The built environment, both at a neighborhood level (
1) and at a direct level closer to the individual (
2), may exert an important influence on mental health and well-being. Physical spaces are the aspect of care most clearly remembered by discharged psychiatric patients (
3). Staff spend most or all of their working time there. Since the 1980s, psychiatric care internationally has increasingly been provided outside hospitals (
4–
6). Nonetheless, substantial numbers of inpatient beds remain—in England in 2007, a total of 9,885 beds were provided for inpatients in 554 mental health wards (
7).
Some research has focused upon the built environment in psychiatric care. Examples include work in the United States (
8–
10) and studies in Australia (
11) and the United Kingdom (
12,
13). These studies reported that a wholesale move to a brand new building or substantial refurbishment of an existing facility had modest effects on outcomes, including staff satisfaction and staff sickness rates. Recent research investigating the built environment in inpatient settings has focused upon risk factors for suicide (
14,
15). One systematic review identified randomized trials investigating how the built environment can enhance environments in health care (
16). Interventions to improve sunlight, windows, odors, and seating had positive effects, but the studies were underpowered and nongeneralizable.
Studies to date have one or more methodological drawbacks, including a small or restricted sample, inability to control for organizational and clinical factors affecting outcomes, nongeneralizable settings or patient groups, and inadequate instruments to measure and describe the built environment. In particular, research has treated the built environment as a single “black box” of features, without identifying individual features of the built environment that influence outcomes. Even well-established instruments, such as the Ward Atmosphere Scale, tend to measure the general perceived “atmosphere” of the ward (
17). Studies that have used specific measures of the built environment have relied on subjective evaluations of quality rather than objective description of individual features of the built environment (
18,
19).
Therefore, research is needed to identify features of the built environment that influence ward outcomes. In this research, we aimed to develop an objective measure of the built environment in psychiatric wards, describe the built environment of a large sample of inpatient psychiatric wards in England, and identify objective features of the built environment that were associated with staff evaluations of their wards.
Methods
Sample description and study design
The study was part of a wider investigation of staff morale on psychiatric wards (
20,
21) and received full ethics committee approval. We collected data from 98 wards within 19 participating mental health provider National Health Service (NHS) Trusts in urban and rural areas of England during 2007–2009. Of those 98 wards, 49 were general adult psychiatric wards, 11 were forensic units, 11 were rehabilitation units, nine were psychiatric intensive-care units, nine were wards for older adults, and nine were wards for child and adolescent patients. Acute wards are first-line admission units for adults aged 18–65. Intensive-care units provide compulsory short-term treatment for patients with a severe illness or behavioral disturbance. Adolescent units treat young people up to the age of 17. Forensic units treat patients requiring extended assessment or treatment in a secure setting, often ordered by courts. Older people’s wards treat adults over 65 years for mental illnesses, including dementia, typically for short periods.
Information about the built environment of the ward was collected by using the Ward Design Checklist (WDC), an 18-item instrument completed during an on-site survey by a nonspecialist researcher. A majority of items are designed to be coded during the on-site survey, and the remainder could also be coded through an analysis of ward design plans. Data about staff satisfaction were collected by paper questionnaires that were distributed to all clinical staff (N=2,655) within each ward. Complete data were available for 1,540 staff, representing a 58% response rate.
Measures
Ward Design Checklist.
The WDC comprises objective measures of the built environment, measures a wide range of built environment characteristics, and is suitable for use by nonexperts in the built environment. Each item measures a distinct, easily identifiable, objective physical quality, as opposed to a broader concept requiring a substantial degree of interpretation. None of these individual items were combined within a composite measure.
The WDC was developed on the basis of previous work measuring features of urban environments of different scales (
22,
23). Potential items were partly derived from previous work by one of the authors, in which psychiatric inpatient staff (N=47) and patients (N=30) in six wards in a South London hospital were surveyed about the aspects of the built environment that were most important to them (Sheehan B, unpublished data, 2012).
All raters received training in describing built environments in psychiatric care. Before the instrument was used across the whole sample, a blind interrater reliability exercise was carried out by two researchers on six wards. All 18 items exhibited moderate to excellent agreement (Cohen’s kappa ≥.4 or high intraclass coefficient [ICC]).
The aspects of ward design covered in the WDC were layout (κ=1.00); size (ICC=1.00); density (ICC=1.00); use of single, double, or multiple-bed rooms (κ=1.00); design of nursing station, if present (κ=1.00); opportunities for observation from a central point (ICC=1.00); facilities for staff (κ=1.00); facilities for patients or visitors (κ=.56–1.00); gender separation (κ=1.00); colors of communal areas (κ=1.00); colors of bedrooms (κ=1.00); flooring in communal areas (κ=1.00); flooring in bedrooms (κ=1.00); links between the ward and hospital (κ=1.00); links to outdoor spaces (κ=1.00); views from bedrooms (κ=.72); views from communal rooms (κ=1.00); and outdoor facilities (κ=.66–1.00).
