Abstract
The problem with the practical application of decision-making regarding release of mentally ill defenders lies in the inherently ambiguous definitions of mental illness and dangerousness, both of which are necessary for the continued commitment of insanity acquittees. In this study, we examined how clinicians make release decisions in a forensic facility, with particular attention paid to how such decision-making may have changed over time. Records were reviewed to determine documented criteria indicating readiness for release. The results indicated that compliance and treatment response were the primary reasons that the patients were judged ready for release. In addition, increasing attention to the use of substances as a risk factor was evident in the records, with substantially more documentation found in the most recent decade. Our data suggest that clinicians view three concerns to be of primary import when making release decisions: responsiveness to and compliance with the treatment, substance use, and risk of violence.
The release of mentally ill offenders is a complex process involving the balancing of two potentially opposing viewpoints: individual liberties and the protection of society. Policies regarding these decisions, because of questions related to civil liberties, are often influenced by the United States Supreme Court. Two cases in recent history exemplify such decisions. In both Jones v. United States1 and Foucha v. Louisiana2 at issue was the requirement that mental illness and dangerousness be present for continued commitment. In Foucha, the Court specified that an antisocial personality was inadequate as evidence of mental illness. (A different opinion arose from a different Court related to the confinement of sexually violent predators. In Kansas v. Hendricks, 521 U.S. 346 (1997), antisocial personality was adequate for continued commitment, although this decision may have been related to dangerousness. In Foucha, the Court, however, did not define what it would consider to be evidence of risk of dangerousness.
Although the assignment of psychiatric diagnoses has been difficult, the development of diagnostic classification schemes (e.g., the Diagnostic and Statistical Manual of Mental Disorders (Text Revision) [DSM-IV-TR]3) and structured diagnostic interviews (e.g., the Structured Clinical Interview for DSM [SCID]4) have been developed to assist in this regard. Unfortunately, the classification of risk of violence has been more elusive. An early review of the literature by Monahan5 found that estimations of future dangerousness based on unguided clinical judgment overclassified patients as high risk.5 In the past two decades, numerous instruments have been developed to assign risk more accurately. For example, Webster and colleagues6 developed a guide for clinicians in assessing risk (HCR-20). Although recent research indicates that the previously noted dismal estimates may have been overstated, researchers have consistently found that the use of structured risk assessments improves the accuracy of such decisions.7–9 Unfortunately, these structured risk assessments are rarely used in release decision-making, presumably secondary to the substantial professional time necessary to complete these assessments.10 (Of note, in Louisiana, research suggests that when clinicians are provided with structured assessments of dangerousness and risk, these factors weigh heavily in decision-making.) Monahan and colleagues11 recently have published a risk tool (the Classification of Violence Risk [COVR]) based on the results of the MacArthur study of violence risk. One advantage to the COVR is the relative lack of professional time needed. Unfortunately, the COVR was developed using a civil psychiatric sample. Studies are currently under way to extend the applicability of this instrument beyond civil patients, but no such data have been published to date.
Follow-up studies of forensic patients conditionally released into the community often show revocation rates that range from 35 to 50 percent.12–14 Much of this research indicates that relatively static factors often are more useful in predicting successful reintegration into the community than factors that may be more amenable to treatment. In a recent study,15 characteristics of conditionally released not guilty by reason of insanity (NGRI) acquittees were evaluated to assess those factors most predictive of success in the community. The results indicated that minority status, substance abuse, and prior criminal history were significantly related to conditional release revocation. In another study, the author found that a combination of several variables accurately predicted success in the community.16 One of these variables included an assessment of clinical improvement while in the hospital. Perhaps because of the complexities inherent in decision-making regarding release, although many studies describe the characteristics of NGRI acquittees ultimately released, very few contain recommendations regarding how to make such release decisions.
In one study, Callahan and Silver17 evaluated the decisions made in four states: New York, Ohio, Connecticut, and Maryland. They found that factors related to the decision to release varied among states. In Ohio and Maryland, the seriousness of the crime was the only measured variable associated with release. In New York, only demographic factors were relevant in predicting release patterns. No studied variables were relevant for release in Connecticut. In a recent study from Louisiana, a measure of violence risk was associated with release decisions, although only after clinicians considered other factors.18
Rather than examine the statistical relationship between patient characteristics and release decisions, we sought to evaluate those factors that clinicians involved in decision-making considered to be an indicator that the patient was ready to be released. This study examined those factors documented by clinicians as relevant when describing readiness for release, as measured by the amount of documentation contained in the patient records. In addition, we sought to determine if the factors clinicians deemed important changed over time, especially in light of the Supreme Court decisions noted earlier and the burgeoning literature on prediction of the risk of violence.
