This article appeared in a journal published by Elsevier. The attachedcopy is furnished to the author for internal non-commercial researchand education use, including for instruction at the authors institution

and sharing with colleagues.

Other uses, including reproduction and distribution, or selling orlicensing copies, or posting to personal, institutional or third party

websites are prohibited.

In most cases authors are permitted to post their version of thearticle (e.g. in Word or Tex form) to their personal website orinstitutional repository. Authors requiring further information

regarding Elsevier’s archiving and manuscript policies areencouraged to visit:

http://www.elsevier.com/copyright

Author's personal copy

Methodological issues in monitoring the use of coercive measures

W.A. Janssen a,⁎, R. van de Sande b, E.O. Noorthoorn a, H.L.I. Nijman c, L. Bowers d, C.L. Mulder e, A. Smit f,G.A.M. Widdershoven g, T. Steinert h

a Kenniscentrum GGNet, Warnsveld, The Netherlandsb Utrecht, University of Aplied Science Utrecht & BavoEuropoort Rotterdam, The Netherlandsc Altrecht, Den Dolder and Radboud University Nijmegen, The Netherlandsd King's College, Institute of Psychiatry, London, United Kingdome Bavo-Europoort and Erasmus University Rotterdam, The Netherlandsf Pro Persona Mental Health Care, Wolfheze, The Netherlandsg VU University Medical Center, Department of Medical Humanities, The EMGO+ Institute for Health and Care Research, Amsterdam, The Netherlandsh Centre for Psychiatry, Ravenburg-Weissenau and Ulm University, Germany

a b s t r a c ta r t i c l e i n f o

Available online 12 November 2011

Keywords:Coercive measuresSeclusionRegistrationAnalysisExpressing outcome

Purpose: In many European countries, initiatives have emerged to reduce the use of seclusion and restraint inpsychiatric institutions. To study the effects of these initiatives at a national and international level, consen-sus on definitions of coercive measures, assessment methods and calculation procedures of these coercivemeasures are required. The aim of this article is to identify problems in defining and recording coercive mea-sures. The study contributes to the development of consistent comparable measurements definitions andprovides recommendations for meaningful data-analyses illustrating the relevance of the proposed frame-work.Methods: Relevant literature was reviewed to identify various definitions and calculation modalities used tomeasure coercive measures in psychiatric inpatient care. Figures on the coercive measures and epidemiolog-ical ratios were calculated in a standardized way. To illustrate how research in clinical practice on coercivemeasures can be conducted, data from a large multicenter study on seclusion patterns in the Netherlandswere used.Results: Twelve Dutch mental health institutes serving a population of 6.57 million inhabitants provided theircomprehensive coercion measure data sets. In total 37 hospitals and 227 wards containing 6812 beds wereincluded in the study. Overall seclusion and restraint data in a sample of 31,594 admissions in 20,934 patientswere analyzed. Considerable variation in ward and patient characteristics was identified in this study. Thechance to be exposed to seclusion per capita inhabitants of the institute's catchment areas varied between0.31 and 1.6 per 100.000. Between mental health institutions, the duration in seclusion hours per 1000 inpa-tient hours varied from less than 1 up to 18 h. The number of seclusion incidents per 1000 admissions variedbetween 79 up to 745. The mean duration of seclusion incidents of nearly 184 h may be seen as high in aninternational perspective.Conclusion: Coercive measures can be reliably assessed in a standardized and comparable way under the con-dition of using clear joint definitions. Methodological consensus between researchers and mental health pro-fessionals on these definitions is necessary to allow comparisons of seclusion and restraint rates. The studycontributes to the development of international standards on gathering coercion related data and the consis-tent calculation of relevant outcome parameters.

© 2011 Elsevier Ltd. All rights reserved.

1. Introduction

In European countries, coercive measures such as seclusion, phys-ical and mechanical restraint or forced medication are widely used inclinical practice. Yet, in many countries these measures are regarded

as controversial in managing violence and aggression in acutely illpsychiatric patients (Janssen et al., 2008; Steinert et al., 2009).

In the literature several debates can be identified about the thera-peutic value of seclusion and restraint (Bowers, Simpson, Alexander,Ryan, & Carr-Walker, 2004; Bowers et al., 2007; Haglund, Von Knorring,& Von Essen, 2003; Kaltiala-Heino, Tuohimaki, Korkeila, & Lehtinen,2003; Lendemeijer & Shortridge-Baggett, 1997; Sailas & Fenton,2003). The current literature on coercive measures reveals an increas-ing number of studies comparing seclusion and restraint rates between

International Journal of Law and Psychiatry 34 (2011) 429–438

⁎ Corresponding author at: GGNet Kenniscentrum, Box 2003, NL-7230 GC Warnsveld.,The Netherlands. Tel.: +31 575 580 766; fax: +31 575 580 873.

E-mail address: wim.janssen@ggnet.nl (W.A. Janssen).

0160-2527/$ – see front matter © 2011 Elsevier Ltd. All rights reserved.doi:10.1016/j.ijlp.2011.10.008

Contents lists available at SciVerse ScienceDirect

International Journal of Law and Psychiatry

Author's personal copy

wards, mental health institutes, regions or countries. Studies in Europeand the US indicate that 10 to 30% of all admitted patients are exposedto seclusion, restraint or forced medication in acute psychiatric wards(Steinert et al., 2009). Such data are hard to compare, since coercivemeasures are rarely systematically recorded (Busch & Shore, 2000;Janssen et al., 2008), nor calculated and analyzed or expressed in a con-sistent way (Bowers, 2000; Janssen, Hutschemaekers, & Lendemeijer,2005; Steinert & Lepping, 2008).

1.1. Issues in comparing

Janssen et al. (2008) and Steinert et al. (2009) reviewed severalcomparative coercive measure studies from Australian, Americanand European hospitals published since 1990. These studies showwidely disparate and non-uniform measured rates of seclusion andrestraint. There is an overall lack of specific background informationsuch as: demographic characteristics of the admitted patients, num-ber of beds. Variation in the daily bed occupation in the respectivehospitals further hampers appropriate comparison. Apart from this,consistency problems in expressing seclusion and restraint rates aswell as heterogeneity of methods impede sound conclusions. In an ef-fort to solve these methodological problems and bridge differences insize of the hospitals, wards and patient throughput, Bowers (2000)discussed epidemiological rates such as number of seclusions per100 admissions or per 100 occupied beds.

The current study elaborates further on the dilemmas of relevantdefinitions and methods for the assessment, calculation, analysis andexpression of coercive measures. In this article seclusion is used asmain item to clarify various definitions and provide examples of calcu-lations. Furthermore, we propose how to use the collected rates on co-ercivemeasures to enable feedback for clinical teams.We also proposehow to use these rates for comparison purposes on levels such as: in-tervention, admission, patient, ward and hospital. Dependent on thelevel, different numerators and denominators are required. In this ar-ticle the different ways to express coercivemeasure rates are illustrat-ed with seclusion data from a multicenter database in Dutch mentalhealth care institutions.

1.2. Legislation issues with respect to coercive measures

Coercive measures are used within the context of national legisla-tion. The Dutch legislation, as well as other European countries werein accordance with the guidelines of the United Nations Principles forthe protection of persons with Mental Illness and the improvement ofMental Health Care (1991: UN-Principles) and the European Conven-tion on Human Rights (The convention of Rome, November 4th 1950,http://www.echr.coe.int). Within mental health, this convention pro-hibits torture or humiliating treatment and asks for respect of privacy.Only in the interest of prevention of crime and danger to society pub-lic authority may interfere with these rights and restrain individualfreedom (Legemaate, Frederiks, & de Roode, 2007).

