRT
One class	−2,601.58	8	5,247.82	—
Two class	−2,565.28	11	5,191.98	< .001
Three class	−2,541.12	14	5,160.40	.667

Note. Model = type of model (one class, two class, or three class); LL = log likelihood ratio test; Par = number of free parameters; BIC = Bayesian information criterion; BLRT = bootstrapped likelihood ratio test (p value).

The two-class model was composed of inmates who displayed no evidence of psychological deterioration (Class 1; n = 248) and inmates who displayed clear evidence of psychological deterioration (Class 2; n = 18). The trajectories for these two classes are depicted in Figure 1. Class 2, the latent class that embodies psychological deterioration, exhibited a significant rise in psychiatric symptomology between Periods 3 and 5 (Hypothesis 1). A GMM conducted on the subsample of participants who were in AS at the time of the study (n = 124) also supported a two-class model, but whereas 18 inmates were assigned to Class 2 in the full sample, only three inmates from the AS subsample were assigned to Class 2.

Figure 1:

Trajectories for the Two Latent Classes Created From BPRS Scores

In validating Class 2 as an indicator of psychological deterioration, both convergent predictors (mental health needs and self-destructive needs) were found to be significantly more prevalent in the backgrounds of inmates in Class 2 (psychological deterioration group) relative to inmates in Class 1 (see Table 3). Conversely, the eight discriminant predictors failed to differentiate between inmates assigned to Classes 1 and 2. Two out of two convergent predictors and zero out of eight discriminant predictors were found to be significantly different, χ²(1) = 4.73, p < .05 (Fisher’s exact probability test = .022), thereby supporting the construct validity of Class 2 as a marker of psychological deterioration.

Table 3:

Convergent and Discriminant Predictors of the Two Latent Classes Formed Using the BPRS

Predictor	Class 1	Class 2	χ²(1)
Convergent predictors
Mental health needs	137(55.2%)	18(100.0%)	13.83**
Self-destruction needs	53(22.5%)	8(47.1%)	5.24*
Discriminant predictors
Gang membership	147(59.3%)	8(44.4%)	1.52
Prior AS/DS placement	56(22.6%)	4(22.2%)	0.00
Substance abuse needs	184(77.3%)	15(83.3%)	0.35
Medical needs	44(17.7%)	4(22.2%)	0.23
Academic needs	99(39.9%)	4(22.2%)	2.22
Vocational needs	206(83.1%)	17(94.4%)	1.60
Intellectual deficit needs	24(9.9%)	1(5.6%)	0.36
Anger management needs	148(60.7%)	11(61.1%)	0.00

Note. Predictor = category (convergent or discriminant) and type (second level) of predictor; Convergent predictors = predictors hypothesized to be more prevalent in Class 2 than Class 1; Discriminant predictors = predictors hypothesized not to differ between Class 1 and Class 2; Class 1 = number and proportion (in parentheses) of participants in Class 1 (n = 248) who received a “yes” rating on the item; Class 2 = number and proportion (in parentheses) of participants in Class 2 (n = 18) who received a “yes” rating on the item; χ²(1) = chi-square test with one degree of freedom. BPRS = Brief Psychiatric Rating Scale; AS/DS = administrative segregation or disciplinary segregation.

p < .05. **p < .001.

Cross-Tab Analyses

An analysis performed on participants in AS revealed that there were significantly more inmates with documented mental health needs in Class 2 (eight out of 60 or 13.3%) than inmates with no documented mental health needs (zero out of 64 or 0.0%), χ²(1) = 9.12, p < .01 (Fisher’s exact probability test = .002). By contrast, inmates with mental health needs who were housed in AS were no more likely to be members of Class 2 (eight out of 52 or 13.3%) than inmates with mental health needs who were in general population (10 out of 95 or 10.5%), χ²(1) = 0.28, p = .60 (Fisher’s exact probability test = .614). The first finding supports the counter-SHU argument from Hypothesis 2 and the second finding supports the counter-SHU argument from Hypothesis 3.

