Comparing the effect of non-medical mechanical restraint preventive factors between psychiatric units in Denmark and Norway.

BACKGROUND
The use of mechanical restraint (MR) is controversial, and large differences regarding the use of MR are often found among countries. In an earlier study, we observed that MR was used twice as frequently in Denmark than Norway.


AIMS
To examine how presumed MR preventive factors of non-medical origin may explain the differing number of MR episodes between Denmark and Norway.


METHODS
This study is a cross-sectional survey of psychiatric units. Linear regression was used to assess the confounding effects of the MR preventive factors, i.e. whether a difference in the impact of these factors is evident between Denmark and Norway.


RESULTS
Six MR preventive factors confounded [∆exp(B)> 10%] the difference in MR use between Denmark and Norway, including staff education (- 51%), substitute staff (- 17%), acceptable work environment (- 15%), separation of acutely disturbed patients (13%), patient-staff ratio (- 11%), and the identification of the patient's crisis triggers (- 10%).


CONCLUSIONS
These six MR preventive factors might partially explain the difference in the frequency of MR episodes observed in the two countries, i.e. higher numbers in Denmark than Norway. One MR preventive factor was not supported by earlier research, the identification of the patient's crisis triggers; therefore, more research on the mechanisms involved is needed.


CLINICAL IMPLICATIONS
None of the six MR preventive factors presents any adverse effects; therefore, units in Denmark and Norway may consider investigating the effect of implementing, the identification of the patient's crisis triggers, an increased number of staff per patient, increased staff education, a better work environment and reduced use of substitute staff in practice.

I n Denmark and Norway, mechanical restraint (MR) can be legally performed when psychiatric inpatients pose a risk to themselves or others (1, 2). The procedure is primarily performed by nursing staff, and involves the use of leather or fi bre belts to secure the patient to a bed. MR exposes patients to potential physical risks (3 -7) and has negative psychological effects for many patients (8 -12).
A Danish review (13) identifi ed numerous MR preventive factors based on studies that target MR prevention. No MR preventive factors found in the review received the highest scientifi c recommendation grade. 1 Furthermore, several of these MR preventive factors might be effective only in controlled research environments.
In an earlier retrospective association study (14), we found three MR preventive factors associated with low numbers of MR episodes in Denmark and Norway (mandatory review, patient involvement and no crowding). In that study, we eliminated the differences between the countries by making adjustments to make the fi ndings more universal. However, focusing on the differences between the countries is important when the aim is to develop effective MR reduction programs in Denmark and Norway.
In a comparative study on the use of coercion in 11 European countries, including Denmark and Norway

Included participants
The study included all psychiatric hospital units in Denmark ( n ϭ 87) and Norway ( n ϭ 96) that treated adult inpatients (18 -65 years) and used MR for approximately 2 years prior to the year of investigation. MR did occur in Norway in settings other than hospitals; however, given that only 1.5% of recorded MR episodes in 2007 occurred in these settings (16) and that hospitals served as the primary focus of this study, these units were not included.

Excluded participants
The special high-security department in Denmark was excluded due to specifi c legislation concerning seclusion.

Excluded outcome data
If MR was used more than 25 times on the same patient during the investigation period, then the episodes of MR Ͼ 25 were excluded. Chance determined the units to which these patients were admitted, but the presence of these patients could signifi cantly affect the units ' MR frequencies, distorting the results.

Data collection: the questionnaire
A questionnaire was developed based on the results from an earlier systematic literature review that focused on factors that might reduce the use of MR (13).
Each MR preventive factor was " translated " into questions to identify the degree to which the factor was present in the units.
Six factors were not included in the present study. Combined intervention programmes, were omitted because it were too complex to measure. The following four factors were omitted because they were not considered measured appropriately by the questionnaire: treatment of substance abuse and psychosis, patients perceived as a resource in their own treatment, pro necessitate [p.n.]/as-needed medication guided by individual patients, and care alliance. Lastly, music used as a diversion and for relaxation were omitted, because it were partially covered by another factor, activity-based nursing.
Cognitive milieu therapy and patient involvement and patient-centred care and patient involvement, were sepa-(15), Norway had the lowest rate of physical coercion (MR, seclusion and physical restraint/holding) with 29 inhabitants per 100,000 experiencing physical coercion. In comparison, Denmark had 122 episodes of MR per 100,000 inhabitants, and Norway had 87 episodes of MR per 100,000 inhabitants in 2007 (16, 17). The same tendency was observed in the subpopulation we investigated earlier (14); Denmark had 72 MR episodes per 100,000 inhabitants, and Norway had 42 MR episodes per 100,000 inhabitants. The lower frequency of MR episodes in Norway might indicate that more effective MR preventive factors have been implemented in Norway compared with Denmark or that MR preventive factors exhibit differing effects between the countries.

