Associations of Patient and Staff Outcomes With Inpatient Unit Designs Incorporating Decentralized Caregiver Workstations: A Systematic Review of Empirical Evidence

Abstract

Objectives:

This systematic literature review synthesizes and assesses quality of research addressing associations of patient and staff outcomes with inpatient unit designs incorporating decentralized caregiver workstations.

Background:

A current hospital design trend is to include decentralized caregiver workstations on inpatient units. A review of literature addressing decentralized unit design is needed.

Methods:

The systematic review methodology was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement. Database searches were conducted for studies published in peer-reviewed journals through October 2017. Included were empirical studies associating patient and/or staff outcomes and unit design with decentralized caregiver workstations. Individual studies were evaluated for quality using established methods, and Grading of Recommendations Assessment, Development and Evaluation (GRADE) and GRADE-Confidence in the Evidence from Reviews of Qualitative Research (GRADE-CERQual) guided rigorous inspection of evidence quality and strength for quantitative outcomes and qualitative findings, respectively.

Results:

The search yielded 1,096 records with 36 full-text articles examined and 12 articles included in the final review. This work was dominated by studies with limited analyses. Staff outcomes have been most widely studied, especially collaboration/communication and walking. Overall, studies exploring decentralized nursing as a design intervention have produced limited results for both staff and patient outcomes. Strength of evidence of the current literature with quantitative methods as a whole was rated very low quality.

Conclusions:

Although varying degrees of caregiver workstation decentralization in inpatient units are now common, the literature addressing the impacts of such designs is of very low quality and shows inconsistency in associated outcomes. Rigorous, well-designed studies with consistently defined design and outcome measures are needed for greater confidence in determining any effects of decentralized unit design.

Keywords

decentralized centralized nursing unit design inpatient unit design systematic literature review evidence-based design evidence quality

In recent years, inpatient units with decentralized caregiver workstations and, often concomitantly, single-bed patient rooms have become common for new and renovated hospital facilities in the United States (Hamm, 2011; Zborowsky, Bunker-Hellmich, Morelli, & O’Neill, 2010). Traditionally, most hospital units included a mix of single- and multibed patient rooms with a single, centralized staff work area (Hamm, 2011; Noskin & Peterson, 2001). Prior to the adoption of electronic health records (EHRs), this centralized work area included storage for paper-based records in addition to providing workspace for charting activities and face-to-face interaction (Wakefield, 2002).

New or renovated U.S. hospital construction incorporating decentralized caregiver stations in inpatient unit designs has typically also included a change from inclusion of both multi- and single-occupancy rooms to solely single-bed patient rooms. Single-bed patient rooms were required in new hospital construction projects in the 2006 edition of the U.S.-based FGI Guidelines for Design and Construction of Hospital and Health Care Facilities (Facility Guidelines Institute, 2006). Thus, any effects of decentralization have often been intertwined with effects of single-bed patient rooms that generally result in a larger unit footprint.

Decentralized unit design strategies aim to increase staff time in direct patient care, to reduce risk of patient falls, and to reduce noisy conditions that may be associated with centralized multiseat work areas (Flynn, 2005; McCarthy, 2004). Many current inpatient unit designs feature decentralized caregiver workstations in windowed corridor alcoves directly adjacent to patient rooms (Hamilton, Swoboda, Lee, & Anderson, 2018) in an attempt to support both EHR documentation and visibility of patients. In practice, decentralized unit models have included a range of designs, from several multiseat caregiver workstations within a single unit to solely individual workstations distributed adjacent to patient rooms throughout a unit (Cai & Zimring, 2011; Hua, Becker, Wurmser, Bliss-Holtz, & Hedges, 2012). Five typical examples are illustrated in Figure 1.

Figure 1.

Levels of caregiver workstation decentralization.

Patient acuity plays a role in the design and number of decentralized caregiver workstations as well. For high-acuity units where patients’ conditions are severe and volatile (e.g., intensive care), it is a high priority to support nurses’ ability to monitor patients continuously and respond quickly to emergent needs (Hamilton et al., 2018). Patient acuity also influences staffing (Lang, Hodge, Olson, Romano, & Kravitz, 2004) and in turn the type and frequency of staff workstations. As nurse staffing levels have been shown to impact both employee and patient outcomes (Lang et al., 2004), this variable, along with patient acuity, may be important to consider in any unit-based comparisons of outcomes.

While impacts of decentralized hospital units on staff and patient outcomes have been a topic of interest in practice and in the literature, this body of evidence has not been assessed as a whole. This systematic review is intended to provide an up-to-date resource for designers and hospital facility decision makers regarding decentralized inpatient unit design’s potential impacts on patients and staff in adult acute and intensive care units. The specific aims are as follows:

report on the current state of the peer-reviewed, empirical evidence base pertaining to design for decentralized caregiving in adult acute or intensive care inpatient units;

identify convergence of findings across studies;

assess the quality of each study and for the body of evidence as a whole for each outcome; and

delineate implications for future practice and research.

Method

Literature Searches and Source Selection

The methods of this systematic review were developed using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Liberati et al., 2009). Two researchers conducted searches and selected articles based upon the defined inclusion and exclusion criteria. Inclusion criteria required empirical studies published as full-text articles in peer-reviewed journals that addressed patient and/or staff outcomes of decentralized caregiving layouts for adult, nonbehavioral health inpatient units. Both quantitative and qualitative studies were considered. Conference abstracts, articles in non-peer-reviewed journals or magazines, simulation studies, and opinion/theory-based articles were excluded.

Database searches for studies published in English, from January 1900 to October 2017, were conducted in MEDLINE/PubMed, CINAHL, and Health Business Elite. Medical Subject Heading terms and key word alternates used in the searches are shown in Table 1. The source selection process, summarized in Figure 2, included an initial review for title relevance, followed by a review of abstracts, and finally an in-depth analysis of full-text articles. To ensure thorough coverage, we also examined references within relevant articles.

Table 1.

Databases and Search Terms.