Supplemental Analysis

Three groups were contrasted longitudinally on the GSI of the BSI. The three groups consisted of inmates with mental health needs assigned to Class 2 (n = 18), inmates with mental health needs assigned to Class 1 (n = 137), and inmates without mental health needs assigned to Class 1 (n = 111). Results indicated that while Class 1 inmates with mental health needs reported significantly more psychological distress than Class 1 inmates without mental health needs, both displayed modest to moderate reductions in self-reported psychological distress over the course of the 1-year study. Participants in Class 2, by contrast, displayed a modest increase in psychological distress over this same time period (see Figure 2). The upward and downward trends were most evident between Time Periods 1 and 2. A similar pattern of results was obtained when the analysis was restricted to inmates in AS.

Figure 2:

BSI Scores Across Five Time Periods for Participants With MH Needs in Class 1 of the BPRS Latent Model, Participants With MH Needs in Class 2 of the BPRS Latent Model, and Participants Without MH Needs

Discussion

The first hypothesis tested in this study held that clinician ratings on the BPRS would identify a group of individuals with significant psychological deterioration and that this group would be defined by mental health and self-destruction needs, but not by other needs (e.g., substance abuse, anger management) and issues (e.g., gang membership, prior placement in administrative segregation or detention). A GMM analysis identified a small latent class that appeared to represent psychological deterioration as indicated by a sharp rise in BPRS scores during the last 6 months of the observation period (see Figure 1). The construct validity of this latent class was verified in a series of convergent and discriminant comparisons whereby prior mental health and self-destruction needs predicted class membership (convergent effects) but gang membership, substance misuse, medical needs, educational needs, vocational needs, intellectual deficit needs, anger management needs, and prior placement in administrative or disciplinary segregation did not (discriminant effects). Moderately strong concurrent correlations between clinician ratings on the BPRS and offender self-report ratings on the BSI further suggested that the BPRS possessed adequate construct validity in the current study. The findings provide support for the first hypothesis indicating that there was a small, but clearly defined group of inmates who suffered significant mental health deterioration over the course of their stay in AS or general population.

The second and third hypotheses were designed to test the notion that restrictive housing predicts psychological deterioration using class assignments from the GMM analysis to divide the sample into groups of individuals with and without serious psychological deterioration. The second hypothesis held that if the psychological deterioration associated with restrictive housing and SHU syndrome (Grassian, 1983, 2006; Kupers, 2016) is independent of prior mental health status, then there should be as many nonmental health inmates as mental health inmates showing signs of deterioration on the BPRS, a measure that assesses nearly all of the symptoms in Grassian’s (1983) definition of SHU syndrome. Contrary to the second hypothesis, significantly more mental health than nonmental health inmates (13.3% vs. 0.0%) experienced psychological deterioration in AS. This finding is inconsistent with the first assumption of SHU syndrome; namely, that SHU syndrome and psychological deterioration following placement in restrictive housing are as likely to occur in individuals without mental health needs as it is in individuals with mental health needs. In fact, of the 111 inmates who did not receive a rating of serious mental health needs during the intake interview, none were assigned to Class 2 (severe deterioration). These results are consistent with the previously mentioned Zinger et al. (2001) and O’Keefe et al. (2013) studies, neither of which found evidence of psychological deterioration in mental health inmates housed in AS.

The third hypothesis tested another bedrock assumption of SHU syndrome theory; namely, that inmates placed in long-term lockdown will experience more deleterious psychological effects than inmates not placed in long-term lockdown. Because none of the participants rated as not having mental health needs at intake displayed psychological deterioration over a period of 1 year, the AS versus non-AS analysis was restricted to inmates who were identified as having mental health needs at intake. The results of this analysis revealed that mental health inmates housed in AS and mental health inmates housed in general population or a special mental health unit achieved equivalent levels of psychological deterioration. Therefore, although inmates with mental health needs were significantly more likely to deteriorate psychologically compared to inmates without mental health needs, those housed in AS were no more likely to deteriorate than those housed in general population or a special mental health unit. This indicates that psychological deterioration and SHU syndrome have less to do with AS then they do with prior mental health difficulties and need. This finding clearly contradicts a central tenet of SHU syndrome theory whereby long-term placement in a supermax or other AS facility should have an effect above and beyond the initial effect of incarceration.

