UNDER THE GAZE OF STAFF: SPECIAL OBSERVATION AS SURVEILLANCE
Objective: Special observation (SO) is a method of preventing psychiatric inpatients
harming themselves or others. This study explored the relationship of SO to patient
conflict behaviours, the use of containment methods by staff, the physical
environment of wards, patient routines and staffing variables.
Method: End-of-shift reports completed by nurses on the frequency of conflict and
containment events were collected for a 6-month period on 136 acute mental health
wards in England during 2004 and 2005.
Results: Intermittent SO (patient checked at specified intervals) was used five times
more frequently than constant SO (patient kept within sight or reach at all times).
Both were positively associated with verbal aggression, absconding, refusing regular
medication, demanding PRN medication and refusing PRN medication, but
intermittent and constant SO were negatively correlated. Intermittent SO was
associated with fewer self-harm incidents. Significant relationships were found
between SO and measures of ward surveillance, door locking and the ease of
observing patients on the wards. Both types of SO were more common when higher
numbers of unqualified staff were on duty.
Conclusion: A complex range of factors are associated with intermittent and constant
forms of SO. Improved ward design, less reliance on unqualified and greater use of
CCTV and other surveillance measures may reduce the need for SO in some cases.
Special observation (SO) is a method of preventing acutely disturbed psychiatric
inpatients from harming themselves or others. It involves assigning an identified
person to the care of the ‘at-risk’ patient for a certain period of time, above the
minimum general level of observation required for all inpatients. SO may be
intermittent, where the patients is checked at specified intervals, or constant, where
the patient is kept within sight or reach at all times . In practice, however, a
confusing range of terms are used to describe the various levels of observation
practice . Reported rates of SO, expressed as a proportion of admissions to acute
psychiatric care, range from 3% , to 47% . Once patients are placed under SO,
its reported duration ranges from a few hours only  to several weeks . The most
frequently cited reason for initiating SO is to prevent self-harm, but it can be used for
many different purposes including the management of aggressive behaviour, refusal
of medication and to prevent absconding [2, 5, 7-9]. Most hospitals in England have
policies for SO, but these are highly variable , as are professional groups that have
authority to initiate or terminate the procedure , and the way in which it is carried
out . Often initiated by doctors, nurses have been found to make modifications to
the prescribed level of SO [11-13].
The therapeutic value of SO has been questioned . Patients have reported
positive aspects to SO such as feeling understood, secure, reduced dysphoria and
reduced suicidal thoughts, but also experiences of feeling isolated, degraded, or
coerced [6, 15]. Nurses can find it stressful to implement . Allocating nurses to
the care of an individual patient means that less time is available for others on the
ward. Therefore, SO often requires extra bank or agency staff which can be
expensive. The cost to the health service in England has been estimated at £45
million a year for intermittent SO and £35 million for constant SO . Despite the
imperative to prevent self-harm and suicide on acute psychiatric wards, suicides do
unfortunately occur while patients are under enhanced levels of observation. Data
for England and Wales suggest that more inpatient suicides occur among those
under intermittent SO (23%) than under constant SO (3%). In contrast,
intermittent SO has been associated with fewer incidents of self-harm and constant
SO with no apparent benefit for patients [18,19]. This paper uses a large, national
dataset to assess the relationship of SO to: patient characteristics and behaviours,
the use of other containment methods, service environment, physical environment,
patient routines, staff characteristics, and staff group variables.
The sample comprised 136 acute psychiatric wards with their patients and staff in 67
hospitals in England during 2004-5. 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. The
sample represented 25% of the estimated total of 551 wards in England.
Instruments and procedure
The Patient-staff Conflict Checklist (PCC-SR), an end of shift report by nurses on the
frequency of conflict and containment events , was collected for a six month
period on all participating wards. At the end of each shift, nurses indicated the
frequency of 19 conflict events including aggressive behaviours, rule breaking,
absconding/eloping, consumption of drugs or alcohol, regular medication refusal,
refusal of pro re nata (PRN) medication and demanding PRN medication. Counts of
ten containment methods were also measured including medication, seclusion,
manual restraint etc.). SO was categorised as intermittent or constant. Staff were
also asked to note whether constant SO included no or minimal interaction with the
patient, for whatever reason.
