Concerns have grown in recent years, from ethical, medicolegal, and clinical points of view, about the use of coercive interventions (1–6). Patients often experience these interventions as inhumane and humiliating, and such interventions can have physical and mental adverse effects and in some cases can be fatal (1,3,4).
A substantial number of patients who use mental health services are admitted to acute psychiatric wards, the setting in which restraint is most frequently used. For some patients, the inpatient admission is their first experience of mental health services. The restraint rate in acute psychiatric wards is reported to range from .1% to 31% (7–12). Despite the frequent use of restraint in this setting, few studies have examined differences between inpatients for whom this intervention is deemed necessary and those who do not experience restraint (13).
A literature search of the Ovid MEDLINE database from 1950 until March 2010 that combined the search terms “psychiatry,” “mental health,” and “mental disorder” with “coercion,” “restraint,” and “seclusion” and with “psychiatric hospital” and “inpatients” yielded 358 articles. Only two of these studies compared adult patients in an acute psychiatric ward who were and were not restrained. The first study, from a psychiatric emergency service in 1979, did not note whether the sample was representative of the catchment area; in addition, data for 164 of 687 patients were lost (14). The study also did not specify the types of restraint used. The second study was a two-year investigation (1996–1997) by our research group (10). It included all restrained patients from an acute psychiatric ward that served a single catchment area. Unfortunately, neither study examined such variables as length of stay, number of admissions, and voluntary or involuntary admission status, and both were limited to acute psychiatric wards.
To overcome these limitations, we designed a study that covered several acute psychiatric wards, with representative samples of patients who were and were not restrained. We included a wider spectrum of variables than previous studies, and the data were collected over two years.
The first objective of this study was to examine differences in several characteristics between patients who were and were not restrained: age, gender, immigrant background, and residence in the catchment area of the acute psychiatric ward; use of acute psychiatric services (number and duration of admissions); psychiatric diagnoses; and voluntary or involuntary admission status. The second objective was to investigate whether any of these variables predicted the use of restraint.
Design, setting, and sample
The study was approved by the Norwegian Directorate of Health, the Regional Ethical Committee, and the Data Inspectorate. The Norwegian Directorate of Health waived the need for informed consent.
The two-year retrospective study used a case-control design. Acute psychiatric wards in three Norwegian hospitals constituted the study setting. The treatment system for all patients was based on catchment area and was publicly funded. One of the acute wards had catchment areas in the capital city (Oslo) only, whereas the two others covered both urban and rural areas. The catchment areas had a total of 570,000 inhabitants. During the study period 3,365 patients were admitted to the three wards.
The sample included all 375 patients who were secluded or otherwise restrained at least once during a two-year period—January 1, 2004, through December 31, 2005. These patients were compared with a control group of patients who did not experience seclusion or restraint (N=374). The control group consisted of three subgroups, one from each of the three wards. The patients were randomly selected from all who had been admitted during the same two-year period. If the patient selected happened to belong to the group of restrained patients, the next patient on the list who had not experienced restraint was chosen.
This study was based on routinely collected data from patients' files and from handwritten restraint protocols. Norwegian psychiatric institutions are required to document each episode of seclusion and restraint and to describe the type, rationale, and duration. The following data were collected from electronic patient files: age, gender, admission date, length of stay, voluntary or involuntary admission status, ICD-10 diagnosis (15), residence in the catchment area (yes or no), number of admissions during the study period, and immigrant background (both parents were born in another country). Some patients were admitted more than once over the two years. For the group of restrained patients, data were from the first admission when restraint was used. For the control group, data were from the first admission of each patient during the study period. The data from patients' files and restraint protocols were merged into a common data form and transferred to SPSS, version 15.0, for statistical analyses.
Voluntary or involuntary status
According to the Norwegian Mental Health Care Act of 1999 (16), an individual can be referred to inpatient mental health care either voluntarily or involuntarily. Involuntary admission is either for observation (when there is doubt about the presence of a severe mental disorder) or for long-term detention (in the case of severe mental illness). In addition, the patient must be a danger to self or others or be in need of treatment. For patients referred involuntarily to a psychiatric ward, the institution must ensure during the first 24 hours that a psychiatrist or clinical psychologist affirms the legal basis for the admission. An involuntarily admitted patient can be retained either on observational status (up to ten days) or under long-term detention. A patient who has been voluntarily referred and admitted cannot be transferred to observational status or to involuntary status unless a force majeure exists. The patient then must be deemed to be in immediate and serious danger to self or others.