In an effort to more fully understand the relationships that surfaced in this study, the BSI self-report data that served as the focus of O’Keefe et al.’s (2013) study were organized into three groups: the 18 Class 2 inmates, all of whom had mental health needs; the 137 inmates with mental health needs who were assigned to Class 1; and the 111 inmates without mental health needs who were also assigned to Class 1. The results of this analysis, as depicted in Figure 2, revealed that inmates with mental health needs reported much higher levels of psychological distress and disorder than inmates without mental health needs. Of greater concern to those interested in determining why some mentally ill prisoners (Class 2) are at greater risk for psychological deterioration than other mentally ill prisoners (Class 1) is the change that occurred in self-reported symptomatology from baseline (Time 1) to 3-month follow-up (Time 2). Mentally ill inmates from Class 1 displayed a significant decrease in symptomatology from Time 1 to Time 2, a pattern that had been reported previously (Taylor et al., 2010), that then continued to decline over the course of the 1-year study. Mentally ill inmates assigned to Class 2, on the contrary, displayed a significant increase in symptomatology from Time 1 to Time 2 which then both rose and fell in a saw-tooth pattern that exceeded baseline levels.

Implications

The BSI results point to a practical implication of the current results. Namely, clinicians working with offenders in supermax and AS settings should carefully monitor symptoms patterns over time, particularly if there is a sharp increase in symptomatology shortly after admission. Comparing the BSI results with patterns displayed in Figure 1 (BPRS model) shows that the clinician-rated symptomatology of Class 2 inmates remained fairly stable until sometime between the sixth and 12th months, as marked by Time Periods 3 and 5. The self-reported symptomatology of inmates in Class 2 of the BPRS model suggests that Class 2 inmates may have begun having problems several months before they were noticed by the clinician. Hence, the self-report data may be a harbinger of future adjustment difficulties that then come to the attention of clinical staff several months later. This insinuates that when clinicians are evaluating inmates in AS, supermax, or long-term control units, they should pay close attention to self-reported problems from inmates and supplement these with observations from staff in daily contact with the prisoner. Practically, all correctional systems require that inmates in both short- and long-term segregation receive periodic psychological reviews. Given the current results, it would seem advisable to conduct these reviews regularly (with perhaps no more than a month between evaluations) and base the reports on multiple sources, such as a file review, offender self-report, and staff observations.

Even though many of the participants in this sample experienced reduced psychological symptomatology over the course of the study, this should not be interpreted to mean that AS is harmless. Even though the clinician ratings and self-report data displayed good convergence, it could still be argued that the current study underestimated the amount of harm inflicted on participants, not only through placement in AS but also by confinement in prison itself. The criminogenic effects of incarceration are, after all, well documented (Cullen, Jonson, & Nagin, 2011; Vieraitis, Kovandzic, & Marvell, 2007). There is also evidence that some correctional systems may have become overly reliant on supermax or AS confinement as a means of controlling disruptive, unruly, and gang-affiliated inmates (Ahalt et al., 2017). Thus, even though long-term confinement in AS may not be as detrimental as was once thought (Pettigrew, 2002), less restrictive alternatives need to be found. Due process concerns are also of significance when it comes to using solitary confinement and restrictive housing to discipline and control inmates. This is something that is addressed in the CDOC protocol where inmates who are being considered for placement in AS not only attend an initial hearing but also receive periodic rehearings that are reviewed and can be reversed by the warden and other high-level Department of Corrections administrators (O’Keefe et al., 2013). From a policy standpoint, these results suggest that AS may be appropriate under some circumstances but that these circumstances need to be clearly defined and except in extreme cases, time-limited.