Search	Databases	Terms
1	MEDLINE/PubMed CINAHL Health Business Elite	Decentrali* AND Nurs*
2	MEDLINE/PubMed CINAHL Health Business Elite	Decentrali* AND Unit Design
3	MEDLINE/PubMed CINAHL Health Business Elite	Decentrali* AND Layout
4	MEDLINE/PubMed CINAHL Health Business Elite	Pod AND Unit Design
5	MEDLINE/PubMed CINAHL Health Business Elite	Pod AND Layout
6	MEDLINE/PubMed CINAHL Health Business Elite	Pod AND Nurs* AND Station
7	MEDLINE/PubMed CINAHL Health Business Elite	Satellite AND Nurs* AND Station

Figure 2.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram of article inclusion and exclusion.

Data extraction included information reported about the study design, sample population and sample size, research approach, patient acuity of unit(s), beds per unit(s), nurse to patient ratio, level of workstation decentralization, location(s) of support spaces/supplies, outcome variable(s), and study findings.

Quality of Studies and Evidence Assessments

All studies included in this review were assessed for quality individually, and evidence quality was assessed across groups of studies by outcome. Individual studies were assessed within each category of research design (e.g., cross-sectional) and rated as good, fair, or poor. All studies had the possibility of being rated good quality. Considerations included specifying inclusion/exclusion criteria, assessing the representativeness of the participants, ensuring appropriate numbers of participants through sample size calculations, performing appropriate statistical tests, and controlling for confounding variables. Quantitative studies were assessed using methods from the National Heart, Lung, and Blood Institute (2018). Qualitative studies were assessed by the methods of Walsh and Downe (2006). For mixed methods studies, the quantitative and qualitative components were assessed separately.

The Grading of Recommendations Assessment, Development and Evaluation (GRADE) framework guided the assessment of quality of evidence by each outcome across all studies that measured an effect (i.e., a statistical effect of unit design on a single outcome; Guyatt et al., 2011). GRADE Pro software (GRADEpro GDT, 2015) was used to develop a GRADE evidence profile table. The GRADE-Confidence in the Evidence from Reviews of Qualitative Research (CERQual) approach was used to assess the quality of evidence by finding for qualitative studies (Lewin et al., 2018). As GRADE and GRADE-CERQual assess evidence synthesis, they could only be applied when an outcome or finding was reported by more than one study.

Results

Quality Assessment of Included Individual Studies

A total of 12 studies were included in the review. All studies were observational. One was a pre–post study with a separate control group and was rated as fair quality. Five were pre–post studies without a control group; four were rated as fair quality and one as poor quality. Four studies were cross-sectional in design. Three of these were rated as fair quality and one as poor quality (see Table 2 for additional details). Among qualitative studies, two were qualitative alone and two utilized mixed methods. The qualitative-only studies were rated as good quality, and the two mixed methods studies were rated as poor and fair for qualitative analysis (see Table 2).

Table 2.

Summary and Quality Assessment of Individual Articles.