Aim
To examine how presumed MR preventive factors of non-medical origin may explain the differing number of MR episodes between Denmark and Norway.

Material and methods
A comparative study was conducted to investigate the degree to which the non-medical MR preventive factors identifi ed in a previous review (13) were implemented in the psychiatric units and whether the chosen MR preventive factors association with the number of MR episodes differed between Denmark and Norway.
The following conditions must be present legally to initiate MR according to the Danish Mental Health Act (translated by the authors) (1): Mechanical restraint may be used only when necessary to prevent a patient from 1) exposing their body or health or the body or health of others to danger, 2) pursuing or in any other way grossly molesting fellow patients, or 3) committing signifi cant acts of vandalism.
These conditions are consistent with Norwegian legislation (translated by the authors) (2): Restraints may be used on a patient only when absolutely necessary to prevent him or her from damaging him or herself or others or to prevent damage to buildings, clothing, inventory or other things. This paper uses Bowers et al. ' s (18) defi nition of MR: The use of restraining straps, belts or other equipment to restrict movement . Note that this defi nition refers only to the restraint of inpatients in psychiatric hospitals.
The MR preventive factors identifi ed by the literature review (13) were included in an online questionnaire and sent to the clinical nurse managers for the purpose of quantifying whether they were implemented in the units. These managers were chosen because they had exclusive Later, the outcome variable (i.e. the number of MR episodes) was re-examined. This procedure was necessary due to the confusing wording of a heading in the Danish questionnaire and the lack of precise numbers regarding the Norwegian patients who experienced MR more than 25 times (14 answer categories instead of the actual number were provided, which made the exclusion imprecise). The re-examination was accomplished by contacting all of the clinical nurse managers in Denmark and four of the clinical nurse managers in Norway. The respondent rate was 94%, and the results lowered the overall frequency of MR by 3.2%. A high correlation coeffi cient ( t Ken,b ϭ 0.89) was found between the two answers, revealing that the confusing headline and the categorization error did not have an important effect.
A description of how the 22 dichotomized MR preventive factors were measured in the Supplementary Appendix to be found online at http://informa healthcare. com/doi/abs/10.3109/08039488.2014.996600.

Background variables
Several conditions might have affected the outcome. These conditions included unit size, bed occupancy, type of bed, reason for the MR, forensic provision, ethnicity other than Danish/Norwegian and type of unit. A description of the background variables included in the study is presented in Table 1.
Other variables, e.g. the unit ' s population base, the country ' s inhabitants and patient fl ow, could also affect the outcome; however, these variables were covered by the previously mentioned covariates. Unit type and unit size acted as proxy covariates [e.g. patient fl ow and population base were included in unit type; inhabitants were included in unit size, where the number of beds per 100,000 inhabitants in each country was very similar (Denmark ϭ 23.96 and Norway ϭ 22.81 beds per 100,000 inhabitants)].
A patient ' s gender is most likely not associated with MR (24, 25); thus, we did not use gender as a covariate.