Limitations

Whereas the CDOC longitudinal study is probably the most methodologically sound study conducted thus far on the psychological effects of incarceration and AS, the study was not without limitations, some of which were quite severe. One data collection limitation of the CDOC longitudinal study is that not all participants remained in the same status (AS vs. non-AS) over the course of the study. Only 100 of the original 270 participants ended the year in the same status that they began and while the identities of these “pure” cases were not available through the ICPSR website, O’Keefe et al. (2011) performed several supplemental analyses that showed no major changes in variable relationships when the sample was restricted to the 100 “pure” cases. Another potential limitation of this study is that the period of follow-up was set at 1 year. Although this is longer than most longitudinal studies on psychological reactions to AS, many inmates spend more than a year in a supermax or control unit cell (see Mears & Bales, 2010). Consequently, the current results may not accurately represent the level of psychological damage that spending a decade or more in solitary confinement can have on an individual (Morgan et al., 2016).

A potential limitation of any secondary data analysis is the prospect of inherited problems. These inherited problems are measurement and design flaws that are part and parcel of the dataset being analyzed. In the current study, one of the more serious inherited problems appeared in the form of several less than ideal measures. These inherited problems included the use of dichotomous scores rather than the standard continuous scores for the LSI-R and use of a dichotomous instead of a continuous measure of prior placement in restrictive housing. However, conducting the GMM analysis on the entire sample of participants mitigates the criticism that the groups formed by the original researchers were biased (Grassian & Kupers, 2011). A further limitation of this study is that postrelease behavior was not examined even though prior research indicates that being released directly from a supermax facility is associated with significantly poorer outcomes than being released from a regular penitentiary after spending time in a supermax (Lovell, Johnson, & Cain, 2007). Finally, because this study was confined to a single state, the results may not generalize to other jurisdictions, particularly those that follow policies and procedures that differ significantly from those found in Colorado.

Conclusion

There has been a great deal written about SHU syndrome and the putative psychological effects of solitary confinement on inmates living in restricted housing (Blanchette, 2001; Brodsky & Scogin, 1988; Haney, 2003; Kupers, 2016). The problem is that these effects have yet to be verified empirically. All but one of the longitudinal studies conducted on this issue (Andersen et al., 2003) failed to corroborate the psychological deterioration to which SHU syndrome is believed to give rise (Chadick et al., 2018; O’Keefe et al., 2013; Zinger et al., 2001; current study). Although the majority of longitudinal studies on restrictive housing and psychological adjustment have failed to support the assertion that restrictive housing necessarily leads to psychological deterioration and, therefore, constitutes cruel and unusual punishment, such housing is still known to create problems for inmates with specific types of problems, particularly those suffering from mental illness (U.S. Department of Justice, 2017). What the current results suggest is that imprisonment, regardless of whether it occurs in general population or restrictive housing, may have a deleterious psychological effect on those with serious mental health problems and therefore warrants further study.

Fettig (2016) contends that the principal drivers of reform in society’s movement to decrease the use of solitary confinement with prison inmates are civil society campaigns, government and professional group alliances, civil rights litigation, correctional leadership, and media coverage. Missing from this list is rigorous empirical studies. Until which time empirical research is on par with the five drivers on Fettig’s (2016) list, the reform movement will continue to be weighed down by opposition from those who work in the field and must find ways to manage the more incorrigible members of the prison population. In closing, it should be noted that the current investigation, like any secondary data analysis, inherits the problems and limitations of the original dataset, in this case the CDOC longitudinal study. There may have been nothing in the current results to support SHU syndrome but given some of the previously mentioned problems with the Colorado data, as well as the importance of the topic, psychological deterioration in inmates assigned to restrictive housing deserves further empirical study using larger samples, longer exposures, and follow-ups that extend beyond incarceration. In the meanwhile, however, the argument that restrictive housing leads to adverse psychological effects remains unverified and the numbers used to support this argument continue to not add up.

Footnotes

Glenn D. Walters, PhD, is a professor in the Department of Criminal Justice at Kutztown University in Kutztown, Pennsylvania, where he teaches classes in criminology, corrections, research methods, and substance abuse and crime. Prior to this, he worked as a clinical psychologist for the Federal Bureau of Prisons. His current research interests include offender assessment and intervention, adjustment to prison, causal mediation analysis, and development of an overarching psychological theory of crime initiation, maintenance, and desistance.

