Disparities in Segregation for Prison Control: Comparing Long Term Solitary Confinement to Short Term Disciplinary Restrictive Housing

Abstract Following a recent study of disparities in solitary confinement (SC) placements in Florida, we examined related disparities in the use of extended restrictive housing in Ohio (SC conditions) while expanding the analysis to short term restrictive housing, a substantially more common prison experience. Analyses of 183,872 incarcerated persons (IPs) revealed substantive disparities in prevalence and incidence of placements in both short- and long-term restrictive housing (RH). Controlling for types of rule violations and other risk indicators, disparities emerged based on an IP’s sex, age, race, education, learning skills, substance abuse risk, and mental health. Many findings are consistent with the Florida study and extend to the more routine short term RH. Similar studies will be critical for generating a body of knowledge that may demonstrate the same types and levels of RH disparities regardless of prison system, contributing greatly to RH policy debates and critical criminological perspectives.

The enormous growth of prison populations across the U.S. during the latter part of the twentieth century led to greater academic attention toward strategies employed by prison systems to maintain order (e.g., Camp et al., 2003;Steiner & Wooldredge, 2009;Wooldredge et al., 2001).One common strategy that has received much of this attention is the use of segregation including solitary confinement (SC) and other forms of restrictive housing (RH) (e.g., Browne et al., 2011;Cloyes et al., 2006;Morris, 2016).However, studies of the effectiveness of SC or RH for improving prison safety have contributed to skepticism over their potential benefits (Briggs et al., 2003;King, 1999;Labrecque & Smith, 2019;Lucas and Jones, 2019;Mears, 2008Mears, , 2010;;Mears & Bales, 2009;Morris, 2016).
Despite this skepticism, RH is used widely across prison systems and in many forms (Beck, 2015;Liman Program & Association of State Correctional Administrators, 2015;Pullen-Blasnik et al., 2021).Prisons rely heavily on short term RH to respond to acute problems within a facility and rely on extended RH rooted in the assumption that long term solitary confinement can improve overall prison safety and the day-to-day operations of facilities (Mears et al., 2021;Shalev, 2009).This prevalence begs critical unanswered questions about how RH is used and the profiles of individuals most likely to serve time in RH.The lack of conclusive evidence that RH provides any benefits to prison operations highlights the potential for unnecessary harm upon the individuals typically housed there.Who, though, is typically housed there?Mears et al. (2021) recently started to answer this question as it relates to incarcerated persons (IPs) placed in solitary confinement in Florida prisons.Here we provide a replication of their analysis for the state of Ohio and then an important extension with a focus on short term RH which, as noted by Mears et al. (2021), is far more common in prison systems for maintaining safety and order.In doing so, we provide a profile of all individuals who experienced any kind of RH during their prison term within a 10-year study window across Ohio's entire state prison system.Our paper is thus an important extension and replication of prior work.Replication studies are relatively rare in criminology journals (McNeeley & Warner, 2015) even though replication follows the logic of scientific inquiry and the need to validate the findings of any one study.For the topic examined here, empirical findings can have ill consequences for prison systems willing to invest heavily in areas targeted for reform (e.g., prison programming, restrictive housing) if those findings are atypical of a more general reality, which is possible by chance.A large portion of our study also provides an original analysis of short term RH, extending Mears et al. (2021) with the ability to examine not only whether their findings for solitary confinement in Florida extend to Ohio, but also whether the profiles of individuals more likely to be placed in solitary confinement extend to the much more common use of short term RH.
Solitary confinement/long term RH is typically central to policy deliberations and reform movements.This centrality makes sense because of the potentially strong deleterious impacts of long-term exposure to single cell confinement with no congregate activities and essentially no human socialization.However, this experience is substantially less common than shorter term uses of RH that often do not involve solitary confinement (Mears et al., 2019).Short term, non-solitary RH stays appear quite similar to these extended RH placements, except of course for duration, the number of cellmates, and terminology (Rubin & Reiter, 2018).Thus we focus on both types of RH because both are used with the intention of improving prison management and both are understudied.

