Background
Absconding from psychiatric facilities is a significant health and security concern especially for forensic patients who are legally mandated to remain in a secure setting. Incidents of absconding can have considerable adverse effects on the community, hospital and patients alike. There is a large social and economic cost of absconding; for example, police are found to be involved in returning between 13% and 33% of absconders to hospital [
1]. Absconding can also have detrimental effects on the hospital and care providers, while relatives of patients and community members may experience a decreased sense of confidence in the psychiatric services being provided [
1,
2]. When considering forensic populations in particular there is often a heightened perception of risk to public safety, highlighted by recent calls for greater restrictions on those with psychiatric illnesses who are in conflict with the law (e.g., in Canada, Bill C-54 [Not Criminally Responsible Reform Act]). There are currently no empirical findings directly comparing the rate of interpersonal violence or offending among civil versus forensic psychiatric patients who have absconded from hospital.
The impact of absconding may be most acutely felt by patients themselves as these incidents may slow recovery by prolonging hospitalization and interrupting treatment efforts. Also, the literature suggests an association between self-harm behaviors and absconding. A recent review found that 25% of all suicides among inpatient psychiatric clients over a 10-year period in England and Wales took place after patients had absconded from the ward [
3], and older studies have reported comparable figures [
4‐
6]. Dickens and Campbell [
7] reported that 16% of absconding events in their sample of psychiatric inpatients in the U.K. involved serious adverse outcomes including self-harm and victimization. In contrast, the vast majority of studies report very low base rates of offending behavior and violence towards others occurring during a patient’s absconsion (e.g., for violence: 1.6% [
8]; 1.4%, [
9]), including absconders who are forensic patients (2.8% [
10]; 3.2% [
11]; 4.6% [
12]; 4.4% [
13], for all absconding events involving interpersonal violence). Within samples of individuals deemed not criminally responsible for their offenses due to mental illness, there is evidence to suggest that those with a history of absconding are more likely to be re-arrested following their hospitalization [
14,
15].
The frequency of absconding is difficult to distill from the literature, given the varying definitions and measurement of this behavior across studies. That said, most studies appear to adopt a broad definition of the behavior, commonly defining absconding as leaving the hospital without permission [
16‐
18] and including a failure to return from an authorized leave [
7,
8,
12,
19‐
21]. In their systematic review of absconding, Bowers and colleagues [
1] reported the mean rate of absconding for general psychiatry (excluding forensic services) was 12.6% of all patients ‘at risk’ (defined as the total number of inpatients at the beginning of the study period plus the number of those admitted in the course of the study), with a range of 2-44%. Studies from the secure forensic hospitals in the U.K. report lower prevalence rates of between 1-4% of all admissions [
10‐
12,
22], and adopt similar definitions of absconding (i.e., any unauthorized absence from the secure hospital, including an escape from within the perimeter walls; 10,13,22). Given the negative impact of absconding events, further investigation is needed to understand the determinants and outcomes of absconding events, and the characteristics and motives of forensic patients who abscond. This could lead to more effective management of risk, by helping to refine models of decision-making surrounding leave and privilege authorizations, and ultimately, readiness for reintegration into the community.
Methods
Setting
This study was conducted at a large urban psychiatric hospital in Toronto, Canada. The forensic program within the hospital is the largest provider of forensic mental health services in the province of Ontario, comprising 180 inpatient beds divided between four medium and four minimum secure units and serving approximately 250 outpatients living in the community. The majority of patients have been found Not Criminally Responsible on account of Mental Disorder (NCRMD) and are under the auspices of the Ontario Review Board (ORB). The ORB is responsible for annual reviews of the status of every person under its jurisdiction and making ultimate decisions regarding the least restrictive placement of the individual (i.e., continued detention, conditional or absolute discharge from hospital) vis-à-vis public safety. In Canada, both summary convictions and indictable offenses are eligible for a NCRMD defense, resulting in a broad range of offenses ultimately receiving this designation. A current snapshot of our patient population (2012-2013) revealed that the most common index offense was assault (58%), followed by uttering threats (20%) and weapons-related charges (15%). Twelve percent of patients had been charged with a sexual offense, while 13% were charged with murder (7%) or attempted murder (6%). Eight percent had only a non-violent (e.g., property) offense for which they were found NCRMD.
