TAWS PAPER 02 OCCUPANCY, ADMISSIONS, CONFLICT AND CONTAINMENT

Similar documents
SPARRA Mental Disorder: Scottish Patients at Risk of Readmission and Admission (to psychiatric hospitals or units)

Northwick Park Mental Health Centre Smoking Cessation Report October Plan. Act. Study. Introduction

Sonali Wayal, Gwenda Hughes, Pam Sonnenberg, Hamish Mohammed, Andrew J Copas, Makeda Gerressu, Clare Tanton, Martina Furegato, Catherine H Mercer

National Child Measurement Programme Changes in children s body mass index between 2006/07 and 2010/11

Finalised Patient Reported Outcome Measures (PROMs) in England

Mental Health Statistics, to

Organ Donation and Transplantation data for Black, Asian and Minority Ethnic (BAME) communities. Report for 2016/2017 (1 April March 2017)

SBIRT IOWA. Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION. Iowa Army National Guard

NHS Diabetes Prevention Programme (NHS DPP) Non-diabetic hyperglycaemia. Produced by: National Cardiovascular Intelligence Network (NCVIN)

Inpatient Mental Health Staff Morale: a National Investigation

Trends in Hospital Admissions For Diabetes Complications

Organ Donation and Transplantation data for Black, Asian and Minority Ethnic (BAME) communities. Report for 2017/2018 (1 April March 2018)

Downloaded from:

Iowa Army National Guard Biannual Report April 2016

SBIRT IOWA. Iowa Army National Guard THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION. Iowa Army National Guard

Statistics is the science of collecting, organizing, presenting, analyzing, and interpreting data to assist in making effective decisions

Organ Donation and Transplantation data for Black, Asian and Minority Ethnic (BAME) communities. Report for 2015/2016 (1 April March 2016)

Suicide Facts. Deaths and intentional self-harm hospitalisations

Economic study type Cost-effectiveness analysis.

STATISTICS 8 CHAPTERS 1 TO 6, SAMPLE MULTIPLE CHOICE QUESTIONS

SBIRT IOWA THE IOWA CONSORTIUM FOR SUBSTANCE ABUSE RESEARCH AND EVALUATION. Iowa Army National Guard. Biannual Report Fall 2015

AVELEY MEDICAL CENTRE & THE BLUEBELL SURGERY

Statistics is the science of collecting, organizing, presenting, analyzing, and interpreting data to assist in making effective decisions

Somerset Phoenix Project: Self-request for support

BROMLEY JOINT STRATEGIC NEEDS ASSESSMENT Substance misuse is the harmful use of substances (such as drugs and alcohol) for non-medical purposes.

National Audit of Dementia

QStatin Update Information

2016 Children and young people s inpatient and day case survey

Substance misuse among young people The data for

Awareness and Beliefs about Cancer (ABC) measure Final version in UK English

Hazardous drinking in 2011/12: Findings from the New Zealand Health Survey

Workplace stress in South African mineworkers

Author's response to reviews

Chronic Hepatitis C The Patient s Perspective

DMRI Drug Misuse Research Initiative

Annex A: Estimating the number of people in problem debt while being treated for a mental health crisis

Cannabis use and adverse outcomes in young people: Summary Report

Mental Health in STH Mike Richmond, Medical Director Mark Cobb, Clinical Director of Professional Services Debate & Note

Dual Diagnosis. Themed Review Report 2006/07 SHA Regional Reports East Midlands

Section 6: Analysing Relationships Between Variables

Alcohol-related Hospital Statistics Scotland 2011/12

NATIONAL INSTITUTE FOR HEALTH AND CLINICAL EXCELLENCE

Drug Recovery Wing pilots programme: a note of advice to the Department of Health on the proposed outcomes evaluation

Discontinuation and restarting in patients on statin treatment: prospective open cohort study using a primary care database

Prescribing for substance misuse: alcohol detoxification. Clinical background

Worcestershire Dementia Strategy

The varying influence of socioeconomic deprivation on breast cancer screening uptake in London

Prescribing of high-dose and combination antipsychotics on adult acute and intensive care wards: Clinical introduction, methodology and glossary.

Types of data and how they can be analysed

SMART Wokingham Young persons Screening and Referral Form

National Drug and Alcohol Treatment Waiting Times

Client ID Number. If no, please tick as appropriate No claim in place Not eligible Employed Sanctioned

BRIEF REPORT OPTIMISTIC BIAS IN ADOLESCENT AND ADULT SMOKERS AND NONSMOKERS

THE ANALYSES TO DETERMINE THE RELATIONSHIP BETWEEN SLEEPING PROBLEMS AND THE HEALTH OUTCOMES OF THE ELDER PEOPLE

National NHS patient survey programme Community Mental Health Survey: Quality and Methodology

FATHER ABSENCE AND DEPRESSIVE SYMPTOMS IN ADOLESCENT GIRLS FROM A UK COHORT

65G Documentation and Notification.