Data analyses
The data analyses were conducted in fi ve steps. First, we analysed the difference in the frequency of MR episodes between the countries. Second, we analysed how the factors were individually associated with the number of MR episodes in Denmark and Norway separately. Third, we determined whether an interaction between the MR preventive factors and the country (second order interaction, SOI) existed. Fourth, we analysed the difference in the prevalence of the MR predictive factors between the two countries using Pearson ' s chi-squared test ( χ 2 ). Fifth, we analysed the confounding effects of the MR preventive factors, i.e. the difference in the effect of the country in rated into three factors: cognitive milieu therapy, patientcentred care, and patient involvement. Although patient involvement is part of cognitive milieu therapy and patient-centred care, it might also occur by itself. Leaving, 22 separate MR preventive factors to be included in the present study.
The questionnaire was developed by conceptualizing and operationalizing the MR preventive factors. To increase validity and reliability, we followed the Question Appraisal System ] that was developed to identify problems in questionnaires, including the cognitive processes inherent to the question-answering process.
The construction and pre-testing of the questionnaire involved an eight-step appraisal of each item (QAS-99): reading, instructions, clarity, assumption, knowledge/memory, sensitivity/bias, response categories and others. 2 The pre-test was conducted by creating a cognitive-test protocol that included probes for each item. The probes covered the aforementioned eight steps and consisted of several types: general (e.g. How did you reach this answer?); specifi c (e.g. Why do you believe MR is an intrusive and coercive measure?); understanding/interpretation (e.g. How do you defi ne physical activity?); safety (e.g. How sure are you of your answer?); paraphrasing (e.g. Can you repeat the question in your own words?); and memory (e.g. How did you know that you had 56 MR last year?).
The pre-test consisted of cognitive interviews (20) of four clinical nurse managers individually using concurrent probing (21), in which verbal probes were asked immediately after a question. This protocol led to the revision of seven questions that were subsequently pretested in a group interview.
Data collection, administration, and processing, as well as error identifi cation and the creation of electronic data fi les (22, 23), were administered using the electronic web-based survey system Enalyzer (2010).
The questionnaire was checked for test -retest reliability by asking three clinical nurse managers to answer the questionnaire a second time after a 3-month interval (The respondents were instructed not to check their previous answers). The estimated mean correlation between all of the fi rst and second answers was reasonable (Kendall ' s Tau b ( t Ken,b ) ϭ 0.75). With regard to the Norwegian portion of the study, the questionnaire was translated and retranslated by JB and then checked by two healthcare professionals who were familiar with the subject. A pretest was conducted to test face validity through cognitive interviews of seven clinical nurse managers. This test led to the revision of eight questions, mostly due to imprecise translations. Moreover, one question was added to the questionnaire due to different staff competences: psychologists in Norway are partially responsible for their patients ' treatment, which is not the case in Denmark. NORD J PSYCHIATRY·VOL 69 NO 6·2015 coeffi cient exp( B ), indicates the difference between the countries in an average number of MR episodes.
All models except the fourth set of analyses ( χ 2 -test) were additionally adjusted for background variables. Additionally, we adjusted for nesting of units in departments (17 departments contained two units each and three departments contained three units each), using Generalized Estimating Equations (GEE).
To account for multiple testing, a P -value Ͻ 0.01 was chosen to indicate signifi cance. All analyses were performed using IBM SPSS Statistics for Windows, Version 19.0 (Released 2010; IBM Corp., Armonk, NY).

Approval and ethics
Questionnaires, such as the one included in the present study, do not require approval from the scientifi c ethical committee system in Denmark or Norway when the respondents are able to provide informed consent or when sensitive information is not collected. The Danish Data Protection Agency (Journal number: 2007-58-0015, PSV-models with and without the corresponding MR preventive factor expressed as a percentage [ Δ exp( B )]; a percentage of greater than 10% was chosen to denote a difference.
Linear regression was used to estimate the associations among the frequency of MR episodes, the countries, and the MR preventive factors. To stabilize the variance, the outcome was logarithmically transformed. Hence, the regression coeffi cients are back-transformed [exponentiated; exp( B )] and should be interpreted multiplicatively. Because the preventive factors were dichotomized, a regression coeffi cient exp( B ) indicates the average number of times MR increased for units where a factor was present compared with units where this factor was absent (e.g. if the result was exp( B ) ϭ 0.42, it should be interpreted as 0.42 times the average number of MR incidents without the MR preventive factor or 58% fewer MR episodes occur with the MR preventive factor than without). Except in the fi rst analysis, where the dichotomized variable was " country " , here the regression Notably, for cognitive milieu therapy and patient-centred care, the confi dence intervals were large, which makes the interpretation less precise. We identifi ed three incidents of interaction among the MR preventive factor, the countries and the frequency of MR episodes: cognitive milieu therapy (SOI, P ϭ 0.00), patient-centred care (SOI, P ϭ 0.01) and alarm systems (SOI, P ϭ 0. Six MR preventive factors were signifi cantly different between the two countries: the identifi cation of a patient ' s crisis triggers ( χ 2 , P ϭ 0.01), patient -staff ratio ( χ 2 , P ϭ 0.00), staff education ( χ 2 , P ϭ 0.00), acceptable work environment ( χ 2 , P ϭ 0.01), substitute staff ( χ 2 , P ϭ 0.00) and separation of acutely disturbed patients ( χ 2 , P ϭ 0.00).
Lastly, we identifi ed six MR preventive factors with confounding effects, i.e. the MR preventive factor acted as a confounder

