Patterns of Seclusion/Restraint Patterns of Seclusion/Restraint Use at a State Psychiatric Use at a State Psychiatric
HospitalHospital
Using Multivariate Statistical Using Multivariate Statistical Models to Identify Consumers at Models to Identify Consumers at
High RiskHigh Risk
Seclusion/Restraint Use at Fulton Seclusion/Restraint Use at Fulton State Hospital State Hospital
Fulton State Hospital (1856)Fulton State Hospital (1856) Maximum, intermediate and minimum security facilitiesMaximum, intermediate and minimum security facilities It is a good proxy measure of serious aggressionIt is a good proxy measure of serious aggression We have good data on S/RWe have good data on S/R 30% of all patient injuries are directly or indirectly related to 30% of all patient injuries are directly or indirectly related to
S/R useS/R use 70% of all staff injuries are directly or indirectly related to S/R 70% of all staff injuries are directly or indirectly related to S/R
useuse Patients who are ‘high utilizers’ of S/R are both dangerous to Patients who are ‘high utilizers’ of S/R are both dangerous to
self and dangerous to others self and dangerous to others Injuries can occur to either patients or staff during the Injuries can occur to either patients or staff during the
aggressive incident which triggers S/R use in the first aggressive incident which triggers S/R use in the first place….place….
Or during the containment process which involves applying S/ROr during the containment process which involves applying S/R Collateral injuries can also occur!Collateral injuries can also occur!
Research on Seclusion/RestraintResearch on Seclusion/Restraint
Hundreds of studies; a multitude of topicsHundreds of studies; a multitude of topics Characteristics of patients most likely to Characteristics of patients most likely to
be secluded or restrained (building a be secluded or restrained (building a ‘profile’ of a typical ‘high risk’ patient)‘profile’ of a typical ‘high risk’ patient)
Hold forth the possibility of Hold forth the possibility of early early identification and interventionidentification and intervention
Disrupt the cycle of violence?Disrupt the cycle of violence?
Review of the LiteratureReview of the Literature
Way and Banks (1990)Way and Banks (1990) 23 Psychiatric Hospitals in the State of 23 Psychiatric Hospitals in the State of
New YorkNew York Reviewed seclusion/restraint records over Reviewed seclusion/restraint records over
a four week perioda four week period 657 patients secluded/restrained, 22, 929 657 patients secluded/restrained, 22, 929
not secluded restrainednot secluded restrained Younger, female, involuntarily hospitalized Younger, female, involuntarily hospitalized
ContinuedContinued
Jonikas et. al. (2004)Jonikas et. al. (2004) Three psychiatric units at a University Three psychiatric units at a University
HospitalHospital Aggregated seclusion/restraint data by Aggregated seclusion/restraint data by
ward, then plotted data over timeward, then plotted data over time Examined the results of an intervention Examined the results of an intervention
program and documented decreases program and documented decreases associated with program introductionassociated with program introduction
Traditional Approaches to S/R data Traditional Approaches to S/R data analysisanalysis
Aggregate (compute means) of S/R data over Aggregate (compute means) of S/R data over patients, wards, time periodspatients, wards, time periods
Compare and contrast groups of patients Compare and contrast groups of patients formed in this manner, for instance….formed in this manner, for instance….
Patients secluded or restrained during a Patients secluded or restrained during a specified time period vs. those not secluded or specified time period vs. those not secluded or restrained restrained
Use of arbitrary cut-off scores to create groupsUse of arbitrary cut-off scores to create groups Plot aggregated S/R data over time….by ward, Plot aggregated S/R data over time….by ward,
treatment program, treatment facility treatment program, treatment facility
An alternative An alternative approach….trajectories of changeapproach….trajectories of change
Traditional approaches have their place, but….Traditional approaches have their place, but…. Rather than immediately aggregating or averaging S/R Rather than immediately aggregating or averaging S/R
data….which obscures individual differencesdata….which obscures individual differences An alternative strategy is to look at individual patient data An alternative strategy is to look at individual patient data
sets, which consist of day-by-day records of the incidence sets, which consist of day-by-day records of the incidence or frequency of S/R use or frequency of S/R use
NotNot aggregated over patients, a single time period, aggregated over patients, a single time period, treatment units, facilities, etc.treatment units, facilities, etc.