Restrictive Housing in Ohio Prisons and Attendant Goals
For proper context, it is important to begin with a description of restrictive housing in Ohio prisons and the purposes for which it was designed.The Ohio Department of Rehabilitation and Correction (ODRC) uses two forms of RH: within-facility, short term RH (SRH) placements and extended RH (ERH) placements. 1 Although terms of "supermax" or "close management" are not formally used by ODRC, the conditions in ERH are comparable to housing in other systems that would be designated with these labels, including the standard use of solitary confinement.Most placements in ERH are the result of a change in classification/security level that then result in a subsequent transfer to one of two maximum security (to use ODRC's nomenclature, this includes security levels 4b and 5) facilities in Ohio, although in more rare instances IPs may start their prison sentence in one of these facilities at that security level.
According to ODRC, both SRH and ERH exist to promote the safety of incarcerated persons and staff.To be clear, however, there are no policy documents articulating how each form of RH should accomplish this goal beyond isolation/separation from general population (e.g., whether placement in RH is intended to deter the person so placed and/or to deter others by example).Relevant wording in formal RH policies is limited to ensuring the safety of incarcerated persons and staff.
Formal policies do offer some guidance surrounding the kinds of individuals that should (and should not) be targets for RH given the larger goal of promoting safety.SRH is used specifically for prison rule violators although members of security threat groups (STGs) are sometimes placed in SRH to prevent escalations in individual or group conflicts even if those specific members have not violated any rules.The explicit purposes of placing individuals in SRH include "security control" at the time of a rule violation or a threatening situation posed by STG members when a shift captain determines that those involved pose an immediate threat to the safety of others or to themselves; "disciplinary control" when a suspected rule violator has been found guilty of the violation and a rule infraction board (RIB) administers SRH as punishment for the offense(s); and "local control" (akin to security control) when a rule violator who has been sent to RH for discipline engages in other rule violations while in RH, thus raising concerns about their threat to safety and resulting in time added to the original sanction imposed by the RIB.
Regarding ERH, ODRC form 53-CLS-04 (12/4/2018) states the following: It is the policy of the Ohio Department of Rehabilitation and Correction (ODRC) to limit the use of Extended Restrictive Housing (ERH) to only those inmates who pose the greatest threat to the safety and security of a correctional facility and cannot be managed safely in general population.ERH is not to be used for punishment and is reserved for 1 In Ohio, short term RH placement involves housing an individual in a cell block separate from the general population but within the same facility, and most often in a two-person cell for no more than one month.Extended RH placement involves moving a person out of their facility and into a maximum security or supermax prison, and always into a single-person cell for several months or more.For an ERH placement for someone already housed in an Ohio maximum security unit/prison, they will stay in that facility (unless sent to the supermax prison in youngstown, OH), in which case they move to a segregation unit within that facility.Privileges permitted to the general population are denied in both SRH and ERH with the exception of up to two visits per month in ERH.Movement outside the cell is limited to 30 minutes (supermax) or from 30 to 120 minutes (all other forms of RH) per day, depending on the facility.Hand and leg restraints are used when a person leaves their cell, unless at recreation.Lighting in the unit is required 24 hours each day and must be adequate enough for staff to make rounds and confirm the individual is safe.The designation "short term" does not always mean "short," however, even though the average time spent in SRH was 16 days in our sample.+ These stays can be extended by weeks in some cases if individuals are deemed by prison personnel to have behaved poorly while in RH.
those whose violent, disruptive, predatory, riotous or other serious misbehavior poses a serious threat to other inmates, staff, the orderly operation of the institution, or the general public.
The focus of ERH on "management" as opposed to "punishment" overlaps with SRH to some extent (similar to "security control" and "local control").Further clarification on ERH is provided in ODRC form 53-CLS-01 (1-7-2019): Control Units/Prisons -The most restrictive housing assignments in the 3-Tier System containing ERH and ERH-T inmates.Control units are designed to prevent violent and disruptive inmates from interfering with the safe operation of prisons.They also are designed to protect the community from those who have the greatest risk of escape (typified by a history of escape).