SCOTTISH STROKE CARE AUDIT DATA COLLECTION QUICK NOTES

Children and Young People - Community Health Services Procurement. NHS Bromley Clinical Commissioning Group

Rebbecca Aust and Nicola Smith

MICHAEL PRITCHARD. most of the high figures for psychiatric morbidity. assuming that a diagnosis of psychiatric disorder has

Douglas County s Mental Health Diversion Program

Exploring the Relationship Between Substance Abuse and Dependence Disorders and Discharge Status: Results and Implications

National Cancer Patient Experience Survey Results. University Hospitals of Leicester NHS Trust. Published July 2016

Consultation on revised threshold criteria. December 2016

The Coventry Wellbeing Report

Making a dementia diagnosis in areas of cultural diversity

Referral trends in mental health services for adults with intellectual disability and autism spectrum disorders

National Cancer Patient Experience Survey Results. East Kent Hospitals University NHS Foundation Trust. Published July 2016

Analysis of Confidence Rating Pilot Data: Executive Summary for the UKCAT Board

Review of Veterinary Epidemiologic Research by Dohoo, Martin, and Stryhn

Alcohol burden on Accident & Emergency Departments: Implications for policy and practice. Dr. Tom Phillips, PhD. The MCA Symposium 16 th November 2016

Help us prioritise research into alcohol-related liver disease

Appendix C. Aneurin Bevan Health Board. Smoke Free Environment Policy

F2: ORS Scores for a single session sessions

Statistics on Drug Misuse: England, 2007

Clozapine in community practice: a 3-year follow-up study in the Australian Capital Territory Drew L R, Hodgson D M, Griffiths K M

Delirium. Quick reference guide. Issue date: July Diagnosis, prevention and management

PROCEDURE Mental Capacity Act. Number: E 0503 Date Published: 20 January 2016

Determinants of hospital length of stay for people with serious mental illness in England and implications for payment systems: a regression analysis

The Nottingham eprints service makes this work by researchers of the University of Nottingham available open access under the following conditions.

Of those with dementia have a formal diagnosis or are in contact with specialist services. Dementia prevalence for those aged 80+

PREVALENCE OF CONDUCT DISORDER IN PRIMARY SCHOOL CHILDREN OF RURAL AREA Nimisha Mishra 1, Ambrish Mishra 2, Rajeev Dwivedi 3

Quartely Report. Ethnithicity? Ethnicity? Page 1 of 18. Any other Ethnic Group. Asian or Asian British any Other. Asian or Asian British Bangladeshi

Prepared by: Assoc. Prof. Dr Bahaman Abu Samah Department of Professional Development and Continuing Education Faculty of Educational Studies

Northern Ireland Registry of Self-Harm Western Area

Part 8 Logistic Regression

Have your say: what do you think about putting fluoride in the water? 1. Do you agree or disagree with the following statements?

Summary of the Dental Results from the GP Patient Survey; July to September 2014

The audit is managed by the Royal College of Psychiatrists in partnership with:

Child and Adolescent Mental Health Service (CAMHS)

Prescribing for substance misuse: alcohol detoxification

Men Behaving Badly? Ten questions council scrutiny can ask about men s health

RESEARCH. Predicting risk of type 2 diabetes in England and Wales: prospective derivation and validation of QDScore

Statistics on smoking: England, 2006

out of crisis and into recovery our community-based approach

Overweight and obesity: where are we and where are we heading?

State of Iowa Outcomes Monitoring System

Northumberland, Tyne and Wear NHS Foundation Trust. Board of Directors Meeting

Transcription:

TAWS PAPER 02 OCCUPANCY, ADMISSIONS, CONFLICT AND CONTAINMENT AUTHORS: Len Bowers DATE FINALISED: 18 December 2007 VERSION NUMBER: 1 CHANGES FROM PREVIOUS VERSIONS: not applicable INTERNET PAPERS This series was developed to report and make available findings from the analysis of the dataset that in themselves did not merit submission in a peer reviewed scientific journal, but may prove useful to other researchers. Full listings of published papers, and a copy of the final report to project funders, can be found on www.citypsych.com, under research/tompkins Acute Ward Study. BACKGROUND A longitudinal study was conducted on 14 acute inpatient psychiatric wards and three Psychiatric Intensive Care Units in a single NHS Trust. The study Trust served a population of 650,000 in three inner London boroughs, each of which had high proportions of ethnic minority residents (approximately 60%, compared to the England and Wales average of 12%), and high levels of social deprivation (all fell within the category of the 10% most deprived areas in the country). Data were drawn from officially collected information on admissions, adverse incidents, workforce deployment and training; researcher collected information included end of shift reports from the wards, repeated interviews of patients, ward managers and consultant psychiatrists, and repeated waves of questionnaires from patients and ward staff. The study was undertaken in two phases, the first retrospective and utilising officially collected data only (2002-04), the second prospective and including both researcher and officially collected data (2004-06). The study brought together four years of officially collected data on adverse incidents on the wards and patient admissions/discharges; two years of prospectively collected research data on conflict and containment on the wards, composed of approximately 15,000 end of shift reports, 119 patient interviews, 77 ward manager interviews, and 43 consultant psychiatrist interviews. In addition a number of questionnaires were collected on a repeated basis throughout the study. Overall response rates for the prospectively collected data were fair, with 45% of the potential total end of shift reports collected. Precise estimates of response rates for other items are not possible to provide, as numbers of staff in post fluctuated over time during the study. For interviews of patients, a 93% response rate was attained, as replacements were sampled for those patients who did not wish to participate. The same process could 1

not be used for staff as, for example, each ward had only one ward manager. Nevertheless the response from ward managers was excellent, with only a few missed interviews (96% reponse rate), whereas interviews with consultant psychiatrists were much more difficult to obtain (46% response rate). These data are complex, cover a large period of time, and can be analysed in many different ways, including: computer-aided content analyses of interviews; cross sectional time-series analysis; pooled cross sectional analysis; mixed method, multiple case comparison; a contrast analysis between patients who receive different types of care logistic regression; and many more. A full account of the methodology can be found in the final report to the funders, on www.citypsych.com. That report includes definitions and descriptions of all variables, with means, standard deviations, etc. for the sample. AIM To assess the relationship of conflict and containment (top and mid range variables) to patient admission and occupancy characteristics, utilising shifts and weeks as the time segments. INSTRUMENTS The Patient-staff Conflict checklist, and officially collected patient data from the Trust's Patient Administration System. See the project final report for more details of both. ANALYSIS PCC data was available for 14,114 valid shifts (both date and shift present). Patient data was imported into MS Access and queried to produce tables of admissions and occupancy by shift. This was then exported and merged with the PCC data in STATA. The patient data available included age, gender, diagnosis and ethnicity. A second file was created by collapsing the former file by ward and week, controlling for shift type, and discarding any weeks for which there were fewer than seven PCCs. This resulted in a file consisting of 954 ward weeks of data. Conflict and containment counts in this dataset were collapsed into means and rounded to provide counts suitable for analysis. Spearman correlations were used for univariate analysis, as dependent variables were skewed. Stepwise Poisson regression was used to model each conflict and containment variable, admissions and occupancy being modelled separately because of collinearity between the two. Other details of the analysis are provided together with the findings below. 2