Thus, the fundamental units of analysis consists of a Thus, the fundamental units of analysis consists of a series of individual patients’ ‘trajectories’ series of individual patients’ ‘trajectories’
Not Not defined based on arbitrary cut-off scores, or a single defined based on arbitrary cut-off scores, or a single value derived from aggregating seclusion/restraint data value derived from aggregating seclusion/restraint data from an arbitrarily selected time framefrom an arbitrarily selected time frame
Data Analysis Methods: A Two Data Analysis Methods: A Two phase processphase process
Looking at data consisting of trajectories Looking at data consisting of trajectories requires a non-traditional statistical approach requires a non-traditional statistical approach (can’t use t-tests!)(can’t use t-tests!)
Growth mixture modeling/Latent class or profile Growth mixture modeling/Latent class or profile analysisanalysis
Phase 1: Find classes or ‘groups’ of subjects Phase 1: Find classes or ‘groups’ of subjects with relatively homogeneous trajectories (if they with relatively homogeneous trajectories (if they actually exist!) actually exist!)
Phase 2: Look for class characteristics that Phase 2: Look for class characteristics that could be used for the purposes of early could be used for the purposes of early identification (and therefore intervention) identification (and therefore intervention)
Current StudyCurrent Study
Fulton State Hospital Fulton State Hospital 622 patients622 patients Admitted after September 2001Admitted after September 2001 At least two months in hospitalAt least two months in hospital S/R data over a two year period, broken down S/R data over a two year period, broken down
into 12 , two-month time intervalsinto 12 , two-month time intervals So imagine yourself this evening, with 622 So imagine yourself this evening, with 622
individual subject graphs laid out on the kitchen individual subject graphs laid out on the kitchen table, asking the question ‘Is there a way I can table, asking the question ‘Is there a way I can sort these into meaningful groups?’sort these into meaningful groups?’
AverageAverage Number of S/R Episodes Per Number of S/R Episodes Per Client (traditional approach)Client (traditional approach)
0
2
4
6
8
10
12
14
S/R
Epis
od
es
Average Number of S/R Average Number of S/R Episodes, by ClassEpisodes, by Class
Average Number of S/R Average Number of S/R Episodes, by ClassEpisodes, by Class
Average Number of S/R Average Number of S/R Episodes, by ClassEpisodes, by Class
ResultsResults
Very strong evidence for the existence of three Very strong evidence for the existence of three ‘classes’ of patients (not just the use of an ‘classes’ of patients (not just the use of an arbitrary cut-off to create ‘groups’)arbitrary cut-off to create ‘groups’)
High-Medium-Low (7%, 23%,70%) High-Medium-Low (7%, 23%,70%) Patients get ‘into’ their respective classes very Patients get ‘into’ their respective classes very
quickly, certainly within the first few months quickly, certainly within the first few months High class consists of only 41 out of 622 patientsHigh class consists of only 41 out of 622 patients High and medium classes appear to diminish in High and medium classes appear to diminish in
terms of S/R use over time, but even after two terms of S/R use over time, but even after two years, they remain somewhat distinct from the low years, they remain somewhat distinct from the low class class
Injury Data and Class MembershipInjury Data and Class Membership
Class Membership (n)Class Membership (n) Mean Injuries During Mean Injuries During Hospital Course (SD)Hospital Course (SD)
Class 1Class 1 (443) (443) 0.87 0.87 (3.05)(3.05)
Class 2Class 2 (138) (138) 9.18 9.18 (11.0)(11.0)
Class 3Class 3 (41) (41) 25.24 25.24 (20.90)(20.90)
Class 3 Patients 75 times more Class 3 Patients 75 times more likely to be Physically Abusedlikely to be Physically Abused
Class Membership Class Membership (N=622)(N=622)
Physical Abuse (N=18)Physical Abuse (N=18)
Class 1Class 1 (443) (443) .2% .2% (1/443)(1/443)
Class 2Class 2 (138) (138) 8% 8% (11/138)(11/138)
Class 3Class 3 (41) (41) 15% 15% (6/41)(6/41)
Risks associated with Risks associated with Trajectory Class Three Trajectory Class Three
Membership Membership
29 times more likely to be injured or cause 29 times more likely to be injured or cause an injuryan injury