Author	Year	Study Design	Population	Sample Size	Environmental Variable(s)	Outcome Variable(s)	Findings Relevant to Design	Overall Quality Rating
Bayramzadeh and Alkazemi	2014	Cross-sectional	Acute care (four units in one hospital)	Centralized: 30 RN Decentralized: 40 RN	Centralized versus decentralized	– Type of communication (SR)	– Higher agreement for most communication being face to face in decentralized units (p = .03) – Lower agreement for most communication being through technology in decentralized units (p = .02)	Fair
Copeland and Chambers	2016	Pre–post with no control group	Acute care (three units in one hospital)	Pre: 26 RN Post: 35 RN	Centralized versus hybrid decentralized plus decentralized support spaces	– Energy expenditure (ped.) – Steps per shift (ped.) – Job satisfaction (SR)	– Nurse job satisfaction-No change – Lower energy expenditure in decentralized (p < .001) – Fewer steps (p = .041) in decentralized	Fair
Fay et al.	2017	Pre–post with no control group and qualitative	Acute care and intensive care (four units in one hospital)	Pre: 24 Staff (walking) 45 Staff (survey) Post: 21 Staff (walking) 98 Staff (survey)	Centralized versus decentralized	– Distance walked per shift (ped.) – Layout supports teamwork (SR), individual work (SR), walking distance (SR) efficiency of patient care (SR) – Room utilization (obs.)	– No statistical comparison of distance – Lower staff perception of layout’s support of teamwork (p < .01), no difference in support of individual work, lower perception of reasonable walking distances (p < .01), and lower perceived efficiency of patient care (p < .01) in decentralized – No statistical comparison for time/utilization	Fair
Friese et al.	2014	Pre–post with no control group	Acute care (one unit in one hospital)	Pre: NS patients Post: NS patients 19 Staff	Centralized versus decentralized (pod teams of two nurses)	– Patient satisfaction (SR) – Patient falls (records) – Patient call light use (records) – Nurse satisfaction (SR) – Nurse use of overtime (records)	– No statistical comparison between unit types for any outcomes	Poor
Gurascio-Howard and Malloch	2007	Cross-sectional and qualitative	Acute care (one unit each at two hospitals)	Centralized: NS Patients 4 RN (shadowed) Decentralized: NS Patients 4 RN (shadowed)	Centralized versus decentralized	– Communication (shadow) – Teamwork (shadow)	– No statistical comparison reported for any quantitative data – Unclear how qualitative data were analyzed	Poor
Hua et al.	2012	Pre–post with a control group	Acute care (four units in one hospital)	Pre: NS patients 31 RN (survey) 42 RN (shadowing) 25 RN (walking) Post: NS patients 68 RN (survey) 43 RN (shadowing) 11 RN (walking) (did not separate out control numbers)	Pre units (mix of centralized and decentralized) versus post units (decentralized hub design) Control unit (Decentralized) constant pre and post	– Patient satisfaction (SR) – Patient falls and length of stay (records) – Nursing team collaboration (SR) – Organizational climate – Job satisfaction (SR) – Communication (shadow) – Steps/distance per shift (ped.) – Nursing time in patient rooms (shadow)	– No statistical comparison was reported for patient satisfaction – No statistical comparison was reported for patient falls or length of stay – No difference in team collaboration – Higher rating of organizational climate post move for moved units (p = .003) – No difference in nurse job satisfaction – 4.9 min per shift more nurse communication with physicians in the new versus old units (p = .03), but only 4% of nurses’ daily communication – Nurses walked 985 fewer steps per shift in the new versus old units (p = .004) – Nurses spent less time in patient rooms in the moved units versus the control unit (p = .03) both pre and post move	Fair
Krugman et al.	2015	Pre–post with no control group	Acute care and intensive care (six units in one hospital)	Pre: 159 RN (survey) NS work activity NS walking Post: 176 RN (survey) NS work activity NS walking	Centralized versus decentralized pod design	– Perceptions of work environment (SR) – Nurse work activity (PDA) – Steps walked per shift (ped.)	– Higher staff satisfaction with physical work environment in the decentralized units (p < .001) – No change in nurse work activity – Pedometers rated as unreliable	Fair
Parker et al.	2012	Cross-sectional and qualitative	Acute care (two units in one hospital)	Centralized: 11 RN (survey) 14 RN (focus group) Decentralized: 39 RN (survey) 5 RN (focus group)	Centralized versus decentralized pod design	– Perceived stress (SR) – Work environment (SR) – Qualitative focus groups	– No differences in perceived stress – No difference in work environment – Qualitative feedback on work environment was mixed in both groups; however, nurses in the centralized units expressed greater sense of teamwork and camaraderie compared to those in the decentralized unit, who felt more isolated	Fair (cross-sectional) Fair (qualitative)
Pati et al.	2015	Pre–post with no control group	Acute care (three units in one hospital)	Pre: 141 (110 RN, 31 support, 3 MD; survey) 2 RN (on each shift; activities, walking) Post: 141 (85 RN, 24 support, 1 MD; survey) 2 RN on each shift (activities, walking)	Centralized versus decentralized with team station	– Nurse frequency of activities (PDA) – Distance walked (ped.) – Acute stress (SR) – Presenteeism (SR) – Staff collaboration (SR)	– No statistical comparison was reported for nursing activities – No statistical comparison was reported for walking distance – No significant change in acute stress – No significant change in presenteeism – Of 126 items on the staff survey, 7 were statistically significantly different between the pre and post groups within each unit	Fair
Real et al.	2016	Qualitative	Acute care and intensive care (six units in one hospital)	35 Healthcare professionals (post move only)	Centralized versus decentralized with central station plus decentralized supplies	– Communication (FG)	– Decentralized nurse stations changed system interdependencies by reducing nurse-to-nurse interactions and teamwork while heightening nurse interdependencies and teamwork with other healthcare occupations – Nursing communities of practices were adversely affected by the decentralized design	Good (qualitative)
Zborowsky et al.	2010	Cross-sectional and qualitative	Acute care (six units in three hospitals)	57 RN total for survey 13 RN total for focus groups NS RN total for observations NS numbers for centralized or decentralized	Centralized versus decentralized (any unit with more than one nursing station of any size or configuration was classified as decentralized)	– Nurse functional space use (obs.) – Patient visibility (obs.) – Nurse perceptions of work environment (SR) – Sound levels (decibel meter)	– Nurses in all three hospitals spent more time on tasks in all behavior categories in centralized versus decentralized units – No difference in sound levels based on nursing station design, although levels exceeded recommendations in all units – No statistical comparison was reported for perceptions of walking distance (survey considered unreliable) – Qualitative analysis indicated nurse interest in support, sense of control, acknowledgment of professional privacy, and balance of needs for separation and connectedness; many indicated a preference for a centralized station along with several decentralized stations	Fair (cross-sectional) Poor (qualitative)
Zhang et al.	2015	Qualitative	Acute care	Post move: 20 (14 RN, 1 LPN, 5 NA)	Centralized versus decentralized with a teaming station Supplies centralized in both units	– Communication and teamwork (FG)	– Communication and teamwork among nurses and nursing staff were compromised in the hybrid design, with themes including sense of physical, social, and professional isolation – Reduction in distraction and improvement in nurse time management in the hybrid design, as well as nurse perception of a quieter patient environment and increased patient safety – Overall, nurses felt that the advantages of hybrid design outweighed the disadvantages	Good (qualitative)

Key. FG = focus group data collection; NS = not specified; obs. = observation; PDA = self-recorded on PDA device; ped. = pedometer-measured; records = hospital data; shadow = recording by a shadowing researcher; SR = self-report; RN = Registered Nurse; LPN = Licensed Practical Nurse; NA = Nursing Assistant.

Staff Outcomes and Evidence Quality by Outcome

Literature on decentralized caregiver workstations in inpatient units has frequently focused on staff outcomes. All 12 of the included studies measured nursing staff outcomes. These included time with patients, documentation time and frequency, time in various nursing activities, environmental and job satisfaction, job-related stress, walking distance, collaboration and communication, noise and noise levels, visibility, and overtime usage.

Literature on decentralized caregiver workstations in inpatient units has frequently focused on staff outcomes. All 12 of the included studies measured nursing staff outcomes.

Time in patient care/patient room

Time spent in patient care or in a patient room was assessed in three studies. One study reported more staff visits to (centralized M = 9.36; decentralized M = 15.64) and time spent in (centralized M = 9 min; decentralized M = 24 min) patient rooms on decentralized versus centralized units during observed shifts (Fay, Carll-White, Schadler, Isaacs, & Real, 2017). Another study reported more time spent with patients in decentralized versus centralized settings through an observational shadow log (Gurascio-Howard & Malloch, 2007). No statistical analyses were performed in either study. A third study compared a multihub design to other centralized or decentralized units (Hua et al., 2012). A control group remained in a decentralized unit with distributed stations in addition to one central station. The multihub design included multiseat stations and charting alcoves near the patient rooms. Time spent with patients in the moved units was significantly lower than in the control unit in both the pre- and the postperiods (p = .031), suggesting no change with the move to a multihub design (Hua et al., 2012). Only one study measured effects using statistical analysis, so no GRADE assessment was possible.