FINDINGS Conflict and containment Rates of different forms of conflict and containment over the study period are presented in Table 1, based on 15,006 PCC-SRs collected. These rates are very close to the national norms from the City-128 study of 136 acute psychiatric wards, indicating that the study wards are representative in relation to their rates of conflict and containment events. 3

Table 1. Mean rate of conflict and containment events per shift (excluding self-harm). Not standardised to occupancy. Mean SD Verbal aggression 0.58 1.18 Physical aggression against objects 0.11 0.44 Physical aggression against others 0.08 0.43 Total aggression 0.77 1.65 Smoking in a no smoking area 0.63 1.16 Refusing to eat 0.19 0.48 Refusing to drink 0.08 0.32 Refusing to attend to personal hygiene 0.39 0.88 Refusing to get up and out of bed 0.16 0.51 Refusing to go to bed 0.16 0.57 Refusing to see workers 0.03 0.21 Total rule breaking 1.64 2.46 Alcohol use (suspected or confirmed) 0.09 0.36 Other substance misuse (suspected or confirmed) 0.10 0.40 Total substance use 0.19 0.62 Attempting to abscond 0.18 0.49 Absconding (missing without permission) 0.06 0.28 Absconding (official report) 0.04 0.24 Total absconding 0.28 0.72 Refused regular medication 0.20 0.47 Refused PRN medication 0.10 0.38 Demanding PRN medication 0.35 0.75 Total medication related 0.66 1.14 Given PRN medication (psychotropic) 0.65 0.97 Given IM medication (enforced) 0.05 0.25 Sent to PICU or ICA 0.01 0.09 Seclusion 0.02 0.15 Special observation (intermittent) 0.52 1.13 Special observation (constant) 0.20 0.56 Show of force 0.08 0.40 Physically restrained 0.05 0.30 Time out 0.12 0.45 Total conflict 3.61 4.40 Total containment 1.71 2.13 When standardised to 20 patient occupancy, these figures are slightly higher, but remain broadly comparable to City 128 norms. All these variables, including the totals, are highly skewed, as most shifts have few or no incidents of conflict, and shifts with more conflict/containment incidents are 4

progressively rarer. As examples, the histograms for total conflict and total containment per shift are displayed in Figures 1 and 2 below. Figure 1. Histogram of total conflict during the shift. Density 0.05.1.15.2.25 0 20 40 60 80 100 TOTFLICT Figure 2. Histogram of total containment during the shift. 5

Density 0.1.2.3.4 0 10 20 30 40 TOTTAIN Patient characteristics Table 2 shows the admission and occupancy data that was matched to the validly dated PCC data (n = 14,114 shifts). The largest single group were patients with schizophrenia. As their proportion of the occupancy was larger than that of admissions, this indicates they had longer than average stays. The next largest group was people with affective disorders, who tended to have shorter than average lengths of stay. Nearly one in ten admissions was for a disorder caused by substance use, but length of stay for this group was shorter. No other group comprised more than 5% of admissions or occupancy. The mean age of admissions was 36.52 years (sd 3.85), and 60.56% were male. Table 3 displays patients' ethnicity, and shows that less than one third of patients were white British, the other main groups being African, other white, other black, Bangladeshi and Caribbean. Differences in proportions between admissions and occupancy are not as large as in the diagnosis table, but a tendency for some (but not all) ethnic minority groups to have had slightly longer stays, and white patients to have had shorter stays is visible. These differences may, of course, be accounted for by demographic and diagnostic differences between ethnic groups. Table 2 Diagnosis 6

Occupancy and admissions from data matched to valid shift PCCs Diagnosis (ICD-10) Admissions Occupancy (days) n % n % F00s (Organic) 4 0.21% 2159 1.04% F10s (Disorders due to substance use) 156 8.26% 8190 3.93% F20s (Schizophrenia) 888 47.01% 131784 63.21% F30s (Affective disorder) 631 33.40% 53408 25.62% F40s (Neurotic disorders) 67 3.55% 2180 1.05% F50s (Physiologically caused) 4 0.21% 875 0.42% F60s (Personality disorder) 75 3.97% 5069 2.43% F70s (Mental retardation) 0 0.00% 19 0.01% F80s (Developmental disorder) 18 0.95% 1030 0.49% F90s (Child/Adolescent onset disorder) 1 0.05% 24 0.01% Other 34 1.80% 1632 0.78% No diagnosis recorded 11 0.58% 2104 1.01% Total 1889 100.00% 208474 100.00% Table 3 Ethnicity Occupancy and admissions from data matched to valid shift PCCs Ethnicity Admissions Occupancy (days) n % n % British 619 32.68% 57680 27.67% African 243 12.83% 29986 14.38% Any other White background 180 9.50% 18363 8.81% Any other Black background 167 8.82% 24888 11.94% Bangladeshi 165 8.71% 23592 11.32% Caribbean 142 7.50% 18590 8.92% Indian 69 3.64% 6106 2.93% Not stated 57 3.01% 3081 1.48% Any other Asian background 55 2.90% 4073 1.95% Any other ethnic group 43 2.27% 5022 2.41% Irish 43 2.27% 3537 1.70% Pakistani 38 2.01% 4523 2.17% White and Black Caribbean 23 1.21% 3398 1.63% White and Black African 16 0.84% 1352 0.65% Any other mixed background 13 0.69% 892 0.43% White and Asian 11 0.58% 1778 0.85% Chinese 10 0.53% 1613 0.77% Total 1894 100.00% 208474 100.00% Relationship of conflict and containment to patient characteristics 7