In legislation on mental health the interpretation of human rightsdiffers per country. Legislation has to deal with the conflicting interestsof the patient as a legal subject as opposed to the patients' interests asan ill person. The law also has to deal with interest of others and societyas a whole. The way these interests are elaborated leads to specific fea-tures in the legislation over different countries (Legemaate et al., 2007).Therefore legislation on mental health differs between countries withrespect to how individual freedom may be restricted. Dutch legislationdistinguishes forced admission and forced treatment. In Dutch law amentally ill patient may be involuntary admitted but may object treat-ment. Forced medication is seen as very invasive and noted as the lastresort. In this context seclusion and restraint are seen as less invasiveand may be used to contain danger. In the UK or Finland the same ag-gressive patient can be involuntary admitted in order to receive enforcetreatment. In these countries, forced medication is valued as less

invasive than seclusion and more in accordance with patient needs(Bowers et al., 2007).

2. Method and materials

2.1. Development of the Argus-scale

In the Netherlands, five mental health institutes together with theDutch mental health inspectorate developed a rating scale to registerthe use of restrictive measures: the Argus-scale. The Dutch mentalhealth inspectorate needed valid and reliable coercive measure datathat could be easily gathered in daily practice. A study on coerciondata in twelve psychiatric hospitals over 2002–2003 showed figuresgathered following implementation of the Mental Health Act in1994 were barely reliable (Janssen et al., 2008).

Themain objective of this rating scale was to record the number andduration of seclusion and restraint measures per patient more in detail.Nurses completed the Argus scale. All patients and all coercive mea-sures were included, regardless of legal status or whether the patientobjected the use of coercive measures or not. Nurses were instructedto rate each coercive measure for each day separately. The exact timeof the start and the end of each coercive measure was also entered inthe database, allowing the calculation of episodes (Fisher, 2003).

2.2. Definitions on coercive measures in the current study

The Argus-scale covers three main coercive measures: seclusion,restraint and enforced medication.

1. Seclusion is defined as bringing the patient into a locked roomwhere he/she is alone and able to move around. The patient is un-able to leave due to a locked door (Steinert & Lepping, 2008) forsome time, irrespective of the circumstances and safety precondi-tions. In the literature, seclusion is related to several underlyingconcepts. Seclusion may be related to a specifically designed area,minimal furnishing, social isolation, egress, compulsion and timeas a delimiter. Many rationalizations may be related to seclusioncontext of a closed ward (Mason & Alty, 1994). The definition wepropose only refers to being locked into a room (seclusion room,isolation room, (stripped) bedroom or otherwise designed area),regardless of the purpose for which this is done. The frequency ofintermitted security checks and closeness of nursing staff duringseclusion differ from country to country.

2. Mechanical andmanual restraint is defined as immobilizing the pa-tient with external mechanical devices or physical force. Mechani-cal restraint refers to the use of belts to fix a patient to a bed or chair(Steinert & Lepping, 2008). Manual restraint refers to immobilizingthe patient on the floor (Steinert & Lepping, 2008) or upright bystaff members, to calm down through specific holding techniques.

3. Enforced medication is defined as application of intramuscularmedication by force or by definite psychological pressure, i.e. an-nouncing intramuscular medication if medication is not takenorally at once (Bowers et al., 2004; Steinert & Lepping, 2008).

2.3. Calculations on coercive measures

In calculating outcome parameters in Argus the time frame of a co-ercive measure is described in two ways:

1. An episode is defined as the period from the beginning (closing thedoor when the patient is in the room or locking the restraint belt)to the moment of discontinuation (opening the locked door or re-moving the restraint belt). In other words, we use the term epi-sode as a synonym of the discrete seclusion, according Crenshawand Francis (1995).

2. An incident, or event (Heilbrun, Golloway, Shoukry, & Gustafson,1995) is derived from the epidemiological term incidence and

430 W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

covers a sequence of episodes without a discontinuation of morethan 24 h. In Argus, an interruption of more than 24 h leads tothe count of a new incident of seclusion, restraint or forced medi-cation. Fig. 1 illustrates these definitions.

Coercive measures are applied to individual patients at wards.Comparisons between wards and hospitals are dependent on thecharacteristics of these patients (Steinert et al., 2006), as well as oncharacteristics of wards (Bowers et al., 2004; Mason & Alty, 1994)or staff on the wards (Janssen, Noorthoorn, van Linge, & Lendemeijer,2007). For the purpose of this study we used patient's characteristicssuch us sex, age, ethnic origin and primary diagnosis next to size andtypes of wards. In the comparisons between wards these variablesneed to be taken into account as denominators, the coercive measuresas numerators.

With respect to the ward characteristics, the number of beds andseclusion rooms were counted. The ward types were categorized infour groups; acute admission wards, long stay wards, residentialwards for the care of geriatric psychiatric patients, and specializedwards. Acute admission wards are defined as specific wards, whichhave to use non-selective admission criteria to serve emergency psy-chiatric services in a specific catchment area. In long stay wards pa-tients tend to stay more than 9 months for intensive treatment andrehabilitation. Residential wards are mainly focused on the promo-tion of quality of life and long term living arrangements for elderlypatients. Specialized wards often serve patient categories in whichdual diagnosis are prevalent (such as: intellectual disability and se-vere mental illness, brain trauma, addiction and severe mental ill-ness). Due to their specializations status those wards also acceptadmissions from other catchment areas. The risk of using coercivemeasures may be expected to vary between several ward types.

2.4. Levels of analyses

To provide understanding in patterns of coercive measures, fivedifferent levels of calculations over several time frames can be used.Fig. 2 illustrates what kind of outcomes is generated by analyses onthese various levels. Each level has specific input variables, requiresspecific calculations, and provides output only comparable withinthe same level. Data from a lower level can be used for the nexthigher level. For the illustration of outcome we used a timeframe of1 year.

2.4.1. Institutes within an environmental contextThe fifth, and highest level is the mental health institute level,

more specifically the specific urban or rural context and varying num-bers of inhabitants in the defined catchment areas. Rural is defined

following OESO classifications as serving less than 150 inhabitantsper km2 (Terluin, Slangen, van Leeuwen, Oskam, & Gaaff, 2005),urban as serving more. The organization of mental healthcare in theNetherlands implies that specific mental health institutes need tocover a specific catchment area. The size of the mental health instituteand the type of patients it serves are variables that need to be takeninto account in interpreting the data.

On the mental health institute level we focused on all coercivemeasureswithin 1 year. The coercivemeasureswere related tomentalhealth institute variables, such as (1) number of inhabitants in themental health institute's region, (2) location of the mental health in-stitute (mainly rural or urban), (3) presence of high secure wards,(4) number of occupied beds of the wards, and (5) number of admis-sions. Correction for the number of inhabitants is necessary to performstandardization to a specific number of admissions. In addition to this,differences in number of admitted patients within certain diagnosis,age, and gender categories may explain differences in seclusion rates.

The results on this level represent the prevalence of coercive mea-sures as numerators, with the number of inhabitants, admissions, in-patient days or treated patients as denominators. The denominatorswere standardized to 100.000 inhabitants and 1000 admissions or in-patients' days.