Documentation frequency and time

Two studies investigated nurse documentation. Pati, Harvey, Redden, Summers, and Pati (2015) reported more frequent charting and documentation in a decentralized versus a centralized model, but no statistical analysis was provided. Another study suggested more total time charting in a decentralized versus centralized environment, 28% versus 21% of shift time, respectively, but no statistical analysis was presented (Gurascio-Howard & Malloch, 2007). Without any statistical analysis, there is no effect on which to apply GRADE.

Nurse activities

Three studies described nurse activities. Pati et al. (2015) reported that time on the unit, documenting, and in the medication room increased, while time in supply storage and at the nurse station decreased after changing to a decentralized design but provided no statistical analysis. In a pre–post study, Krugman, Sanders, and Kinney (2015) categorized nurse activities as value-adding, non-value-adding, and necessary and found a statistically significant decrease in value-adding activities and an increase in non-value-adding activities, over time (p = .009) in one unit with no change in the other five units (Krugman, Sanders, & Kinney, 2015). Zborowsky, Bunker-Hellmich, Morelli, and O’Neill (2010) found that nurses on centralized units spent significantly more time using telephones (centralized M = 10.3 min, SD = 6.7; decentralized M = 2.4 min, SD = 3.1; p < .0001), computers (centralized M = 18.2 min, SD = 3.4; decentralized M = 7.6 min, SD = 5.6; p < .0001), and in other administrative duties (centralized M = 17.5 min, SD = 4.6; decentralized M = 8.6 min, SD = 6.6; p < .0001) than nurses on decentralized units. Substantial heterogeneity in definition and measurement prevented a quality assessment using the GRADE approach.

Work environment and job satisfaction

Two studies presented findings on the work environment. From premove to postmove, Krugman et al. (2015) found an improvement in nurses’ perceptions of their physical work environment using a self-authored survey (centralized M = 07.42, SD = 4.39; decentralized M = 122.81, SD = 12.27; p < .001). Friese et al. (2014) measured satisfaction with a decentralized pod design on a self-authored questionnaire. Of the responding nurses, 31.6% indicated better communication with physicians and nursing assistants, 84.2% agreed/strongly agreed unbalanced patient acuity across assignments was an issue, 78.9% said they were more efficient in answering call lights, and 68.4% agreed/strongly agreed pods should remain the unit’s model of care (Friese et al., 2014). However, with no centralized comparison group, these findings are only descriptive.

Four studies measured job satisfaction. Two of these used the Demand-Control-Support Questionnaire (DCSQ). One study reported no significant difference between the decentralized and centralized units (centralized M = 4.40, SD = .34; decentralized M = 4.26, SD = 0.35; p = .26; Parker, Eisen, & Bell, 2012). These authors used focus groups for further explanation and suggested that work efficiency and access to patients was greater on centralized compared to decentralized units (Parker et al., 2012). Zborowsky et al. (2010) also employed the DCSQ and found no statistical difference between unit types, but the response rate was deemed too low (5–31%) to be representative. Reponses to open-ended questions indicated both types of units were dissatisfied with space, walking distances, supply areas, sound levels, workflow, visibility, furniture, equipment, crowding, and climate (Zborowsky et al., 2010). Copeland and Chambers (2016) asked nurses to rate their job satisfaction on a 0–10 scale and reported no difference between designs (centralized M = 7.4, SD = 1.8; decentralized M = 8.0, SD = 1.3; p > .05). Responses to open-ended questions suggested that nursing stations make work difficult in centralized units, while in decentralized units, noise from the nursing stations was eliminated, but teamwork was negatively affected by reduced coworker contact (Copeland & Chambers, 2016). Hua, Becker, Wurmser, Bliss-Holtz, and Hedges (2012) measured job satisfaction on a subscale of the Nursing Team Collaboration Survey and found no difference with the change from centralized and decentralized units to a decentralized, multihub unit. However, nurses with more than 3 years at the hospital reported increased satisfaction in the decentralized multihub unit, while satisfaction declined in nurses with <3 years tenure (p = .015; Hua et al., 2012).

The GRADE approach was not used for work environment satisfaction because of heterogeneity in the measurement. The GRADE quality assessment of the evidence on job satisfaction’s association with nursing unit design was very low (Table 3). Since all four studies were observational, the starting GRADE assessment was low. Based on risk of bias, imprecision, and indirectness, quality was further downgraded, and no upgrade criteria were applicable.

Table 3.

GRADE Evidence Quality for Staff Outcomes.

Staff Outcome	No. of Studies	Study Design	Risk of Bias^a	Inconsistency^b	Indirectness^c	Imprecision^d	Upgrade Criteria^e	Quality Assessment^f
Job satisfaction	4	Observational	Very serious	Not serious	Very serious	Very serious	None	Very low
Job-related stress	3	Observational	Very serious	Not serious	Very serious	Very serious	None	Very low
Walking distance	4	Observational	Very serious	serious	Very serious	Very serious	None	Very low
Collaboration/communication (questionnaire)	3	Observational	Very serious	serious	Very serious	Very serious	None	Very low
Collaboration/communication (observation)	2	Observational	Very serious	Not serious	Very serious	Very serious	None	Very low

Note. GRADE = Grading of Recommendations Assessment, Development, and Evaluation.

^a Risk of bias was determined by assessing failure to apply appropriate eligibility criteria (i.e., the inclusion of a control group), flawed measurement of either exposure or outcome, failure to adequately control confounding variables, or incomplete follow-up (i.e., attrition; Guyatt et al., 2011). ^bInconsistency was defined as a large and unexplained inconsistency of results across studies (Guyatt et al., 2011). ^cIndirectness was determined by assessing differences in interventions (e.g., decentralized nurse station is defined differently across studies); differences between population and sample, differences in outcome measures, and indirect comparison (i.e., no head-to-head comparison of two or more interventions of interest; Guyatt et al., 2011). ^dImprecision was defined by the lack of a confidence interval for effect size, or a risk ratio under 0.75 or over 1.25 (Guyatt et al., 2011). ^eUpgrade criteria were a large magnitude of effect, the presence of a dose response gradient, and/or plausible confounding that could increase the estimated effect. Down grade criteria precede consideration to rate up the quality of evidence. Upgrade criteria should rarely be applied if limitations are present in any down grade criteria (Guyatt al., 2011). ^fThe GRADE quality assessment rating addresses evidence quality only for outcomes addressed in more than one study (Guyatt et al., 2011).