At the time that this study was conducted, the process of granting privileges and leave from the hospital was based on clinical team discussions, and ultimately decided by the patient’s psychiatrist and approved by the unit manager. Clinical variables, including the patient’s current mental status and their community reintegration needs/readiness were considered when deciding whether and what level of privilege was granted. The upper limit of the privilege (i.e., amount of community access permitted) is set by the patient’s ORB disposition.
Design
A case-control design was used. All current forensic patients who had at least one incident of absconding from hospital within the previous 24-months were identified (n = 57). An equally-sized control group was formed in order to conduct group-based comparisons along specified demographic and clinical variables. The control group was individually matched on age (within five years, and most [73%] within three years), sex, and security level within the hospital (residing on the same unit as their absconding counterpart, or on a unit with the same level of security). We also took care to ensure that no patients in the control sample had any history of absconding; otherwise, it is possible that our two groups would have simply differed on the timing of their absconding behavior rather than its presence or absence. With just one exception, every patient in the absconding group was matched successfully to a control patient in this manner.a Consistent with prior research, we defined absconding as any unauthorized absence from the hospital. This included breaching the security of an inpatient unit, accessing hospital grounds or the community without permission, or being absent for longer than permitted. The study was approved by the institutional ethics review board prior to the commencement of data collection.
Measures
A comprehensive coding scheme was developed to gather all relevant demographic and clinical information for each patient group. For patients who had an incident of absconding in the past 24 months, additional data pertaining to the month prior to the event (e.g., medication change or non-compliance, change in mental status, substance use, voiced ideation/intent to abscond), events transpiring during the unauthorized absence (e.g., involvement in or experience of violence, substance use), as well as characteristics of the absconsion itself (e.g., method of leave, duration, location traveled to, form of return to hospital) were recorded. All data were collected from the patient’s health record, including assessment and treatment reports, legal documents, as well as daily progress notes completed by nursing staff and other members of the clinical team. Data pertaining to absconding events in particular were collected and cross-referenced across three separate sources (daily progress notes, incident reports, and required email communications when a patient absconds). Information pertaining to patient motivations was also collected from the daily progress notes. These notes summarized the interaction that took place with the patient upon their return to the unit, including patients’ responses to being asked directly about why they absconded. Notes from biweekly meetings occurring between the patient and his or her attending psychiatrist were also accessed, as they often contained additional information about the absconding incident.
The Historical, Clinical and Risk Management-20 (HCR-20; [
32]) was used to compare absconding versus non-absconding patients, as well as to predict the occurrence of an absconding incident. The HCR-20 is a 20-item violence risk assessment scheme for use with adults who have a history of violent behavior as well as mental illness and/or personality disorder. The items appearing on the HCR-20 may be grouped thematically into historical/static risk factors, clinical/current concerns, and future-oriented/risk management variables, and are coded on a 0 (not present), 1 (possibly or partially present), and 2 (definitely present) point scale.
The Psychopathy Checklist, Revised (PCL-R; [
33]) was used for similar purposes. It is a 20-item symptom construct rating scale designed to measure the interpersonal, affective, and behavioral characteristics of psychopathic personality disorder in adults. The items appearing on the PCL-R are scored on a 0, 1, 2 scale reflecting trait presence and severity.