Both shift type, occupancy and admissions were related to conflict and containment, as shown in Tables 4 and 5. The only variable not associated with shift type was selfharm, although this is a tentative result as this item was measured in two different ways at different times during the study. However self-harm in the City-128 study also behaved differently from other conflict variables. Most of these events are lower during the night shift, except for medication related conflict, and total containment. All variables except self-harm are significantly associated with occupancy, i.e. the number of patients present on the ward on each shift who are therefore available to perform these behaviours or be subject to containment. However the direction of these associations is not always positive, with total containment and total aggression being inversely associated with occupancy. Admissions during the shift were also significantly related to conflict and containment, except for rule breaking. Untangling these relationships is complex because wards are of different sizes, or to state this another way occupancy is confounded by ward, and ward associations with conflict and containment could appear in the analysis as occupancy relationships with conflict and containment. Table 4 Mean rates of conflict and containment by shift type, with Kruskal-Wallis equality of proportions test Shift Total Total rule aggression breaking Total substance use Total absconding Total medication related Total selfharm (z score) Total conflict Total containment Morning 0.816 1.952 0.174 0.329 0.508 0.001 3.851 1.635 Afternoon 0.937 1.648 0.228 0.351 0.494 0.019 3.734 1.688 Night 0.587 1.339 0.173 0.172 0.951-0.022 3.280 1.808 Chi square 64.41 104.01 14.8 129.41 331.41 0.86 18.46 61.98 df 2 2 2 2 2 2 2 2 p <0.001 <0.001 <0.001 <0.001 <0.001 0.576 <0.001 <0.001 Table 5 Spearman correlations between conflict and containment and occupancy and admissions Conflict & containment Occupancy Admissions rho p rho p Total aggression -0.019 <0.001 0.030 <0.001 Total rule breaking 0.120 <0.001 0.002 0.765 Total substance use 0.101 <0.001 0.023 0.006 Total absconding 0.078 <0.001 0.049 <0.001 Total medication related 0.094 <0.001 0.016 0.063 Total self-harm (z score) 0.006 0.445 0.016 0.055 Total conflict 0.104 <0.001 0.024 0.005 Total containment -0.067 <0.001 0.043 <0.001 Overleaf are Tables 6 and 7, showing the univariate relationships between patient characteristics, and the conflict and containment variables. Admission characteristics 8

are associated with other variables, most strangle with total containment and with total absconding. Occupancy variables are much more strongly related to conflict and containment, however this analysis is problematic because most of the occupancy variables are themselves associated with total occupancy during the shift, and total occupancy is the number of people who are available to be aggressive, abscond etc. 9