2.4.2. Ward levelThe fourth level refers to the characteristics of a specific ward.

Most wards within a mental health institution normally vary inbed-size, age category, treatment focus, level of security and averageduration of stay (Janssen et al., 2007). These variables need to betaken into account when comparing rates of the use of coercive mea-sures at multi-ward level. Thus, at the specific ward level the coercivemeasures were related within a time span of 1 year to (1) the numberof daily occupied beds, (2) the number of admissions, and (3) thenumber of occupied beds, or more specifically the number of bedhours excluding the hours of leave, for each ward separately.

Comparisons on this level are interesting for wards with similarcharacteristics and may be most helpful for designing specific inter-ventions targeted on specific patient groups. Even though the calcula-tions of the numerators are similar to the calculations on mentalhealth institute level, specific wards may be more interested in learn-ing of specific patterns. Examples are: the duration and numbers ofpatients exposed to seclusion related to ward admission patterns inspecific time frames. In addition to this, relative risks for patients tobe exposed to coercive measures at different wards of a mental healthinstitute may help to understand patterns of incidence and duration.On the mental health institute level and ward levels single or repeat-ed occurrence of coercive measures within single patients is an

xxxxxxxxxxxxrestraint

16

15

14

13

12

11

10

9870-6

23

22

21

20

19

18

17

16

15

14

13

12

11

10

9870-6

hour

xxx

Day 1

xxxxxxseclusion

Day3

Day 2

Episode (number and duration)

Day (number and duration)

Incident (number and duration)

Patient (number)

Fig. 1. Notifications on intervention level (seclusion and restraint measures) within patient admission.

431W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

important distinctive variable. More severely mentally ill patientsmay be associated with a higher risk for aggression and subsequentlybe exposed to repeated coercive measures (Bowers, 2000; Janssen etal., 2005; Martin et al., 2005).

2.4.3. Patient levelThe third level is the patient level. This level implies an elabora-

tion of specific patient characteristics within the ward's populationeither secluded or not. Analyses on this level are important for epide-miological research and hospital management. Analysis at this levelallows calculation of associations between static and dynamic char-acteristics of seclusion and restraint by means of stratification(Ahmed & Lepnurm, 2001; Betemps, Somoza, & Buncher, 1993;Bower, McCullough, & Timmons, 2000; Fisher, 1994; Kaltiala-Heino,Korkeila, Tuohimäki, Tuori, & Lehtinen, 2000; Kaltiala-Heino et al.,2003; Steinert et al., 2006; Way & Banks, 1990). For example, do pa-tients with specific DSM-IV axis I or axis II disorders have more orless risk of being exposed to seclusion or restraint? In addition tothis: age, sex, ethnic background as well as GAF scores may betaken into account in comparing individual risk to average risk ofbeing exposed to coercive measures. For example, what is the riskin terms of days in seclusion per number of admission days of a pa-tient with for example a psychotic disorder? What is the risk of anolder patient to be repetitively restrained? Identification of patientswith an extraordinary high number of seclusion incidents may be rel-evant for the Mental Health Inspectorate as well as for medical direc-tors in their hospital work package.

2.4.4. Patient's admission levelThe second level refers to the patient's admission on a specific

ward and/or mental health institute. Frequency, time of the day andduration of coercive measures can be correlated with local contextual

variables such as location of the ward, ward culture or the possibilityto manage difficult patients' behavior either in their own ward or in amore specialized ward.

2.4.5. Separate measuresFinally, the first level is the level of separate seclusion and re-

straint episodes and incidents per individual patient. This informationmay be used in treatment plan evaluations and more in general to de-velop strategies to manage challenging behavior in a less restrictivebut safe way. National Inspectorates of Health will be focused onthis information during routine audits. The data gathered at thislevel provides individual information which can be crucial to reflecton the information derived from the third level and answer questionswhy specific patients or patient groups have distinct chances to be ex-posed to coercive measures.

2.4.6. Relevance of the levelsRatios and figures within the fifth level and the fourth level are in

general the same, as well as the ratios and the figures in the threelowest levels. The figures within levels 4 and 5 are important for theinstitute management officers, while the figures in the three lowerlevels are relevant for nurses, therapists, ward management and es-pecially the medical director in order to comply with information re-quired by the mental health inspectorate. The first represent generalfigures while the second represent outcome at an individual basis.

2.5. Data collection

Four large Dutch mental health institutes started collecting datawith the Argus-scale in 2007. Other mental health institutes followedin subsequent years. Currently up to 30 mental health institutes nowparticipate on a voluntary basis in the development developing a

Levels Denominators Numerators Correction factors Ratios and figures Comparison goals

(number of ….) (number of ….) (number of ….)

5. Hospital Occupied beds Started seclusions Ethic minority patients Incidents / 100.00 inhabitants International

Inhabitants of region Started restraints Admission beds Incidents /1000 admissions National Benchmark

Admissions Given PRN Medication Long stay beds Days / 1000 admission days National ranking

Patients Patients in diagnostic groups Hours / 1000 inpatient hours Monitoring at hospital level

Days in seclusion Competitor hospitals in same area Number of patients

Hours in seclusion Number of incidents

Days in seclusion

Hours in seclusion

4. Ward Occupied beds Started seclusions Staff occupancy Incidents /1000 admissions Benchmark between same wards

Admissions Started restraints Ward facilities (comfort rooms/ PICU) Days / 1000 occupied days Monitoring at ward level

Bed hours Given PRN Medication Ward types Hours / 1000 inpatient hours Team management information

Days in seclusion Ward size Number of patients Quality control

Hours in seclusion Number of incidents

Days in seclusion Number of episodes

Episodes in seclusion Relative risk between wards

3. Patients Patients in age groups Patients involved in seclusion Staff per patient per restrictive measure: Mental health inspectorate control

Patients of certain gender Patients involved in restraints Space per patient % age Medical directorate

Patients in diagnostic groups Patients receiving PRN med. % male/female Patients advocates

Patients with high or low GAF Diagnostic categorisation % per ethnic groups Ward psychiatrist

Started seclusions % per diagnosis

Days secluded % per higher of lower GAF

2. Admission Number of re-admissions Started seclusions Number of incidents Ward psychiatrist monitoring

Length of stay Number of days Ward nurses monitoring

Number of hours Ward managers monitoring

Mean duration per days Mental health inspectorate control

Mean duration per episode

1. Intervention Seclusion episodes Evaluation treatment plan

Restraint episodes Mental health inspectorate control

Episodes of receiving PRN med.

Fig. 2. Levels of analysis.

432 W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

national benchmark coercion database, which covers about 95% of allmental health institutes in the Netherlands. As far as we know, thedatabase currently is the largest database on coercive measures inEurope. After applying for participation in this initiative each institutewas previously audited and during the process of data gatheringmonitored by two research supervisors, the first and third author ofthis paper.

The data presented in the current study were collected in 2008 or2009 and covered all the wards of each mental health institution for afull year. Inclusion was therefore limited to the first twelve partici-pants. The reliability of the Argus rating scale was determined bycomparing the results to other sources of information such as thedata collected for the Dutch Inspectorate of the Ministry of Health,nursing records and medical charts in four hospitals over 1 month.The comparison between Argus-scale and other sources revealed per-centages of agreement between 85 and 96% and a Cohen's Kappa's be-tween 0.64 and 0.92.

