Nursing staff numbers and their relationship to conflict and containment rates on psychiatric wards—A cross sectional time series Poisson regression study

Article Outline

• Abstract
• 1. Background
• 2. Methods
  • 2.1. Sample
  • 2.2. Instrument
  • 2.3. Analysis
• 3. Findings
• 4. Discussion
• 5. Limitations and conclusions
• References
• Copyright

Abstract 

Background

The link between positive outcomes and qualified nurse staffing levels is well established for general hospitals. Evidence on staffing levels and outcomes for mental health nursing is more sparse, contradictory and complicated by the day to day allocation of staff resources to wards with more seriously ill patients.

Objective

To assess whether rises in staffing numbers precede or follow levels of adverse incidents on the wards of psychiatric hospitals.

Design

Time series analysis of the relationship between shift to shift changes over a six month period in total conflict incidents (aggression, self-harm, absconding, drug/alcohol use, medication refusal), total containment incidents (pro re nata medication, special observation, manual restraint, show of force, time out, seclusion, coerced intramuscular medication) and nurse staffing levels.

Settings

32 acute psychiatric wards in England.

Methods

At the end of every shift, nurses on the participating wards completed a checklist reporting the numbers of conflict and containment incidents, and the numbers of nursing staff on duty.

Results

Regular qualified nurse staffing levels in the preceding shifts were positively associated with raised conflict and containment levels. Conflict and containment levels in preceding shifts were not associated with nurse staffing levels.

Conclusions

Results support the interpretation that raised qualified nurse staffing levels lead to small increases in risks of adverse incidents, whereas adverse incidents do not lead to consequent increases in staff. These results may be explicable in terms of the power held and exerted by psychiatric nurses in relation to patients.

Keywords: Mental health nursing, Inpatients, Conflict, Containment, Staffing levels, Violence, Self-harm, Safety, Coercion

What is already known about the topic?

What this paper adds

Back to Article Outline

1. Background 

The international evidence on nurse staffing levels in general hospitals has been reviewed (Lankshear et al., 2005) and shows that higher numbers are associated with lower mortality and lower rates of medical complications. These findings have since been confirmed by more recent studies in the UK (Rafferty et al., 2007) and Switzerland (Schubert et al., 2009), but not in Belgium (Van den Heede et al., 2009).

There is much less literature on psychiatric nurse staffing levels and outcomes, with some studies being rather old and findings contradictory. Finding a relationship between staffing numbers and outcomes is complicated by the fact that psychiatric hospital managers deploy staff selectively to wards that experience higher levels of patient disturbance and acuity (Kellam et al., 1967). These mechanisms could produce a misleading correlation between higher staffing levels and poorer outcomes in terms of disturbed behaviour. Using longitudinal analysis of officially collected data from a single NHS Trust, we have previously demonstrated an association between the presence of regular staff on the ward and lower rates of incidents of physical aggression and self-harm (Bowers et al., 2005a). Other positive findings have been associations between higher staffing levels and amore therapeutic wards (Moos, 1972); intensive staffing and lower lengths of stay (Becker, 1969); and lower rates of violent incidents (Chou et al., 2002, Lanza et al., 1994). However other research has found no connection between staff/patient ratios and outcomes (Ellsworth et al., 1979); more staff and more violence (Owen et al., 1998); and more staff and increased interaction between them, rather than between staff and patients (Sandford et al., 1990). Finally, in a study utilising 22 wards, although high staffing levels were associated with lower readmission rates, substantially better predictions of care outcomes could be made when the amount of attention patients received from staff was taken into account (Coleman and Paul, 2001). Thus staffing utilisation may be as important as overall staff numbers.