The PCC–SR form was also used to collect data on patients: age, gender, ethnicity,
diagnosis, reason for admission and postcode. A Bongar Lethality Scale  was
completed for all incidents of self-harm or attempted suicide. Information on the ward
physical environment and the policies in operation was collected on a site visit by a
researcher and a form completed by the ward manager. Ward observability was
assessed through calculating a single score representing the complexity of the
environment, with higher scores representing more complicated, difficult to observe
wards. Additional instruments used included the Attitudes to Containment Measures
Questionnaire (ACMQ) ; Attitude to Personality Disorder Questionnaire (APDQ)
 ; Ward Atmosphere Scale (WAS) ; Team Climate Inventory (TCI) ;
Multifactor Leadership Questionnaire (MLQ) ; Maslach Burnout Inventory (MBI)
. The scales have been widely used and have well established validity and
reliability. Over 45,000 PCC-SRs were collected with an average response rate of
67%. A full description of the study methodology and other instruments used can be
found elsewhere .
Variables from all sources were divided into seven logical domains: patient variables
(features of the population admitted to the ward during the study, collected via the
PCC-SR, plus mean responses to the patient ACMQ); service environment variables
(the presence or absence of related community services likely to impact on ward
operation, plus the availability of seclusion and/or psychiatric intensive care, as
reported by ward managers); physical environment (composite rating of quality and
an objective measure of complexity, both taken by project research assistants);
conflict (all conflict variables on the PCC-SR); containment (all containment variables
on the PCC-SR, plus intensity of security policies as reported by ward managers
including banned items, searching of patients, restrictions on patient access to
certain areas, drug/alcohol monitoring and access to security guards); staff
characteristics (numbers of nurses and allied staff on duty from the PCC-SR, their
demographic features collected together with the MBI, and numbers of other
disciplines overall as reported by the ward manager); and staff group factors (mean
questionnaire scores from the ACMQ staff, APDQ, MBI, MLQ, TCI and WAS).
Multilevel random effects modelling was conducted using Stata v.11 to control for
clustering by Trust (a self-governing organisational unit within the National Health
Service), ward and shift. Counts of intermittent and constant special observation
were dependent variables, with number of beds on each ward as the exposure
variable. The models were produced through a staged process of backward
selection, dropping the least significant variable at each stage. Variables from each
domain were used to build a separate initial model, and then the significant variables
were used to construct a final comprehensive model using the same process of
backward selection. Coefficients are expressed as incident rate ratios (IRR). There
was little appreciable difference between models for constant SO with or without
engagement. Since nearly all the recorded constant SO was with engagement, a
statistical model for constant SO without engagement is not presented.
Intermittent was the most common form of SO, with a mean of 1.70 (sd=2.40) events
per shift on a standardised 20 bed ward. The mean for constant SO with
engagement was 0.35 (sd=0.73) per shift. Very few shifts utilised constant SO
without engagement (mean=0.09; sd=0.51), so this form of SO was included only as
an independent variable in further analysis.
The results of multilevel regression models are shown in Tables 1 and 2. Only
intermittent SO retained any patient variables in the final models, and was associated
with a greater proportion of younger patients and fewer patients admitted for causing
harm to others. The number of admissions during a shift was a significant predictor
in both final models.
Aspects of the ward physical environment were related to SO. Intermittent SO was
associated with more complex and difficult to observe wards, while constant SO was
associated with fewer windows in the doors of single rooms.
Similarities were found in the types of patient behaviours associated with intermittent
and constant SO. Both were positively associated with verbal aggression,
absconding, refusing regular medication, demanding PRN medication and refusing
PRN. They were also associated with rule breaking behaviours, although the type of
infringements varied. Incidents of physical aggression against persons and objects
were positively associated with constant SO, but the former was negatively
associated with intermittent SO. There was a negative correlation between
intermittent SO and incidents of self-harm.
The models show a negative association between intermittent and constant SO (with
or without engagement). In terms of other containment methods, coerced and PRN
medication were both associated with SO although coerced medication was not
retained in the final intermittent SO model. Both forms of SO were linked with show
of force, but seclusion was predictive of constant SO only. Being sent to PICU or
ICA was correlated with constant but not intermittent SO.