2.6. Statistical procedures

Most of the figures presented here are descriptive in nature. Oddsratios were used to calculate the chance of seclusion within severaldiagnostic groups, wards or mental health institutes.

3. Results

3.1. Sample

In this study, 12 mental health institutes provided their full coer-cion data set, all together 37 psychiatric hospitals. The mental healthinstitutes were spread all over the country, five in an urban area andseven in a rural area. All participating institutes had a non-selectiveadmission policy in their catchment area. In total, the institutesserved a population of 6.57 million inhabitants, about 41% of thetotal population in the Netherlands. A total of 227 wards were includ-ed with a total of 6812 beds covering 31,594 admissions of 20,934 pa-tients either once or more often.

3.2. Characteristics of the participating institutes

Table 1 shows the main characteristics of the participating insti-tutes. In this table considerable variation in availability of seclusionrooms may be identified. While most institutions had about one se-clusion room available for 50 patients, institute number 6 had lessthan one seclusion room for every 100 patients. Furthermore, sub-stantial differences in patient characteristics such as age, gender,and ethnic background between the institutes can be identified. Con-siderable variation was also present in the available data on diagnos-tic characteristics. All these features need to be taken into account inthe interpretation of the outcomes. This issue will be explored morein-depth in the discussion section.

3.3. Ward level findings

In the data set of each of the twelve participating institutes, oneacute admission ward was selected to illustrate the characteristicsat the ward level (Table 2). The size of these acute admission wardsvaried between 11 and 50 beds. All wards had at least one seclusionroom; availability differed between one per 4 beds up to one per 17beds.

Large variations were identified between wards in diagnostic cat-egories. For example, while in the acute admission ward of institutenumber 6 no diagnosis on DSM Axis I was recorded for 1% of the pa-tients, whereas this was the case for 51% of the patients admitted onthe selected acute admission ward in institution 5. The prevalence ofDSM-IV schizophrenia codes reached towards 13.1% (acute admission

ward in institute number 12) but was much higher in an acute admis-sion ward in institute number 7 (47.8%). Mood disorders DSM-IVcodes differed between 4.5% (acute admission ward, institute number5) and 25.6% (acute admission ward, institute number 9). Drug abusewas registered infrequently as a primary diagnosis in some instituteswhereas others showed percentages up to 14.4% (acute admissionward, institute number 12).

3.4. Various levels compared

The figures and epidemiological ratios of seclusion incidents aswell as episodes at institute level, ward level, and patient level, aredepicted in Table 3.

On the institute level, substantial differences were found in theduration of seclusion hours per 1000 inpatient hours: these varyfrom 0.3 to 18. The number of seclusion days per 1000 inpatientdays also varies strongly between the institutes (range 1–50), muchstronger than the rate of seclusion incidents per 1000 admissions(range 79–745) or the number of incidents per 100.000 inhabitants(range 0.31–1.60). Notably, the percentage of incidents for individualpatients with more than one seclusion incident does not explainthese differences, as these percentages are comparable between theinstitutes (mean 62.8; range 53.2–79.4).

On the level of the acute admission ward (i.e. the fourth level), 94to 896 patient admissions took place during 1 year (Table 2). Withinthis group of admitted patients 14 up to 57 patients experienced atleast one or more seclusion incidents (Table 3). The number of (seclu-sion) days per 1000 occupied bed days varied between 14 and 134, astaggering difference.

The relative risk of being exposed to seclusion varied strongly onthe acute admission wards compared to the other wards of the insti-tute. For example, patients admitted to the acute ward in institutenumber 2 had a 10 times higher chance to be exposed to seclusionthan on the other wards in the same institution. On the other hand,patients admitted to the admission ward in mental health institution1 had the same chance to experience seclusion as those staying onother wards in this mental health institution (Table 3).

On the patient level, a number of interesting findings emerge. Pa-tients with the absence of a DSM-IV-Axis I code had a much higherrisk for experiencing seclusion (OR=4.98) on the acute admissionward of institute 5 than on the admission ward of institute number11 (OR=0.18). Patients with a DSM-schizophrenia code had morethan 4 times more chance to experience seclusion on the acute ad-mission ward of institute number 6 (OR=4.10) than on the acute ad-mission ward of institute number 5 (OR=0.72). In the institutenumber 4 and 5 the majority of the secluded patients were between18 and 40 years of age (62.5 to 70%), while in institute number 2, 6and 10 the majority of the secluded patients were older than 40(58%, 61.1% and 63.3% respectively).

On the admission level, information of random patients is pre-sented. This shows great differences in the seclusion figures betweenpatients during their admission. The last column shows the means ofseclusion incident figures at patient level throughout the full data-base. The mean number of seclusion hours per incident was nearly184.

3.5. Summary of findings

In short, the most striking difference can be identified in the largevariability in the seclusion figures between the participating insti-tutes, with respect to the incidence ratios (incidents per 1000 admis-sions or per 100.000 inhabitants) as well as the prevalence ratios(days per 1000 occupied beds or hours per 1000 inpatient hours).By international comparison, the number of seclusion incidents per1000 admissions as well as the absolute duration and the durationof seclusions per 1000 admission hours in the Netherlands is

433W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

Table1

Characteristicsof

thepa

rticipatinginstitutes

andad

mittedpa

tien

ts.