Results
A fl owchart of the participating units can be found in Figure 1, and the annual number of MR episodes can be found in Table 2 . A difference in MR episodes was observed between the countries in a model that included the background variables [exp( B ) ϭ 1.92, P ϭ 0.01] (Fig. 1, Table 2). Table 3 indicates that a single MR preventive factor, staff education [exp( B ) ϭ 0.34, P ϭ 0.00], was significantly associated with a lower frequency of MR episodes (RCT), Putkonen et al. (29) reported the same tendency, where seclusion and restraint time decreased by 51% after implementing six core strategies to prevent seclusion and restraint. In both studies, the effect was attributed to variables other than the patients ' crisis triggers. In a one-way case-crossover designed study, Fluttert et al. (30) reported that implementing the early recognition method (ERM) decreased the number of seclusions by 54%. The ERM is a technique for patients and staff to work systematically with the patients ' triggers (early warning signs), crisis intervention and coping strategies; this is the same identifi cation procedure we asked the clinical nurse managers about in the questionnaire. Overall, the earlier research supports the fi nding that, identification of a patient ' s crisis triggers, is part of explaining the difference in MR use between the two countries. The mechanism involved in this process is probably that patients and staff earlier recognize sign of a forthcoming crisis, and together develop counteractions to avoid escalation of the crisis, thereby reducing the risk of a MR episode. Early prevention could constitute one of the most effective strategies to reduce MR episodes. Second, patient -staff ratio: units in which the staffto-patient ratio was higher than or equal to three-to-one on average vs. those units where this ratio was lower than three-to-one. In Denmark, the effect was small, possibly due to the small number of units where the staff-topatient ratio was lower than three-to-one on average. The factor explains a large [ Δ exp( B ) ϭ Ϫ 11%] portion of the difference in MR episodes between the countries. The effects of the staff-to-patient ratio were partially supported by earlier research. Smith et al. (31) found a 66% decrease in restraint episodes. However, because several factors contributed to the effect, it is unclear what portion of the MR reduction was attributable to the 50% increase in staffi ng. Janssen et al. (32) found that seclusion was used more commonly when the number of patients per staff was greater in long-term wards. Donat (33) reported a signifi cant association between increased staffi ng levels and decreased MR and seclusion numbers. However, Owen et al. (34) noted that the relative risk of violence increased with more nursing staff. In addition, Bowers et al. (35) reported that increased levels of aggressive incidents were associated with increased staffing. Nonetheless, both studies with opposite outcomes did not precisely target MR or seclusion but instead targeted violence and aggression; the results could be infl uenced by staff employment policies, wherein staffi ng levels are increased in units with more aggressive incidents. Our study most likely eliminated this problem by introducing " type of unit " and " type of bed " as confounders. Overall, the earlier research specifi c dealing with the same conditions does support the fi nding that, the staff to patient ratio, is part of explaining the difference in MR use between the two countries. The mechanisms involved in Norway, exp( B ) ϭ 0.54), except more separation of acutely disturbed patients, which was associated with increased MR episodes (Denmark, exp( B ) ϭ 1.05; Norway, exp( B ) ϭ 1.94).