Regulations and types of restraint
The Norwegian Mental Health Care Act (16) defines and regulates use of the following “restraint procedures”: pharmacological restraint, mechanical restraint, and seclusion. They may be used regardless of the patient's legal status, but they are not allowed to be part of the treatment plan, and less restrictive interventions must first have proven unsuccessful. Pharmacological restraint refers to single doses of medications with an antipsychotic or sedative effect that are given by injection or taken orally. Mechanical restraint refers to different types of belts (for restraint in bed or used outside bed for arms and feet only). Seclusion refers to detention for a short period (up to two hours) behind a locked or closed door without a staff member present.
Statistical analyses and categories
Length of stay was merged into three categories: four days or fewer, five to 15 days, and 16 to 279 days. Number of admissions was merged into three categories: one, two, and three or more admissions. The three diagnostic categories were substance use disorders and psychosis, comprising psychoactive substance use (F.10–19), schizophrenia, schizotypal and delusional disorders (F.20–29), and manic or bipolar disorder (F.30–31); nonpsychotic disorders, comprising nonpsychotic mood disorders (F.32–39) and neurotic, stress-related, and somatoform disorders (F.40–49); and disorders of adult personality and behavior (F.60–69) and “other” disorders (organic mental disorders [F.00–09], behavioral syndromes [F.50–59], intellectual disability [F.70–79], disorders of psychological development [F.80–89], disorders of childhood and adolescence [F.90-98], and unspecified). Patients' admission status was dichotomized as voluntary or involuntary.
For comparisons of patients who were and were not restrained, we used cross-tabulation with chi square tests (Pearson and linear by linear). To analyze the impact of age, gender, immigrant background, residence in the catchment area, admission status, length of stay, diagnosis, and number of admissions on the probability of being restrained, we applied a two-step binary logistic regression analysis in which being restrained at least once during the study period was the dependent variable. The first step was a separate analysis for each independent variable (unadjusted models). The second step was a binary logistic regression analysis that adjusted for age, gender, immigrant background, residence in the catchment area, and the clinical variables. Correlation tests between all variables were conducted to control for possible collinearity between the independent variables. No significant interactions were found, and all the correlations were below .70 and above −.70. Analyses used SPSS, version 15.0, and the significance level was set at .05.
As shown in Table 1, compared with the group of nonrestrained patients, the group of restrained patients included significantly larger percentages of men, of patients with an immigrant background, and of patients who resided outside the catchment areas of the acute psychiatric wards. In addition, the restrained group had a larger proportion of patients with a primary diagnosis of a substance use disorder or psychosis and a smaller proportion of patients with a nonpsychotic disorder. The restrained patients also had significantly longer hospitalizations (median 13 days), more admissions (median five days), and fewer voluntary admissions. None of the patients in the nonrestrained group had more than six admissions, whereas 30 of the restrained patients (8%) had. The number of patients whose legal status changed from involuntary to voluntary did not differ significantly between the groups.
Unadjusted logistic regression analysis
Table 2 summarizes the results of unadjusted logistic regression models. Being restrained at least once during the study period was the dependent variable. Compared with patients who had a primary diagnosis of a nonpsychotic disorder, patients in the group with a primary diagnosis of disorder of adult personality and behavior and “other” diagnoses had more than five times the odds of experiencing restraint (OR=5.3, p<.001). Patients with a substance use disorder or psychosis had nearly eight times the odds of experiencing restraint as those with a nonpsychotic disorder (OR=7.7, p<.001). Compared with patients who had the shortest hospital stays (zero to four days), those who had the longest stays (16 to 279 days) had four times higher odds of experiencing restraint (OR=4.0, p<.001). Compared with patients who had one admission during the study period, those with two admissions had nearly three times higher odds of being restrained (OR=2.7, p<.001) and those with three to 23 admissions had five times higher odds (OR=5.5, p<.001). Compared with patients whose referral was voluntary, those with an involuntary referral had nearly ten times the odds of being restrained (OR=10, p<.001).
Adjusted logistic regression analyses
In an analysis that used the same dependent variable (being restrained at least once during the study period), we adjusted for age, gender, immigrant background, and residence in the catchment area, as well as for all clinical variables—primary diagnosis, length of stay, number of admissions, and voluntary or involuntary status. As shown in Table 2, the four clinical variables were found to be independent predictors of the use of restraint. Compared with patients with a nonpsychotic disorder, patients in the group with a primary diagnosis of disorder of adult personality and behavior and “other” diagnoses had more than three times higher odds of being restrained (OR=3.5, p<.001) and patients with a substance use disorder or psychosis had nearly three times higher odds (OR=2.8, p<.001). Patients with the longest hospital stays (16 to 279 days) had nearly three times higher odds than those with the shortest stays (zero to four days) of experiencing restraint (OR=2.9, p<.001). Compared with patients who had one admission, those with two admissions had nearly three times higher odds of being restrained (OR=2.8, p<.001), and patients with three to 23 admissions had nearly six times higher odds (OR=5.9, p<.001). Compared with patients whose referral was involuntary, those with a voluntary referral had significantly lower odds of being restrained (OR=.11, p<.001).