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Totals

12

34

56

78

910

1112

Urban

ization

Rural

Rural

Rural

Rural

Urban

Rural

Urban

Urban

Urban

Rural

Urban

Rural

Num

berof

inha

bitantsin

catchm

entarea

390.00

070

0.00

060

0.00

051

2.00

065

6.00

065

0.00

090

0.00

040

0.00

038

2.00

050

0.00

060

0.00

028

0.00

06.57

0.00

0Num

berof

hosp

italswithininstitutions

25

45

24

33

32

22

37Ty

peof

wards

inthestud

yNum

berof

admission

wards

46

63

46

44

43

52

51Num

berof

long

stay

wards

33

35

83

310

34

53

53Num

berof

reside

ntialw

ards

forelde

rly

37

13

104

44

52

22

47Num

berof

specializ

edwards

15

65

207

34

33

154

76To

taln

umbe

rof

wards

1121

1616

4220

1422

1512

2711

227

Num

berof

beds

includ

edin

theda

taba

se26

577

164

079

884

277

553

358

241

230

952

136

468

12Num

berof

observed

newly

admittedpa

tien

ts12

6022

7410

5611

8925

1615

4434

7015

4212

4317

5320

5310

3420

934

Bedoc

cupa

ncyinclud

eddu

ring

2008

or20

0996

.725

295.53

017

3.32

725

8.83

923

5.63

727

1.14

327

7.10

819

0.27

713

7.29

930

2.37

531

0.16

513

3.55

126

8197

6To

tala

dmission

s18

2830

0112

0521

0833

5821

7545

3328

8433

0728

5931

3622

8532

697

Num

berof

seclus

ionroom

sav

ailable

1718

2116

196

2520

229

2711

211

Mea

nag

eof

alla

dmittedpa

tien

ts42

.347

4547

.144

.646

39.1

43.7

47.0

40.3

40.4

47.1

43.8

%Male

44.8

38.8

51.5

56.8

61.3

43.0

72.0

53.1

46.1

52.3

61.6

52.8

54.6

%Ethn

icorigin

7.4

54.2

57.2

629

.236

.911

.06.8

10.6

10.8

13.5

Prim

arydiag

nosis

%Diagn

osis

unkn

own

9.8

23.3

5.8

30.7

31.4

2.6

0.2

16.6

17.8

19.8

9.8

9.5

15.1

%Psyc

ho-soc

iald

isorde

rs6.5

134.3

9.1

2.3

4.2

7.0

0.3

1.8

11.9

3.2

5.9

5.7

%Anx

iety

disorders

12.2

9.5

9.0

9.7

3.1

10.3

6.4

1.2

1.4

9.1

5.7

13.6

6.8

%Dep

ressivedisorders

9.6

18.5

13.8

8.7

5.8

18.6

7.4

2.4

15.9

11.0

9.3

12.6

10.8

%Bipo

lardisorders

7.9

49.2

3.1

2.8

7.3

7.0

7.4

9.1

4.9

4.7

7.5

6.0

%Schizo

phrenicdisorders

38.9

13.4

38.5

13.3

18.4

30.0

33.8

64.0

33.6

19.9

34.4

24.5

29.8

%Psyc

ho-organ

icdisorders

9.5

13.8

4.1

5.6

4.7

7.1

10.5

1.0

7.7

5.7

3.3

6.3

6.8

%Drugab

use

2.8

1.9

6.1

13.9

29.3

3.1

3.0

5.3

3.5

11.4

12.2

12.8

9.2

%Pe

rson

alitydisorder

01.1

9.2

5.6

2.2

5.9

17.9

1.8

9.1

6.2

39.9

43.3

12.8

434 W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

definitely higher than in some European countries (Janssen et al.,2008; Steinert et al., 2009).

4. Discussion

In this article we addressed several methodological issues whenregarding data on the use of coercive measures at a national, institute,ward or patient level. The aim of the study was to contribute in thedevelopment of consensus on how to record and analyze rates of co-ercive interventions. Data of 12 mental health institutes were used toillustrate the possibilities of comparing seclusion and restraint figuresat several levels when using a standardized coercive measure ratingscale (Argus).

It is important to use consistently one time frame, since varyingtime frames complicate comparisons. A time frame of 1 year seemsto be preferred, because shorter time frames are sensitive to seasonalvariation. In local feedback reporting to staff on the wards however,shorter time frames may be used to raise the awareness of currentpatterns in coercive measures and also reflect on specific to beachieved in coercion reduction programs. For each level, from thehospital down to the individual level, we discuss how to deal withcomparison and interpretation of outcomes for reporting purposes.

4.1. Institute level (fifth level)

Using absolute numbers, such as number of coercive measures ornumber of patients experiencing coercive measures, obviously doesnot result in meaningful comparisons between hospitals. Using abso-lute numbers for a comparison within a hospital over several yearsalso results in some shortcomings whereas this does not take into ac-count the size, composition and changes of the institute's infrastruc-ture (Bowers, 2000). During the current changes in healthcareinstitute sizes may vary over time and the admission and dischargerates of patients are not stable within years. Ward types may not bediscrete and comparable categories across mental health institutes.Comparable wards may admit different groups of patients in differentcircumstances.

The number of incidents of seclusion and restraint per 100.000 in-habitants in a hospital's catchment area might be a more powerfulway to compare hospitals, given that this is a more stable number.However, this epidemiological calculation only provides a soundcomparison if (1) patients from a defined catchment area are alwaystreated in the respective hospital, which probably is the case in somecountries but not in others and (2) no other institutes, hospitals orresidential homes, apply such coercive measures in the respectivecatchment area.

The calculation of the number of seclusion incidents per 1000 ad-missions to compare outcomes was previously recommended byBowers (2000). This calculation strategy suggests that new patients atacute psychiatric wards are most likely to be exposed to coercive mea-sures. On the level of hospitals, this is only a part of reality. It does nottake into account that patients on long-term care wards are sometimessecluded for long periods in countries, like the Netherlands. Using bothlong term and newly admitted patients as denominator over a period ofa year has the same problems as using newly admitted patients only.The percentage of patients with repeated coercive measures as well asthe mean number of incidents within this group of patients providesbetter understanding of the possible degree of overestimation. Theoverestimation in the incidence per 1000 admissions seems to be simi-lar for all hospitals.

The number of seclusion days per 1000 occupied beds is useful forwards having a low turn-over or a population with minimal turn-overrates. However, this rate is not directly sensitive to throughput andthe figure ignores patients on leave, again a problem for all hospitals.Also, variations in rates of incidents by this method may be due tovariations in length of stay and the severity of the patient's mentalTa

ble2

Characteristicsof

admission

wards

(one

perinstitute)

andad

mittedpa

tien

ts.

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Totals

12

34

56

78

910

1112

Num

berof

beds

available

1110

5027

2231

3440

3028

2136

340

Num

berof

admission

s94

139

170

168

154

271

388

896

155

653

242

793

4129

Abs

olutenu

mbe

rof

occu

pied

beds

.37

6829

8618

570

9882

8052

1146

712

373

1445

411

260

9434

7185

1269

812

2129

Mea

nda

ysof

stay

withinan

admission

4021

109

5952

4232

1673

1430

1630

Num

berof

seclus

ionroom

sav

ailable

32

32

31

26

14

24

33Age %

18–40

years

51.9

43.9

52.3

49.7

54.5

51.0

43.2

53.7

52.3

43.0

64.5

50.9

50.3

%41

–65

years

46.2

50.7

44.4

34.4

44.8

47.7

54.0

40.5

47.4

54.7

35.5

48.7

46.6

%66

–olde

r1.9

5.4

1.9

9.2

0.7

1.1

2.9

5.7

0.0

2.5

00.5

2.8

%Male

55.8

60.8

46.8

54.8

74.7

57.0

46.9

57.1

55.8

46.4

5046

.252

.1%Ethn

icorigin

9.6

5.5

3.3

8.3

5.1

09.1

43.2

44.2

6.6

12.4

12.0

17.1

Prim

arydiag

nosis

%Diagn

osis

unkn

own

17.3

14.9

6.2

27.4

51.9

0.7

12.1

18.1

9.3

26.1

7.9

10.9

16.4

%Psyc

ho-soc

iald

isorde

rs4.8

4.1

6.0

7.7

03.6

2.1

1.8

1.2

7.8

3.3

8.6

4.8

%Anx

iety

disorders

7.7

3.4

9.7

1.2

0.6

5.3

4.6

4.0

2.3

10.9

9.1

22.8

9.1

%Dep

ressivedisorders

5.8

12.2

15.7

11.9

4.5

9.9

12.6

7.1

25.6

16.2

8.7

14.8

12.1

%Bipo

lardisorders

9.6

10.8

12.0

3.6

2.6

9.6

9.1

4.3

12.8

5.4

8.3

5.8

6.7

%Schizo

phrenicdisorders

46.2

44.6

36.1

30.4

23.4

38.7

47.8

47.0

25.6

21.1

45.0

13.1

32.9

%Psyc

ho-organ

icdisorders

1.0

2.0

5.6

4.8

04.3

3.0

1.9

2.3

2.9

1.2

4.2

2.9

%Drugab

use

1.9

7.4

4.6

11.3

14.3

3.0

8.7

8.2

5.8

6.7

8.3

14.4

8.8

%Pe

rson

alitydisorder

5.8

0.7

4.2

01.3

21.5

31.9

7.6

15.1

33.2

37.6

47.5

23.6

435W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

Table3

Ratios

andfigu

resof

seclus

ions

perinstitute.