Relationships between SO and measures of ward surveillance and security were
notable. Intermittent SO was associated with less searching and use of closes circuit
television (CCTV) monitoring, but more door security, door locking and use of alarms.
Door locking was also positively correlated with constant SO, as was the banning of
items. Staffing variables were more closely associated with constant than intermittent
SO, but both types showed an association with higher numbers of unqualified staff on
This large and representative study of acute wards in England found that intermittent
SO was reported around five times more frequently than constant SO with
engagement, while SO without engagement was little used. These findings suggest
that the benefits of engaging with patients have been recognised. There are few
studies of intermittent SO with which to compare the results. Intermittent checks on
all patients has been used as an alternative to constant SO  and as part of a
package of measures to reduce absconding . Consistent with our findings, a
Scottish study reported 15 minute observations to be more frequently used than both
general and constant observation , prompting the authors to question whether
staff were being over-cautious with some patients.
The preference for intermittent SO may have been influenced by resource
constraints. Higher staffing levels were more strongly associated with constant
rather than intermittent SO, probably because the former is much more resource
intensive. The unit cost of constant SO has been estimated to be at around three
times higher than intermittent SO . One study calculated that constant SO
represented up to 20% of the total nursing budget .
The results indicate that SO is frequently conducted by less qualified staff. Greater
numbers of unqualified staff was correlated with more SO. It did not make a
difference whether qualified nurses were regular or bank staff as both these variables
were negatively correlated with constant SO. Previous research has found SO to be
regarded as an unpleasant and low status activity which can be delegated to junior or
untrained staff [11, 33,34], sometimes against official hospital policy . SO use is
only likely for those patients posing the highest risk to themselves or others, yet this
contrasts sharply with the delegation of intensive psychiatric nursing care to those
least well trained to conduct it. This could have an adverse impact on team
functioning, since greater team cohesion and was associated with less constant SO.
It may be that where multidisciplinary teams made decisions about observation SO
was used less frequently .
There was an inverse correlation between intermittent and constant SO, indicative of
one form of SO being purposefully chosen over another during a shift. The reasons
for initiating SO are wide ranging  and this is confirmed by the present analysis.
Although the study did not measure antecedents to SO directly, the findings indicate
that both forms of SO were usually associated with patient behaviours rather than
broader patient characteristics. Moreover, many of the same conflict items were
associated with intermittent and constant SO, specifically verbal aggression,
absconding, refusing regular and PRN medication, demanding PRN medication and
breaking ward rules. The decision to use intermittent or constant SO in response to
these behaviours would seem, therefore, to reflect other factors. Previous studies
have found large and unexplained differences in SO use between wards and
clinicians in the same hospital [19, 31, 36], exacerbated by the scarce empirical
evidence to support the effectiveness of SO, but also reflecting varying clinical
cultures between wards rather than systematic and consistent assessment.
Physical aggression against others and objects was associated with constant SO.
Violent behaviour is likely to require more intensive supervision of patients in order to
keep themselves and others safe. Over half of violent incidents on wards have
been found to be followed by SO , and SO has been used as a substitute for
seclusion in the management of violent patients in a psychiatric intensive care unit
. Unfortunately, the study cannot determine whether constant SO is reserved for
these most urgent patterns of behaviour or that such incidents occur while patients
are under constant SO. Alternatively, while patients are being observed, others on
the ward may be more likely to become aggressive. The association between
constant SO and patients being sent to psychiatric intensive supports the conclusion
that patients under constant SO have the most disturbed and problematic behaviour,
which may require transfer from the ward if SO and other strategies are not proving
successful. This explanation is less likely to apply to seclusion, which was also
associated with constant SO, because hospital seclusion policies may recommend
staff constantly observe patients placed in seclusion.