In 2005, we conducted the City-128 study, during which we collected end of shift reports on patient outcomes for six months from a random sample of 136 English acute psychiatric wards. The outcomes of interest were rates of conflict (aggression, rule breaking, alcohol/drug use, absconding, medication refusal and self-harm/suicide) and containment (pro re nata medication, intermittent observation, constant observation, coerced intramuscular medication, show of force, manual restraint, seclusion and time out). A series of cross sectional multilevel analyses of this dataset have been undertaken, and the statistically significant findings in relation to nurse staffing numbers are displayed in Table 1. These findings take into account patient characteristics, the physical environment, ward routines, staff demographics and staff groups factors/attitudes on the sample wards. With the exception of medication related issues and self-harm rates, these findings suggest a positive association between patient aggression frequency and staff numbers on duty during the shift. These findings are most consistent for qualified nursing staff, and further show a positive relationship between the numbers of such staff on duty and the more severe forms of containment.

Table 1. Cross sectional results related to nurse staffing, + indicating a positive correlation and − indicating an inverse correlation.

Findings reported in Bowers et al., 2008, Bowers et al., 2009a, Bowers et al., 2009b and Baker et al. (2008).

It is possible that these associations have arisen because staff are redeployed to wards where patients are in an agitated and disturbed condition. We therefore decided to analyse our dataset longitudinally to see whether rises in staffing numbers preceded or followed levels of disturbance on the wards.

Back to Article Outline

2. Methods 

2.1. Sample 

The sample comprised 136 acute psychiatric wards with their patients and staff in 67 hospitals within 26 NHS Trusts (organisational units with common clinical policies and investment levels) in England, proximate to three regional centres during 2004–2005. Acute psychiatric wards were defined as those that primarily serve acutely mentally disordered adults, taking admissions in the main directly from the community, and not offering long-term care or accommodation. Wards that were organised on a speciality basis, or that planned to change population served, location, function, or which were scheduled for refurbishment during the course of the study were excluded. Each centre identified all eligible wards within reasonable travelling distance of their research base. It was initially intended to randomly sample wards, with replacement for refusals to participate. However the geographical dispersion of wards outside of London meant that to achieve the requisite sample size, two centres had to recruit all available wards within practical reach for data collection. In London, it was possible to randomly sample from a list of 112 wards. The 136 acute psychiatric wards that participated in the study represented 25% of the estimated total of 551 wards in England. The study was approved by the NW Multi-centre Research Ethics Committee.

2.2. Instrument 

The Patient-staff Conflict Checklist (PCC-SR), an end of shift report by nurses on the frequency of conflict and containment events (Bowers et al., 2005b), was collected for a six month period on all participating wards. On entry to the study, ward nursing staff received training in the use of the PCC-SR, and each ward was provided with a handbook giving definitions of items. In recent tests based on use with case note material, the PCC has demonstrated an inter-rater reliability of 0.69 (Bowers et al., 2005b), and has shown a significant association with rates of officially reported incidents (Bowers et al., 2006). The PCC-SR also contained items reporting the numbers of nursing staff on duty at the commencement of each shift, broken down into regular qualified, regular unqualified, bank and agency qualified, bank and agency unqualified, and student nurses. For this analysis total conflict (the sum of all conflict events during a shift) and total containment (the sum of all containment events) were calculated.

2.3. Analysis 

A total of 45,989 PCC-SRs were collected during the study, representing an average response rate of 67% of all possible shifts that could be reported. As we wished to conduct a longitudinal analysis, we selected the data only from those wards that submitted 80% or more PCC-SR returns (32 of the 136 participating wards, returning 15,449 PCC-SRs), so as to make a near continuous time series for each ward. Values for missing shifts were not interpolated. The analysed wards had slightly higher bed numbers than those with lower response rates (mean 22.21 vs. 20.69, t=2.01, p=0.046); a marginally lower nurse to bed ratio (0.91 vs. 1.01, t=2.24, p=0.027); no difference in staff vacancy rates, the proportion of nurses qualified, the proportion of staff of an ethnic minority background, the proportion of staff male, the proportion of staff over 30 years of age, total conflict or total containment rates.