Data analysis
Statistical tests of difference (t-test [Mann-Whitney U for variables with non-normal distributions], χ
2) were used to compare the demographic and clinical profiles of absconding versus non-absconding patients, as well as between patients who absconded from secure/supervised settings versus non-directly supervised passes occurring on hospital grounds or in the community. We conducted Cox regression analyses to assess whether specified variables predicted the occurrence of an absconding incident, and if so, to identify the magnitude of association between each predictor variable and absconding. Predictor variables were tested in blocks of conceptually-related factors (e.g., clinical variables, risk-related variables). The Cox model is ideal as it allows for the inclusion individuals who have not yet experienced the outcome of interest (i.e., absconding) by the completion of the study. Lastly, qualitative thematic analysis of patient motives for absconding was undertaken to identify distinct profiles of absconding patients. To do this, the first three study authors (TW, SP, and SF) independently read all of the available clinical information surrounding a client’s absconsion (i.e., sources of information described in the ‘Measures’ section), including documentation of the patient’s self-reported motives. We then each rated what we judged to be the primary motivation(s) underlying the behavior; that is, the one or two variables that appeared to be functionally or causally related to their absconding. We subsequently met to discuss our ratings for each case. Based on this discussion, four distinct and non-overlapping profiles of absconding behavior were created, and each incident of absconding was assigned into one of the four groups. We were able to achieve perfect agreement at this stage; that is, for every case we agreed on the person’s group membership reflecting the primary motivation(s) driving the absconding behavior.
Discussion
Absconding from forensic psychiatric units is an issue which causes significant concern, not only in the community, but on hospital units. Despite findings in the literature, including results from the current study, to suggest that these incidents are relatively low in frequency and risk, they are perceived as significant breaches of public confidence and hospital oversight. It is therefore imperative to examine absconding events in greater depth to provide empirical evidence as to the characteristics of individuals who abscond from secure settings, as well as the predictors of and circumstances surrounding the incidents. This will facilitate accurate assessments of risk prior to granting a person leave, as well as the implementation of effective risk management interventions that address the specific circumstances and motivations of absconding events.
We employed an inclusive definition of absconding to capture all relevant dimensions of this behavior. The rate of absconding we detected (14.4%) was somewhat higher compared to prior studies, particularly those coming from secure settings, but here it is relevant to note that this study represents one of the first to examine this behavior comprehensively in a rehabilitating forensic sample, located in a hospital within a major urban center. We found that absconding events were generally of brief duration, and that the ultimate level of public endangerment posed by those who absconded was low. Specifically, no member of the public was harmed and no new criminal charges arose. The absence of adverse outcomes does not represent grounds for satisfaction, however, though it does suggest that persons being given privileges (such that a greater opportunity existed for them to abscond) did not present an imminent risk for violence to others or themselves. Beyond overt public safety risks such as when a patient reoffends in the community, there is a corrosive effect of absconding, a slowing of progress and delayed progression through rehabilitation goals, including reputational risk to the hospital.
Over half of all absconding patients returned on their own, with smaller proportions returning with the assistance of police (28%), family (7%), or hospital staff (6%). With respect to the duration of absconding events in this sample, it is comparable with what has been reported in the literature, but somewhat lengthier than the data from secure hospitals in the U.K. where many patients were reported to be apprehended within minutes of fleeing [
10]. This difference is likely reflective of the fact that the majority of absconding events occurring in most samples, including this one, arose after the patient had already been granted leave from the unit or hospital. There will be a natural time delay in apprehending a person on an unescorted pass in comparison to patients who flee while under direct staff supervision. It may also relate to the relative ease geographically to gain access to the wider community in this (urban) setting.
The most striking difference between patients who had versus had not absconded was that the former group had spent significantly more time under the auspices of the forensic mental health system. Further, within the absconding group there were often significant time lapses between the date of admission for the index hospitalization and the date of first absconsion. This is consistent with at least one study conducted in a secure setting [
11] where the average length of stay prior to a patient’s first absconsion was two months, and where the majority of events occurred within six or more months of admission. In contrast, the literature from non-forensic populations documents that absconding events tend to occur sooner, often within days or weeks of admission [
23]. Unlike admissions to acute psychiatric units, entry into the forensic system is different in that many patients anticipate hospitalization that can last several years. As individuals come to understand and experience the often lengthy constraints on their liberty, even in the face of self-perceived improvements in their mental state, absconding events may become more frequent.