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Institution

Totals

Leve

l1

23

45

67

89

1011

12

Institution

Num

berof

patien

tsin

seclus

ion

115

156

232

163

196

172

220

202

113

172

208

7520

24Num

berof

seclus

ionincide

nts

374

423

560

475

1079

638

276

370

280

281

376

308

5440

%Pa

tien

tswithrepe

ated

incide

nts

29.9

25.2

45.8

22.8

34.0

27.8

25.8

31.7

29.2

23.8

28.7

27.6

30.0

Mea

nN

ofincide

ntsin

patien

tswith

repe

ated

incide

nts

3.8

3.8

4.6

3.9

6.3

4.7

3.4

3.3

3.4

3.6

3.2

3.1

4.0

%Incide

ntswithinpa

tien

tswithmore

than

oneincide

nt61

.656

.479

.453

.276

.664

.557

.360

.158

.453

.456

.354

.162

.8

Seclus

ionincide

ntspe

r10

0.00

0inha

bitants

1.03

0.70

0.90

0.90

1.60

0.98

0.31

0.92

0.73

0.56

0.62

1.10

0.82

Seclusionincide

ntspe

r10

00ad

mission

s20

414

174

539

942

829

379

128

225

160

183

134

259

Seclus

ionda

yspe

r10

00oc

cupied

beds

5013

910

51

845

143

77

12Se

clus

ionho

urspe

r10

00inpa

tien

tho

urs.

186

105

50.3

4.52

12.2

5.5

1.7

9.6

3.3

6.0

Adm

ission

ward

(one

perinstitute)

Num

berof

patien

tsin

seclus

ionat

the

admission

ward

1456

3627

5631

2957

1252

3347

38

Num

berof

seclus

ionincide

nts

5410

665

6014

668

3784

1818

494

104

85Se

clusionincide

ntspe

r10

00ad

mission

s57

475

138

241

194

825

138

858

611

628

238

813

143

4Num

berof

seclus

ionda

ys50

637

925

828

737

634

020

729

715

674

081

528

238

7Se

clus

ionda

yspe

r10

00oc

cupied

bed

days

134

127

1428

8530

1745

1478

113

2259

Num

berof

episod

esseclus

ion

1078

1101

439

310

146

252

525

1925

285

2110

571

641

782

Relative

risk

toseclud

epa

tien

tsof

admission

ward

0.95

11.1

1.55

2.26

6.35

1.50

1.48

7.72

1.12

12.5

2.6

1.7

4.2

Patien

t%withintheseclud

edov

erthefullinstitute

Age %

18–40

years

57.2

37.7

55.6

62.5

7038

.736

.246

.254

.637

.759

.357

.550

.7%41

–65

years

42.9

55.4

38.9

37.5

3012

.921

.743

.838

.563

.337

.835

.937

.7%66

–olde

r0

3.6

2.8

00

45.2

37.0

10.1

7.0

02.9

6.2

10.4

%Male

64.3

69.6

47.2

62.5

7458

.270

.351

.460

.153

.360

.465

.160

.8%Ethn

icorigin

14.2

3.6

2.8

8.3

63.2

14.1

38.1

16.8

13.1

16.2

19.5

12.7

Prim

arydiag

nosis

%Diagn

osis

unkn

own

(ORad

mission

andothe

rwards

)14

.3(0

.79)

35.7

(3.09)

2.8(0

.71)

20.8

(0.69)

64(4

.98)

019

.6(0.68)

4.6(0

.79)

7.0(1

.19)

23.0

(1.14)

1.6(0

.18)

12.8

(0.92)

17.3

(1.19)

%Psyc

ho-soc

iald

isorde

rs7.1

1.8

012

.50

01.8

3.1

5.6

2.5

0.8

5.6

2.9

%Anx

iety

disorders

7.1

011

.10

03.2

2.9

6.6

0.7

7.4

4.5

12.8

4.5

%Dep

ressivedisorders

010

.78.3

04

3.2

3.3

15.3

9.8

7.4

14.1

4.6

7.1

%Bipo

lardisorders

14.3

8.9

16.7

4.2

212

.918

.59.7

15.4

9.8

8.2

11.8

11.0

%Schizo

phrenicdisorders

(ORad

mission

andothe

rwards

)42

.9(0.87)

37.5

(2.45)

44.4

(1.52)

37.5

(1.43)

20(0

.72)

51.6

(4.10)

43.5

(1.71)

50.5

(1.13)

41.3

(1.03)

34.4

(2.21)

45.7

(1.72)

26.7

(1.03)

38.2

(1.70)

%Psyc

ho-organ

icdisorders

00

5.6

00

04.0

1.0

4.9

2.5

2.1

3.6

2.0

%Drugab

use

03.6

8.3

258

06.5

9.2

2.8

8.2

6.6

19.5

7.9

%Pe

rson

alitydisorder

14.2

1.8

2.8

02

29.1

15.3

27.8

12.6

22.1

39.9

49.2

16.9

%GAF:

>60

28.5

7.2

22.4

16.7

143.2

4.3

4.6

31.5

16.2

2.9

20.5

12.8

40–60

49.9

50.2

36.1

62.3

4819

.445

.922

.238

.567

.562

.549

.245

.5b40

21.3

42.6

41.5

2138

77.4

49.7

73.2

30.1

16.3

34.5

30.3

41.6

Adm

ission

Num

berof

seclus

ionincide

nts

91

11

24

34

21

22

2.35

Num

berof

days

217

24

29

1015

66

145

218

Durationin

hours

3341

:50

14:30

47:00

22:05

194:25

22:00

158:00

57:58

73:00

116:45

25:00

4:00

183:54

Mea

ndu

ration

perda

y(h

h:mm)

15:25

7:15

11:45

11:02

21:35

2:06

10:31

9:67

12:11

8:21

5:00

2:00

4:54

Num

berof

episod

es56

03

63

104

4227

3650

62

82Mea

ndu

ration

ofep

isod

es(h

h:mm)

6:00

4:50

7:50

7:20

19:27

5:30

3:46

2:09

2:01

2:35

4:12

2:00

2:14

436 W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

illness. Using seclusion hours per 1000 bed hours is a more preciseand valid measure and may be preferred in the evaluation of suchdata. Having Bowers' (2000) recommendations in mind and usingthe method with caution and consistently we believe that this meth-od can be used for comparisons between hospitals. Identifying theproportion of patients with more than one seclusion incident mayhelp in identifying the degree of overestimation. However, this over-estimation does not influence interpretation when reporting aboutthe hospitals within a nationwide benchmark.

4.2. Ward level (fourth level)

Using absolute numbers on ward level is even more problematicthan on hospital level, because the impact of fluctuations in wardsize and throughput of patients is greater.

The number of seclusion days per 1000 occupied beds also hasmore drawbacks on this level than on hospital level. The outcomesare rather sensitive for outliers such as single patients with an ex-traordinary high number of episodes. This is important to considerwhen studying outcome at ward level. Especially on acute admissionwards, the variations in the number of occupied beds is sensitive tothe daily or weekly throughput of patients, the length of stay on theward and number of discharges from the ward. In general, the bed oc-cupancy on long-term treatment wards and wards for the elderly ismore stable. Another, more important, drawback in using the numberof days as a numerator lies in the inherent suggestion that the patientremains in seclusion for the full 24 h, while in reality there are largevariations of time spent in seclusion per day. Our data showed seclu-sion episodes in the Netherlands usually occurred at largely varyingintervals during the day (means between 2 up to 21 h a day). Thenumber of seclusion hours per 1000 inpatient hours is a refinementin comparison to the seclusion days per 1000 occupied beds.