Methods
This research was approved by the Human Subjects Committee at Napa State Hospital (NSH), the state (of California) Committee for the Protection of Human Subjects, and the University of California-Davis (UCD) School of Medicine Institutional Review Board. A waiver of informed consent was granted.
Subjects
This study included a random sample of all persons released from Napa State Hospital (NSH) between November 13, 1974, and March 1, 2006, under the penal code commitment Not Guilty by Reason of Insanity (NGRI). A database tracking the outcome of all conditionally released patients was accessed. The sample was selected for another study designed to evaluate factors associated with outcome in the California conditional release program (CONREP). Subjects were randomly sampled from seven categories of outcome. Five groups had release revocation for one of the following reasons: dangerousness, psychiatric decompensation, substance use, noncompliance, and reoffending. The remaining two groups were graduates of CONREP (patients who were released from court jurisdiction) and those still active in the program.
NSH is an inpatient psychiatric facility with approximately 1,200 beds located in northern California. In 1997, the hospital shifted its focus of treatment from providing services to a primarily civil population to providing treatment to a primarily forensic population. Currently, 80 percent of the beds at NSH are dedicated to patients under various forensic commitments. These commitment types include incompetent to stand trial (IST), NGRI, mentally disordered offender (MDO), and a small number of other types. In 1986, California established by statute a conditional release program (CONREP) that was designed to provide community support and treatment to forensically committed patients. The selected sample was limited to NGRI patients, as CONREP is more likely to provide services to these patients (there is no statutory requirement that MDOs be released to CONREP). Because NSH shifted its focus of treatment in the 1990s, most patients sampled were from that period.
Procedure
Data were collected with a structured chart review instrument developed by two of the authors (B.E.M., D.B.). When available, the following records were reviewed: physician discharge summaries, nursing discharge summaries, and conditional release summary reports. Research assistants reviewed all records. Only those admissions occurring after the initial NGRI commitment were tracked (i.e., prior civil or IST commitments were not reviewed). The records from the patient's longest hospital stay at NSH were used to determine those factors associated with the decision to release.
The information coded included basic demographic data and clinical information, including diagnosis at the time of discharge and commitment offense. Coders categorized documentation of readiness for release into six general areas: compliance with treatment, treatment responsiveness, insight, substance-related problems, aggressive behavior, and any use of structured risk assessments. The procedure for coding release decisions included a careful review of the discharge documentation to evaluate whether any of the referenced criteria were reviewed as evidence of readiness for release. For each summary item, one point was assigned for each notation included in the summary. For example, the treatment compliance summary item was composed of four distinct criteria. For every criterion, a score was assigned: present, 1; absent, 0. If all items were noted, the individual received an average score of 1 on this summary item; if only two of the four items were noted, the individual received an average score of 0.5 on this summary item; if no notations were found regarding treatment compliance (or any other summary item), the score was 0. Thus, for each of the six summary items for each patient, the score could range from 0 to 1. Table 1 presents the total number of items for each category.
Items by Category
A variety of statistical methods were used to evaluate patterns in decision-making and changes in the patterns over time, including analyses of variance and chi-square analyses. All analyses were conducted using SPSS 15 software.
Results
Characteristics of the Offenders
Table 2 presents the demographic and clinical characteristics of the sample. As can be seen, the overall sample was primarily Caucasian (63%) and overwhelmingly male (85%). Most patients (44%) had a diagnosis of schizophrenia, with a large number having either bipolar disorder (16%) or schizoaffective disorder (19%). Eleven percent had substance misuse as the primary Axis I disorder. More than 60 percent of the sample had at least completed high school, with 87 percent never married, separated, or divorced. The patients committed a range of offenses, with 47 percent having been found NGRI for assault, and 17 percent committed for murder or manslaughter.
Demographic/Clinical Characteristics of the Sample
The majority of patients (43%) were released from NSH in the 1990s, with 24 percent released before the 1990s and 33 percent in 2000 or later. Length of stay for the studied admission varied between three months and 10 years, with the average length of stay for the studied admission of 3.8 years. The average total length of stay for all admissions (post index offense) was 10.5 years. In general, the longest admission tended to be the first NGRI admission after the index offense, with an average length of stay of 5.1 years.