Evidence on the Use of Restrictive Housing
It is important for both theory and policy discussions to develop an understanding of correlates to RH placements (see, e.g., Motiuk & Blanchette, 2001;O'Keefe, 2008).What types of IPs are more likely to be placed in RH?This question is particularly important in the context of shorter term, disciplinary segregation placements because they are far more common than extended and supermax-like placements (i.e., 34.4% of all IPs admitted during our study window and 31.1% of those admitted and released during our study window served at least one stint in SRH, versus 2.4% and 1.7% in ERH, respectively).Related studies can provide insights into the microlevel processes that inform prison decisions to place IPs in SRH and reveal any disparities in its use based on demographics (age, sex, race, ethnicity) and other background factors (education level and learning skills, substance abuse risk, mental health, criminal history).These insights are, among other things, critical for assessing whether any collateral harms of RH practices might be unequally distributed across the prison population.
To underscore the salience of this issue, it is important to consider how RH is 'prison within prison' (where SRH is used for rule violators in a specific prison population and ERH is used for those deemed safety risks across an entire state prison system).Sykes (1958) noted that an imprisoned individual's loss of liberty is twofold whereby a person is first sent to a prison facility and then restricted to where they can move within that facility.By definition, RH represents the extreme form of the latter.Prisons are inherently coercive organizations by nature of forcing the confinement of convicted persons, severely restricting their civil rights and freedom of movement, treating their needs as subordinate to the efficiency of staff routines, and saddling many with concerns over use of force by staff in response to perceived disobedience.These deprivations can be greatly intensified for individuals placed in restrictive housing, especially ERH.Aside from the harsher confinement conditions of SRH relative to the general prison population, these conditions vary in intensity across Ohio prisons based on a number of factors, particularly unit officers' engagement in problem solving, respect towards those incarcerated, and tolerance of verbal abuse; noise levels contingent on the composition of those held in a unit at any given time; and the age and architecture of a facility whereby the cells in older RH blocks (formerly "solitary confinement") are smaller and subject to greater fluctuations in physical climate.ERH conditions also vary in similar ways across the few facilities designed for that purpose although cells are larger, even in the older facility.It is challenging to defend the human costs of RH, and impossible to do so in the absence of evidence favoring its utility for prison safety, let alone for subgroups of incarcerated persons who may face much higher odds of placement in RH.
Focusing on long term restrictive housing in Florida prisons, specifically "extended solitary management" or "ESM" (supermax-like conditions of confinement), Mears et al. (2021) conducted perhaps the most rigorous study to date of individual level correlates to ESM in a pool of over 190,000 state prisoners.Our goal here is to extend directly from their work-to advance a "parallel line of research…that examines the factors associated with placement into other types of housing" (Mears et al., 2021(Mears et al., :1512; see also Wildeman & Andersen, 2020)-by examining many of the same predictors of ERH in a different prison system while also extending this analysis to SRH.As such, we extrapolate their arguments to an understanding of SRH placements because the arguments can be logically extended.
An underlying theme to the predictors of RH examined by Mears et al. (2021), as well as others, is that factors which present IPs as potentially greater threats to safety and management, whether real or stereotyped, are more likely to show up in RH populations given the purpose of RH to promote safety and good order in the general prison population.Individual demographics are relevant in this regard, as men and younger IPs are potentially at higher risk of violence than women and older IPs, respectively (Lovell et al., 2020;Mears and Bales, 2009;Mears et al., 2021;Toman, 2022), and racial/ethnic minorities might be stereotyped by certain staff as more violent than White IPs (Tasca and Turanovic, 2018;Wooldredge & Cochran, 2022), in conjunction with some evidence that Black IPs are more likely to be confronted for violent rule infractions (Harer and Steffensmeier, 1996;Steiner & Wooldredge, 2015).
There is a relatively large body of research on the effects of RH placements on an individual's mental health but far fewer studies of mental health effects on the odds of RH placement.Related to the latter group of studies, recent work by Siennick, Picon, Brown, and Mears et al. (2021) of Florida prisons, and Simes et al. (2022) of Pennsylvania prisons provide strong evidence of mental health as a predictor of SC (see also Mears et al., 2021).These results are consistent with the idea that IPs with poor mental health are more likely to commit and be cited for rule violations (Felson et al., 2012) and that individuals with poor mental health are either stigmatized in a prison setting (Simes et al., 2022) or staff are concerned that particular individuals with poor mental health might interfere more with facility order and/or antagonize others to behave aggressively towards them (Siennick et al., 2022).On the other hand, good intentions might underlie some of these RH placements if certain staff believe it is necessary to ensure an individual receives more mental health attention.Other scholars have also found evidence of mental health effects on RH placements (examples include Clark, 2018;Dellazizzo et al., 2020;Labrecque and Mears, 2019;Lanes, 2011;Severson, 2019;Warfield, 2013).
Substance abuse risk might also increase the odds of RH placement if it impacts how staff perceive IPs (i.e., as more or less of a risk to safety and order) (Mears et al., 2021;Motiuk & Blanchette, 2001).Also, in the face of limited recovery services programs to address the needs of drug offenders, staff may rely more on RH to manage those who engage in drug violations and also have histories of drug abuse by separating them from general population and reducing their access to drugs.
RH placements might also be impacted by an IP's education level and adult learning skills, whether those factors are linked to actual misconduct or whether they feed staff perceptions of IPs at higher risk of committing rule infractions in the future (Mears et al., 2021;Motiuk & Blanchette, 2001;O'Keefe, 2008).It is possible that persons with weaker learning skills might have greater difficulties with understanding rules and directions, and the rules of all state prison systems include "offenses" such as not following direct orders, showing disrespect to staff (which can be as simple as ignoring them), and being "out of place" or not being where they are supposed to be at particular times of the day.
Perhaps most intuitively, we would expect any reliable indicators of "true" risk for misconduct, in conjunction with misconduct itself, to be the strongest predictors of placement in either SRH or ERH (see Mears et al., 2021, for their argument).Risk indicators might include gang membership and activity level, the most serious offense incarcerated for, sentence length (correlated to the severity of the offenses incarcerated for), and number of prior prison sentences.Regarding specific types of rule violations, IPs confronted for and found guilty of violent offenses might have the highest odds of RH placement, given that violent offenses may be most problematic for maintaining prison safety, which is the explicit goal of RH.These odds could also be very high for individuals found guilty of drug offenses given the problems with the proliferation of drugs in prison.Moreover, IPs with more of these types of offenses as well as more rule violations in general would likely face even higher odds of RH placement given a track record of failure to abide by facility rules.
There are other potentially relevant predictors of RH placements not examined by Mears et al. (2021), including whether IPs have jobs while incarcerated, visitation with outsiders, and program participation during incarceration.All of these factors could potentially be considered in decisions to send rule violators to RH if they are treated as reflections of an IP's social integration and investment in conventional goals.Specifically, prison industry jobs are arguably the most coveted among IPs because they offer the highest wages and involve semi-skilled labor (Anderson et al., 2022), and holding a prison industry job is an outcome of establishing trust over time (Wooldredge & Cochran, 2022).As such, IPs holding these jobs may have lower odds of being placed in RH.Similarly, participation in cognitive-based reentry programs (e.g., education and recovery services) might demonstrate to decision makers an IP's investment in positive change while also reflecting an IP's generally lower risk to the disruption of prison order (Long et al., 2019).The frequency of visits with outsiders could operate similarly but with the added incentive to keep IPs with more visits out of RH in order to avoid disruptions to contacts with family members (Anderson et al., 2022).We do not speak here to whether these in-prison experiences affect RH placements (as the relationships could operate in either direction) but examine them to provide additional knowledge on the "profiles" of IPs who experience RH.Consistent with Mears et al. (2021), "The(se) models are not meant to capture causal pathways into ESM but rather to identify potential disparities in placements" (p.1506).

Method
The analysis described here was designed to answer the following questions: What factors predict (a) whether an Ohio IP is ever placed in SRH and ERH during his/her sentence, (b) whether those with such placements served two or more stints during their sentence, (c) whether they experienced these placements within the first 6 months of their sentences, and (d) whether IPs sent to ERH experienced at least one year of confinement.Placements in SRH and ERH were examined separately.Length of confinement (≥ 1 year) was examined by Mears et al. (2021) because of their focus on SC.Here we examine the same outcome for ERH but not SRH because the latter terms are short, by definition.Some but very few IPs spent at least one year in SRH over the course of their sentences.

Sample
The focus here on profiles of IPs most likely to be placed in RH was part of a larger study of 224,288 IPs admitted to Ohio prisons during a 10-year period .Facilities include all 32 public and private state prisons in Ohio in operation at any point during the 10-year period.The larger project included the population of admissions during this period although we treated these cases as an analytical "sample" of prison admissions under the assumption that it is a representative cross-section of persons admitted to Ohio prisons before and after the study window.However, we cannot speak to the specific time frame of generalizability (e.g., three years before through three years after, versus 20 years before through the present).
This analysis includes all individuals admitted to and released from OH prisons within our study window, which encompasses 183,872 IPs or 82% of the admissions pool (the remaining 18% were admitted during this period but remained in prison as of December 31, 2016).We selected all IPs who completed their sentences in order to describe the totality of certain events throughout a sentence such as worst mental health status during a prison term, whether s/he was able to work in prison industry, total visits, total rule violations, etc.