Table 6 Univariate associations by Spearman correlations between conflict and containment variables, and admissions and their characteristics. Significant associations are highlighted in red. Total medication Total self-harm (z Total aggression Total rule breaking Total substance use Total absconding related score) Total conflict Total containment r p r p r p r p r p r p r p r p Number of admissions during shift 0.030 0.000 0.003 0.766 0.023 0.006 0.050 0.000 0.016 0.063 0.016 0.055 0.023 0.005 0.043 0.000 Number of male admissions 0.021 0.011-0.002 0.809 0.011 0.200 0.033 0.000 0.011 0.179 0.034 0.000 0.014 0.087 0.037 0.000 African 0.004 0.664-0.006 0.464 0.000 0.973 0.023 0.006 0.002 0.816 0.013 0.136-0.002 0.814 0.014 0.092 Any other Asian background -0.019 0.027-0.027 0.001-0.003 0.740 0.008 0.350-0.001 0.860 0.029 0.001-0.028 0.001 0.024 0.005 Any other Black background 0.023 0.005 0.000 0.964 0.009 0.264 0.000 0.994-0.005 0.580 0.005 0.555 0.011 0.174 0.012 0.138 Any other White background 0.012 0.139 0.009 0.272 0.005 0.581 0.020 0.018 0.003 0.704 0.014 0.092 0.014 0.095-0.013 0.127 Any other ethnic group 0.020 0.020 0.003 0.716 0.002 0.804 0.001 0.946-0.006 0.500 0.012 0.149 0.005 0.573-0.009 0.309 Any other mixed background -0.002 0.824 0.005 0.520 0.016 0.055 0.015 0.070-0.007 0.378-0.005 0.519 0.000 0.988 0.001 0.876 Bangladeshi 0.006 0.507 0.009 0.274 0.014 0.090 0.002 0.857 0.004 0.624 0.023 0.005 0.006 0.471 0.032 0.000 British 0.011 0.207 0.003 0.728 0.013 0.111 0.030 0.000 0.021 0.014-0.011 0.191 0.019 0.024 0.027 0.001 Caribbean 0.019 0.027 0.001 0.859 0.002 0.829 0.011 0.202 0.001 0.861 0.007 0.376 0.009 0.269 0.008 0.351 Chinese 0.005 0.526 0.001 0.863-0.010 0.229-0.006 0.440 0.002 0.787 0.009 0.303 0.001 0.902-0.004 0.645 Indian -0.002 0.839-0.010 0.223 0.003 0.694 0.012 0.154 0.019 0.021 0.010 0.217-0.002 0.794 0.035 0.000 Irish 0.003 0.745 0.004 0.656-0.001 0.931 0.014 0.098 0.005 0.553 0.003 0.752 0.010 0.248 0.005 0.571 Not stated 0.004 0.640 0.022 0.010 0.025 0.003 0.032 0.000-0.015 0.076-0.021 0.014 0.021 0.013-0.017 0.044 Pakistani -0.003 0.751-0.003 0.765-0.008 0.346-0.002 0.845 0.003 0.714 0.015 0.080-0.005 0.514 0.008 0.316 White and Asian 0.020 0.015-0.001 0.873-0.003 0.695 0.026 0.002 0.009 0.297 0.001 0.903 0.012 0.154 0.024 0.005 White and Black African 0.004 0.654-0.010 0.224-0.001 0.936 0.016 0.062 0.006 0.489-0.014 0.093 0.004 0.636 0.005 0.577 White and Black Caribbean 0.009 0.274 0.004 0.637 0.016 0.064 0.007 0.378 0.001 0.888-0.016 0.061 0.005 0.573 0.017 0.048 F00s (Organic) 0.006 0.498-0.001 0.950-0.006 0.447-0.008 0.334 0.005 0.533-0.014 0.098 0.003 0.705 0.001 0.866 F10s (Disorders due to substance use) 0.013 0.109-0.016 0.057 0.015 0.076 0.022 0.010 0.013 0.137 0.007 0.404 0.005 0.563 0.023 0.005 F20s (Schizophrenia) 0.025 0.003 0.007 0.410 0.011 0.180 0.026 0.002 0.005 0.541 0.017 0.040 0.019 0.026 0.010 0.258 F30s (Affective disorder) 0.013 0.124 0.007 0.420 0.019 0.024 0.037 0.000 0.009 0.309 0.004 0.623 0.017 0.044 0.040 0.000 F40s (Neurotic disorders) -0.013 0.116-0.013 0.117-0.008 0.336 0.013 0.121-0.002 0.787 0.010 0.246-0.011 0.186 0.012 0.144 F50s (Physiologically caused) 0.008 0.317-0.008 0.350 0.006 0.497 0.012 0.156 0.004 0.627-0.001 0.864 0.002 0.851 0.015 0.077 F60s (Personality disorder) 0.003 0.684 0.003 0.766-0.003 0.743 0.005 0.577 0.015 0.080-0.004 0.671 0.002 0.831 0.020 0.018 F80s (Developmental disorder) 0.003 0.747-0.012 0.147 0.004 0.672 0.015 0.076 0.005 0.548-0.010 0.226 0.003 0.702 0.007 0.440 F90s (Child/Adolescent onset disorder) 0.008 0.360 0.014 0.107 0.021 0.012 0.016 0.057 0.008 0.352 0.004 0.615 0.013 0.135-0.009 0.260 Other 0.004 0.610-0.004 0.639-0.001 0.949 0.006 0.456-0.003 0.698 0.011 0.191-0.004 0.645-0.002 0.815 No diagnosis recorded -0.006 0.476 0.017 0.050 0.011 0.179-0.013 0.109 0.011 0.204-0.013 0.124 0.010 0.234 0.020 0.019 10

