The relationship between the nursing practice environment and five nursing‐sensitive patient outcomes in acute care hospitals: A systematic review

Abstract Aim To synthesize the available evidence on the relationship between the nursing practice environment in acute care hospitals and five selected nursing‐sensitive patient outcomes (mortality, medication error, pressure injury, hospital‐acquired infection and patient fall). Design A quantitative systematic review of literature was conducted using the PRISMA reporting guidelines (PROSPERO: CRD42020143104). Methods A systematic review was undertaken up to October 2020 using: CINAHL, MEDLINE and Scopus. The review included studies exploring the relationship between the nursing practice environment in adult acute care settings and one of five selected patient outcomes using administrative data sources. Studies were published in English since 2000. Results Ten studies were included. Seven studies reported that a favourable nursing practice environment reduced the likelihood of mortality in acute care hospitals, but estimates of the effect size varied. Evidence on the association between the nursing practice environment and medication administration error, pressure injury and hospital‐acquired infection was mixed.

Nursing-sensitive patient outcomes are outcomes that are relevant to nurses' scope of practice where evidence has linked nursing inputs or interventions with patient outcomes (Sim et al., 2018;Stalpers et al., 2015). Nurses are not wholly responsible for these patient outcomes but they have a significant influence on the prevention of these outcomes occurring in hospitalized patients (Sim et al., 2018). Patient mortality, medication error, pressure injury, hospital-acquired infection and patient fall are the most frequently examined patient outcomes (Doran, 2011;Dubois et al., 2017). The most commonly examined of these is mortality, which has been studied in relation to the NPE in a large number of primary studies (Lee & Scott, 2018). Only a small number of studies examine the association between the NPE and medication errors, pressure injuries, hospitalacquired infections and patient falls. Research on the relationship between the NPE and these patient outcomes can provide nurses with actionable evidence to advocate for improvements in the NPE and potentially mitigate the risk of these preventable adverse events.

| BACKG ROU N D
A variety of data sources and methods have been historically used to examine patient outcomes, which impacts on the ability to pool results and understand causation (Stalpers et al., 2015). Nurses' perceptions of frequency and/or severity of adverse events are commonly used to examine patient outcomes (Lee & Scott, 2018). Some studies use administrative data such as hospital discharge datasets, risk management or incident reporting systems or annual reports of clinical indicators (Sim et al., 2019;Stalpers et al., 2015). The Nursing Outcomes Classification (Moorhead et al., 2018) and standardized nursing languages are also being increasingly used to document nursing diagnoses, interventions and outcomes within electronic medical records (Oreofe & Oyenike, 2018). This heterogeneity in data sources leads to inconsistencies in measurement and is a challenge for synthesizing research findings.
The existing systematic reviews examining the NPE and patient outcomes combine data from nurse-perceptions of the frequency of adverse events and administrative data sets (Bae, 2011;Lake et al., 2019;Lee & Scott, 2018). Only one systematic review included studies collecting patient outcomes (delirium, malnutrition, pain, patient falls and pressure injuries) from administrative data sources and they did not consider all of the NPE attributes (Stalpers et al., 2015).
Most reviews examine the association between patient outcomes and nursing characteristics, such as nurse staffing, skill mix, nurse education levels, nurse satisfaction, nurse burnout and intention to leave (Bae, 2011;Lake et al., 2019;Lee & Scott, 2018;Stalpers et al., 2015). Only two reviews have investigated the link between the NPE and patient outcomes, and their findings were not conclusive (Bae, 2011;Lee & Scott, 2018). This review, therefore aims to: (a) review and synthesize the available evidence from administrative datasets on the relationship between the nursing practice environment and five selected nursing-sensitive patient outcomes (mortality, medication error, pressure injury, hospital-acquired infection and patient fall) and (b) describe how the nursing practice environment influences theses five patient outcomes.

| Research question
What is the influence of the nursing practice environment on five nursing-sensitive patient outcomes (mortality, medication error, pressure injury, hospital-acquired infection and patient fall) among adults in acute care hospitals using administrative data sources?

| Design
This systematic review was conducted using the guidelines and procedures outlined in the Preferred Reporting Items for Systematic Review and Meta-Analysis reporting guidelines (PRISMA; Moher et al., 2015) (See File S1). The review protocol was registered (CRD42020143104) in the international prospective register of systematic reviews (PROSPERO) at the University of York Centre for Reviews and Dissemination on April 28, 2020.

| Search methods
The initial search was carried out in June 2019 and updated in October 2020 using three online bibliographic databases; Cumulative Index to Nursing and Allied Health Literature (CINAHL) Plus with full text, MEDLINE with full text and Scopus. The following search terms were used for the main independent variable: "practice environment" OR "nurs* work environment" OR "PES-NWI" OR "practice environment scale" OR "nurse work index". A separate search was performed for each outcome as described in Table 1. The search strategy was reviewed by a health librarian to improve the efficacy of the search.