Consistent with the extant literature across civil and forensic samples, patients in this study who absconded were more likely to have a history of problematic substance use, as well as a history of attempted absconding (although this latter variable did not emerge as a unique predictor in the context of regression analysis). Indeed, the finding that one-third of all absconding events in this study involved the use of substances while absent from hospital, in violation of conditions specified in patients’ review board dispositions, suggests that this is a salient variable in the clinical picture of these patients.
Patients who absconded were also estimated to be at higher risk for future violence, as indicated on a structured professional risk assessment measure completed by the individual’s treatment team. Use of structured professional judgment tools such as the HCR-20, while designed to inform judgments of risk for future violence, may also be of use in assessing individuals who are at risk for absconding. This is clinically intuitive, given that absconding events often appear to reflect general noncompliance, impulsivity, active psychiatric symptoms, or an antisocial orientation, which are themselves risk variables identified in the HCR-20. Further, the HCR-20 has been shown to be useful in predicting and managing risk for a wide array of adverse outcomes (e.g., non-violent reoffending, hospital readmission; [
35,
36]), suggesting that it is not limited to interpersonal violence. That said, it is also the case that a tool such as the HCR-20 may be of limited utility in making predictions about violent behavior occurring during an absconding event. This is largely due to the fact that the base rate of violence during an absconsion tends to be extremely low, and was effectively nil in the current study. On the other hand, risk management strategies based on the HCR-20, designed to prevent and reduce the likelihood of harm, would conceivably be relevant across authorized and unauthorized community access scenarios.
We found that the majority of absconding events were characterized by expressed feelings of boredom and frustration. In their interviews with 52 patients residing on acute psychiatric admission wards in London, Bowers and colleagues [
8] found boredom to play a salient role in patients’ decisions to abscond, alongside other variables such as feeling frightened and confined, or needing to complete tasks and household responsibilities. Interestingly, and in contrast to prior work in non-forensic samples [
30], we did not find that fear or safety concerns played a significant role in patients’ decisions to abscond. It is likely that being in the forensic system confers particularly strong motivations related to feelings of frustration and despair when faced with a lengthy hospitalization and the curtailment of liberties in the community [
11]. This may be further compounded by the fact that expressed feelings of frustration and boredom are challenging clinical issues for treatment teams to address. Nevertheless, the current findings suggest that larger systemic and environmental issues could be examined so as to create a consistent, transparent and respectful milieu which can contribute to a sense of legitimacy the individual attributes to their detention.
The next most common motivating influence we found in this sample pertained to absconders’ psychiatric symptoms. Individuals in this group appeared to act in response to auditory hallucinations or delusional beliefs, sometimes pursuing ostensibly goal-directed behavior directly linked to active psychotic symptoms. This aligns partially with the motivations reported by some of the patients in the Bowers et al. [
8] study, but differs in that patients reporting psychotic motivations in that study concurrently expressed non illness-related reasons for absconding. Our finding also contrasts with the study by Dolan and Snowden [
11] in which only one individual expressed a psychotic motivation for their behavior. In the current study, there appeared to be clear proximal risk indicators for this group: in the month prior these patients were experiencing active symptoms of their illness with notable mental status instability, medication changes, and/or missed medication. These variables indicate important changes in patient’s level of clinical stability, and may also reflect deteriorations in the therapeutic relationship between the patient and the treatment team. These variables thus appear relevant to the assessment of risk for absconding, particularly among this subgroup of patients, and could well serve as choice points for intervention prior to an individual being granted privileges to leave the unit.