Table 6 Univariate associations by Spearman correlations between conflict and containment variables, and occupancy characteristics. Significant associations are highlighted in red. Total aggression Total rule breaking Total substance use Total absconding Total medication related Total self-harm (z score) Total conflict Total containment r p r p r p r p r p r p r p r p Occupancy mean age -0.043 0.000-0.009 0.296 0.039 0.000 0.052 0.000 0.030 0.000 0.079 0.000 0.002 0.842-0.038 0.000 Number of males -0.017 0.043 0.142 0.000 0.139 0.000 0.088 0.000 0.053 0.000 0.033 0.000 0.114 0.000-0.123 0.000 African 0.042 0.000 0.098 0.000 0.065 0.000 0.086 0.000 0.056 0.000-0.064 0.000 0.105 0.000-0.020 0.017 Any other Asian background -0.021 0.013-0.065 0.000 0.001 0.945 0.111 0.000 0.038 0.000 0.030 0.000-0.014 0.102 0.094 0.000 Any other Black background 0.056 0.000-0.009 0.285 0.036 0.000 0.088 0.000 0.016 0.054 0.108 0.000 0.032 0.000-0.071 0.000 Any other White background 0.028 0.001 0.054 0.000 0.013 0.119 0.035 0.000-0.025 0.003 0.140 0.000 0.033 0.000-0.195 0.000 Any other ethnic group 0.042 0.000 0.143 0.000-0.008 0.367 0.020 0.018 0.096 0.000-0.002 0.805 0.136 0.000 0.011 0.205 Any other mixed background -0.015 0.081 0.000 0.963 0.023 0.007-0.006 0.502 0.002 0.817 0.025 0.003-0.019 0.028-0.026 0.002 Bangladeshi -0.060 0.000 0.077 0.000 0.011 0.175-0.091 0.000 0.045 0.000 0.002 0.853 0.017 0.047 0.117 0.000 British -0.042 0.000 0.054 0.000 0.096 0.000 0.013 0.122 0.075 0.000-0.038 0.000 0.051 0.000-0.036 0.000 Caribbean 0.021 0.013 0.035 0.000 0.067 0.000 0.077 0.000 0.008 0.361 0.080 0.000 0.043 0.000-0.070 0.000 Chinese -0.019 0.026-0.012 0.138 0.038 0.000 0.026 0.002-0.021 0.012 0.017 0.048-0.021 0.011-0.153 0.000 Indian -0.057 0.000-0.149 0.000-0.055 0.000 0.032 0.000 0.001 0.925 0.007 0.386-0.111 0.000 0.069 0.000 Irish -0.053 0.000 0.006 0.451-0.004 0.614-0.046 0.000 0.043 0.000-0.088 0.000-0.005 0.566-0.016 0.057 Not stated -0.002 0.788 0.086 0.000 0.076 0.000 0.123 0.000-0.062 0.000-0.102 0.000 0.076 0.000-0.154 0.000 Pakistani -0.066 0.000-0.084 0.000-0.062 0.000 0.004 0.619 0.005 0.539-0.066 0.000-0.071 0.000 0.132 0.000 White and Asian -0.042 0.000-0.123 0.000-0.005 0.519 0.084 0.000 0.036 0.000 0.237 0.000-0.055 0.000 0.165 0.000 White and Black African 0.024 0.004-0.047 0.000-0.044 0.000 0.036 0.000 0.021 0.013-0.028 0.001-0.010 0.244 0.076 0.000 White and Black Caribbean 0.045 0.000-0.021 0.012-0.029 0.001-0.024 0.005 0.010 0.246-0.189 0.000-0.010 0.222 0.049 0.000 F00s (Organic) 0.001 0.886 0.065 0.000 0.053 0.000 0.063 0.000 0.021 0.012 0.053 0.000 0.066 0.000-0.119 0.000 F10s (Disorders due to substance use) -0.005 0.562-0.052 0.000-0.008 0.313 0.086 0.000 0.040 0.000 0.032 0.000-0.004 0.596 0.070 0.000 F20s (Schizophrenia) -0.026 0.002 0.142 0.000 0.090 0.000 0.055 0.000 0.034 0.000 0.055 0.000 0.091 0.000-0.172 0.000 F30s (Affective disorder) -0.026 0.002-0.025 0.003 0.038 0.000 0.037 0.000 0.062 0.000-0.041 0.000 0.004 0.659 0.098 0.000 F40s (Neurotic disorders) -0.029 0.001-0.015 0.081 0.014 0.106 0.023 0.007 0.037 0.000 0.011 0.200-0.004 0.670 0.046 0.000 F50s (Physiologically caused) -0.004 0.657-0.008 0.357-0.025 0.003 0.020 0.016-0.003 0.739-0.044 0.000 0.003 0.752 0.020 0.020 F60s (Personality disorder) -0.005 0.547-0.028 0.001-0.017 0.040 0.055 0.000 0.038 0.000-0.054 0.000 0.011 0.198 0.114 0.000 F70s (Mental retardation) -0.002 0.843 0.013 0.113-0.008 0.319-0.002 0.794 0.030 0.000 0.018 0.028 0.015 0.081-0.004 0.609 F80s (Developmental disorder) 0.059 0.000 0.013 0.136 0.009 0.267 0.041 0.000 0.040 0.000-0.069 0.000 0.035 0.000-0.039 0.000 F90s (Child/Adolescent onset disorder) -0.019 0.022-0.005 0.556 0.020 0.020-0.011 0.193 0.000 0.980 0.021 0.014-0.010 0.236 0.018 0.029 Other -0.038 0.000 0.013 0.136 0.077 0.000-0.029 0.001 0.001 0.944 0.012 0.138-0.004 0.600-0.075 0.000 No diagnosis recorded -0.005 0.550 0.162 0.000 0.034 0.000-0.045 0.000 0.037 0.000-0.042 0.000 0.106 0.000 0.056 0.000 11

Two modelling exercises were then conducted with conflict and containment as dependent variables, using poisson regression: one for admissions and the second for occupancy. In these models, allowance was made for the clustering of results by ward, and shift type was included to control for the effects of differences between morning, afternoon and night shifts. The number of potential variables was decreased by collapsing all ethnicities and diagnoses comprising less and 3% of admissions into 'other' categories, and these were then dropped from the analyses as a reference category. A stepwise procedure was used obtain the final models, with p = 0.05 as the inclusion criteria. In the models of occupancy data, in order to separate the effect of individual variables from that of occupancy itself, both ethnicity and diagnoses were expressed as proportions for each shift, and total occupancy was used as the exposure variable. The models resulting from this exercise are displayed in Tables 7 & 8 below. Table 7. Models of conflict and containment in relation to admission characteristics IRR 95% CI p Total aggression F40s (Neurotic disorders) 0.568 0.341 0.947 0.030 Night shift 0.671 0.550 0.819 <0.001 Total rule breaking Afternoon shift 0.847 0.736 0.974 0.020 Night shift 0.688 0.543 0.872 0.002 F10s (Disorders due to substance use) 0.753 0.576 0.983 0.037 F40s (Neurotic disorders) 0.657 0.450 0.959 0.029 Total substance use Afternoon shift 1.309 1.151 1.489 <0.001 African ethnicity 0.707 0.524 0.954 0.023 F10s (Disorders due to substance use) 1.363 1.069 1.738 0.012 F30s (Affective disorder) 1.258 1.081 1.465 0.003 Total absconding Night shift 0.505 0.381 0.669 <0.001 F10s (Disorders due to substance use) 1.586 1.088 2.313 0.016 F30s (Affective disorder) 1.389 1.124 1.717 0.002 Total medication related Night shift 1.895 1.528 2.352 <0.001 Indian ethnicity 1.367 1.036 1.803 0.027 Total conflict F40s (Neurotic disorders) 0.712 0.571 0.887 0.003 Total containment Indian ethnicity 1.428 1.116 1.827 0.005 F10s (Disorders due to substance use) 1.217 1.030 1.438 0.021 F30s (Affective disorder) 1.164 1.057 1.281 0.002 F60s (Personality disorder) 1.234 1.043 1.460 0.014 Table 8. Models of conflict and containment in relation to occupancy characteristics. 12