Discussion
Three potential MR preventive factors displayed very different effects in the two countries, which was demonstrated by signifi cant interactions among the MR preventive factor, the countries and the frequency of MR episodes. These preventative factors included cognitive milieu therapy (SOI, P ϭ 0.00), patient-centred care (SOI, P ϭ 0.01) and alarm systems (SOI, P ϭ 0.00). None of the factors had a signifi cant difference in prevalence ( χ 2 , P ϭ 0.09, 0.72, and 0.50). Therefore, it was diffi cult to interpret the differences in the effect based on the country [ Δ exp( B )] due to the interaction. It was also diffi cult to determine why these factors exhibited a different effect in the two countries. One explanation of why cognitive milieu therapy and patient-centred care were effective in Denmark could be that these two factors were well described as methods and included a treatment philosophy in Denmark. Therefore, in Denmark, these descriptions of care and treatment actually covered phenomena such as patient involvement and cooperation; this was, to our knowledge, not the case in Norway in 2010. Another potential explanation is that the limited number of units in Norway that had implemented the three MR preventive factors could be special units with fewer MR episodes. Although the three MR preventive factors were signifi cantly associated with an increased number of MR episodes in Norway, we would not recommend that Norway implement less of these three factors given the small number of units that implemented the factors. Further research is needed to explore the different effects of these three MR preventive factors.
Six MR preventive factors exhibited a signifi cant difference between the countries with regard to the frequency of MR episodes [ Δ exp( B )].
First, identifi cation of a patient ' s crisis triggers: units wherein the initial nurse evaluation of at-risk patients always contained a written evaluation of aggressive, selfmutilating or suicidal behaviours (e.g. crisis triggers, unique calming techniques, and coping strategies) vs. units where the initial evaluation only occasionally or never contained a written evaluation. This factor explains a large [ Δ exp( B ) ϭ Ϫ 10%] portion of the difference in MR episodes between the countries. The identifi cation of a patient ' s crisis triggers was supported by earlier research. In a well-designed before and after study, Jonikas et al. (28) observed a signifi cant decrease (between 49% and 99%) in physical restraint episodes after patients were interviewed to determine their crisis triggers and personal crisis management strategies. In a reasonably designed cluster randomized controlled trial association between work-related factors and some types of violence. Some of the other examined factors are also part of the overall work environment, such as staff training, staff/patient ratio, etc. Overall, the research that supports the association between MR and work environment includes many parts of a healthy work environment, and it supports the fi nding that the work environment partly explains the difference in MR use between the two countries Whatever part of the environment that creates a positive work environment, the wellbeing of the staff is most certainly affecting the wellbeing of the patients, so when staff experiences a more positive work environment the patient probably experience a more positive environment.
Fifth, substitute staff: units in which substitute staff was used less than 50 times (8-h shifts) over the year vs. units where these employees were used greater than or equal to 50 times. The factor explains a large [ Δ exp( B ) ϭ Ϫ 17%] portion of the difference in MR episodes between the countries. The effects of substitute staff were supported by earlier research. Bonner et al. (8) found that both patients and staff discriminated between permanent and regular staff. Bowers et al. (44) reported that adverse incidents were more likely to occur during weeks where regular staff was absent. In another study, Bowers et al. (35) noted that increased levels of aggression were associated with an increased use of bank and agency staff. None of the studies was an RCT, but all the studies indicate similar results, supporting less use of substitute staff. The reason for this fi nding could be that substitute staff lacks knowledge of the unit structures, processes and patients, thereby creating insecurity and disturbances in the everyday life of the unit among some of the most vulnerable patients. This could be a stressing factor for the patients creating more aggressive or self-destructive behaviour -hence, more MR episodes.
Sixth, one factor, separation of acutely disturbed patients, had different results: units that were able to separate acutely disturbed patients from other patients vs. units that were unable to separate patients. The factor explains a large [ Δ exp( B ) ϭ 13%] portion of the difference in MR episodes between the countries. This negative association was not supported by earlier research; in fact, the opposite result was indicated. In a qualitative content analysis study, Meehan et al. (45) noted that patients proposed that the separation of acutely disturbed patients would reduce the number of aggressive incidents. Overall, this result is not supported by earlier research. Therefore, more research is needed to explore the mechanisms involved in placing one or few patients in a separate and isolated part of the unit.

Limitations
Several limitations must be considered when examining the results.
higher staffi ng levels could be that the staff would be more able to observe, interact and corporate with the patients, probably reduce the number of stressed staff, thereby calming the environment, having more personal resources to positive reinforcement, support, and communication, hopefully creating a more empowering and recovery orientated environment for the patients.
Third, staff education: units wherein the staff received basic training for at least 3 1/2 years vs. units where the staff had less than 3 1/2 years of training. The factor explains a large portion [ Δ exp( B ) ϭ Ϫ 51%] of the difference in MR episodes between the countries. Staff education was supported by earlier research. Janssen et al. (32) reported that seclusion was used more commonly in units with less educated staff in the long-term wards. Gim et al. (36) found a signifi cant difference between psychiatric trained and non-psychiatric trained staff in identifying and managing precipitants of violence. In a large 20-year analysis of 1,565 mental health workers, Flannery et al. (37) found that staff who were victims of patients ' aggressions were less formally educated and less trained. None of the studies was an RCT, but otherwise well designed, so it is reasonable to deduce, that earlier research supports the fi nding that, staff education, is part of explaining the difference in MR use between the two countries. Higher basic education among staff is probably leaving the staff with higher analytic and action competences, even in crisis situations, thereby reducing the number of MR episodes.
Fourth, acceptable work environment: units where a workplace environment assessment was present, no or only insignifi cant physical or mental work-environment problems existed, and the overall work environment was evaluated as acceptable vs. units where the workplace environment was not assessed, more workenvironment problems existed, and the overall work environment was evaluated as poor. This factor explains a large [ Δ exp( B ) ϭ Ϫ 15%] portion of the differences in MR episodes between the countries. An acceptable work environment is somewhat complex because the work environment is a mix of leadership, degree of order, organization, staffi ng, training, crowding, etc. De Benedictis et al. (38) reported that several organizational factors, e.g. team climate, staff perception of aggression and interaction of team-members, were associated with increased use of seclusion and restraint. Hanrahan et al. (39) observed a relationship between psychiatric nurses ' work environments and inpatients ' environments. van der Schaff et al. (40) reported a number of building design features that had an effect on seclusion and restraint. Virtanen et al. (41) indicated that patient overcrowding was associated with violent assaults towards employees. Huckshorn (42) found that leaders played a critical role in reducing seclusion and restraint. Camerino et al. (43) found a signifi cant MR, but it is important to acknowledge that this type of comparative research does not allow for cause -effect conclusions. Thus, we cannot claim that the implementation of these potential preventive measures in Denmark and Norway would reduce the number of MR episodes. More high-quality studies are needed to investigate the identifi cation of a patient ' s crisis triggers, an increased number of staff per patient, increased staff education, better work environment, reduced use of substitute staff, and separation of acutely disturbed patients.