Data and Measures
Variables for the analysis were derived from administrative data systematically collected by the ODRC between 2007 and 2016 (i.e., we created variables for the statistical analyses described herein by transforming the administrative data in its original form).The applicable administrative data files were uploaded to a secure server for our personal access by the ODRC.The administrative data included information across the following domains: demographics (sex, birth year, race, ethnicity), incarceration history in Ohio, felony offense levels incarcerated for (F1A through F5), all mental health assessment scores and dates during the study window, security threat group (gang) activity, admission date, release date for those released during the study window and expected release date for persons held beyond 2016, security classification, reading and math scores on the Comprehensive Adult Student Assessment Systems (CASAS) aptitude test, prison GED program completion, recovery services program participation and withdrawal/completion dates, prison industry job participation, substance abuse risk score (from the Texas Christian University Drug Screen II assessment; Institute of Behavioral Research, 2007), visitation dates during confinement, dates and types of all rule violations committed during the study window, dates of all RH placements, dates of all other movements within the Ohio system and the facilities moved from and to, and dates of an individual's return to prison for a new crime or parole revocation (through May 2021). 2  All variables for the analyses presented here are described in Table 1 although some of these require more details for assessment.Regarding mental health scores, ODRC's Training and Education Manager for Behavioral Health operations provided the following background: Following the disposition section of the "Mental Health Evaluation (DRC5161)," an IP is assigned a mental health classification, and an "Independently Licensed Mental Health Professional then completes a Mental Health Classification (DRC5286) on all offenders who have had a mental health evaluation completed." This assessment has been in use since the beginning of our study window and has never been validated.For classification purposes, ODRC designates an IP's mental health as follows: no mental health issue; mental health issue but not an intellectual and developmental disability (psychotherapy track); and seriously mentally ill (chronic care track)."Psychotherapy track" includes a DSM-IV diagnosis, the offender 2 Full information was available for all data fields except CASAS literacy and math scores and TCU substance abuse score.Literacy scores are missing for 1.7% of the sample, and math scores are missing for 2.3%.Predicted values for these scores were computed from regression models with indicators of participation in different reentry approved programs and the specific type(s) of offense(s) for which incarcerated (derived from statutory offense codes), created from additional data provided by ODRC, and these values were imputed for missing cases (Pearson R > .60).Missing values on the TCU score are due primarily to individuals with no documented substance abuse histories and drug offenses (estimated at roughly one-third of the sample by ODRC researchers).However, there are additional cases with missing TCU scores for unknown reasons, and altogether there are 37% missing values on this scale.We were unable to discern whether blank entries for these cases are due to no documented substance abuse histories and prior drug offenses versus "other" reasons.Otherwise, persons without substance abuse histories and drug offenses would have been assigned the lowest value on the TCU scale.Unlike for CASAS scores, predicted values from regression models were not used to replace these missing values because of poor prediction (Pearson R < .40).Instead, the mean TCU score for all available cases was used to replace the missing values.This approach may have weakened the "true" estimates for this variable and so those estimates should be interpreted with caution. is receiving mental health services which either include or do not include medication prescription, and "the offender's acuity functional level is not impaired as demonstrated in a pattern of high risk behavior.""Chronic care track" includes offenders with the SMI designation-"a substantial disorder of thought or mood which significantly impairs judgement, behavior, and capacity to recognize reality or cope with the Notes: interval/ratio variables include age, # prior prison sentences, sentence length, substance abuse risk score, casas scores, # visits during sentence, total rule violations during sentence, total violent offenses during sentence, and total drug offenses during sentence.all other variables are binary (coded 0,1).the mean of a binary variable coded 0 and 1 is the proportion of cases scoring "1", and the standard deviation is the square root of the product of the proportions of 0's and 1's (with values closer to .5 reflecting a more even distribution of cases between the two values).
that has not been validated and/or missing cues elicited by certain IPs that would have led to their reassessment of mental health during incarceration. 3ecovery services programs are rehabilitative substance abuse programs involving progression through a curriculum that seeks to address documented substance abuse problems assessed via the Texas Christian University Drug Screen II instrument.ODRC designates its recovery service programming as reentry-approved.
GED prison programs are the only mandated prison programs in Ohio.Anyone without a high school diploma or GED at admission is required to participate in the GED program for up to 6 months.Participation is voluntary thereafter if the individual has not yet earned a GED.
The CASAS aptitude test administered at intake is designed to assess the relevant real-world basic skills of adult learners (Comprehensive Adult Student Assessment Systems, 2015).CASAS tests provide the research-based measurements of skills relevant to effective functioning at work and in life such as reading comprehension, mathematics, writing skills, and workforce preparation.Every individual sent to an Ohio prison is required to take reading and mathematics tests at intake.The reading and math scores each range between 150 to 260, with higher scores representing higher aptitude.In practice, the reading scale is divided into five levels from A (Beginning Literacy, 200 and lower) to E (Advanced Secondary, 245 and higher).
Security threat group activity is recorded at four levels in Ohio: not a gang member/no participation in gang activities, passive (inactive gang member), active (active membership in a gang but not a threat to order and safety), and disruptive (engages in gang activities within prison that pose threats to order and safety).Here we examined known gang activity at prison intake so it could be included in both sets of analyses described below.
Sentence length is measured in months and reflects the sentence administered by the court.Some IPs were either released before the administrative sentence termination date or after the date, but the original scale is retained because it provides an additional indicator of risk and offense severity.