IRR 95% CI p Total aggression Night shift 0.678 0.561 0.818 <0.001 Any other black background 9.003 2.622 30.916 <0.001 Mean age 0.906 0.835 0.983 0.018 Total rule breaking Afternoon shift 0.847 0.733 0.978 0.024 Night shift 0.674 0.533 0.853 0.001 Indian ethnicity 0.013 0.001 0.253 0.004 F10s (Disorders due to substance use) 0.106 0.016 0.703 0.020 Total substance use Afternoon shift 1.336 1.176 1.517 <0.001 Indian ethnicity 0.047 0.002 0.972 0.048 Total absconding Night shift 0.500 0.377 0.663 <0.001 Bangladeshi ethnicity 0.076 0.022 0.267 <0.001 Total medication related Night shift 1.857 1.498 2.303 <0.001 Total conflict Indian ethnicity 0.088 0.008 0.937 0.044 Mean age 0.955 0.913 0.998 0.043 Total containment Any other white background 0.225 0.057 0.890 0.034 Mean age 0.925 0.876 0.977 0.005 F10s (Disorders due to substance use) 7.574 1.635 35.076 0.010 F30s (Affective disorder) 4.059 1.165 14.148 0.028 F60s (Personality disorder) 14.305 1.199 170.640 0.035 These relationships are not straightforward to interpret, especially with respect to ethnicity, but also with regard to diagnosis. For both diagnoses and ethnicities are not randomly distributed across the different wards, as the wards represent local communities and the psychiatric teams that serve them. Some feature of teams or wards might thus be confounded with the proportions of an ethnicity or a diagnosis. They reflect the univariate analyses, in that plainly containment rates are more closely allied to admission and occupancy characteristics than conflict. However they give no grounds for thinking that ethnic minority patients are more likely to be subject to containment methods. Gender is noticeably absent from any of the models. The same analysis was completed after aggregating the data by week, retaining only those ward weeks where more than 7 PCCs had been returned (n = 954). The results are displayed in Tables 9 and 10. Table 9. Models of conflict and containment in relation to admission characteristics (ward week as time segment) 13

IRR 95% CI p Total aggression Nil significant Total rule breaking Nil significant Total substance use Nil significant Total absconding F10s (Disorders due to substance use) 55.409 6.944 442.113 <0.001 F20s (Schizophrenia) 6.628 1.545 28.435 0.011 Total medication related Any other black background 0.040 0.013 0.117 <0.001 F10s (Disorders due to substance use) 22.518 6.510 77.885 <0.001 F60s (Personality disorder) 50.928 7.701 336.800 <0.001 Total conflict Indian ethnicity 0.153 0.026 0.894 0.037 F30s (Affective disorder) 2.312 1.218 4.390 0.010 Total containment Bangladeshi 5.060 1.465 17.475 0.010 British 0.503 0.124 0.310 0.816 Indian 25.093 4.877 129.106 <0.001 F10s (Disorders due to substance use) 9.006 3.408 23.797 <0.001 F30s (Affective disorder) 2.854 1.098 7.419 0.031 F40s (Neurotic disorders) 6.705 2.324 19.347 <0.001 F60s (Personality disorder) 12.011 1.321 109.226 0.027 Table 10. Models of conflict and containment in relation to occupancy characteristics (ward week as time segment) 14

IRR 95% CI p Total aggression Any other black background 24.377 7.235 82.139 <0.001 Indian 0.060 0.004 0.991 0.049 Mean age 0.907 0.834 0.986 0.022 Proportion male 0.655 0.468 0.916 0.014 F30s (Affective disorder) 2.621 1.137 6.041 0.024 Total rule breaking Indian ethnicity 0.007 0.0002 0.167 0.002 F10s (Disorders due to substance use) 0.041 0.003 0.546 0.016 Total substance use Indian ethnicity 0.0002 0.000 0.012 <0.001 Total absconding Bangladeshi ethnicity 0.018 0.002 0.159 <0.001 Total medication related Nil significant Total conflict Indian ethnicity 0.040 0.004 0.445 0.009 Total containment Any other white background 0.194 0.047 0.795 0.023 Mean age 0.918 0.870 0.968 0.002 F10s (Disorders due to substance use) 11.978 1.905 75.319 0.008 F30s (Affective disorder) 4.192 1.119 15.708 0.033 To facilitate comparisons across the different sets of poisson modelling results, these are summarised in Table 11. 15

Table 11. Poisson modelling results summary on the relationship between conflict/containment and patient characteristics ADMISSION/SHIFT IRR ADMISSION WEEK IRR OCCUPANCY/SHIFT IRR OCCUPANCY/WEEK IRR Total aggression Total aggression Total aggression Total aggression F40s (Neurotic disorders) 0.568 Nil significant Night shift 0.678 Any other black background 24.377 Night shift 0.671 Any other black background 9.003 Indian 0.060 Mean age 0.906 Mean age 0.907 Proportion male 0.655 F30s (Affective disorder) 2.621 Total rule breaking Total rule breaking Total rule breaking Total rule breaking Afternoon shift 0.847 Nil significant Afternoon shift 0.847 Indian ethnicity 0.007 Night shift 0.688 Night shift 0.674 F10s (Disorders due to substan 0.041 F10s (Disorders due to substance 0.753 Indian ethnicity 0.013 F40s (Neurotic disorders) 0.657 F10s (Disorders due to substan 0.106 Total substance use Total substance use Total substance use Total substance use Afternoon shift 1.309 Nil significant Afternoon shift 1.336 Indian ethnicity 0.0002 African ethnicity 0.707 Indian ethnicity 0.047 F10s (Disorders due to substance 1.363 F30s (Affective disorder) 1.258 Total absconding Total absconding Total absconding Total absconding Night shift 0.505 F10s (Disorders due to substance 55.409 Night shift 0.500 Bangladeshi ethnicity 0.018 F10s (Disorders due to substance 1.586 F20s (Schizophrenia) 6.628 Bangladeshi ethnicity 0.076 F30s (Affective disorder) 1.389 Total medication related Total medication related Total medication related Total medication related Night shift 1.895 Any other black background 0.040 Night shift 1.857 Nil significant Indian ethnicity 1.367 F10s (Disorders due to substance 22.518 F60s (Personality disorder) 50.928 Total conflict Total conflict Total conflict Total conflict F40s (Neurotic disorders) 0.712 Indian ethnicity 0.153 Indian ethnicity 0.088 Indian ethnicity 0.040 F30s (Affective disorder) 2.312 Mean age 0.955 Total containment Total containment Total containment Total containment Indian ethnicity 1.428 Bangladeshi 5.060 Any other white background 0.225 Any other white background 0.194 F10s (Disorders due to substance 1.217 British 0.503 Mean age 0.925 Mean age 0.918 F30s (Affective disorder) 1.164 Indian 25.093 F10s (Disorders due to substan 7.574 F10s (Disorders due to substan 11.978 F60s (Personality disorder) 1.234 F10s (Disorders due to substance 9.006 F30s (Affective disorder) 4.059 F30s (Affective disorder) 4.192 F30s (Affective disorder) 2.854 F60s (Personality disorder) 14.305 F40s (Neurotic disorders) 6.705 F60s (Personality disorder) 12.011 16