Analysis of other containment measures showed a consistent relationship between
SO and coerced and PRN medication. This could mean that patients receiving more
medication require greater observation during or after it is given. Previous studies
have reported that SO is sometimes used in combination with oral or IV
tranquilisation , and that majority of patients have additional (mainly intra-
muscular) medication during constant SO . Another explanation, as the
correlations between conflict behaviours and SO suggest, could be that patients
under SO are more likely to also refuse regular or PRN medication. This might well
involve staff using a show of force to encourage patients to accept their medication,
before resorting to coercion (under manual restraint). Alternatively, while nurses are
busy undertaking SO it becomes easier for other patients to refuse medication in
order to gain attention. This explanation may be less likely because both intermittent
and constant SO were associated with medication refusal, but staff would be
expected to have more time available for other patients during the former.
The results point to the role of ward environment in influencing SO use. A more
complex physical layout was related to greater intermittent SO, possibly because of
difficulty conducting even routine observation of patients on some wards. On the
other hand, a lack of windows in the doors of single rooms may mean that
maintaining a continuous nursing presence is the only practical method of
observation under some circumstances. Ward design has been by identified as
impeding observation in instances of patients committing suicide while under SO [17,
39]. These studies report that patients were more likely to be under intermittent than
constant SO at the time of suicide, and suggest that the former method of observing
at risk patients may be unsafe. However, self-harm was negatively correlated with
intermittent SO in the present study. Given the potentially serious implications for
patient safety, our results suggest that interactions between ward design, observation
practice and patient safety requires further research attention.
One solution to observing patients in poorly designed wards is to use CCTV.
Cameras have been used to detect and verify assaults [40, 41], but the implications
of CCTV use on regular psychiatric wards have not been evaluated. One study
describes the introduction of night-time CCTV on a low secure unit as complementing
rather than replacing traditional methods of observation which provide opportunities
for interaction with patients and confer feelings of safety . This does not appear to
have been the case in the present study. We found less frequent SO when wards
used CCTV and searched patients, indicating that these strategies were being used
as alternative methods of achieving a broader surveillance function. The finding was
most apparent for intermittent SO, perhaps because this was considered as primarily
a means of checking the physical whereabouts of patients and so most akin to
passive surveillance which could also be conducted by CCTV. Alternatively, CCTV
and patient searches could have been regarded as effective and efficient means of
managing self-harm, violence and absconding risk on some wards, to the extent that
SO was required less often.
Keeping patients in locked wards would seem to require more SO, possibly to deal
with the negative consequences of locking ward doors. Staff and patients can view
locked doors such as means of creating a safe environment and preventing patients
from harming themselves and others, but disadvantages include a more volatile ward
environment and depression associated with lack of patient freedom . Although
a frequently cited reason for the use of SO and for locking the ward door is to prevent
absconding by patients, both were positively correlated with SO. The locking of the
ward door is sometimes justified by the argument that SO can thereby be reduced,
however our evidence shows this benefit is not routinely achieved. Other security
measures such as banning items and use of alarms were also positively correlated
with SO, controlling for the levels of conflict behaviours on the ward. It is therefore
possible that SO is being used more than necessary by risk averse and anxious staff.
The principle limitation of this study is its cross-sectional design, which means
observed correlations do not allow firm conclusions about the direction of causality.
Diagnostic information about the patients only included the proportion of patients
suffering from schizophrenia. There is a possibility that some variables were
identified as significant by chance because of the large number of variables entered
in the analyses. However the large scale of the study, the inclusion of potential
confounding variables, and the statistical allowance made for the clustering of
responses by organisation and ward, increase the accuracy and the reliability of the
The study found that in England’s acute psychiatric wards most SO is intermittent in
form, with some evidence that constant SO is used for the most severe incidents and
behaviours. However, there were similarities in some of the behaviours associated
with intermittent and constant SO, particularly with regard to absconding and
medication issues, which are difficult to explain. In this sense, the study supports
previous findings that the functions of SO are confused . The decision to use
intermittent or constant SO for these behaviours may actually reflect variations in
local clinical practices, philosophies and resource availability. There would,
therefore, seem to be scope for national and hospital policies to clarify the
circumstances under which intermittent or constant SO should be initiated. That
said, improved knowledge of what works best and for whom is still required for this
intervention to be delivered with greater confidence and consistency, and for more
efficient management of staffing and financial resources. For example, the use of
both SO and locking ward doors to reduce the risk of patients absconding is
unnecessary as there alternative packages of anti-absconding measures which have
been implemented successfully [45, 46].