Statistical Analyses
The analysis proceeds in two parts.First, we examine the profiles of IPs that experience different forms of RH.We differentiate RH subgroups into the following eight categories: no SRH or ERH, any SRH stays (with no ERH stays), 2+ SRH stays (with no ERH stays), SRH placement within 6 months of admission to prison, any ERH stays, 2+ ERH stays, ERH placement within 6 months of admission to prison, and ERH placement(s) totaling one year or more.The four ERH measures are consistent with the four outcomes examined by Mears et al. (2021), and here we apply their first three ERH outcomes to SRH.The SRH categories do not include individuals who experienced ERH to better isolate the factors associated with SRH.Due to the high prevalence of SRH, it is not possible to focus only on ERH IPs who did not experience SRH.This segment of the analysis includes all 183,872 individuals admitted and released during the 10-year period, broken down by the subgroups noted above.
The second part of the analysis focuses on relevant predictors of placements in RH, treating the last seven outcomes above as dependent variables in multivariable binary logit models, which provide insight into when and how prisons use the practice.(Findings for the indicator of "no SRH or ERH" are redundant with findings for any SRH stays and any ERH stays.)The multivariable analysis provides a more rigorous assessment of potential disparities that emerge in SRH and ERH placements via statistical control.Following Mears et al.'s (2021) analysis of SC, the models of placement in SRH and ERH are estimated with the full sample (183,872 IPs) whereas all subsequent models are estimated with the pool of individuals placed in each type of RH (54,676 IPs for SRH; 3,038 IPs for ERH).
The profile (bivariate) analyses include both background characteristics (demographics, prior prison terms, substance abuse score at intake, etc.) as well as experiences that unfold over the course of a prison term (visitation, work in prison industry, worst mental health status during sentence, etc.).These provide potentially important insights into the types of experiences that overlap with various RH subgroups.However, factors reflecting in-prison experiences are excluded from the logit models predicting membership in each of these subgroups because of the problems those variables create for causal order (e.g., RH placements necessarily interfere with jobs, visitation, and program participation).Those models thus only include factors recorded at intake, including mental health scores.Each multivariable model was estimated for the pool of 183,872 IPs admitted and released between 2007 and 2016.
The identities of all persons admitted to Ohio prisons during the 10-year study window were anonymized by ODRC.In conjunction with analyses of agency data only, recruitment of IPs for the study was not necessary.Approval for the larger study and the specific analyses described here was granted by the University of Cincinnati Institutional Review Board (2016-8256), and the ODRC Institutional Review Board in conjunction with the Ohio Department of Health Institutional Review Board (IRB00002180).
The Data and Methods Appendix, available on the journal website, provides additional information on the study in conjunction with a codebook and the computer syntax used to generate all findings presented here.

Negotiating the Challenges of Administrative Data
It is important to acknowledge that the validity of our findings largely depend upon the validity of ODRC's administrative data used for the analyses.We thus find it important to describe the collaboration and quality control steps taken to use ODRC's administrative records.A great deal of work was involved on both our part and by ODRC research staff to check the validity of these data, make corrections where needed, and ultimately use only the data fields considered most valid by ODRC research staff.These efforts were made possible by a statewide record keeping system for the variables examined here.The ODRC Office of Research manages centralized data bases which include information on individuals' demographics and social backgrounds, incarceration histories, crimes for which incarcerated, system intake scores (on substance abuse risk, mental health, and adult literacy), risk assessment scores (although the assessment subsequently changed from RAP to ORAS several years later), security threat group status (prison and street gangs), facility transfers, movements into and out of the different types of RH, rule violations including criminal offenses during the current sentence, prison programs and status of completion (wait-listed versus participated in versus completed), jobs during incarceration, visitations, information on returns to prison after release, and the corresponding dates for all dynamic fields (rule violations, RH movements, re-assessments, facility movements, program participation, visitations, and other events throughout a prison sentence).
We had discussions with ODRC research staff at the beginning of the larger project on what they considered to be the most reliable data fields within their centralized data bases.These discussions led us to exclude certain fields, such as when incarcerated persons held specific types of in-prison jobs (aside from prison industry, which is very well documented) and specific years of education or education level aside from high school degree or GED.We were also able to cross-reference some of the static fields across data sets to check consistency (e.g., person's sex versus sex of population in prison held; GED or high school completed prior to intake versus GED class taken during current sentence; anticipated release date versus date of placement into transitional control).In instances of conflicting information (under 3% of all cases for any of these fields), we coded the field from the most reliable data set for that field as suggested by research staff (e.g., prison facilities held for a person's sex, education program file for high school degree or GED at intake, facility movement file for release date where final move indicates system discharge).Bogus coding errors (insertion of characters or numerical codes falling outside the range of legitimate values) were not an issue for the variables examined here because they are caught and edited by ODRC in the centralized record system.With that in mind, it is important to acknowledge that these data bases are fed by personnel with potentially varying attention to detail, such as a shift captain versus a line officer entering information on rule violations, or duplicate information entered by someone who ignored a previous entry in the system (e.g., duplicate movement dates).While many of the potential problems associated with this heterogeneity can be checked (e.g., duplicate entries on movement, inconsistencies in data fields across data bases), we could not fully assess the quality of data entry by all those involved.
A broader concern about the ability of administrative records to make sense of on-the-ground prison operations also warrants attention.We consider the administrative data examined here to be suitable for our goals because we are providing descriptions of individuals more likely to be placed in SRH and ERH.With these data we are unable to disentangle the degree to which the overrepresentation of any one group reflects deliberate biases on the part of prison staff.The latter is arguably the more interesting question although the existence of overrepresented groups in RH must first be established and, if established, would be consistent with deliberate bias.Answering the tougher question of intentional discrimination requires field observations across multiple prisons, staff focus groups, and/or broader survey data reflecting both staff and incarcerated persons experiences with and perceptions of biases in RH decisions.