7 times more likely to be abused/neglected7 times more likely to be abused/neglected 75 times more likely to be physically 75 times more likely to be physically
abusedabused
Past History of Abuse/Neglect in Past History of Abuse/Neglect in Childhood/Adolescence Childhood/Adolescence
Sexual Abuse Physical Abuse Psychological Abuse
Class 1 24% 32% 19%
Class 2 25% 30% 10%
Class 3 54% 51% 20%
Past History of Aggressive Past History of Aggressive Behavior in InstitutionsBehavior in Institutions
Class Membership (n)Class Membership (n) Mean Score on 5-point, Mean Score on 5-point, Likert scaleLikert scale
Class 1Class 1 (443) (443) 1.94 (1.40) ‘seldom’1.94 (1.40) ‘seldom’
Class 2Class 2 (138) (138) 3.25 (1.63) 3.25 (1.63) ‘Occasionally’‘Occasionally’
Class 3Class 3 (41) (41) 4.34 (1.24) ‘Frequently’4.34 (1.24) ‘Frequently’
Phase 2: Predictors of Class Phase 2: Predictors of Class Membership upon AdmissionMembership upon Admission
Age (younger = higher risk)Age (younger = higher risk) Admission code (voluntary by guardian, Admission code (voluntary by guardian,
transfers from jail / DOC)transfers from jail / DOC) Previous hospitalizationPrevious hospitalization Marital status (never married)Marital status (never married) Employment history (never worked)Employment history (never worked) Diagnoses: Borderline PD, Antisocial PD, & Diagnoses: Borderline PD, Antisocial PD, &
Intermittent Explosive Disorder; Alcohol/Drug Intermittent Explosive Disorder; Alcohol/Drug (protective)(protective)
Accuracy of PredictionAccuracy of Predictionat Admissionat Admission
Predicted Predicted classclass
Actual classActual class LowLow MediumMedium HighHigh
LowLow 418418 2323 11
MediumMedium 9393 3636 66
HighHigh 1111 1919 1010
•N = 617 clients; 75.2% hit rate
Predictors of Class Membership after Predictors of Class Membership after Two MonthsTwo Months
First two months of S/R useFirst two months of S/R use Admission code (voluntary by guardian, Admission code (voluntary by guardian,
transfers from jail / DOC) transfers from jail / DOC) Age (younger age)Age (younger age) Previous hospitalizationPrevious hospitalization Marital status (never married)Marital status (never married) Intermittent Explosive DisorderIntermittent Explosive Disorder
Accuracy of Prediction Accuracy of Prediction after Two Monthsafter Two Months
Predicted Predicted classclass
Actual classActual class LowLow MediumMedium HighHigh
LowLow 438438 55 00
MediumMedium 4848 8181 99
HighHigh 33 77 3131
•N = 622 clients; 88.4% hit rate
Implications for Policy, Patient Care, & Implications for Policy, Patient Care, & Facility DesignFacility Design
Use predictive equations to categorize new admissions (or currently Use predictive equations to categorize new admissions (or currently hospitalized patients) and thereby permit early identification of hospitalized patients) and thereby permit early identification of intramural & treatment needsintramural & treatment needs
Develop interventions that are specific to individuals in the High Develop interventions that are specific to individuals in the High Class (administrative segregation, diversion to other facilities?)Class (administrative segregation, diversion to other facilities?)
Provide specialized training for staff who work with these individuals Provide specialized training for staff who work with these individuals (training in dispute resolution skills?)(training in dispute resolution skills?)
Fast-track those who are in the Low ClassFast-track those who are in the Low Class Inform security continuum adaptations corresponding to future Inform security continuum adaptations corresponding to future
physical plant changesphysical plant changes Keep in mind the distinction between placement decisions (which Keep in mind the distinction between placement decisions (which
are limited by the current architecture of the security continuum on are limited by the current architecture of the security continuum on the FSH campus) and treatment/transfer decisions the FSH campus) and treatment/transfer decisions
……as well as the distinction between extramural vs. intramural as well as the distinction between extramural vs. intramural security needs (e.g.., need for a secure perimeter, as opposed to security needs (e.g.., need for a secure perimeter, as opposed to internal security issues)internal security issues)