Setting
We collected data from all of the Norwegian adolescent acute psychiatric inpatient units that are approved for involuntary admissions (N = 16), which included a total of 126 beds (mean 7.4, SD 2.9, range 2–14). These units provide inpatient mental health care mainly for adolescents aged 13–17 years, but they accept admission of younger adolescents if needed. Some adolescents are more than 17 years old at the time of discharge. During the study period, all of the units accepted around-the-clock emergency admissions. The units are distributed throughout Norway, and each unit has a uniquely defined catchment area. As a rule, drug-addicted adolescents are cared for by the child protection service. Three of the 16 acute psychiatric inpatient units were locked when needed, and the other 13 were permanently locked or had one permanently locked ward.
Data collection
We collected data on all of the inpatients in the included units who experienced restraint from January 1, 2008 through December 31, 2010. We collected the data retrospectively during a nine-month period from August 2011 to May 2012. The first author visited all of the institutions and collected data on restraint episodes, demographic characteristics, and clinical variables. Information about restraint episodes was collected from routinely used handwritten restraint protocols. Other data were collected from the electronic patient records. The total number of admitted adolescent patients during the study period was retrieved from the electronic patient administrative system at each unit.
Definitions of restraint in the Norwegian Mental Health Care Act
The Norwegian Mental Health Care Act regulates the practice of restraint procedures in Norway [
22]. Staff members must consider less restrictive interventions first, and they cannot use restraint as a treatment. The following types of restraint may be used: (a)
mechanical restraints, which inhibit the patient’s freedom of movement, including belts and straps and clothing specially designed to prevent injury; (b)
seclusion, which refers to detention for a short period of time behind a locked or closed door without a staff member present; (c)
pharmacological restraint, which refers to single doses of medicines that have a short-term effect and are used to calm or sedate a patient; and (d)
physical holding, which refers to any technique in which staff members physically restrain a patient without using tools. Mechanical restraints and locked seclusion are not allowed for patients under the age of 16. Restraint can be used during either voluntary or compulsory admissions. All psychiatric institutions in Norway are obligated by law to have a restraint protocol in which each restraint episode is registered. The protocol describes the type and duration of the restraint and the reason for its use. Independent and authorized control commissions regularly checks all registrations in these protocols.
In this study, we did not include episodes of restraint that were needed for compulsory feeding in cases of severe anorexia (1896 restraint episodes distributed across 21 patients). The Norwegian Mental Health Act also regulates the use of compulsory feeding for patients with anorexia. These episodes are often included in the restraint protocols because wards may use mechanical restraints or physical holding to conduct forced feeding. However, whether or not these episodes are included in the protocols varies between wards.
Data and variables
The dependent variable was the number of restraint episodes for each patient from all the admissions during the whole study period. The number of restraint episodes was categorized as 1, 2–4, 5–9, and ≥10. For adolescent patients with more than one admission in the three-year period, we collected data on the patient’s social and mental health characteristics from the most recent admission.
Social characteristics. We defined immigrant background as having two foreign-born parents, and coded this Yes or No. The variable living arrangement at the time of the most recent admission was coded in four categories: living with both parents (biological or adoptive), living with one parent (with or without stepparent), living in foster care or institution, and other. The variable current involvement with the child protection service was coded Yes or No.
Mental health characteristics. The local clinical teams assessed the adolescent patients’ conditions and coded their
main psychiatric disorder using the Axis One (clinical psychiatric syndromes) in the multiaxial ICD-10 classification of child and adolescent psychiatric disorders from the World Health Organization [
23]. Using this information, we grouped the adolescent patients into one of five categories based on their most recent admission (the ICD-10 codes are in parentheses): (1)
No Axis One disorder; (2)
psychotic (F20–29)
or pervasive developmental disorder (F84); (3)
manic episode or bipolar affective disorder (F30, F31.0–F31.2, F31.6–31.9); (4)
internalizing disorder (depression F31.3, F32–33; anxiety F40–41, F93, F94; OCD F42; stress related F43; dissociative F44); (5)
externalizing disorders (substance use F10, F12, F19; personality F60, F69; hyperkinetic F90; conduct F91–92; tics F95).
Global psychosocial functioning was routinely rated by the clinicians at admission using the Children’s Global Assessment Scale (CGAS) [
24]. We used the CGAS score from each patient’s most recent admission. The CGAS measures general functioning, with scores ranging from 1 (needs constant supervision) to 100 (superior functioning). We divided CGAS scores into three groups (tertiles): 1–34, 35–44, and 45–75. We did not measure the interrater reliability of the CGAS for this study. However, the interrater reliability of the CGAS in routine use was found to be moderate (intraclass correlation coefficient, .61) in a large study of clinicians in Norwegian outpatient child and adolescent mental health services [
25].
Treatment characteristics. We divided the number of admissions in the study period into three groups (tertiles): 1, 2–3, and ≥4 admissions. We defined the length of admission as the number of days for the most recent admission and we divided this into three groups (tertiles): 1–4, 5–21, and ≥22 days. We coded involuntary admission as Yes if the patient was involuntarily admitted during the study period. We defined concomitant use of restraint as the use of pharmacological restraint in combination with any of the other types of restraint, and it was coded Yes when it occurred.
We developed a risk index score using the patient characteristic variables that were significantly associated with the number of restraint episodes (as indicated by the multivariate regression analysis). The categories for the variables were scored as 0, 1, or 2 (depending upon the number of possible response categories), with higher scores representing a stronger positive association with the number of restraint episodes. These scores were summed to make the risk index score. Because of the retrospective design of this study, and the fact that some of the variables required the completion of inpatient care, the prospective use of this risk index score at the patient level is limited. However, it may be useful to compare groups of adolescents admitted to inpatient care.
Statistical analysis
Descriptive statistics are presented as frequencies and percentages. Zero-truncated Poisson regression analysis was applied to analyze the impact of gender, social, mental health, and treatment characteristics on the number of restraint episodes. We did not include
age in our regression analyses; because each patient’s date of birth and exact age at the date of admission were unknown (only the age attained during the calendar year was available). In addition, adolescents must be at least 16 years old to be involuntarily admitted and to be restrained by mechanical means or seclusion. We performed univariate analyses for the independent variables: gender, immigrant background, living arrangement, current involvement with the child protection service, main psychiatric disorder, global psychosocial functioning (CGAS score), number of admissions in the study period, length of admission, involuntary admission, and concomitant use of restraint. Variables with
p < .20 in the analysis were selected for inclusion in the multivariate model. Variables that were not statistically significant (
p ≥ .05) in the multivariate analysis were deleted (largest
p values first) until all of the remaining variables were significantly associated with the outcome. We used robust standard errors for the parameter estimates, as recommended by Cameron and Trivedi [
26]. The effects are presented as incidence rate ratios (IRR) with 95% confidence intervals and p values. We tested the final model for multicollinearity by calculating the variance inflation factors for each of the independent variables. Estimated model fit is presented as pseudo R
2 (explained variance). Because of the large number of missing values in CGAS (70 adolescents had no information), we reran the analyses omitting CGAS from the model to check for selection bias. We used the Goodman and Kruskal’s rank correlation statistics to measure the strength of association between the risk index score and the observed number of restraint episodes.
A p value <.05 was regarded as statistically significant. All statistical analyses were conducted using SPSS (version 18.0) and Stata [
27,
28].