Notes
The review was a a systematic review combining qual-1.
itative and quantitative research, so the highest recommendation grade could be: qualitative; at least one high-quality meta-synthesis, systematic review, or metasummary, or a body of high-quality evidence, or quantitative; at least one meta-analysis, systematic review or one or a body of high-quality RCT. STEP 1: READING: Determine if it is diffi cult for the 2.
interviewers to read the question uniformly to all respondents. STEP 2: INSTRUCTIONS: Look for problems with any introductions, instructions or explanations from the respondent ' s point of view. STEP 3: CLARITY: Identify problems related to communicating the intent or meaning of the question to the respondent. STEP 4: ASSUMPTIONS: Determine whether there are problems with assumptions or the underlying logic. STEP 5: KNOWLEDGE/MEMORY: Check whether respondents are likely to not know or have trouble remembering information. STEP 6: SENSITIV-ITY/BIAS: Assess questions for sensitive nature or wording, and for bias. STEP 7: RESPONSE CATEGO-RIES: Assess the adequacy of the range of responses to be recorded. STEP 8: OTHER: Look for problems not identifi ed in Steps 1 -7 (19).
First, some confounding covariates might be missing from the analyses. We believe that we included the most important covariates or covered the effects of others by proxy (46), e.g. patients who experience MR could be more aggressive or self-mutilating in one of the countries -hence, acting as a confounder. However, one earlier study that assessed aggressive behaviour (47) did not report large differences between Denmark and Norway. Additionally, some of the covariates (number of forensic patients, type of bed and the reason for the MR) could eliminate portions of the difference. Therefore, it is not likely that the potential differences in aggressive or selfmutilating patients would infl uence the results. Second, recall bias is possible, especially for the Danish data; the clinical nurse manager had to recall all MR preventive factors as well as the conditions and number of MR episodes from 2 years ago. The outcome variable (i.e. the number of MR episodes) were validated or re-examined, and this did not reveal signifi cant recall bias with regard to the outcome variable. Additionally, recall bias could affect the results if the MR preventive factors had only been implemented for a portion of the year. Thus, this recall bias would most likely only increase the effect on patient involvement. Nevertheless, we must consider the possibility of recall bias with regard to some variables.
Third, three participants answered the questionnaires twice during a 3-month interval to check the test -retest reliability. Although three participants are insuffi cient to calculate reliability properly, the results indicate no substantial difference in the measured constructs between the two administrations.
Fourth, selection bias -the non-responders ' reasons for failure to reply were reviewed, and no systematic differences were observed between responders and nonresponders. Reviewing the Danish National Register on coercion, we found that 49% of responding units were " responsible " for 52% of the MR episodes in the same population, suggesting minimal differences in MR episodes between responders and non-responders; however, no proper drop-out analysis was conducted. Thus, the results most likely can be generalized to the population of both countries. Transferring the results to other countries, however, would require a thorough examination of their cultures, legislation and other relevant factors.
Finally, when data were reanalysed to include patients who experienced MR more than 25 times, the results changed only marginally. Therefore, the exclusion of the patients who experienced MR more than 25 times did not affect results.

Conclusion
The comparison of these MR preventive factors may contribute to better understanding of the phenomenon of