Figure 1 presents the average standardized scores (z-scores) for the six areas documented by clinicians as reasons why the patient was ready for release. Scores above the mean of 0 indicate that these items were more likely to be documented in the specific decade; scores below the mean indicate that the items were less likely to be documented in the records for that particular decade. As can be seen in the figure, documentation in the records of patients most recently released was superior to the previous two decades. In this decade, treatment compliance and substance use evidenced the largest differences from previous decades, indicating that more records contained notations regarding these two areas. In all cases, documentation was more extensive in the current decade, although aggressive behavior was the most salient indicator in the first decade, with insight and substance use most prominent in the second decade. This figure depicts the increased importance of substance use over time. In decade 1, documentation of substance abuse was significantly below the mean, essentially at the mean in decade 2, and well above the mean in decade 3.
Z-score differences in indicators between decades.
Table 3 presents the unstandardized mean scores for each indicator between decades. Scores close to 1.0 indicate that most charts contained documentation of all items in the indicator, whereas scores closer to 0 suggest that few charts contained notations about the indicator. As can be seen from this table, compliance and treatment responsiveness were the factors most prominently mentioned as indicating readiness for release. Substance use increased in importance over the three studied decades, although in examining both comorbid diagnoses and documentation of substance usage, there was no indication of progressively increased substance use over the three decades for any studied substances.
Differences in Unstandardized Scores on Release Decision Documentation and Admissions Between Decades
The admission patterns differed between decades as well. In general, admissions were longer in the first decade, especially compared with decade 2. Post hoc testing indicated that number of admissions and all length-of-stay variables were different in decade 1 from those in decade 2. The only difference between decades 1 and 3 was total length of stay. Stays during decades 2 and 3 were essentially equal and shorter than those in decade 1. The length of the studied admission was longer in decade 3 than in decade 2, but it was comparable with decade 1.
Discussion
The procedures for the release of offenders with psychiatric illness vary widely in different jurisdictions. In most instances, however, the factors for making these decisions are similar and are driven by Supreme Court rulings: whether there is a mental illness and whether, because of this mental illness, the person is dangerous. The careful assessment of these two factors is necessary to protect the civil liberties of the individual. Continued involuntary commitment with no just cause is viewed as inherently wrong. In the United States, the procedure for making the decision to release varies from state to state.17 While the procedures may vary by statute (for example, Louisiana requires a review panel to make the determination of readiness), it is rare that guidance is provided regarding what information to use to assess an individual's need for continued commitment, or, by natural extension, to approve their release. One notable exception is Missouri, where the statute specifies what is to be considered in judging readiness for release.19
The literature indicates that mental health professionals using unguided clinical decision-making are no more accurate than chance in predicting the risk of future violence.5 Our data revealed interesting patterns that are relevant for deciding to release and may reflect trends in forensic facilities. In the early years, before the Foucha decision, releases were more likely to be based exclusively on mental health status, with some attention paid to current (while hospitalized) aggressive behavior. Our data suggest that minimal attention was paid to risk of future violence. In the 1990s and beyond, substance use appeared to be of more importance, and more attention was paid to assessing future risk.
The increased attention to substance use is especially noteworthy, as our data indicate no difference in actual substance use by the patients between decades. One factor that may be associated with this increased attention is the possibility that patients in recent years may be successfully using voluntary intoxication as an NGRI defense, even though it is statutorily prohibited. Although no statistics support this speculation, anecdotally it appears that it may be true, especially since California passed its three-strikes law in 1994. This law requires that an individual be sentenced to 25 years to life for the conviction of a third felony following two previous so-called strikes (convictions for specific serious or violent felonies). In evaluating the effectiveness of this law, it appears that many of the third-strikers are sentenced for drug offenses.20 In addition, in another unpublished study,21 we found that many court evaluators did not adequately assess the role that substances played in the commission of the index offense. However, in addition to this statutory change in California, there has been considerable research published in the past two decades regarding the linkage between substance use and violent behavior. For example, in the Epidemiologic Catchment Area survey, the use of substances in general and alcohol in particular increased substantially the risk of violence in patients with or without a mental illness.22,23 Similar results have been found in other studies.24,25 Thus, this increased attention to the use of substances as an indicator of readiness for release may be multiply determined. With the transition to a forensic facility occurring in the late 1990s, clinicians are becoming more sophisticated in their understanding of the role of substances in aggressive behavior, bolstered by increased evidence in the literature.