Job-related stress

Three studies measured stress among nurses. One study used the Perceived Stress Scale (Cohen, Kamarck, & Mermelstein, 1983) to measure perceptions of stress and found no significant difference (centralized M = 3.39, SD = 0.26; decentralized M = 3.31, SD = 0.27; p = .36; Parker et al., 2012). Pati et al. (2015) used the Stress and Arousal Adjective Checklist (King, Burrows, & Stanley, 1983) and found no significant difference between the units (centralized M = −5.57, −1.42, −2.16; decentralized M = −5.55, −1.04, −1.77; p > .05). Hua et al. (2012) measured stress on a subscale of the Nursing Team Collaboration Survey and found no significant difference between pre- and postmove or between the control group and the moved units.

The GRADE quality assessment of the evidence on the association between decentralized nursing and job-related stress was very low (Table 3). Observational studies start the GRADE assessment as low. Based on risk of bias, imprecision, and indirectness, quality was further downgraded. No included studies for this outcome met the upgrade criteria.

Walking distance

Five studies measured walking distance. In one study, nurse average walking distance increased significantly during a 12-hr shift for two of the three units after moving to a decentralized design (centralized M = 2.05, 2.50, 2.84 miles; decentralized M = 3.79, 2.89, 3.50 miles; p = .001, .175, .004) based on pedometer measures (Pati, Harvey, Redden, Summers, & Pati, 2015). Additionally, three questions suggested a statistically significant increase in perceived time walking (p < .05) from centralized to decentralized in all three study units (Pati et al., 2015). A mixed methods study reported walking distances to be higher for nurses postmove (centralized M = 3.75 miles; decentralized M = 4.00 miles), but no statistical analysis was reported (Fay et al., 2017). Nurses also perceived walking distances to be less reasonable on the decentralized unit (Fay et al., 2017). Krugman et al. (2015) found no difference in walking distances based on unit design (centralized M = 7,318 steps; decentralized M = 7,713 steps, p > .05). However, reported outliers and variability in pedometer performance suggest questionable validity of their results. Copeland and Chambers (2016) reported that nurses took fewer steps (centralized M = 8,363, SD = 2,549 steps; decentralized M = 8,093, SD = 2,173 steps; p = .041) on decentralized versus centralized units (Copeland & Chambers, 2016). A final study found that the move to decentralized multihub units significantly reduced steps compared to the premove units (p = .004), while steps in the decentralized control unit did not change (Hua et al., 2012).

Assessment of the quality of evidence on the relationship between walking distance and nursing unit design was very low (Table 3). The four included studies that measured an effect were observational; thus, the starting GRADE assessment was low. Quality was further downgraded based on very serious risk of bias, serious limitations due to consistency, and very serious limitations of imprecision and indirectness. None of the included studies met the upgrade criteria.

Collaboration and communication

Collaboration and related indicators such as communication and teamwork were assessed in five studies using quantitative measures and qualitatively in three studies. Two types of quantitative measurements were used: survey questions or direct observation of interactions.

Four studies used survey questions to measure collaboration and communication. Hua et al. (2012) used a validated survey and found no statistically significant change from pre to post or for the decentralized control unit. Two studies indicated a significant decrease in agreement that the unit layout supports teamwork after decentralization: one assessing three units (centralized M = 4.72, 4.63, 4.15; decentralized M = 2.25, 3.18, 2.86; p = .05; Pati et al., 2015) and one overall (centralized M = 4.21, decentralized M = 2.94, p = .00; Fay et al., 2017). Finally, Bayramzadeh and Alkazemi (2014) found that nurses on decentralized compared to centralized units had lower disagreement that most communication was face-to-face in their unit (centralized M = 2.15, SD = 0.87; decentralized M = 1.77, SD = 0.68, p = .034). Nurses on the decentralized compared to centralized units also had higher disagreement that most communications between nursing stations occur through technology (centralized M = 3.10, SD = 0.79; decentralized M = 3.60, SD = 0.91, p = .017; Bayramzadeh & Alkazemi, 2014).

Three studies reported direct observation of interactions between staff and other providers. In one study, the Clinical Work Measurement Tool was used (Hua et al., 2012). Nurse communication did not change significantly in most categories including information, assistance, consultation, education, and patient information but did change significantly for social communication comparing a decentralized multihub design with premove designs (premove 7.75% of total shadowing time; postmove 4.91% of total shadowing time; p = .004; Hua et al., 2012). These authors also reported mixed results on frequency of communication. Nurse–doctor communication at nursing stations increased postmove compared to premove with no change in the decentralized control unit, while the overall information exchange between doctors and nurses in the new decentralized multihub design decreased, pre to post, with no change in the control unit (Hua et al., 2012). Zborowsky et al. (2010) observed significantly fewer medical staff consultations (centralized M = 10.1, SD = 5.7; decentralized M = 3.8, SD = 3.8; p < .0001) and social interactions (centralized M = 4.0, SD = 4.2; decentralized M = 1.1, SD = 1.7; p = .0012) in the decentralized models compared to centralized models. Another study shadowed four registered nurses in each environment during one 8-hr shift and found that nurses in the decentralized unit spent a larger proportion of their time in all types of communication except providing patient information to other caregivers (centralized 31%; decentralized 83%), but no statistical analysis was performed (Gurascio-Howard & Malloch, 2007).

Three qualitative studies addressed communication and collaboration. One study in a decentralized environment suggested that learning to work as a team was an advantage (Zhang, Soroken, Laccetti, De Castillero, & Konadu, 2015). Other studies indicated that centralized models may be preferred for professional communication (Friese et al., 2014; Parker et al., 2012). Two studies identified lower social support and a sense of isolation as issues that need to be addressed in decentralized environments (Real, Bardach, & Bardach, 2016; Zhang et al., 2015).