Cross sectional time series mixed effects Poisson regression was used to construct models relating staffing numbers in the preceding 9 shifts with total conflict rates, and then total conflict rates for the previous 9 shifts with staffing numbers. In this way it could be demonstrated whether conflict leads to the later deployment of more staff to the ward, or vice versa. Similar models were constructed for total containment and staffing. NHS Trust and ward were accounted for as nested hierarchical levels in the regression equation, and the number of beds on the ward was used as an indicator of the number of patients, and set as the exposure variable. All variables were entered and significant results reported, no stepwise elimination was applied. Total conflict, total containment, and numbers of nurses on duty were all associated with shift type (a.m., p.m. or night), day of the week and number of admissions during the shift. These three variables were therefore entered in all the models constructed to control for their effects. Comparisons between groups were conducted using Kruskal–Wallis tests, and correlations using Spearman's test, as most of the data was skewed. All analyses were conducted using STATA version 11.

Back to Article Outline

3. Findings 

The mean numbers of staff on duty by shift type are displayed in Table 2. The modal number of staff on morning and afternoon shifts was two regular qualified nurses and two regular unqualified staff (healthcare assistants). The modal number on a night shift was one regular qualified nurse and one unqualified staff. Shifts where bank/agency staff or student nurses were on duty were in a minority, and only a tiny number of night shifts (226, 4%) had a student nurse present at the beginning of the shift. Differences in staffing numbers between shifts were statistically significant (chi square=4362, df=2, p<0.001). Total staff numbers varied by day of the week, (chi square=325.14, df=6, p<0.001), being highest on a Wednesday (mean 5.09) and lowest on a Sunday (mean 4.51). The mean number of beds in this subsample of 32 wards was 22.2 (sd 3.7), most catered for both genders (24, 75%) and most operated a system of internal rotation by which the whole staff team rotated through all three shifts to provide cover (23, 72%).

Table 2. Mean numbers of staff on duty by shift type.

The median number of conflict events in a shift was three (55% of shifts three or less, mean rate 4.39, sd 4.89), and containment two (51% of shifts two or less, mean rate 3.26, sd 3.33). Total conflict varied significantly: by shift (chi square=164.31, df=2, p<0.001), being highest on the morning shift (mean 4.88 incidents) and lowest at night (mean 3.69 incidents); by day of the week (chi square=25.13, df=6, p<0.001), being highest on a Wednesday (mean 4.75 incidents) and lowest on a Sunday (mean 4.05 incidents); and was associated with admissions during the shift (r_s=0.02, p=0.002). Total containment varied significantly: by shift (chi square=7.75, df=2, p<0.021), being highest on the afternoon shift (mean 3.37 events) and lowest in the morning and at night (mean 3.21 and 3.20 events respectively); and was significantly correlated with admissions (r_s=0.28, p<0.001); but did not vary significantly by day of the week.

Table 3 shows the result of regressing staffing numbers on total conflict and total containment. Each row of the table represents the results of one model. For example, the first row represents the regression of numbers of regular qualified nursing staff on duty and total conflict rates. Each column of the table represents the number of lags before the dependent variable, total conflict. Regular qualified staff numbers were significantly associated with total conflict nine shifts later, with an IRR (incident rate ratio) of 1.03. In other words, for every one extra member of regular qualified staff on duty nine shifts before, one additional conflict incident is 3% more likely, and so on for the remaining columns. The top half of the table shows models with total conflict and total containment as the dependent variables, the bottom half those with staffing variables. Numbers of regular qualified nursing staff were most systematically and consistently related to subsequent levels of conflict and containment. There were also some patchy relationships between bank and agency unqualified staff and subsequent conflict and containment. Bank and agency qualified staff showed positive, inverse, and no relationships at different lags.

Table 3. Time series regression of nine lags of staffing numbers on total conflict and total containment, and lags of conflict and containment on staffing variables (incident rate ratios).

When the same equations were then reversed, with lags of total conflict, then total containment regressed on staffing numbers (the bottom half of Table 3), virtually no statistically significant results were found, levels of conflict and containment were not related to subsequent staffing numbers. The only indication this might be the case was over the short term for bank and agency unqualified staff.