Staff questionnaire.
We designed a new three-item scale to rate staff satisfaction with the built environment. Respondents were asked to rate the physical environment of the ward in terms of overall design, fitness for purpose, and maintenance of safety on a 5-point scale, from 1, very poor, to 5, very good. The three items addressed aspects of design that staff identified as important in an earlier survey and deliberately excluded service or management issues (Sheehan B, unpublished data, 2012).
An exploratory factor analysis of the validity and reliability of the three items suggested a one-factor solution was most suitable, with the single factor explaining 74% of the variance and communality of each item exceeding .65. The internal consistency of the three items was also high (Cronbach’s α=.825). Having supported the existence of a valid and internally consistent measure, we created a mean score across these three items to use as a single measure of this construct.
Using the format of the NHS Staff Survey (
24) wherever possible, we also asked staff structured questions about sociodemographic characteristics, including age, sex, and ethnic background; their profession; and length of service in both mental health services and in their current ward. Psychological symptoms were measured by using the 12-item version of the General Health Questionnaire (GHQ-12) (
25).
Analysis
The analysis comprised two principal stages. First, we described the properties of each ward by using the WDC variables and the demographic characteristics of the staff. Second, we built a series of regression models to test the strength of association between objective elements of the built environment and staff ratings of satisfaction with the built environment. Because staff ratings of satisfaction with the built environment varied both within wards and between wards, we conducted multilevel regression analysis by using SPSS, version 19, statistical software.
Though the employee-level sample was large, the relatively small sample size at the ward level (N=98 wards) necessitated a step-by-step approach in the order of entering predictors to reduce the number of ward-level variables that were simultaneously assessed in any one analysis—hence, avoiding “overfitting” the model to the sample. We conducted the multilevel regression analysis in three steps. Initially, we entered the background demographic variables with potentially confounding effects that we wished to control. Next, we used a second set of five models to control for demographic variables with statistically significant effects at the first step. Each model added one set of variables measuring the broad types of ward features, including layout and shape; staff facilities; patient facilities; appearance; and connections with the hospital or outdoors and outdoor facilities. In the third and final step we added together the ward features from all subsets that had proved to be statistically significant in their respective analyses at step 2.
We used the p<.05 level of statistical significance in testing the fixed effects of built environment features, with two-tailed tests applied throughout. Because of the small sample size at the ward level, and subsequent low power in detecting ward-level effects, we collapsed multicategory WDC items to dichotomies for the purpose of analysis, for example, spoke versus nonspoke layout design and soft versus hard floor coverings.
Results
Wards and staff
Table 1 lists both the staff and the ward characteristics. Over half of the responses were by nurses; other staff who responded included nursing assistants, psychiatrists, occupational therapists, social workers, and psychologists. A total of 27% of 1,451 respondents (N=389) scored >4 on the GHQ-12 (a conventional threshold for diagnosing potential case-level anxiety or depression).
Table 2 shows staff satisfaction with the physical environment. There was no difference between nurses and other groups in satisfaction with overall design, but nurses rated ward environment lower on ensuring safety (t=2.10, df=1,496, p=.036) and on fitness for purpose (t=2.52, df=1,478, p=.012).
Built environment of psychiatric wards
Table 3 shows the features of the wards’ built environment. Ward size varied substantially (range 307–2,789 square meters). Over half of patient room doors could be directly observed from the optimal viewing point on the wards. Spoke designs (L shaped, cruciform, or V shaped) were more common than corridor designs. Single room accommodation was most common. Personal bathrooms were provided for a quarter of rooms. Most wards linked directly to an outside space. Nearly all had a formal nursing station, usually a glass box or an enclosed space with a window facing the ward. Half of wards provided views of only buildings, and half had views of other features, such as gardens, courtyards, or green areas. In most wards, bedrooms and communal areas had neutral or pastel colors, and about half had soft, carpeted flooring.
Built environment features and staff satisfaction
An initial “unconditional” model containing no predictor variables (model 1) assessed the variability within and between wards (
Table 4). It found that 25% of the variability in employee satisfaction with the built environment was composed of between-ward variation (ICC=.25). This substantial ICC statistic showed clear clustering of employee observations by ward and, therefore, a need to use multilevel regression.
Model 2 added potential control variables as predictors, which together reduced the unexplained variation between employees within wards by 4%, from .51 to .49; similarly, they explained 7% of the variation between wards. Racial-ethnic background, occupational group, and ward type each exhibited unique statistically significant effects upon employee satisfaction with the built environment, and, therefore, we retained these variables for the next stage of the modeling process.