Relationship of conflict and containment to over and under occupancy For 83% of shifts, wards were at full occupancy or below, the rest of the time they were over occupied. In the study Trust, over occupation was dealt with in four ways. The first, which would not show in our data, was to make a full transfer of patient(s) to another ward which was not currently full. Such transfers add complexity of the management of care, forcing the responsible community clinicians to work with different wards, attend different meetings on different days to do so, and running the risk of poor communication of the patients current care plan from one ward to another. The second would be to send patients on leave to their home address, sometimes requesting them to attend the ward during the day, utilising the ward as a sort of day care facility. This method would also not show up in our data, as patients on leave were not counted in our occupancy statistics. The third, was to send patients to sleep over on another less full ward overnight, and return for care during the day. This resulted in more patients being on the ward during the day than there were beds. Although it resolved some workload and communication issues, it could still mean disjunctions in care overnight, and meant that some patients, potentially tired or drowsy from medication side effects, had no bed to rest on during the day, and no easy access to their personal possessions. The fourth solution was to allow patients to sleep on sofa's or camp beds on the ward. This meant patients at least had continuity of care, albeit with nowhere to rest during the day. Both the third fourth and sometimes second solutions led to overcrowding on the wards during the day, with more patients present than the ward had been designed for. Crisis teams were set up during the course of the study (at two hospitals at about week 64, and at the the third at about week 92), and indeed, occupancy fell possibly as a result of this. However a new policy allowing transfers between the three hospitals was implemented at about week 100, and this also may have contributed to the fall. Smoothed data shows a statistically significant decline (see Figure 3). Mean occupancy across all the wards within weeks was generally low, i.e. the wards were overall under occupied, and any crises in occupancy tended to be brief and confined to small parts of the overall service provision. Figure 3. Smoothed over and under occupancy figures by week 17

lowess overunderocc week -5-4 -3-2 -1 40 60 80 100 120 140 week The nature of the relationship of over/under occupancy and conflict and containment is not easy to determine. Inspection of the scattergrams appears to imply a curvilinear relationship, with a peak of conflict at maximum occupancy, and lower conflict with both extreme over and under occupancy (see Figure 4). However this is an illusion based on the fact that more shifts were at near to full occupancy, therefore there was more chance of a shift with an extreme number of conflict events being reported. Figure 4. Scattergram of total conflict by over and under occupancy 18

TOTFLICT 0 20 40 60 80 100-10 -5 0 5 10 overunderocc Curve fitting tests were carried out using the curvefit procedure of SPSS. Although many nonlinear formulae provided a statistically significant fit to the data (as did in every case a linear model), r-squared values were uniformly small (<= 0.005), therefore none were to be preferred to a standard linear model. The data therefore provided no evidence that overcrowding led to increases in conflict or containment over and above those caused by simply having more patients on the ward who could behave in a difficult fashion. DISCUSSION Occupancy variables appear to be more predictive than admission variables, in that there are generally more and stronger associations. In addition, although patient characteristics seen through the lens of admissions are very similar to those seen through occupancy, they are not the same. And as can be seen for the findings reported here, results using admissions and occupancy do differ, although there are some repetitive similarities. The City 128 analysis was based on admissions only, as no occupancy data was available, and results from that analysis will be biased towards admission rather than occupancy associations. The main finding here is that containment seems to be more associated with patient characteristics than conflict is. Maybe this is about staffs risk assessments and use of preventive containment. However this was not the same in City 128 data, where univariate analysis showed conflict more related to patient characteristics and multivariate no real difference. These two sets of results do not match, even on this issue. 19

Overcrowding is not specifically related to conflict and containment, except in the sense of providing more patients who can engage in difficult behaviour. Finally, the question arises as to which dataset contains the greater amount of error, affecting the results of the above reported analysis: the official data inputted largely by ward clerks with diagnoses recorded by junior doctors, or the PCC data collected by ward nursing staff. CONCLUSIONS Occupancy and admissions in terms of both numbers and characteristics, are related to conflict and containment use. However those relationships are neither strong nor entirely consistent, suggesting that other factors may be more important. The nature of the relationship appears to be linear, or at least curve fitting does not produce a significantly superior fit, suggesting that overcrowding does not lead to exponential rises in conflict or containment. CREDIT The data collection on which this paper is based was funded by NIHR SDO. REFERENCES 20