Back to Article Outline

4. Discussion 

Data from 32 wards submitting 15,449 end of shift reports were subjected to time series analysis to discover whether rises in staffing levels preceded or followed rates of total conflict and total containment. The results showed that only numbers of regular qualified staff were systematically related to conflict and containment rates over time. Moreover, rises in the numbers of nurses preceded rather than followed increases in conflict and containment. Previous cross sectional analysis had shown positive correlations between qualified nurse numbers and many conflict and containment items. The results reported in this paper undermine the explanation that rises in conflict and containment rates lead to deployment of more staff to the wards concerned. Instead they suggest that higher nurse numbers lead to more of these adverse events.

The seriousness and gravity of conflict and containment rates as outcomes of nursing care should not be underestimated. Patient aggression can cause staff injuries, time off sick (Hillbrand et al., 1996), and post traumatic stress disorder (Needham et al., 2005). By definition, self-harm causes patient injury, and in England there are over a hundred psychiatric inpatient suicides every year (Appleby et al., 2006). Medication refusal (Kasper et al., 1997), the consumption of drugs and alcohol (Phillips and Johnson, 2003), and absconding (Bowers et al., 1998) all interfere with treatment and prolong lengths of stay. Containment is often aversive and unpleasant for patients (Whittington et al., 2009) and staff (Olofsson et al., 1995). Mean annual costs (for staff time only) for all conflict behaviours across England's in-patient psychiatric wards exceed £72 million per annum, whilst containment costs an estimated £106 million per annum (Bowers et al., 2007).

However these are not the only possible outcomes of psychiatric nursing care. Length of stay, the level of patients’ symptoms, and readmission rates, for example, may also serve as outcome indicators. If investigated, these might a show a different and more flattering relationship to nurse staffing numbers. In addition, the relationships demonstrated in this study, whilst statistically strong are not so strong in terms of effect size. For example, within the shift concerned, one extra qualified nurse on duty raised the risk of an additional conflict incident by 4%. Given that the mean rate of total conflict was 4.39, seven extra nurses would need to be supplied to secure one extra conflict event. Because of the large sample number, confidence intervals around the IRRs in Table 3 are tight, those for the within shift regular qualified staff (1.04) being 1.02–1.05 for example. Whilst the effect is small, the causal direction is depicted very starkly by the failure to find increases in staffing consequent upon total conflict or containment rates.

There might be particular reasons for a positive connection between qualified staff numbers and total conflict. Qualified staff are perhaps those who are more likely to seek to engage with very mentally ill patients. It is they who are most likely to have to refuse patient requests (thus eliciting anger) or who are more likely to instruct patients to comply with ward rules or render resented assistance with self-care. Finally, more junior staff, when faced with an agitated, demanding and threatening patient, perhaps more likely to refer the dispute up to a qualified member of staff, rather than confront the patient or set limits themselves. All these are well known to be amongst the immediate antecedents of violent incidents on psychiatric wards (Quanbeck et al., 2007, Daffern and Howells, 2007).

Not all types of conflict showed this positive relationship with staffing numbers. Cross sectional analysis showed some findings in the other direction, with more qualified nurses being associated with lower rates of self-harm, medication refusal, alcohol intoxication, and the giving of pro re nata medication. These four items may suggest that qualified nurses are effective at reducing certain types of patient distress through their availability.

These findings do not in any way constitute a basis for reductions in qualified nurse staffing levels. Decreasing the proportion of qualified staff might negatively affect assessment, information giving, therapeutic activity and physical care, none of which were measured in this research. In addition, sufficient numbers of nurses needs to be available to supervise patients safely, deliver treatment prescribed by medical staff, and deal with and ameliorate more serious incidents when they occur.

Back to Article Outline

5. Limitations and conclusions 

The 32 wards with the high response rate may differ in some way from other wards, and the findings may thus not be completely generalisable to the total population of acute psychiatric wards in England. It is also possible that some form of response bias is operating, for example that shifts with more qualified nurses and duty are likely to observe and report more adverse incidents, however it is unclear how this would have given rise to temporal relationships in one direction but not the other. In addition, if qualified staff were associated with more accurate reporting, controlling in the analysis for response rates should weaken or remove the associations found. However when we repeated our analyses with the response rate included, the results remained unchanged.