| Inclusion/exclusion criteria
Research studies written in English, and published since 2000, were included if: (a) the study population included nurses working in acute care hospitals where the NPE was measured using a validated instrument; (b) the outcomes examined include adults with one of the five patient outcomes being investigated (mortality, medication error, pressure injury, hospital-acquired infection and patient fall); and (c) data on the outcome of interest were empirically obtained from administrative data sources.

| Search outcomes
A total of 2,122 studies were retrieved from online databases and stored using EndNote X9 (Clarivate Analytics, 2019). After removing duplicate titles, 1,845 studies were subject to title and abstract review by one researcher (TA). Twenty-nine studies were subsequently subject to full-text review which was undertaken independently by two researchers (TA and JS). Finally, 19 studies were excluded with reasons documented in Figure 1. A total of 10 studies were therefore included in the systematic review.

| Quality appraisal
The Appraisal tool for Cross-Sectional Studies (AXIS; Downes et al., 2016) was used to assess the quality of all included studies.
The AXIS tool includes 20 questions which are used to assess the research credibility and integrity of cross-sectional studies in seven domains (Downes et al., 2016). Each question has three options: Yes, No or Unknown (Downes et al., 2016) and the appraisal process does not produce a numerical score.
Two researchers (TA and JS) independently appraised the methodological quality of the selected studies and then discussed and agreed on the final judgement of quality for each study. Ratings of good, moderate or poor were assigned to each study. Overall, all studies were of good quality but with common issues identified in two domains; six studies did not describe the process used for evaluating non-responder bias and eight studies did not provide details about differences in characteristics between responders and nonresponders. There were no studies excluded based on quality appraisal. More information about the quality appraisal of all included studies is in File S2.

Mortality
Mortality OR "failure to rescue" Medication error "medication error" OR "medication administer* error" Pressure injury "pressure inj*" OR "bed sore" OR "pressure ulcer" OR "decubitus ulcer" Hospitalacquired infection "healthcare associated infection" OR "nosocomial infection" OR "hospital acquired infection" OR "central line associated infection" OR "ventilat* associated bloodstream infection" OR "surgical site infection" OR "catheter associated urinary tract infection" Patient fall "fall*" OR "patient fall*" and measurement criteria (instrument used to assess the NPE). The association between the NPE and each patient outcome (mortality, medication error, pressure injury, hospital-acquired infection and patient fall) was examined by extracting the regression coefficient, adjusted or unadjusted odds ratio (OR) and associated confidence intervals for each outcome in each study. If an odds ratio or regression coefficient was not reported, the correlation coefficient was extracted.

| Synthesis
The synthesis without meta-analysis (SWiM) guidelines for systematic reviews (Campbell et al., 2020) were used to guide data synthesis. The SWiM reporting guideline is intended to guide reporting of the synthesis of effect estimates when meta-analysis is not used (Campbell et al., 2020). Meta-analysis was not used in this review because of the large differences in the outcomes measures. Also, for each of the five patient outcomes, the population, study design and findings from included studies were analysed and synthesized. Measures of association between the NPE and each outcome, and statistical significance, were summarized.
Finally, we described the limitations in the methodological approaches of included studies, and described how these could bias the results and limit generalizability.

| Characteristics of included studies
Of the ten included studies, nine were conducted in the USA and one study was conducted in South Korea (Cho et al., 2015). No studies were undertaken in Australia, the United Kingdom or in Europe to assess the influence of the NPE on the selected patient outcomes using administrative data sources. No studies used the Nursing Outcomes Classification or standardized nursing languages. All included studies used a cross-sectional design.

| Nursing practice environment
Nine studies examined the NPE using the Practice Environment Scale-Nursing Work Index (PES-NWI) and one study used the 42-item Perceptions of Nurse Work Environment scale (Stone et al., 2007). These two tools were derived from the same original instrument (Choi et al., 2004), the Nursing Work Index (Kramer & Laurin, 1989). The Perception of Nurse Work Environment Scale comprises four subscales; professional practice; nursing management; staffing and resources adequacy; and nursing process (Choi et al., 2004). The Perceptions of Nurse Work Environment scale used a four-point Likert scale (1 = strongly disagree to 4 = strongly agree), where higher mean composite scores indicate a favourable NPE, and scores were aggregated to the hospital level (Stone et al., 2007).
The PES-NWI consists of five subscales; nurse participation in hospital affairs; nursing foundations of quality care; nurse manager ability, leadership and support of nurses; staffing and resource adequacy; and collegial nurse-physician relationships. For the studies which used PES-NWI, six studies used all subscales to measure the NPE (Cho et al., 2015;Fasolino & Snyder, 2012;Flynn et al., 2012;Friese et al., 2008;Ma & Park, 2015;Olds et al., 2017), and two studies used four subscales (Aiken et al., 2011;McHugh et al., 2016). One study examined three subscales (nursing foundation for quality of care; nurse manager ability, leadership, and support; and collegial/ physician relations; Aiken et al., 2008). All studies using the PES-NWI used a four-point Likert scale (1 = strongly disagree to 4 = strongly agree), and scores were aggregated to either the unit level (Flynn et al., 2012;Ma & Park, 2015) or hospital level (Aiken et al., , 2011Cho et al., 2015;Fasolino & Snyder, 2012;Friese et al., 2008;McHugh et al., 2016;Olds et al., 2017).
Several approaches were taken to determine whether the NPE was favourable. Three studies used the mean score of the composite PES-NWI scale; where a higher mean composite score indicated a favourable NPE (Fasolino & Snyder, 2012;Flynn et al., 2012;Stone et al., 2007). This approach was used by Fasolino and Snyder