We then fitted models 3.1–3.5. For these models, the predictors corresponded to the five subsets of WDC items (items 1–6, layout and shape; item 7, staff facilities; items 8 and 9, patient or visitor facilities; items 10–13, appearance; and items 14–18, connections between the ward and the hospital and the outdoors and outdoor facilities. The layout and shape variables together explained a further 14% of variation between wards. Working in wards with corridors was associated with lower levels of satisfaction with the built environment than working in wards without corridors (B=–.24, p<.05, 95% CI=–.46 to –.03). Working in wards in which patients had a personal bathroom was associated with higher levels of employee satisfaction than working in wards in which patients did not have a personal bathroom (B=.24, p<.05, 95% CI =.01–.47).
Of variables measuring patient facilities, only the existence of a ward-specific treatment room had a statistically significant effect upon employee satisfaction with their physical environment, the effect being negative (B=–.19, p<.05, 95% CI=–.37 to –.01). None of the variables measuring ward appearance, staff facilities, and connections between the ward and the hospital and the outdoors and outdoor facilities had a statistically significant effect; they explained just a further 4%, 5%, and 0% of variation between wards.
When the three significant ward design variables were entered together in model 4, the negative effect of a ward-specific treatment room was no longer statistically significant at the p<.05 level. The effects of corridor-type design and personal bathrooms remained statistically significant (
Table 4). These three variables together explained a further 22% of the total between-wards variance when added to the model containing just the control variables.
Table 4 reports the variance in staff satisfaction with the built environment explained by models 1–4.
Discussion
This study is the first to use objective measures of the built environment to examine the association between design features and staff satisfaction in a large sample of inpatient wards.
The presence of noncorridor design and personal bathrooms had a strong positive association with staff ratings of the built environment. Some features that might be expected to be important to staff, such as views from the ward, colors, flooring, observability of patients, and characteristics of nursing stations, did not exhibit statistically significant effects. Features that appeared to reflect more modern design principles were associated with positive views of the built environment. For example, noncorridor designs, such as spoke designs and courtyard arrangements, are likely to appear less institutional and more modern and to lend themselves to easier observation of patients. The provision of personal bathrooms tends to improve the dignity, privacy, and safety of patients, which inpatient staff work hard to maximize. Any design feature that can improve these patient concerns can only make the ward an easier place to work or stay.
The lack of association between ability to observe from central areas and staff rating of the physical environment was surprising; a preoccupation with ability by staff to observe psychiatric inpatients has been a feature of design since the earliest burgeoning of asylum building (
26). The associations established in this study suggest that staff may be influenced primarily by built-environment features that enhance patient well-being (single rooms and noncorridor design) rather than by enhancement of staff experience, either directly (direct observation and nursing-station type) or indirectly (flooring, types of view, and aesthetic choices).
This study had a number of methodological disadvantages. Cross-sectional designs can show associations rather than suggest causal mechanisms. The response rate of 58% was similar to many surveys among ward staff. Nevertheless, the possibility of selection bias remained. However, we enrolled a large and diverse sample of English NHS psychiatric wards, achieved a substantial sample size, and reported GHQ-12 scores among staff that were comparable to previous reports in such settings (
27). We separately tested for association between staff GHQ-12 scores and staff satisfaction with the built environment, and no independent association was found. Our instrument did not measure some aspects of the wards that may be relevant to staff satisfaction with the built environment. For example, we omitted ward age and treatment orientation, given that we focused only on objective and directly measureable aspects of the built environment. A further possibility was that built-environment features influenced staff stress. However, we found that GHQ-12 scores among staff of different wards varied by only 2.5%, meaning that there was no possibility of significant differences in GHQ-12 scores in wards with a different built environment (a ward-level feature). It was also beyond the scope of this study to address patient experience. Previous research has shown that patient and staff experience of the ward’s built environment may differ substantially (
8,
19), and we cannot infer that patients and staff on the 98 wards studied would experience built-environment features in the same way.
Our findings indicate that the built environment can be described reliably. Future research on environments for psychiatric care can include measures of built-environment features. Clinicians who use such spaces tend to come from disciplines, for example, medicine and nursing, that value empirical research results above expert opinion or tradition. Improved research methods can encourage more rigor in testing hypotheses about the influence of the built environment on patient as well as on staff outcomes.
Conclusions
It is possible to reliably describe built environments in psychiatric care. This research suggests that two built-environment features that are becoming more common, namely, noncorridor designs and personal bathrooms, are likely to contribute to staff satisfaction with the ward built environment. Future research can use tools to describe the built environment and establish its influence on important staff and patient outcomes.
Acknowledgments and disclosures
The project was funded by the National Institute for Health Research (NIHR) Services and Delivery Research Programme (08/1604/142). The views and opinions expressed are those of the authors and do not necessarily reflect those of the program, the NIHR, the NHS, or the Department of Health.
The authors report no competing interests.