Profiles of IPs Placed in RH
Table 1 presents univariate descriptive statistics that illustrate the profiles of the no-RH and RH groups noted above.Here we note some of the key findings from the group comparisons.First, men have a higher likelihood than women of being placed in SRH, and also within the first 6 months of their sentences, and this sex difference becomes slightly greater with more placements in SRH.Moreover, men are the only ones placed in ERH even though these units are available in the largest state facility for women.
Second, IPs in SRH and ERH are generally younger than those never placed in RH, and the ERH group is younger (on average) than the SRH group.The average age at admission to prison for the no-RH subgroup is 34.1 years whereas the average ages at admission for the RH groups are all under 30.The youngest group includes IPs with 2+ ERH placements over the course of their sentences (25.6 years).
Third, each SRH and ERH group contains a substantially higher proportion of non-Latino Black IPs, and the ERH groups include the largest proportions of Black IPs.The no-RH subgroup is 38% Black whereas all RH groups are over 45% Black.The ERH, ERH 2+, and ERH ≥ 1 year groups are 53, 56, and 55% Black, respectively.
Fourth, IPs placed in both forms of RH have fewer years of education as a group relative to IPs never sent to SRH and ERH, with even greater differences when moving from SRH to ERH.IPs with high school diplomas made up 20% of the group never placed in RH, versus 17% of the SRH group and 14% of the ERH group.The group of IPs serving two or more terms of ERH were the least educated with 12% having high school diplomas.However, we see no substantive differences across these groups on learning skills based on CASAS scores, with overall reading scores dropping only from 233 to 231 when moving from no SRH or ERH to ERH 2+, respectively, and math scores dropping from 228 to 225.
Fifth, IPs in SRH and especially those in ERH have more serious offending profiles.These groups have substantially longer sentences, substantially larger proportions of gang members regardless of activity level, and larger proportions with the most serious felony convictions (felony 1s and 2s).On the other hand, they do not have larger numbers of prior incarcerations, with fewer prison terms served (on average) among IPs with more stints in either SRH or ERH relative to individuals with no RH placements.Taken together, these observations suggest that the severity of past criminality trumps the longevity of criminality for predicting RH placements.
Sixth, mental health emerges here as a strong correlate of all RH groups except for the group placed in ERH within 6 months of admission to prison, but especially 2 or more stays in SRH and ERH.For example, 9% of no-RH IPs are classified as chronic care track (per ODRC) compared to 17% of each of the SRH 2+ and ERH 2+ groups.
Seventh, IPs in SRH and ERH accumulate far more infractions than no-RH IPs, with the highest counts of violence, drug offenses, and rule violations overall found in the ERH groups, except for ERH within the first 6 months of incarceration (with higher counts for SRH within the first 6 months).In conjunction with the other comparisons described above, the profiles of IPs with the highest odds of placement in RH include young, less educated, minority men at greater risk of both poor mental health and posing a threat to safety.These profiles are further magnified when moving from short term to long term RH (i.e., greater representations of young minority men with poor mental health and higher risk to prison safety).
Important to note is that some of the group comparisons were counter to expectations, with RH groups including higher proportions of IPs working in prison industry, more visits during a sentence (on average), and lower proportions of IPs who did not participate in a recovery services program.All of these could reflect differences in sentence lengths, however, with longer sentences corresponding with higher risk of RH placement in conjunction with more time to (a) become eligible for prison industry work, (b) accumulate visits, and (c) complete a recovery services program.

Higher-Order Predictors of RH
Table 2 presents results from multivariable logistic regression models of each type of RH, with only factors recorded at prison intake as predictors (due to issues related to causality, as noted earlier).An IP's sex was excluded from the models of ERH because, as noted above, no women were placed in ERH during the study window.Also, Felony 1A's were combined with Felony 1's because treating 1A's as a separate category produced inflated estimates and standard errors.Across all models, findings illustrate potentially critical disparities in RH placements consistent with several of the profile factors identified at the first stage of the analysis.
Sex differences in the odds of SRH persisted in the multivariable analyses, with men facing 45% higher odds of placement in SRH relative to women.Among IPs placed in SRH, the odds of serving two or more stints are 20% higher for men (p < .001)and the odds of placement within the first 6 months of admission are 9% higher for men (p < .01),even when controlling for risk factors recorded at intake such as known gang activity, severity of the incarcerated offense, prior imprisonment, sentence length, and substance abuse risk.
Significant racial disparities also exist in SRH and ERH placements.Non-Latino Black IPs are significantly more likely to experience both SRH and ERH compared to Non-Latino White IPs, and these odds are higher for ERH versus SRH (e b = 1.13 versus 1.17, respectively).The odds of an ERH placement are even higher among Black Latino IPs relative to White IPs (e b = 2.24), suggesting an even greater impact when considering an IP's race in conjunction with ethnicity.Among individuals placed in SRH, Non-Latino Black IPs also face significantly higher odds of two or more such placements relative to White IPs, as well as higher odds of being sent to SRH within 6 months of admission to prison.These last two disparities do not hold among IPs placed in ERH, however.
Known gang activity at intake emerges as one of the strongest predictors of SRH placements and especially ERH placements.In particular, disruptive gang members are over 18 times more likely than non-gang members to be placed in ERH versus twice as likely to be placed in SRH.Both active and passive gang members, on the other hand, were six times more likely than non-gang members to ever serve time in SRH, versus four times and three times more likely (respectively) to serve time in ERH.
Among persons ever placed in SRH, any type of gang activity was a significant and strong predictor of serving two or more stints in SRH (p < .001;e b ≥ 1.94), and a significant predictor of being placed in SRH within the first 6 months from admission to prison (p < .001;e b > 1.3 for active and disruptive members).By contrast, among those ever placed in ERH, only disruptive and passive gang members were significantly more likely to serve two or more terms (p < .001;e b > 2.2), and all three groups had substantially lower odds of being placed in ERH during their first 6 months in prison (p < .001;e b < 0.55).None of these groups were significantly more likely than non-gang members to serve at least one year in ERH.
Regarding other indicators of safety risk, each additional prior prison term increases the odds of placement in SRH by 10% and the odds of placement in ERH by 7% (p < .001).Among those ever placed in SRH or ERH, each additional prison term increases the odds of two or more SRH placements by 7% versus 22% for two or more ERH placements (p < .001).Prior imprisonment is unrelated to the remaining two ERH outcomes, yet significantly and inversely related to placement in SRH within 6 months of prison intake.The most serious felonies convicted for are statistically significant predictors across the board (save F4s in the ERH models only), with estimates in expected directions except for RH placements within 6 months of prison intake.Among IPs ever placed in either form of RH, those convicted of felonies more serious than F5s are significantly less likely to be placed in SRH or ERH within 6 months of their admission to prison (p < .001).Also related to safety risk, one unit increases on both sentence length and substance abuse risk score correspond with significant and comparable increases in the odds of either SRH or ERH placements.
Higher levels of both education and reading/math skills (CASAS) generally coincide with significant declines in the odds of each SRH outcome, the exception being a nonsignificant estimate for readings scores in the model predicting two or more placements in SRH during incarceration.However, the CASAS scores were  nonsignificant predictors of the ERH outcomes, and a higher level of education corresponded with significant declines in the odds of ERH but not in the odds of the other ERH outcomes.Finally, poor mental health at prison intake maintains significant and substantive impacts on the odds of either type of RH placement among all IPs (p < .001;e b range = 1.32 to 2.04), and on the number of stints served in each among IPs placed in either type (p < .001;e b range = 1.41 to 1.73), with the strongest effects on SRH among IPs with the worst mental health (chronic care track).The mental health estimates are significant across all three SRH models but only the first two ERH models displayed in Table 2. Also, differences in estimates between IPs classified as psychotherapy track versus chronic care track are substantive for SRH but not for ERH.Both groups maintained similarly higher odds of both ERH placement and the number of ERH placements relative to IPs with no mental health issues.