| Mortality
Mortality was measured in seven studies (Aiken et al., , 2011Cho et al., 2015;Friese et al., 2008;McHugh et al., 2016;Olds et al., 2017;Stone et al., 2007). Mortality was inconsistently defined across studies. Three studies used 30-day mortality (where death occurred either in hospital or postdischarge within 30 days of admission) (Aiken et al., 2011;Cho et al., 2015;Olds et al., 2017). Three studies reported 30-day inpatient mortality Friese et al., 2008;Stone et al., 2007), and one reported 30-day inpatient survival (McHugh et al., 2016). All studies reported that a favourable NPE was associated with a reduction in the likelihood of mortality, but estimates of the contribution of the NPE to the reduction in mortality varied. Cho et al., (2015)  In intensive care unit patients, Stone et al., (2007) found no association between mortality and the NPE (OR = 0.97, 95% CI: 0.90-1.02).

| Medication error
Two of the included studies examined the relationship between the NPE and medication error, with one study assessing association using correlation (Fasolino & Snyder, 2012) and the other using hierarchical linear regression, accounting for clustering within hospitals (Flynn et al., 2012). In both studies, medication error was defined as number of medication errors per 1,000 patient bed days. Fasolino and Snyder (2012) identified that NPE had a weak positive association with medication error rate (r = .15, p < .01) instead of the expected inverse correlation. Flynn et al., (2012), found no association between the NPE and medication error rates (β = −0.26, t = −0.87, p < .388).

| Pressure injury
The relationship between the NPE and pressure injury was examined in two studies (Ma & Park, 2015;Stone et al., 2007). In both studies, pressure injury was defined as the number of patients with any stage of pressure injury per 1,000 discharges, and the likelihood of a pressure injury developing was modelled. Ma and Park (2015) found a significant relationship between the NPE and development of pressure injury, where each one-unit improvement in the composite NPE score led to reduced odds of developing a pressure injury by 29% (OR = 0.71, 95% CI: 0.55-0.91, p = .013). Stone et al., (2007) found no relationship between the NPE and the likelihood of developing a pressure injury (OR = 1.06, 95% CI: 0.83-1.37).

| Hospital-acquired infection
Hospital-acquired infections were reported in only one of the ten studies (Stone et al., 2007). This study measured three types of

| Patient fall
None of the included studies examined patient falls.