This exploratory analysis has demonstrated a small but significant positive relationship between qualified psychiatric nurse numbers and total conflict and containment rates. Further research is now required to show why this exists, and investigate whether changes in nursing practice can change this relationship to its opposite in a comprehensive and enduring fashion.

Conflict of interest: None declared.

Funding: The data collection for this study was funded by the National Institute for Health Research (NIHR) SDO programme, and the analysis supported by an NIHR Programme Grant for Applied Research scheme (RP-PG-0707-10081). The views expressed are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

Ethical approval: NW Multi-centre Research Ethics Committee.

Back to Article Outline

References 

1. Appleby L, Shaw J, Kapur N, Windfuhr K, Ashton A, Swinson N, et al. Avoidable Deaths: Five Year Report of the National Confidential Inquiry into Suicide and Homicide by People with Mental Illness. Manchester: University of Manchester; 2006;
   • View In Article
2. Baker J, Bowers L, Owiti J. Ward features associated with high rates of medication refusal by patients. General Hospital Psychiatry. 2008;31:80–89
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (170 KB)
   • CrossRef
3. Becker RE. Staffing level and treatment effectiveness. British Journal of Psychiatry. 1969;115(521):481–482
   • View In Article
4. Bowers L, Allan T, Simpson A, Jones J, Van Der Merwe M, Jeffery D. Identifying key factors associated with aggression on acute in-patient psychiatric wards. Issues in Mental Health Nursing. 2009;30:260–271
   • View In Article
   • CrossRef
5. Bowers L, Allan T, Simpson A, Nijman H, Warren J. Adverse incidents, patient flow and nursing workforce variables on acute psychiatric wards: the Tompkins Acute Ward Study. International Journal of Social Psychiatry. 2005;53(1):75–84
   • View In Article
6. Bowers L, Flood C, Brennan G, LiPang M, Oladapo P. Preliminary outcomes of a trial to reduce conflict and containment on acute psychiatric wards: City Nurses. Journal of Psychiatric and Mental Health Nursing. 2006;13:165–172
   • View In Article
   • MEDLINE
   • CrossRef
7. Bowers L, Van Der Merwe M, Nijman H, Hamilton B, Noorthoorn E, Stewart D, et al. The practice of seclusion and time out on English acute psychiatric wards: the City-128 study. Archives of Psychiatric Nursing. 2009;24(4):275–286
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (187 KB)
   • CrossRef
8. Bowers L, Whittington R, Nolan P, Parkin D, Curtis S, Bhui K, et al. The relationship between service ecology, special observation and self-harm during acute in-patient care: the City-128 study. British Journal of Psychiatry. 2008;193(5):395–401
   • View In Article
9. Bowers L, Whittington R, Nolan P, Parkin D, Curtis S, Bhui K, et al. The City 128 Study of Observation and Outcomes on Acute Psychiatric Wards. Report to the NHS SDO Programme. London: NHS SDO Programme; 2007;
   • View In Article
10. Bowers L, Douzenis A, Galeazzi G, Forghieri M, Tsopelas C, Simpson A, et al. Disruptive and dangerous behaviour by patients on acute psychiatric wards in three European centres. Social Psychiatry and Psychiatric Epidemiology. 2005;40:822–828
    • View In Article
    • MEDLINE
    • CrossRef
11. Bowers L, Jarrett M, Clark N. Absconding: a literature review. Journal of Psychiatric & Mental Health Nursing. 1998;5(5):343–353
    • View In Article
12. Chou KR, Lu RB, Mao WC. Factors relevant to patient assaultive behavior and assault in acute inpatient psychiatric units in Taiwan. Archives of Psychiatric Nursing. 2002;16(4):187–195
    • View In Article
    • Abstract
    • Full-Text PDF (97 KB)
    • CrossRef
13. Coleman JC, Paul GL. Relationship between staffing ratios and effectiveness of inpatient psychiatric units. Psychiatric Services. 2001;52(10):1374–1379
    • View In Article
14. Daffern M, Howells K. Antecedents for aggression and the function analytic approach to the assessment of aggression and violence in personality disordered patients within secure settings. Personality and Mental Health. 2007;1(2):126–137