References

Ahalt

Haney

Rios

Fox

M. P.

Farabee

Williams

(2017). Reducing the use and impact of solitary confinement in corrections. International Journal of Prisoner Health, 13, 41-48.

Allison

(2001). Missing data: Quantitative applications in the social sciences (Vol. 136). Thousand Oaks, CA: SAGE.

Andersen

H. S.

Sestoft

Lillebæk

Gabrielsen

Hemmingsen

(2003). A longitudinal study of prisoners on remand: Repeated measures of psychopathology in the initial phase of solitary versus nonsolitary confinement. International Journal of Law and Psychiatry, 26, 165-177.

Andrews

D. A.

Bonta

(1995). The Level of Service Inventory–Revised. Toronto, Ontario, Canada: Multi-Health Systems.

Arrigo

B. A.

Bullock

J. L.

(2008). The psychological effects of solitary confinement on prisoners in supermax units. International Journal of Offender Therapy and Comparative Criminology, 52, 622-640.

Association of State Correctional Administrators-Liman. (2015, August). Time-in-cell: The ASCA-Liman National Survey of Administrative Segregation in Prison 2014. New Haven, CT: The Liman Program, Yale Law School, Association of State Correctional Administrators.

Blanchette

(2001). Characteristics of administratively segregated offenders in federal corrections. Canadian Journal of Criminology, 43, 131-144.

Briggs

C. S.

Sundt

J. L.

Catellano

T. C.

(2003). The effect of supermaximum security prisons on aggregate levels of institutional violence. Criminology, 41, 1341-1376.

Brodsky

Scogin

(1988). Inmates in protective custody: First data on emotional effects. Forensic Reports, 1, 267-280.

10.

Chadick

C. D.

Batastini

A. B.

Levulis

S. J.

Morgan

R. D.

(2018). The psychological impact of solitary: A longitudinal comparison of general population and long-term administratively segregated male inmates. Legal and Criminological Psychology. Advance online publication. doi:10.1111/lcrp.12125

11.

Cullen

F. T.

Jonson

C. L.

Nagin

D. S.

(2011). Prisons do not reduce recidivism: The high cost of ignoring science. The Prison Journal, 91, 485-655.

12.

Derogatis

(1993). Brief Symptom Inventory (BSI): Administration, scoring, and procedures manual (3rd ed.). Minneapolis, MN: National Computer System.

13.

Fettig

(2016, October/November). What is driving solitary confinement reform? Correctional Law Reporter, 28, 33-46.

14.

Grassian

(1983). Psychopathological effects of solitary confinement. American Journal of Psychiatry, 140, 1450-1454.

15.

Grassian

(2006, January). Psychiatric effects of solitary confinement. Washington University Journal of Law & Policy, 22, 324-383.

16.

Grassian

Friedman

(1986). Effects of sensory deprivation in psychiatric seclusion and solitary confinement. International Journal of Law and Psychiatry, 8, 49-65.

17.

Grassian

Kupers

(2011). The Colorado study vs. the reality of supermax confinement. Correctional Mental Health Report, 13(1), 1-11.

18.

Greenwood

Burt

(2001). Validity of the Brief Psychiatric Rating Scale within a forensic inpatient hospital. Expert Evidence, 8, 15-30.

19.

Haney

(2003). Mental health issues in long-term solitary and “supermax” confinement. Crime & Delinquency, 49, 124-156.

20.

Hodgins

Côté

(1991). The mental health of penitentiary inmates in isolation. Canadian Journal of Criminology, 33, 177-182.

21.

Jackson

(2001). The psychological effects of administrative segregation. Canadian Journal of Criminology, 43, 109-116.

22.

Kupers

T. A.

(1999). Prison madness: The mental health crisis behind bars and what we must do about it. San Francisco, CA: Jossey-Bass.

23.

Kupers

T. A.

(2016). The SHU post-release syndrome: A preliminary report. Correctional Mental Health Report, 17, 81-95.

24.

Lovell

Johnson

L. C.

Cain

K. C.

(2007). Recidivism of supermax prisoners in Washington state. Crime & Delinquency, 53, 633-656.

25.

McLachlan

Peel

. (2000). Finite mixture models. New York: John Wiley.