Discussion and Conclusions
This study is a response to calls by scholars for more empirical and systematic analysis of RH practices (e.g., Garcia, 2016;Mears, 2013;Morgan et al., 2016;Shalev, 2009), and constitutes a replication and extension of Mears et al. 's (2021) study of the profiles of IPs in Florida prisons with the highest odds of being placed in solitary confinement.
Here we examined similar profiles for IPs in Ohio prisons with the highest odds of being placed in solitary confinement (ERH) and SRH.Our results provide a comprehensive analysis of how one prison system (Ohio) uses its two main forms of RH when pursuing its stated goal of "safety."The results have important implications for research and policy.

Implications for Research
Several "extra-legal" disparities in RH placements were uncovered in the multivariable analyses even when controlling for safety risk, and some of these disparities grew in magnitude as the severity of RH increased (i.e., from SRH to ERH).These disparities included significantly higher odds of both SRH and ERH placements for men (with no women placed in ERH), younger IPs, Black IPs (with even higher odds when moving from any SRH placement to any ERH placement), less educated IPs (with comparable odds for either SRH and ERH), with weaker learning skills (but for SRH placements only), at higher risk for substance abuse (based on TCU scores), and those with poorer mental health (with markedly higher odds of SRH for individuals on the chronic care track).
These findings raise the importance of exploring disparities in decisions that precede and potentially influence decisions to place IPs in RH, as well as decisions that could further exaggerate group differences between SRH and ERH placements.For example, differential rule enforcement by officers could magnify disparities in the use of SRH if non-white IPs have higher odds of being written up for certain offenses in conjunction with higher odds of being placed in SRH either at the time of the incident (by shift captains) or after being found culpable by a rule infraction board.The results also further underscore the need to develop a rigorous understanding of the intended and unintended impacts of RH placements across prison populations (e.g., Wildeman & Andersen, 2020) as well as the potential for disparate impacts due to differential exposure to SRH and ERH for already disadvantaged and high risk groups.
Scholars might also consider how male facilities differ from female facilities in RH usage.We found that females who served their entire sentences during the 10-year study window were never placed in ERH.Research is needed that can discern the extent to which sex-based differences in how prisons use RH stem mainly from practical considerations (e.g., reduced bed space pressures in female SRH units; a limited ERH physical infrastructure) versus philosophical differences in the strategies used to manage incarcerated men and women.By extension, research is needed that can then more closely consider the relative impacts of these strategies across facility types.
Research should also focus as much attention on shorter and more routine usages of restrictive housing as it does on supermax incarceration and long term solitary confinement, despite the fact that those shorter, within-facility stays do not necessarily include solitary or single-cell confinement.Restrictive housing, with or without the solitary confinement component (e.g., Rubin & Reiter, 2018), exposes individuals to the most challenging and adverse conditions of incarceration.Any exposure to such conditions may harm individuals and undermine long term correctional goals, including goals of rehabilitation and successful reentry, and more generally the goal of operating carceral institutions in ways that are perceived by IPs and society members as procedurally just.Logically, then, research that focuses on the more common forms of RH may be just as if not more valuable for informing policy deliberations as studies of the most extreme versions of it.