It is also of concern that some of the most acutely ill patients are looked after
constantly by the least qualified staff. SO is not a passive intervention. Nurses
employ a range of skills to develop therapeutic relationships with patients under SO
, but reliance on unqualified staff is likely to undermine its potential effectiveness.
In fact, our findings are consistent with others  that the presence of more qualified
and experienced nursing staff on wards may actually reduce the need for constant
The choice to use intermittent or constant SO does not relate solely to patterns of
patient behaviour. The ease with which patients can be observed on the wards
seems to influence the necessity for SO. As many wards are not ideally designed for
observing patients, CCTV, searching and other methods of increasing patient
visibility could be regarded as alternative means of ensuring patient and staff safety.
The case for these forms of surveillance may be strengthened if SO is being routinely
undertaken by less qualified staff and being implemented under financial constraints.
Potentially, freeing up staff time from conducting observations would enable more
engagement with patients and therapeutic activities. Further research into the use of
electronic surveillance is required, but it must consider the role of other forms of
containment if the benefits or otherwise are to be properly established.
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. All views are those of the authors, not the funding body or the
Department of Health. The authors have no conflict of interest to declare, nor do they
have any relevant financial interest in this manuscript.
 Department of Health. Mental Health Nursing: Addressing Acute Concerns.
Standing Nursing and Midwifery Advisory Committee, Department of Health: London,
 Bowers L, Gournay K, Duffy D. Suicide and self-harm in inpatient psychiatric
units: a national survey of observation policies. J Adv Nurs 2000; 32: 437-44.
 Tardiff K. Emergency control measures for psychiatric inpatients. J Nerv Ment
Dis 1981; 169: 614-618.
 Bowers L, Simpson A, Alexander J. Patient-staff conflict: Results of a survey on
acute psychiatric wards. Soc Psychiatry & Psychiatr Epidemiol 2003; 38: 402-8.
 Shugar G, Rehaluk R. Continuous observation for psychiatric inpatients: a critical
evaluation. Compr Psychiatry 1990; 1: 48-55.
 Jones J, Lowe T, Ward M. Inpatients’ experiences of nursing observation on an
acute psychiatric unit: a pilot study. Ment Health Care 2000; 41: 125-129.
 Dennis S. Close observation: how to improve assessments. Nurs Times 1997;
 Langenbach M, Junaid O, Hodgson-Nwaefulu CM, et al. Observation levels in
acute psychiatric admissions. Eur Arch Psychiatry Clin Neurosci 1999; 249: 28-33.
 Neilson P, Brennan W. The use of special observations: an audit within a
psychiatric unit. J Psychiatr Ment Health Nurs 2001; 8: 147-55.
 Hodgson C, Kennedy J, Ruiz P, et al. Who is watching them? A study of the
interpretation of the observation policy in a mental health unit. Psychiatr Bull 1993;
 Duffy D. Out of the shadows: a study of the special observation of suicidal
psychiatric in-patients. J Adv Nurs 1995; 21: 944-50.
 Aidroos N. Nurses' responses to doctors' orders for close observation. Can J
Psychiatry 1986; 31: 831-3.
 Kettles A, Paterson K. Flexible observation: guidelines versus reality. J
Psychiatr Ment Health Nurs 2007; 14: 373-381.
 Cutcliffe J, Barker P. Considering the care of the suicidal client and the case for
'engagement and inspiring hope' or 'observations'. J Psychiatr Ment Health Nurs
2002; 9: 611-21.
 Cardell R, C Pitula. Suicidal inpatients’ perceptions of therapeutic and
nontherapeutic aspects of constant observation. Psychiatr Serv 1999; 50: 1066-
 Flood C, Bowers L, Parkin D. Estimating the Costs of Conflict and Containment
on Adult Acute Inpatient Psychiatric Wards. Nurs Econ 2008; 26: 325-4.
 Meehan J, Kapur N, Hunt I et al. Suicide in mental health in-patients and within 3
months of discharge: national clinical survey. Br J Psychiatry 2006; 188: 129-34.
 Bowers L, Whittington R, Nolan P et al. The relationship between service
ecology, special observation and self-harm during acute in-patient care: the City-128
study. Br J Psychiatry 2008; 193: 395-401.