Assessment of the quality of evidence was evaluated using GRADE for quantitative methods (separately for surveys and observations) and GRADE-CERQual for qualitative methods. Bayramzadeh and Alkazemi (2014) was not included in GRADE, since they measured a different construct than the other studies. Three studies were included for the survey measurement and two were included for observations. The GRADE assessment was very low for both measurement methods (Table 3). The starting GRADE assessment was low since studies were observational. Quality was downgraded based on very serious risk of bias, and very serious limitations in imprecision and indirectness and no upgrade criteria applied. The assessment of the qualitative evidence on communication and collaboration produced moderate confidence when combining the results from three studies (Table 4). All studies reported similar findings; however, one study was less robust than the others, resulting in minor concerns about methodological limitations and minor concerns regarding adequacy of data across the body of evidence.

Table 4.

GRADE-CERQual Confidence in Qualitative Findings.

Review Finding	Studies Contributing to the Review Finding	Assessment of Methodological Limitations^a	Assessment of Coherence^b	Adequacy of Data^c	Overall CERQual Assessment of Confidence^d	Explanation of Judgment
Decentralized nursing unit designs reduce peer visibility and communication, therefore reducing collaborative teamwork and peer support	Parker et al. Real et al. Zhang et al.	Minor concerns (two studies with minor and one study with substantial methodological limitations)	No concerns (all three studies reported similar findings)	Minor concerns (two studies had substantial findings for this outcome)	Moderate confidence	This finding was graded as moderate confidence due to minor concerns with methodological limitations and few studies regarding this outcome

Note. GRADE = Grading of Recommendations Assessment, Development, and Evaluation; CERQual = Confidence in the Evidence from Reviews of Qualitative Research.

^a The extent to which there are concerns about the design or conduct of the primary studies that contributed evidence to an individual review finding (Lewin et al., 2018). ^bAn assessment of how clear, well-supported, and compelling the fit is between the data from the primary studies and a review of findings that synthesizes that data (Lewin et al., 2018). ^cAn overall determination of the degree of richness and quantity of data supporting a finding (Lewin et al., 2018). ^dThe extent to which the body of evidence from the primary studies supporting a review finding is applicable to the context (i.e., perspective or population, phenomenon of interest, setting) specified in the review question (Lewin et al., 2018).

Perception of noise and objectively measured sound levels

Observed outcome measurements for noise were quantitative and qualitative perceptions of sound and objective sound levels. One study asked about nurse satisfaction with noise level on the unit on a 4-point scale and reported a significant increase between the pre- and postperiods (pre M = 2.53; post M = 2.85; p < .001; Krugman et al., 2015). A qualitative study also suggested that satisfaction with noise levels increased after a change from a centralized to a decentralized hybrid model with a central station plus distributed alcove stations (Zhang et al., 2015). Another study measured unit noise levels objectively with decibel meters. Zborowsky et al. (2010) found no difference in noise levels on centralized versus decentralized nursing units, and decibel levels exceeded recommendations on all units. Differences in the constructs precluded an assessment of quality of the evidence using the GRADE framework.

Visibility

Visibility includes the staff’s ability to see patients, rooms, beds, and/or peers from their workstation. Three studies measured some type of visibility. Zborowsky et al. (2010) counted patient beds and patient rooms in direct sight from the nursing station and the number of monitors used in the nursing station and found that neither centralized nor decentralized stations offered superior visibility to patient rooms or beds. Rather, they noted the importance of the placement of nursing stations in relation to patient rooms and the number of vital sign monitors used in the nurse stations. Fay, Carll-White, Schadler, Isaacs, and Real (2017) counted nurses’ peer-to-peer lines of sight as well as patient visibility on centralized (pre) units and decentralized (post) units. They found that in the centralized acute and progressive care units, nurses could see each other, but not patient beds from the workstations, while on the decentralized units, nurses could see one to two patients and potentially multiple peers (Fay et al., 2017). In the centralized intensive care environment, glass doors allowed nurses at the centralized station visibility of more patients than the decentralized environment, but peer visibility was similar in both designs (Fay et al., 2017). Krugman et al. (2015) assessed visibility using a single survey item assessing satisfaction with hallway visibility for call lights and patients ambulating. The mean score on a 4-point scale was 2.47 on a centralized unit and 2.41 on a decentralized unit; no statistical analysis was conducted. The authors noted that a curved hallway in the decentralized unit affected long-range visibility (Krugman et al., 2015). Included literature regarding room, bed, patient, and peer visibility is merely descriptive and does not include comparative analysis. Therefore, there is no evidence of effect, and GRADE was not applicable.

Nurse overtime usage

Only Friese et al. (2014) measured nurses’ use of overtime. Over the 12-month study period, overtime usage decreased from 1,350 min in the centralized design to 1,168 min in the decentralized pod design (Friese et al., 2014), but no statistical analysis was conducted.

Patient Outcomes and Evidence Quality by Outcome

Four studies reported specific patient outcomes related to decentralized inpatient unit design, including length of stay (LOS), hospital-acquired pressure ulcers, call light usage, falls, and patient satisfaction.

LOS and hospital-acquired pressure ulcers

Two outcomes, LOS and hospital-acquired pressure ulcers, were included in one study, with only aggregated data available, rendering inferential analysis impossible (Hua et al., 2012). None of the included studies assessed how unit design relates to LOS or hospital-acquired pressure ulcers.

Call light usage

One study found that the number of call lights per month decreased over the study with the highest number of calls in the month prior to the change to decentralized nursing, but no statistical analysis was performed (Friese et al., 2014).

Falls

Three studies explored patient falls by unit design. One study reported that the monthly fall rate dropped from 4.8 to 3.7 falls per 1,000 patient days after switching from a centralized model to a decentralized pod model (Friese et al., 2014). However, no statistical comparisons were performed. Another study was unable to conduct an analysis due to availability of aggregated data only (Hua et al., 2012). A qualitative study found that nurses felt that the decentralized design would help prevent patient falls (Zhang et al., 2015). With no statistical analysis of quantitative data, there is no measure of the effect of decentralized nursing on falls.