Seclusion hours per 1000 inpatient hours may be a more sensitivemeasure for short-term changes in the duration of seclusion duringpatient stay. The number of seclusion days per 1000 occupied bedsand the number of seclusion hours per 1000 inpatient hours are veryuseful for internal comparisons and comparisons between wardswithin a single hospital. Nurses can relate these figures to their dailyexperience.

In line with Bowers (2000) we found that the number of seclusionincidents per 1000 admissions is not sensitive for the length of stay.On acute admission wards, the variation in length of stay is smallcompared to other types of wards, but still the mean length of stayvaries from a few days up to around 30 days. The number of patientsat risk to be exposed to coercive measures is small in relation to thenewly admitted patients, allowing the use of this figure for compari-sons between acute psychiatric admission wards. Especially for hospi-tals aiming at the reduction of seclusion and restraint, this rate islikely to be of great relevance. Teams on acute admission wards canuse these rates to evaluate their success in lowering the number of se-clusion incidents. However, using these rates on other types of wardsraises interpretation problems as described above at hospital level.Further studies and more data are necessary for relevant interpreta-tions of these figures in other patient categories like in child and ad-olescent psychiatry, forensic psychiatry, geriatric-psychiatry andintellectual disability care.

The number and duration of seclusion episodes per day is verysensitive to change, and may show effects of ward policies on seclu-sion. In hospital number 2 for example, the ward policy is aimed ondownsizing a seclusion episode as far as possible. Results showedthat patients were in seclusion shorter, on the other hand the somepatients were secluded more often, which led to an increase in seclu-sion episodes per seclusion incident. On other wards the policy was toshorten the seclusion episode depending on whether the patientwould be calm enough and the team able to handle patient's behavioron the ward. These different policies may lead to a lower number of

seclusion episodes within one seclusion incident. These examples il-lustrate the effect of (implicit or explicit) choices on when and howto seclude a patient. On ward level, figures on episodes are helpfulin evaluation of policies.

Determining the relative risk for seclusion between wards in a sin-gle hospital may be helpful in illustrating differences between wards.The relative risk for a patient to be secluded on the respective ward isimportant for nurses to reflect on. This phenomenon may be relatedto the hospital's seclusion and restraint policy whether or not to con-centrate seclusion in one or two wards in the hospital. However, therelative risk cannot be calculated if all seclusion incidents take placeon one specific crisis ward in a hospital. Another problem is that therelative risk reflects changes of the number of seclusions both onthe ward under consideration and on the other wards in the hospital.This may be misleading: if the seclusion figures during a defined pe-riod of time on one ward are stable, while they increase on another,the relative risk decreases. Combining the relative risk with the abso-lute number of seclusion incidents, may provide in our experience apresentation of results in a clear and understandable way for mentalhealth professionals working on these wards.

4.3. Patient level (third level)

On the patient level, figures on age, sex and diagnosis provide adeeper understanding in what groups of patients are exposed to ahigher risk of being secluded in comparison to other patient groupsin a hospital or on a ward. Again, specific ward policies may lead todifferences. Our data confirm the findings of Martin et al. (2005),Martin, Bernhardsgruetter, Goebel, and Steinert (2007), Betemps etal. (1993) andWay and Banks (1990) that, overall, patients diagnosedwith schizophrenia are more likely to be exposed to seclusion. Also,we observed patients in whom a DSM-IV code was absent or in pro-cess had a relative higher chance to be secluded. These patientsoften had too short an admission to obtain any diagnosis.

The differences between wards were remarkable in this respect.Again, relative risk is a simple and understandable parameter to dem-onstrate these differences. Presenting the odds ratio when monitor-ing the wards may be used as a warning sign since high odds ratiosimply problems within a specific group of patients. Focused attentionto this group by the team and management may help to decrease theuse of seclusion.

4.4. Admission (second level) and intervention (first level)

Both levels provide the basic information. The information on ad-mission level is useful to evaluate events during a single patient's ad-mission. It can be used in weekly, monthly or other periodicalevaluations. The information on intervention level is useful for con-trolling the registered coercive measures. It contains information onan individual level but is inappropriate for the use in hospital reportsor scientific publications. In our point of view, seclusion incidents(Heilbrun et al., 1995) have more clinical relevance than (discrete)seclusion episodes (Fisher, 1994, 2003) as they can be related tostaff–patient or patient–patient conflicts. For comparison purposesdata at this level are difficult to interpret. It is, however, informationthat the Dutch Health Care Inspectorate considers highly relevant. Al-though from an individual patient rights perspective this may be seenas justified, as a tool for quality control, this is less the case.

In line with Bowers (2000), we recommend using data covering aperiod of at least 1 year in order to calculate and report meaningfuland comparable rates of incidents. Shorter periods are too sensitivefor seasonal fluctuations and variations due to individual patients.

The comparisons between institutes and wards showed a greatvariability in the seclusion figures, with respect to the incidence ratios(incidents per 1000 admissions or per 100.000 inhabitants) as well asthe prevalence ratios (days per 1000 occupied beds or hours per 1000

437W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438

Author's personal copy

inpatient hours). Such calculations have not been done before andneed further research into the origins of these differences.

5. Conclusion

Bowers (2000) argues that it is important to carefully considerwhich outcome addresses answers on specific research questions.We agree that there is no general solution available. We state that itis important to combine analyses at a macro (hospital) as well as ona micro (the patient) level, in fact the micro level findings are crucialfor data accuracy and bear validity for daily practice at the wards.With respect to our data, we suggest that different levels of analysisneed different numerators and denominators. The distinctions madein this paper can assist researchers to choose methods and presenta-tion of relevant findings carefully.

In the case of internal hospital or ward reports, absolute numbers,number of seclusion hours per 1000 inpatient hours, relative risk andodds ratio are helpful to evaluate patterns of the use of seclusion andrestraint. For acute admission wards only, the number of incidentsper 1000 admissions can be used for evaluation and internal compar-ison goals. For external reports and comparisons, figures on hospitallevel number of incidents per 100.000 inhabitants, number of seclu-sion or restraint days per 1000 occupied bed or, more precise, seclu-sion or restraint hours per 1000 inpatient hours are useful.

To compare the same types of wards between different hospitals,outcomes like seclusion days per 1000 occupied beds or seclusionhours per 1000 inpatient hours, or the relative risk and mean dura-tion of seclusion and restraint per day or episode could be used.Data on the composition of the population of patients and the charac-teristics of secluded and restrained patients is helpful as backgroundinformation to interpret the outcomes on the ward level. Inclusionof such data in a single nationwide database provides the possibilityfor ward to learn from each the variety of coercive measures from dif-ferent perspectives. Information on the admission level is primarilyuseful for individual evaluation goals and for quality control activitiesby the medical director or the inspectorate of health.

We propose to use the definitions, data and analysis techniqueswhich we have presented, in further research and reports to evaluateresults on seclusion and restraint use on different levels. Providing anoverview of the rates of all coercive measures on all the levels men-tioned enables a deeper understanding into the actual use of coercivemeasures. The demonstrated approach can serve as a framework forcomparisons between wards, hospitals and countries.