| D ISCUSS I ON
The aim of this study was to synthesize the literature on the relationship between the NPE and five selected nursing-sensitive patient outcomes (mortality, medication error, pressure injury, hospitalacquired infection and patient fall) from administrative data sources.
Most studies reported that mortality was less likely in hospitals with a favourable NPE, but the considerable variance in the results suggests the magnitude of this relationship is unclear. This inconsistency may result in part from a discrepancy in the operational definitions of mortality used within included studies, differences in populations and settings or from inconsistencies in measurement/classification of favourable practice environments.
Surprisingly, none of the included studies examined the impact of the NPE on falls or falls resulting in patient harm. Many studies that use falls as a patient outcome examine nurses' perceptions of the numbers of falls as a proxy measure (Ausserhofer et al., 2013;Duffield et al., 2011;Swiger et al., 2018). However, studies that used nurse perceptions on the frequency and/or severity of adverse events were excluded from this review. The lack of empirical data examining the association between the NPE and falls should be addressed in future research because patient falls are a significant and frequent adverse event in hospitals (Manojlovich et al., 2016).
Within the Australian healthcare system, 26,060 falls occur annually at a cost of $9.8 million each year to the Victorian and New South Wales states alone (Morello et al., 2015).
In this review, two studies examined medication errors. Fasolino and Snyder (2012) found that a favourable NPE led to higher rates of reported medication errors while Flynn et al. (2012) found no association between these variables. Reporting bias may influence the reliability of results on medication errors for two reasons. Firstly, under-reporting of medication error is well documented in the literature (Pfeiffer et al., 2013) with the reasons including fear of consequences (Yung et al., 2016) and lack of knowledge about using the reporting system (Rutledge et al., 2018). Secondly, a favourable NPE may improve the safety culture and thus increase the reporting rate (Yoo & Kim, 2017). Therefore, increased rates of medication errors may result from increased reporting of medication errors rather than any change in the NPE. In addition, nurses make up only one part of the team that are responsible for medication errors. Nurses are involved in administration errors but prescribing and dispensing errors usually relate to other healthcare professionals (Doran, 2011).
Neither study reported on the phase in which the error occurred (Fasolino & Snyder, 2012;Flynn et al., 2012). This means the errors may or may not have been due to the nurse incorrectly administering the medication and the NPE may not have had any association with medication error rates. In future research, measures of medication error at the administration stage would provide better insight into the relationship between the NPE and medication errors resulting from nursing care. Alternative sources of data, such as chart review or direct observation, could also be used to capture medication errors; which may be more reliable (Flynn et al., 2012).
Pressure injuries have been documented as an indicator of high quality nursing care (Rodgers et al., 2020a(Rodgers et al., , 2020bStalpers et al., 2015). This review found mixed findings on whether the NPE was associated with pressure injury (Ma & Park, 2015;Stone et al., 2007). One study found that a favourable NPE reduced the odds of developing a pressure injury (Ma & Park, 2015). In contrast, Stone et al., (2007) found that the NPE was not associated with pressure injury development. These differences may relate to the different populations and research settings in each study. The study by Stone et al., (2007) was conducted among elderly people in intensive care units, where critically ill patients are at higher risk of developing skin breakdown (Coyer et al., 2017). Both studies reported that an increase in nursing hours per patient day reduced the rate of pressure injuries (Ma & Park, 2015;Stone et al., 2007). This indicates that nurse staffing and resource adequacy may be the key element of the NPE that influences the development of pressure injuries. Further studies are needed to strengthen the evidence on the relationship between the NPE and pressure injury and may include using risk adjustments to minimize confounding factors and control for nurse staffing.
The most common types of hospital-acquired infections are central line-associated blood stream infection, catheter-associated urinary tract infection and ventilator-associated pneumonia (Doran, 2011;Javadi Bashar, 2019). Mixed results were reported in the only included study that examined this outcome (Stone et al., 2007). Stone et al. (2007) (Stone et al., 2007). Further research is also required to explore hospital-acquired infections in non-critical care settings.

| CON CLUS ION
This review has identified that an improved NPE was associated with a reduction in mortality in seven studies. Due to variations in how mortality was defined and measured within included studies a metaanalysis was not conducted. Other outcome measures (medication error, pressure injury and hospital-acquired infection) only included small numbers of studies, and the association between the NPE and these patient outcomes was varied. No studies examined the association between the NPE and falls. More research needs to be done to understand the relationship between the NPE and these nursingsensitive patient outcomes in acute care hospital settings.
Improving the NPE is important for both nurses and patients.
Nurses provide continuous care for hospitalized patients and this review demonstrates that a favourable NPE is associated with a reduction in mortality. Nurse Managers should actively promote improvements to the NPE to reduce patient mortality and potentially improve other patient outcomes. Supportive leadership and effective management practices should focus on ensuring appropriate evidence-based nurse staffing, nurse participation in hospital decision-making, effective collegial nurse-physician relationships and that the culture is focused on providing high quality patient care. Concepts such as nurse staffing and resource adequacy may be the key driver for improving some nursingsensitive patient outcomes such as pressure injuries (Shin et al., 2019).
A favourable NPE may also improve the safety culture, which in-turn leads to a non-punitive response to errors and enables nurses to report and then learn from error (Chiang et al., 2017). Nursing leaders play an important role in improving the NPE and reducing preventable adverse events in hospital settings.

ACK N OWLED G EM ENTS
The authors wish to thank Ms Adrienne Corradini, Social Sciences Librarian at the University of Wollongong, for assistance in designing the search strategy.

CO N FLI C T O F I NTE R E S T
The authors report no conflict of interest.

AUTH O R CO NTR I B UTI O N S
TA, JS, LL: Substantial contributions to conception and design, or acquisition of data, or analysis and interpretation of data; Drafting of the manuscript or revising it critically for important intellectual content; Final approval of the version to be published and take responsibility for the content; Agreed to be accountable for all aspects of the work.

PROS PE RO R EG I S TR ATI O N
The review protocol was registered (CRD42020143104) in the international prospective register of systematic reviews (PROSPERO) at the University of York Centre for Reviews and Dissemination on April 28, 2020.

E TH I C A L A PPROVA L
This was a systematic review of previously published research and did not include human participants. It was therefore exempt from an ethics committee review.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data used to support the findings of this study are available from the corresponding author upon reasonable request.