Implications for Policy
Recall our description of ODRC's policies related to the use of SRH and ERH for promoting the safety of incarcerated persons and staff.According to this logic, any disparities in RH use might result from prison facilities targeting particular groups who pose greater safety risks, and so disparities could reflect group differences in risk (as measured by the prison system).It is not clear though that these disparities clearly link practices for maintaining safety and order with the actual threats to safety and order within a prison facility.RH use is common across prison facilities and IPs' experiences, but does the over-representation of younger IPs, minorities, less educated IPs, and those facing mental health issues actually improve the conditions of confinement for the general prison population?The zero-order correlations (r) between each of these factors and the number of violent offenses committed during incarceration are roughly .10(non-Latino Black), −0.04 (high school degree), −0.03 (for each of the reading and math scores on the CASAS), .05(mental health chronic care track), and −0.20 (age), and are weaker still with total drug offenses and total rule violations overall.Aside from age, these correlations are very weak and translate into proportions of shared variance with total violent offenses ranging from .0009 to .01.The correlation for age is modest at best, accounting for .04 of the variance in violent offenses.The magnitude of these correlations with documented violence during confinement does not account for the levels of over-representation of these groups in SRH and/ or ERH.In short, the degree to which these groups are over-represented in RH are not closely linked, it would appear, to a potential for violence reduction, either via incapacitation or otherwise.
Indeed, there is a growing knowledge base suggesting that RH is not substantively effective for reducing violence and other rule violations (Briggs et al., 2003;King, 1999;Labrecque & Smith, 2019;Lucas and Jones, 2019;Mears, 2008Mears, , 2010;;Mears & Bales, 2009;Morris, 2016).While disparities are cause for concern regardless of RH effectiveness, such research further underscores that the "extra-legal" disparities identified here are not serving a useful management function and may only be generating harsher conditions of confinement for certain groups.In short, the policy response may not be addressing underlying causes of the problem it seeks to address.Prison systems should prioritize comprehensive evaluation of safety and order problems across its many sources, including organizational characteristics, that may lead to disorder and violence and that also may require a response different than a close reliance on segregated housing units.
Prison systems and states should weigh more heavily the potential long term costs of using RH, including the financial costs to DOCs as well as the social costs to those placed in RH given the disparities in placements found here.Selective use of RH that serves to disadvantage some groups more than others might reduce the odds of successful reentry for those groups by interrupting or preventing participation in potentially useful reentry programs while also increasing the cynicism of minority groups toward legal authority.The higher odds of SRH and ERH placements in addition to higher odds of multiple RH placements during a sentence for IPs with poor mental health might restrict their access to mental health resources and programs that could ultimately benefit them upon reentry by improving their quality of life.States should consider more closely other mechanisms that might be used to reduce reliance on RH that might avoid potential individual and system harms.A proper cost-effectiveness evaluation surrounding RH usage would consider not only short and long term implications for misconduct and aggregate prison safety, but also any detrimental impacts RH practices might impose on other correctional goals that fall under the purview of the prison system, including recidivism and other social outcomes for individuals.Related to the above, prison systems need to systematically evaluate alternative responses to people with mental illness that fail to comply with prison rules.At the same time, policymakers and court actors should reevaluate sentencing practices that lead to "warehousing" these individuals in prison systems without adequate services for their needs.The correspondence between poorer mental health and the higher odds of placement in either SRH and ERH found here is consistent with a trend identified in prior research that various forms of RH serve as temporary and largely ineffective solutions to "treating" mental illness among correctional populations (Shames et al., 2015;Simes et al., 2022).Prison systems should evaluate the underlying factors that lead to such a trend, including the possibility that mental health treatment resources in carceral institutions are insufficient and also that state courts disproportionately place people with serious mental illness in prisons and jails in response to criminal offending rather than more treatment-based alternatives.
Finally, prison facilities should closely monitor potentially problematic points of discretion in the day-to-day operations to manage prison safety and order.To our knowledge, few self-assessments of such points of discretion in prison operations have occurred (e.g., decisions regarding citations for rule violations, segregation at the time of the incident, and segregation as punishment after guilt is established).In the context of RH usage, which decision points are the most contentious?Where are inequalities most likely to appear?What are the potential causes of them?Identifying and addressing any problems identified via such assessments is likely to have important impacts in terms of informing future policy changes and improving the fairness and effectiveness of prisons.

Study Limitations
Any implications of the findings presented here should be considered with the study's limitations in mind.First, results stem from empirical analyses of one state's prison system from one 10-year period of time.Although we do think that the time period is long and recent enough so that the results should have substantive relevance for quite some time, and we were able to follow over 80% of our sample through their entire sentences, any characteristics of Ohio's prison system that make it unique during this time period will undermine the ability of these results to generalize to other places and time periods.
Second, the analyses do not include, nor are they augmented by, perspectives and insights of ODRC correctional officers and staff or incarcerated people.Our results stem from analysis of administrative records only.Although this provides many strengths, including the fact that many of the data points are systematically tracked and collected, which avoids many problems that would otherwise have been imposed by missing data, it also means that our results are limited to what we can observe in the administrative records.The interpretations and implications of these results could only be improved with systematic assessment of qualitative and survey data from those who work and reside in Ohio prison facilities.
Third but not necessarily final, our measure of mental illness is limited.Mental illness is a critical factor and point of focus in the RH literature.Our measure of mental health was limited in that it lacked nuance and that it was based only on official designations made by ODRC.On the one hand, this limitation does not detract from our investigation of whether the actions of prison staff involving RH placements reflect the information at their disposal, including how a specific prison system defines "mental health" (see also Simes et al., 2022).On the other hand, some extant studies of mental health discussed earlier in the report have utilized more detailed mental health measures (e.g., Siennick et al., 2022).
The analysis described here not only reinforces some of the key findings of Mears et al. (2021) for solitary confinement, it extends their findings to the far more common use of short term restrictive housing.Most telling is that the disparities uncovered for ERH in both studies were largely common to SRH in Ohio while also reflecting similar magnitudes in group differences.We encourage others to pursue similar studies in other states to investigate the stability of these disparities across different prison contexts.A body of knowledge that replicates disparities in RH to the same degree regardless of context would be a compelling theme that decisionmakers cannot ignore and which could make a major contribution to critical criminological perspectives.

Table 1 .
Profiles of iPs never placed in rH during sentence versus placed in srH only versus placed in erH; all iPs admitted and released 2007-2016.
ordinary demands of life within the prison environment and which manifested by substantial pain or disability."They may have any DSM-IV diagnosis.A limitation of the mental health data is possible misclassification due to the use of an instrument

Table 2 .
Binary logistic regression models of srH and erH placements (odds ratios [e b ] and standard errors of regression estimates [s.e.b ] reported).