Patient satisfaction

Three studies reported measurement of patient satisfaction. Hua et al. (2012) reported on 4 items from the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS, 2018). Questions asked about nurses treating the patient with courtesy and respect, nurses explaining things in a way the patient could understand, patients receiving help as soon as they wanted, and pain control. Average satisfaction ratings were higher on all 4 items after a move to a decentralized multihub nursing model from centralized and decentralized models, but due to aggregate data, no statistical analysis was possible (Hua et al., 2012). A second study also reported higher patient satisfaction with prompt response to calls in decentralized versus centralized settings, but no statistical analysis was performed (Gurascio-Howard & Malloch, 2007). Friese et al. (2014) reported satisfaction with overall nursing care on a 0–100 point scale. Over the study period, satisfaction with nursing care as measured by the Press-Ganey patient satisfaction survey changed from 84.8 to 86.4, but no statistical analysis was reported (Friese et al., 2014). For all studies, only descriptive data were reported for patient satisfaction. With no statistical calculations, there is no evidence of effect.

No studies exploring patient outcomes reported statistical analysis

Since the GRADE assessment assumes a measurement of effect, it cannot apply to patient outcomes. Only one qualitative study included patient outcomes, precluding use of GRADE-CERQual.

Discussion

Overall, evidence is limited as to the effect of inpatient unit designs with decentralized workstations both on patient and on staff outcomes. No studies measured an effect of centralized versus decentralized unit design on patient outcomes; only descriptive data have been reported. Therefore, there is currently no evidence to support claims of differences in patient outcomes based on decentralized unit design. A few more studies assessed staff outcomes data. Results were mixed, with some positive and negative effects. However, for all outcomes included in the GRADE assessment, the evidence was rated as very low quality. Overall, study quality of the quantitative studies was fair to poor with high potential for bias. Some good quality qualitative studies have been done, but these do not provide substantial evidence on their own. Thus, higher quality research is needed to produce convincing evidence to support outcomes-oriented inpatient unit design decision-making.

This systematic review revealed several important issues with the current research on inpatient unit design and the ability to draw conclusions based on this body of work. Several study design issues should be considered in developing future studies: (1) inclusion of more data describing specific unit designs and variations, (2) designation of a control group, (3) engagement in appropriate statistical analysis methods, (4) addressing other possible bias, and (5) standardization of outcome measures. Each of these considerations is detailed below.

Several study design issues should be considered in developing future studies: (1) inclusion of more data describing specific unit designs and variations, (2) designation of a control group, (3) engagement in appropriate statistical analysis methods, (4) addressing other possible bias, and (5) standardization of outcome measures.

Inclusion of Specific Design Variables

Implementing high-quality studies is challenging in the design field, since projects are specific to the space and community in which they are planned. Different footprints and layouts may affect walking distances and other outcomes more than broad definitions of centralized and decentralized design. Therefore, we must be careful in overly generalized assessments and, if possible, assess outcome changes based on specific design measures and unit differences or groupings of such. Relevant variables could include not only nurse station location but also other design features such as supply and medication room locations, as well as operational variables such as staffing ratios, and methodologies for nurse assignment to patient rooms. If more studies are done that include outcome data along with detailed design and operational information, stronger and more applicable findings are possible.

Designating a Control Group

Another challenge with studies of the built environment is that there is often a lack of a true control group. Pre–post study designs are becoming more common, but groups differ in not only the design but also the time period when measurements are taken. Having a separate control group, such as in the included study by Hua et al. (2012), would help determine the effects of the design rather than the confounding effects of other events or changes over time.

Performing Appropriate Statistical Analysis

Of substantial concern in the current literature is the fact that few studies actually performed statistical analysis. The omission was partly due to use of aggregate data that were not at a level to be able to compare specific units. Many studies only reported mean values and did not assess variability either through providing variance estimates or confidence intervals. Even providing this basic information would allow for a more complete assessment of data collected.

For those studies that did perform statistical analysis, the methods used were generally overly simplistic. Use of t tests and chi-square tests does not control for confounding, the hierarchical nature of the data, and potential biases in observational studies. For studies with multiple units or hospitals, a hierarchical analysis would be an appropriate way to combine the data overall rather than reporting a within-unit or within-hospital analysis only. Simple statistical methods also do not account for well-known potential confounding factors such as age, gender, and tenure for nurses or acuity, age, gender, and socioeconomic status for patients. Seasonal variation in outcomes may also need to be explored.

Addressing Other Possible Bias

Selection bias and measurement error also likely influenced study results in this review. These issues can be addressed by clearly identifying the population of interest and comparing it to the study sample and using well-established measurements, such as surveys that are valid and reliable.

Standardization of Outcome Measures

A recommendation for the field is to agree on a set of valid and reliable measurement tools to allow for comparison and future meta-analyses. Data sources should be thoroughly vetted to ensure the best possible data and understand any threats to reliability and validity. The process of standardization is being addressed in the health sciences through such collaborative efforts as The National Institutes of Health Toolbox. This set of well-designed and validated tools has been identified to facilitate data sharing and comparison across studies (Weintraub et al., 2013).

Other issues in study design are not so easily solved, such as when multiple factors change at the same time. In decentralized nursing, most units not only changed the nursing set-up but also moved to single patient rooms so that any effect cannot be solely associated with a single change. Such limitations should be openly discussed and acknowledged in research articles.

The field of evidence-based design requires, and must demand, rigorous, accurate, and effective study design and research methods. While characteristics of the field make it difficult to conduct randomized controlled trials, well-designed quasi-experiments or natural experiments are possible and would produce much stronger evidence than what emerged in this review. Inclusion of control groups, attention to internal validity, and appropriate analyses controlling for confounding variables will strengthen the body of evidence in the future.

Implications for Practice

The current body of literature does not provide definitive evidence to warrant either a centralized or decentralized inpatient unit design.

There are limited, but consistent, qualitative findings that indicate issues with teamwork, support, and communication in decentralized environments, warranting further study.

Practitioners should consider potential impacts of organizations' staffing models and adoption of technology in designing inpatient units.

It is critically important to further build the evidence base in this area by integrating well-designed studies with consistent environmental, organizational, and outcome measures in hospital design projects.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was funded by HDR.

ORCID iD

Jeri Brittin, PhD, Allied ASID, EDAC

Supplemental Material

Supplemental material for this article is available online.