26.

Mears

D. P.

Bales

W. D.

(2010). Supermax housing: Placement, duration, and time to reentry. Journal of Criminal Justice, 38, 545-554.

27.

Morgan

R. D.

Gendreau

Smith

Gray

Labrecque

R. M.

MacLean

. . . Mills

J. F.

(2016). Quantitative syntheses on the effects of administrative segregation on prisoners’ well-being. Psychology, Public Policy, and Law, 22, 439-461.

28.

Motiuk

L. L.

Blanchette

(2001). Characteristics of administratively segregated offenders in federal corrections. Canadian Journal of Criminology, 43, 131-143.

29.

Muthén

Asparouhov

(2008). Growth mixture modeling: Analysis with non-Gaussian random effects. In Fitzmaurice

Davidian

Verbeke

Molenberghs

(Eds.), Advances in longitudinal data analysis (pp. 143-165). Boca Raton, FL: Chapman & Hall/CRC Press.

30.

Muthén

(1998-2007). Mplus user’s guide (5th ed.). Los Angeles, CA: Author.

31.

Nylund

K. L.

Asparouhov

Muthén

B. O.

(2007). Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study. Structural Equation Modeling, 14, 535-569.

32.

O’Keefe

M. L.

Klebe

K. J.

Metzner

Dvoskin

Fellner

Stucker

(2013). A longitudinal study of administrative segregation. Journal of the American Academy of Psychiatry and the Law, 41, 49-60.

33.

O’Keefe

M. L.

Klebe

K. J.

Stucker

Sturm

Leggett

(2011, January). One year longitudinal study of the psychological effects of administrative segregation (NCJRS 232973). Washington, DC: U.S. Department of Justice.

34.

Overall

J. E.

Gorham

D. R.

(1962). The Brief Psychiatric Rating Scale. Psychological Reports, 10, 799-812.

35.

Pettigrew

C. A.

(2002). Technology and the Eighth Amendment: The problem of supermax prisons. North Carolina Journal of Law & Technology, 4, 191-215.

36.

Schwarz

(1978). Estimating the dimension of a model. Annals of Statistics, 6, 461-464.

37.

Suedfeld

Ramirez

Deaton

Baker-Brown

(1982). Reactions and attributes of prisoners in solitary confinement. Criminal Justice and Behavior, 9, 303-340.

38.

Taylor

P. J.

Walker

Dunn

Kissell

Williams

Amos

(2010). Improving mental state in early imprisonment. Criminal Behaviour and Mental Health, 20, 215-231.

39.

U.S. Department of Justice. (2017, July). Review of the Federal Bureau of Prisons’ use of restrictive housing for inmates with mental illness. Washington, DC: Office of the Inspector General, U.S. Department of Justice.

40.

van Beek

Vuijk

P. J.

Harte

J. M.

Smit

B. L.

Nijman

Scherder

E. J. A

. (2015). The factor structure of the Brief Psychiatric Rating Scale (expanded version) in a sample of forensic psychiatric patients. International Journal of Offender Therapy and Comparative Criminology, 59, 743-756.

41.

van der Knaap

L. M.

Alberda

D. L.

Oosterveld

Born

M. P

. (2012). The predictive validity of criminogenic needs for male and female offenders: Comparing the relative impact of needs in predicting recidivism. Law and Human Behavior, 36, 413-422.

42.

Ventura

Lukoff

Nuechterlein

K. H.

Liberman

R. P.

Green

M. F.

Shaner

(1993). Brief Psychiatric Rating Scale (BPRS) expanded version (4.0): Scales, anchor points, and administration manual. International Journal of Methods in Psychiatric Research, 3, 227-244.

43.

Vieraitis

L. M.

Kovandzic

T. V.

Marvell

T. B.

(2007). The criminogenic effects of imprisonment: Evidence from state panel data, 1974-2002. Criminology & Public Policy, 6, 589-622.

44.

Zinger

Wichmann

Andrews

D. A.

(2001). The psychological effects of 60 days in administrative segregation. Canadian Journal of Criminology, 43, 47-83.