In line with two previous studies (9,14), our study found that inpatients who experienced restraint were younger and more likely to be male and to have an immigrant background and a psychotic disorder. However, in this study, age, gender, and immigrant background lost their predictive power when the analysis controlled for four clinical variables that the two previous studies did not take into account: long hospital stay, multiple admissions, involuntary referral, and diagnosis of a personality disorder, psychosis, or a substance use disorder. High odds ratios indicated strong independent predictive power for each of these variables.
The importance of involuntary legal status and a long hospital stay is in line with the results of a large study by Way and Banks (17) of 23 public psychiatric state hospitals for adults in the United States. These authors examined data for all patients admitted during one month in 1984 and compared restrained patients (N=657) and patients who were not restrained (N=22,939). They noted several characteristics of restrained patients: age less than 26 years, stays from 30 to 365 days, involuntary legal status, female gender, and a diagnosis of mental retardation. Their findings are not directly comparable with ours because their study included patients from all types of hospital units, whereas we studied only patients from acute psychiatric wards. Furthermore, it is not quite clear how they defined restraint. Despite these differences, both studies concluded that restrained patients had longer hospital stays than those who were not restrained and were also younger and were more often involuntarily admitted. These similarities are interesting because the studies were conducted at different times and in different cultures, with different health care and legislative systems.
How should our results be interpreted as predictors of restraint? The diagnoses reflect severity of illness and, indirectly, the risk of violence. Several studies have demonstrated that violence risk is higher among patients with severe personality disorder or psychosis (18), especially when these patients also have drug use disorders (19). Involuntary admission is also related to illness severity and especially to lack of insight. Obviously, patients admitted against their own will are more prone to have conflicts with staff members. Multiple admissions and admissions of long duration may also reflect deficiencies in insight and lack of a treatment alliance. Kreyenbuhl and colleagues (20) have underlined the risk of treatment dropout among patients who have had negative experiences with the mental health system. The large number of admissions among restrained patients may also be attributable to problems with receipt of aftercare. Persons with severe mental illness may be rehospitalized when community support is lacking or the treatment system's response to their needs is insufficient (21).
In any case, use of restraint must be regarded as an indication of mutual distrust that may be hard to overcome. In a 2007 review article, Gaskin and colleagues (22) focused on the need to improve several systemic factors, such as leadership and staff education, and they recommended establishing emergency response teams and monitoring seclusion episodes. They also pointed to individual-related factors, such as changing the therapeutic environment and facility and treating patients as active participants. Identifying patients at risk of restraint may allow clinicians to approach these patients at an early stage and listen to their accounts of previous treatment experiences. Staff could then employ strategies to improve trust and reduce the risk of use of restraint. Such strategies might include development of individual treatment models for these patients (23), with a focus on the patient's involvement (24,25). Staff could also work together with the patient to identify “triggers” (26,27) and early warning signs (28) of potentially violent behavior. Such trauma-informed care could reduce the need for restraining these patients and provide a safer environment and fewer traumas for both patients and staff (29).
Our study has several methodological strengths. To our knowledge, among studies in this area, it has the largest sample of patients from acute psychiatric wards. In addition, the sample is representative of acutely hospitalized patients because the wards had total responsibility for their catchment areas—that is, there were no other options for psychiatric hospitalization in the catchment areas. A further strength is that the policy of mandatory registration of all restraint episodes makes it likely that we included all episodes of use in our study.
Our study also has some limitations. The primary diagnoses that we used were given to patients by hospital clinicians, and we were not able to test their reliability. However, with the large sample and the fairly crude categorization used, it seems unlikely that this limitation seriously influenced the results. Another limitation is that all three types of restraint procedures were analyzed together, which makes it difficult to generalize the findings to a specific type of restraint. Furthermore, our data registration was limited to a two-year period, and we thus lacked information about patients' prior admissions and restraint experiences during these admissions.
In this study, use of restraint was predicted by a long hospital stay, multiple admissions, involuntary referral, and a diagnosis of a personality disorder, psychosis, or a substance use disorder, even when we controlled for confounding variables such as age, gender, immigrant background, and residence in the catchment area. The findings indicate that some patient characteristics that predict use of restraint might be stable over time and across countries. Identifying patients at risk of restraint may enable clinicians to intervene early, listen to patients' accounts of previous treatment experiences, and employ strategies to improve trust and reduce risk of use of restraint.
The authors thank Nihal Perera, B.Sc., for helping to develop the database for the restraint data.
The authors report no competing interests.