References

Bayramzadeh

Alkazemi

M. F.

(2014). Centralized vs. decentralized nursing stations: An evaluation of the implications of communication technologies in healthcare. Health Environments Research & Design Journal, 7, 62–80.

Cai

Zimring

(2011). Nursing culture and performance: The impact of nurse station typology on nurses’ informal communication and learning. World Health Design, 61, 60–67.

Cohen

Kamarck

Mermelstein

(1983). A global measure of perceived stress. Journal of Health and Social Behavior, 24, 385–396.

Copeland

Chambers

(2016). Effects of unit design on acute care nurses’ walking distances, energy expenditure, and job satisfaction: A pre-post relocation study. Health Environments Research & Design Journal, 10, 22–36.

Facility Guidelines Institute. (2006). Guidelines for design and construction of healthcare facilities. Retrieved from https://www.fgiguidelines.org/guidelines/earlier-editions/

Fay

Carll-White

Schadler

Isaacs

Real

(2017). Shifting landscapes: The impact of centralized and decentralized nursing station models on the efficiency of care. Health Environments Research & Design Journal, 10, 80–94.

Flynn

(2005). Healthcare boom: Nursing stations for the 21st century. Building Design and Construction, 1–5. Retrieved from https://www.bdcnetwork.com/healthcare-boom-nursing-stations-21st-century

Friese

C. R.

Grunawalt

J. C.

Bhullar

Bihlmeyer

Chang

Wood

(2014). Pod nursing on a medical/surgical unit: Implementation and outcomes evaluation. The Journal of Nursing Administration, 44, 207–211.

GRADEpro GDT. (2015). GRADEpro guideline development tool [software] (developed by evidence prime, inc.). McMaster University, Retrieved from https://gradepro.org

10.

Gurascio-Howard

Malloch

(2007). Centralized and decentralized nurse station design: An examination of caregiver communication, work activities, and technology. Health Environments Research & Design Journal, 1, 44–57.

11.

Guyatt

Oxman

A. D.

Akl

E. A.

Kunz

Vist

Brozek

… Jaeschke

(2011). GRADE guidelines: 1. Introduction—GRADE evidence profiles and summary of findings tables. Journal of Clinical Epidemiology, 64, 383–394.

12.

Hamilton

D. K.

Swoboda

S. M.

Lee

Anderson

D. C.

(2018). Decentralization: The corridor is the problem, not the alcove. Critical Care Nursing Quarterly, 41, 3–9.

13.

Hamm

L. E.

(2011). Improving the effectiveness of nursing work area design in inpatient care units. (Master of Science). Available From All Theses. (1095). Retrieved from https://tigerprints.clemson.edu/cgi/viewcontent.cgi?article=2095&context=all_theses

14.

Hospital Consumer Assessment of Healthcare Providers and Systems. (2018). Retrieved from http://www.hcahpsonline.org/en/

15.

Hua

Becker

Wurmser

Bliss-Holtz

Hedges

(2012). Effects of nursing unit spatial layout on nursing team communication patterns, quality of care, and patient safety. Health Environments Research & Design Journal, 6, 8–38.

16.

King

Burrows

Stanley

(1983). Measurement of stress and arousal: Validation of the stress/arousal adjective checklist. British Journal of Psychology, 74, 473–479.

17.

Krugman

Sanders

Kinney

L. J.

(2015). Part 2: Evaluation and outcomes of an evidence-based facility design project. The Journal of Nursing Administration, 45, 84–92.

18.

Lang

T. A.

Hodge

Olson

Romano

P. S.

Kravitz

R. L.

(2004). Nurse-patient ratios: A systematic review on the effects of nurse staffing on patient, nurse employee, and hospital outcomes. Journal of Nursing Administration, 34, 326–337.

19.

Lewin

Booth

Glenton

Munthe-Kaas

Rashidian

Wainwright

… Noyes

(2018). Applying GRADE-CERQual to qualitative evidence synthesis findings: Introduction to the series. Implementation Science, 13, 2.

20.

Liberati

Altman

D. G.

Tetzlaff

Mulrow

Gotzsche

P. C.

Ionnidis

J. P. A.

… Moher

(2009). The PRISMA statement for reporting systematic reviews and meta-analyses of studies that evaluate health care interventions: Explanation and elaboration. PLoS Med, 6, e1000100.

21.

McCarthy

(2014). Healthy design. The Lancet, 364, 405–406.

22.

Noskin

G. A.

Peterson

L. R.

(2001). Engineering infection control through facility design. Emerging Infectious Diseases, 7, 345–357.

23.

Parker

Eisen

Bell

(2012). Comparing centralized vs. decentralized nursing unit design as a determinant of stress and job satisfaction. Journal of Nursing Education and Practice, 2, 66–76.

24.

Pati

Harvey

T. E.

Jr Redden

Summers

Pati

(2015). An empirical examination of the impacts of decentralized nursing unit design. Health Environments Research & Design Journal, 8, 56–70.

25.

Real

Bardach

S. H.

Bardach

D. R.

(2016). The role of the built environment: How decentralized nurse stations shape communication, patient care processes, and patient outcomes. Health Communication, 32, 1557–1570.

26.

The National Heart Lung and Blood Institute. (2018). Study quality assessment tools. Retrieved from https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools

27.

Wakefield

(2002). The changing “shape” of the nursing station. Contemporary Nurse, 13, 148–157.

28.

Walsh

Downe

(2006). Appraising the quality of qualitative research. Midwifery, 22, 108–119.

29.

Weintraub

Dikmen

S. S.

Heaton

R. K.

Tulsky

D. S.

Zelazo

P. D.

Bauer

P. J.

… Gershon

R. C.

(2013). Cognition assessment using the NIH toolbox. Neurology, 11, S54–S64.

30.

Zborowsky

Bunker-Hellmich

Morelli

O’Neill

(2010). Centralized vs. decentralized nursing stations: Effects on nurses’ functional use of space and work environment. Health Environments Research & Design Journal, 3, 19–42.

31.

Zhang

Soroken

Laccetti

De Castillero

E. R.

Konadu

(2015). Centralized to hybrid nurse station: Communication and teamwork among nursing staff. Journal of Nursing Education and Practice, 5, 34–41.