In the “goal-directed” profile, individuals were likely to have voiced their desires to complete a specific directed activity, but did not have the approval to do so at the time. During the absconsion, most of these patients completed their self-identified goal and then returned to the unit without difficulty. This group appears similar to Bowers et al.’s [
8] patients who absconded due to needing to complete everyday household tasks and chores. A review of the documentation surrounding these events did not always provide the reasons for the refusal of passes so that the individual could accomplish their goal. It is also notable that in approximately one-third of incidents, interspersed across the motivational groups, the patient openly expressed thoughts or intentions to abscond from hospital. These statements require further examination as proximal risk indicators of absconding. Here, conducting interviews with clinical and front-line staff may be particularly informative in determining how or why breakdowns in communication occur between staff and patients, and interventions that can be implemented to improve this.
At present, there exists no structured decision-making tool to assess a patient’s risk for absconding. Hilterman, Philipse, and Graaf [
37] published a tool assessing the risk of violence on discharge or unauthorized leave in the community, but not of the risk of absconding itself. This is problematic, particularly in light of the finding that forensic hospitals show considerable variation with respect to the criteria they rely on to make decisions about a patient’s readiness for community access [
38]. Bowers, Alexander, and Gaskell [
39] suggested identifying patients at higher risk for absconding based on the presence of specific variables shown to be predictive of absconding in the research base. Then, intensified resources such as one-to-one nursing time and increased family visits are encouraged to promote the open discussion of worries or concerns that might be further contributing to the person’s risk of absconding. One of the major benefits of this study is the identification of four broad motivations for absconding, which can further facilitate efforts to target those variables that are associated with absconding risk (e.g., mental status and medication effectiveness in the symptomatic group) in clinical decision-making and care planning.
There are limitations to the current study. First, our measure of absconding rate was a coarse one; although it is consistent with prior studies and recommended as one of the better options available [
1,
34], we were unable to supplement it with more fine-grained measures (e.g., number of absconsions on unsupervised day passes divided by the total number of day passes granted) due to the unavailability of this type of data. Second, information which formed the basis of our analyses were taken from the electronic health record, and supported by other legal documentation. We did not specifically interview patients about why they absconded. Therefore, the available information was that which staff documented; interviews with absconders and front line staff may have assisted in providing more detailed information surrounding the motivational aspect of the behavior. Further, this was a retrospective study, and we did not examine the impact of absconding prospectively (e.g., in terms of length of stay in the system, clinical progress, or experiencing other adverse outcomes such as violence or victimization in the community). Additional longitudinal analyses may be able to detect other trends in absconding behavior and identify the impact of changing political or environmental issues on this behavior over time. Importantly, our sample size was modest, and became smaller when conducting analyses based on motivation. Consequently, power to detect effects, particularly those small in size, was also modest, and underscores the need to replicate the current findings in larger samples of forensic patients.
Given that so few studies have examined absconding events from forensic settings, further research is needed to replicate the significance and clinical utility of the variables identified in this study that were associated with a heightened likelihood of absconding. Additionally, replication of the motivational subtypes using interview-based techniques is needed. Further examination of interventions which address the specific variables known to influence absconding behavior is also essential; as noted by Bowers and colleagues [
1], “there are no thoroughly convincing, well designed, rigorously carried-out trials of interventions to reduce absconding” (p. 350). The current study can provide the basis for an intervention study tailored to the clinical profiles and reported motivations of those who abscond from forensic rehabilitation settings. In fact, as a consequence of this work, we made substantial changes to our program’s policies outlining the granting of privileges and we will report the outcome of this policy implementation in a separate paper.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
TW and SP conceptualized the aims and design of the study. SF carried out the data collection, while TW, SP and SF contributed to the data coding. SP performed the statistical analyses. All authors contributed to the interpretation of the data and the study write-up. TW, SP, and SF drafted the manuscript, and AIFS revised it critically for intellectual content. All authors read and approved the final manuscript.