Our data suggest that only in the past decade has there been significant attention paid to both mental health and future dangerousness. In fact, not surprisingly, very few clinicians used structured assessments of either risk of violence or psychiatric symptoms in making release decisions. In only a very few instances were these assessments used and only in the past few years, even though such assessments have been widely available for 10 to 15 years (e.g., the PCL-R was developed in 1980 and revised in 1991 and again in 2003).26–28 There are multiple reasons why this pattern may have evolved. In general, although actuarial assessments have shown superiority in the statistical prediction of violence, they are labor-intensive assessments that are not routinely conducted in forensic facilities.29 In addition, these assessments have been criticized as being only relevant when looking at summary data. When applied to specific individuals, the accuracy of certain actuarial risk assessments decreases substantially.30 Finally, as our facility has become more sophisticated in managing and treating forensic patients, the use of applicable assessments has increased.
Our data suggest that compliance with psychiatric treatment also plays a significant role in the decision-making process, especially recently. This is not surprising, as individuals evidencing higher levels of compliance with less extensive psychiatric histories may be viewed as less of a risk for relapse. Research suggests that improving aftercare attendance is a critical area for reducing relapse and rehospitalization rates in individuals with chronic mental illness.31,32 In addition, overall compliance with treatment regimens has been strongly related to better outcomes for psychiatric patients.33
An interesting phenomenon observed in our data was the variability of length of stay. Not surprisingly, patients released in earlier decades had a longer total length of stay. The reason for this is fairly obvious: they have had 20 more years to be readmitted. What was most interesting was that the length of the first admission and the studied admission were both shorter in the 1990s, especially compared with the 1980s. This phenomenon may be explained by an increased sophistication over the years in treating such patients. However, this increased sophistication also appears to have an untoward effect. With the use of more structured risk assessments and a greater focus on dangerousness, length-of-stay variables appear to be increasing in the current decade.
While it could be argued that the observed changes in noted factors is an artifact of regulatory changes in documentation, NSH has been a state facility in all studied decades, and certain documentation requirements remain consistent. However, the content of, for example, discharge summaries, is probably related to those factors that the individual writer either understands thoroughly or views as important. We believe that as NSH transitioned from a civil to a forensic psychiatric facility, the clinicians' awareness of factors important for such patients increased. The increased attention in discharge summaries to substance use is a particularly strong example of this. With the rapidly accumulating and overwhelming evidence that substance use is related to aggressive behavior in both mentally ill and non-mentally ill individuals, documentation of this important concern increased. Furthermore, with the transition to a forensic facility, training regarding those factors most relevant to the treatment of forensic patients has increased, as well as case law related to the continued commitment of mentally ill offenders.
Very few studies have been published examining those factors associated with decision to release. Two studies published from data gathered in Louisiana indicate that risk assessment is a critical factor in making such decisions and that when clinicians are provided with risk assessment information, the data are factored into the decision-making process. What remains unclear is the process of making those decisions. Our data suggest that clinicians weigh two issues: remediation of mental illness and the likelihood of continuation of such remediation on an outpatient basis, and risk of dangerousness. More recently, the interrelationship with substance use appears to play a role in such decisions. In mentally ill offenders, the balancing of these factors is critical. Although studies indicate that the mentally ill are more likely to commit acts of violence, such increase is minimal.22 Substance use and abuse plays a much stronger role.24 Thus, a decision-making tool that combines the assessment of symptoms of severe Axis I psychiatric disorders and those of substance use with standard risk assessments is needed. It may be that the COVR will be useful in incorporating these constructs; however, research in forensic patients is sorely needed. Until the COVR is validated in forensic patients, our data suggest that three areas must be adequately assessed in a structured way: treatment responsiveness, substance abuse, and future risk.
There are several limitations to this study. The primary one is its retrospective design and reliance on medical records to determine the reasons clinicians decided that patients were ready to be released. The amount and quality of documentation describing such readiness varied, especially for patients who were released in the earlier decades. If a patient has been released from the facility and does not return within seven years, documents are purged. However, key documents, such as discharge summaries, are never purged. In addition, rather than increased emphasis, it may be that documentation simply has improved in the past decade, especially as the facility becomes more adept at navigating the criminal justice system and understands the requirements for continued commitment. Finally, a relatively small number of patient records were examined, although in the years examined, only 437 patients were discharged from NSH. Thus, our sample represents almost 20 percent of the population of releases.
Conclusions
The use of structured assessments to assist in clinical decision-making is not new to mental health. As resources continue to dwindle, the demand for mental health delivery systems to provide treatment will continue to increase and data-based decision-making may well become mandated. The management of forensic patients is no exception to this trend in health care utilization. The need for data-driven decisions in forensic systems tasked with making release decisions is even more critical. Our data suggest that as facilities have become more sophisticated and research has increased on how to make such decisions, more structured assessments are used, although this has been a relatively recent phenomenon. Our data indicate that it is imperative that such assessments contain factors related to treatment response and substance use.
- American Academy of Psychiatry and the Law