    • View In Article
15. Ellsworth RB, Collins JF, Casey NA, Schoonover RA, Hickey RH, Hyer L, et al. Some characteristics of effective psychiatric treatment programs. Journal of Consulting and Clinical Psychology. 1979;47(5):799–817
    • View In Article
    • CrossRef
16. Hillbrand M, Foster HG, Spitz RT. Characteristics and cost of staff injuries in a forensic hospital. Psychiatric Services. 1996;47(10):1123–1125
    • View In Article
17. Kasper JA, Hoge SK, FeuchtHaviar T, Cortina J, Cohen B. Prospective study of patients’ refusal of antipsychotic medication under a physician discretion review procedure. American Journal of Psychiatry. 1997;154(4):483–489
    • View In Article
18. Kellam S, Goldberg S, Schooler N, Berman A, Schmelzer J. Ward atmosphere and outcome of treatment of acute schizophrenia. Journal of Psychiatric Research. 1967;5(2):143–163
    • View In Article
19. Lankshear A, Sheldon T, Maynard A. Nurse staffing and healthcare outcomes: a systematic review of the international research evidence. Advances in Nursing Science. 2005;28(2):163–174
    • View In Article
20. Lanza M, Kayne H, Hicks C, Milner J. Environmental characteristics related to patient assault. Issues in Mental Health Nursing. 1994;15(3):319–335
    • View In Article
    • MEDLINE
    • CrossRef
21. Moos R. Size, staffing, and psychiatric ward treatment environments. Archives of General Psychiatry. 1972;26:414–418
    • View In Article
22. Needham I, Abderbalden C, Halfens R, Fischer J, Dassen T. Non-somatic effects of patient aggression on nurses: a systematic review. Journal of Advanced Nursing. 2005;49(3):283–296
    • View In Article
    • MEDLINE
    • CrossRef
23. Olofsson B, Norgberg A, Jacobsson L. Nurses’ Experience with using force in Institutional Care of Psychiatric Patients. Nordic Journal of Psychiatry. 1995;49(5):325–330
    • View In Article
    • CrossRef
24. Owen C, Tarantello C, Jones M, Tennant C. Violence and aggression in psychiatric units. Psychiatric Services. 1998;49(11):1452–1456
    • View In Article
25. Phillips P, Johnson S. Drug and alcohol misuse among in-patients with psychotic illnesses in three inner-London psychiatric units. Psychiatric Bulletin. 2003;27(217):220
    • View In Article
26. Quanbeck C, McDermott BE, Lam J, Eisenstark H, Sokolov G, Scoatt CL. Categorization of aggressive acts committed by chronically assaultive state hospital patients. Psychiatric Services. 2007;58(4):521–528
    • View In Article
27. Rafferty A, Clarke SP, Coles J, Ball J, James P, McKee M, et al. Outcomes of variation in hospital nurse staffing in English hospitals: cross-sectional analysis of survey data and discharge records. International Journal of Nursing Studies. 2007;44(2):175–182
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (161 KB)
    • CrossRef
28. Sandford DA, Elzinga RH, Iversen R. A quantitative study of nursing staff interactions in psychiatric wards. Acta Psychiatrica Scandinavica. 1990;81:46–51
    • View In Article
    • CrossRef
29. Schubert M, Clarke SP, Glass TR, Schaffert-Witvliet B, De Geest S. Identifying thresholds for relationships between impacts of rationing of nursing care and nurse- and patient-reported outcomes in Swiss hospitals: a correlational study. International Journal of Nursing Studies. 2009;46(7):884–893
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (347 KB)
    • CrossRef
30. Van den Heede K, Sermeus W, Diya L, Clarke SP, Lesaffre E, Vleugels A, et al. Nurse staffing and patient outcomes in Belgian acute hospitals: cross-sectional analysis of administrative data. International Journal of Nursing Studies. 2009;46(7):928–939
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (183 KB)
    • CrossRef
31. Whittington R, Bowers L, Nolan P, Simpson A, Neil L. Approval ratings of inpatient coercive interventions in a national sample of mental health service users and staff in England. Psychiatric Services. 2009;60(6):792–798
    • View In Article