References

Ahmed, A. G., & Lepnurm, M. (2001). Seclusion practice in a Canadian forensic psychi-atric hospital. The Journal of the American Academy of Psychiatry and the Law, 29,303–309.

Betemps, E. J., Somoza, E., & Buncher, C. R. (1993). Hospital characteristics, diagnoses andstaff reasons associated with use of seclusion and restraint. Hospital and CommunityPsychiatry, 44, 367–371.

Bower, F. L., McCullough, C. S., & Timmons, M. E. (2000). A synthesis of what we knowabout the use of physical restraints and seclusion with patients in psychiatric andacute care settings. The Online Journal of Knowledge Synthesia for Nursing, 7, 2.

Bowers, L. (2000). The expression and comparison of ward incident rates. Issues inMental Health Nursing, 21, 365–374.

Bowers, L., Simpson, A., Alexander, J., Ryan, C., & Carr-Walker, P. (2004). Cultures ofpsychiatry and the professional socialization process: The case of containmentmethods for disturbed patients. Nurse Education Today, 24, 435–442.

Bowers, L., Van der Werf, B., Vokkolainen, A., Muir-Cochrane, E., Allan, T., & Alexander,J. (2007). International variation in attitudes to containment measures for dis-turbed psychiatric inpatients. International Journal of Nursing Studies, 44, 357–364.

Busch, A. B., & Shore, M. F. (2000). Seclusion and restraint: A review of recent literature.Harvard Review of Psychiatry, 8, 261–270.

Convention for the Protection of Human Rights and Fundamental Freedoms.1950. Rome. http://www.echr.coe.int/NR/rdonlyres/D5CC24A7-DC13-4318-B457-5C9014916D7A/0/ENG_CONV.pdf

Crenshaw, W. B., & Francis, P. S. (1995). A national survey on seclusion and restraint instate psychiatric hospitals. Psychiatric Services, 46, 1026–1031.

Fisher, W. A. (1994). Restraint and seclusion: A review of the literature. The AmericanJournal of Psychiatry, 51, 1584–1591.

Fisher, W. A. (2003). Elements of successful restraint and seclusion reduction programsand their application in a large, urban, state psychiatric hospital. Journal of Psychi-atric Practice, 9, 7–15.

Haglund, K., Von Knorring, L., & Von Essen, L. (2003). Forced medication in psychiatriccare: Patient experiences and nurse perceptions. Journal of Psychiatric and MentalHealth Nursing, 10, 65–72.

Heilbrun, K., Golloway, G. G., Shoukry, V. E., & Gustafson, D. (1995). Physical control ofpatients on an inpatient setting: Forensic vs. civil populations. The Psychiatric Quar-terly, 66, 133–145.

Janssen, W. A., Hutschemaekers, G. H. M., & Lendemeijer, H. H. G. M. (2005). Restraintuse quantified [Dwang cijfermatig in beeld]. In T. Abma, G. Widdershoven, & B.Lendemeijer (Eds.), Restraint in psychiatric hospitals: The quality of patients free-domrestricted interventions (pp. 67–76). Dwang en drang in de psychiatrie; kwaliteitvan vrijheidsbeperkende interventies (pp. 67–76). Utrecht: Lemma.

Janssen, W. A., Noorthoorn, E. O., de Vries, W. J., Hutschmaekers, G. H. M., Lendemeijer,H. H. G. M., &Widdershoven, G. A. M. (2008). The use of seclusion in the Netherlandscompared to countries in and outside Europe. International Journal of Law and Psychi-atry, 31, 463–470.

Janssen, W. A., Noorthoorn, E. O., van Linge, R., & Lendemeijer, H. H. G. M. (2007). Theinfluence of staffing levels on the use of seclusion. International Journal of Law andPsychiatry, 30, 118–126.

Kaltiala-Heino, R., Korkeila, J., Tuohimäki, C., Tuori, T., & Lehtinen, V. (2000). Coercionand restrictions in psychiatric inpatient treatment. European Psychiatry, 15,213–219.

Kaltiala-Heino, R., Tuohimaki, C., Korkeila, J., & Lehtinen, V. (2003). Reasons for usingseclusion and restraint in psychiatric inpatient care. International Journal of Lawand Psychiatry, 26, 139–149.

Legemaate, J., Frederiks, B. J. M., & de Roode, R. P. (2007). International developments[Internationale ontwikkelingen]. Rijswijk: Ministerie van Volksgezondheid, Welzijnen Sport.

Lendemeijer, B., & Shortridge-Baggett, L. (1997). The use of seclusion in psychiatry: Aliterature review. Scholarly Inquiry for Nursing Practice, 11, 299–315.

Martin, V., Bernhardsgruetter, R., Goebel, R., & Steinert, T. (2007). The use of mechan-ical restraint and seclusion in patients with schizophrenia: A comparison of thepractice in Germany and Switzerland. Clinical Practice and Epidemiology in MentalHealth, 3, 1.

Martin, V., Kuster, W., Baur, M., Bohnet, U., Hermelink, G., Knopp, M., et al. (2005). In-cidence of coercive measures as an indicator of quality in psychiatric hospitals:Problems of data recording and processing, preliminary results of a benchmarkingstudy [DieUinzidenz von Zwangsmassnahmen als qualitätsindikator in psychia-trischenKkliniken: Probleme der Datenerfassung und –verarbeitung und ersteErgebnisse]. Psychiatrische Praxis, 32, 1–9.

Mason, T., & Alty, A. (1994). Seclusion and mental health: A break with the past. London:Chapman & Hall.

Sailas, E., & Fenton, M. (2003). Seclusion and restraint for people with serious mentalillnesses. (Cochrane Review). In: The Cochrane Libary, Issue 1 2003. Oxford.

Steinert, T., & Lepping, P. (2008). Legal provisions and practice in the management ofviolent patients. A case vignette study in 16 European countries. European Psychi-atry, 24, 135–141.

Steinert, T., Lepping, P., Bernhardsgrütter, R., Conca, A., Hatling, T., Janssen, W., et al.(2009). Incidence of seclusion and restraint in psychiatric hospitals: An interna-tional comparison. Social Psychiatry and Psychiatric Epidemiology, doi:10.1007/s00127-009-0132-3.

Steinert, T., Martin, V., Baur, M., Bohnet, U., Goebel, R., Hermelink, G., et al. (2006).Diagnosis-related frequency of compulsory measures in 10 German psychiatrichospitals and correlates with hospital characteristics. Social Psychiatry and Psychi-atry Epidemiology, doi:10.1007/s00127-006-0137-0.

Terluin, I. J., Slangen, L. H. G., van Leeuwen, E. S., Oskam, A. J., & Gaaff, A. (2005). Therural economy in the Netherlands, an exploration of definitions, indicators, institutionsand policies [De plattelandseconomie in Nederland; Een verkenning van definities,indicatoren, instituties en beleid]. Den Haag: LEI http://www.lei.dlo.nl/publicaties/PDF/2005/4_xxx/4_05_04.pdf

Way, B. B., & Banks, S. M. (1990). Use of seclusion and restraint in public psychiatrichospitals: Patient characteristics and facility effects. Hospital & Community Psychi-atry, 41, 75–81.

438 W.A. Janssen et al. / International Journal of Law and Psychiatry 34 (2011) 429–438