| Abstract Background: The impact of registered nurse (RN) staffing on patient care quality has been extensively studied. Identifying additional modifiable work environment factors linked to patient care quality is critical as the projected shortage of approximately 250,000 RNs over the next 15 years will limit institutions’ ability to rely on RN staffing alone to ensure high-quality care. Purpose: We examined the association between RNs’ ratings of patient care quality and several novel work environment factors adjusting for the effects of two staffing variables: reported patient-to-RN ratios and ratings of staffing adequacy. Methodology: We used a cross-sectional, correlational design and a mailed survey to collect data in 2009 from a national sample of RNs (n = 1,439) in the United States. A multivariate logistic regression was used to analyze the data. Findings: Workgroup cohesion, nurse–physician relations, procedural justice, organizational constraints, and physical work environment were associated with RNs’ ratings of quality, adjusting for staffing. Furthermore, employment in a Magnet hospital and job satisfaction were positively related to ratings of quality, whereas supervisory support was not. Practice Implications: Our evidence demonstrates the importance of considering RN work environment factors other than staffing when planning improvements in patient care quality. Health care managers can use the results of our study to strategically allocate resources toward work environment factors that have the potential to improve quality of care. | Hospitals in the United States are under increasing financial pressure (Centers for Medicare and Medicaid Services, 2010) to provide high-quality patient care. Two strategies that hospitals can use to bolster patient care quality are (a) ensuring adequate registered nurse (RN) staffing (Kane, Shamliyan, Mueller, Duval, & Wilt, 2007) and (b) improving other aspects of RNs’ work environments (Institute of Medicine, 2004) to support patient care delivery by RNs. A robust body of research relates RN staffing to a number of patient care quality indicators, such as patient mortality, length of stay, and multiple adverse events (Harless & Mark, 2010; Kane et al., 2007; Sales et al., 2008). In these studies, data on RN work environment factors other than staffing are not included, because administrative databases do not capture them. Addressing this problem is critical, because the projected workforce shortage of approximately 250,000 RNs in the United States over the next 15 years (Buerhaus, Auerbach, & Staiger, 2009) will limit institutions’ ability to rely only on RN staffing to ensure high-quality patient care. Front line staff RNs’ assessments of their work environment and patient care quality have been related to actual patient outcomes, such as 30-day mortality (Hansen, Williams, & Singer, 2011; Tourangeau et al., 2006). Furthermore, RNs are the largest group of health care providers in hospitals (Institute of Medicine, 2011). Assessing their views on patient care quality is one of several essential components for helping health care managers to better focus quality improvement efforts. There are few U.S. studies identifying which factors are associated with RN-rated patient care quality (Aiken, Clarke, & Sloane, 2002; Aiken, Clarke, Sloane, Lake, & Cheney, 2009; Friese, 2005; Patrician, Shang, & Lake, 2010) and patient outcomes (Aiken et al., 2009; Friese, Lake, Aiken, Silber, & Sochalski, 2008; McHugh, Kutney-Lee, Cimiotti, Sloane, & Aiken, 2011; Stone et al., 2007), adjusting for RN staffing. In the few that exist, except for Friese (2005) and Kim, Capezuti, Boltz, and Faichild (2009), researchers use a summed score of various work environment factors as a measure of the overall work environment quality (Aiken et al., 2002, 2009; Friese et al., 2008; Patrician et al., 2010; Stone et al., 2007). This makes it difficult to determine what the independent effects of the factors are on the quality of care, which consequently hampers strategic investment in work environment improvement. We aim to extend the existing evidence in four ways. First, we examine the association between RN-rated patient care quality and physical work environment, workgroup cohesion, and RN personality traits, which have previously received little research attention but may influence RNs’ ratings of quality (Carayon, 2009). Second, we use a measurement model that synthesizes work environment factors that have been linked to patient care quality in the past (Estabrooks, Midodzi, Cummings, Ricker, & Giovannetti, 2005; Gurses, Carayon, & Wall, 2009; Kenaszchuk, Wilkins, Reeves, Zwarenstein, & Russell, 2010; Laschinger, 2008; Patrician et al., 2010; Stone et al., 2007) but have not been studied together. Third, we use data from a national sample of RNs. With the exception of the study of U.S. army hospitals by Patrician et al. (2010), the related studies use data from individual U.S. states (Aiken et al., 2009; Friese, 2005; Kim et al., 2009), Canada (Laschinger, 2008), or European countries (Poghosyan, Clarke, Finlayson, & Aiken, 2010). Using a well-specified measurement model and a national RN sample can increase our confidence in the findings regarding which RN work environment factors, in addition to RN staffing, are related to RN-rated patient care quality. Fourth, we identify the independent effects of different work environment factors instead of using a summed score of overall work environment quality. This is important for helping managers to focus their improvements. |
Back to Top | Conceptual Framework A work system design model (Carayon, 2009; Carayon et al., 2006) was used to select variables. It extends Donabedian’s structure–process–outcome model of quality by including structural components, such as physical work environment, and emphasizing the relationship between employee and patient outcomes (Carayon et al., 2006). The model captures the complexity of RN work environments by specifying factors across several components—person (e.g., education), task (e.g., autonomy), organization (e.g., nurse–physician relations), physical work environment (e.g., lighting), and tools and technologies (e.g., lack of supplies)—hypothesized to be associated with patient, employee, and organizational outcomes. Carayon explains that, “According to the work system model, tasks are performed by an individual who uses tools and technologies; the tasks are performed in a physical environment and under organizational conditions” (Carayon, 2009, p. 317). Balancing these factors in work environment redesign is likely to optimize work system performance. Optimal outcomes can be achieved by either removing the factors of the work system that negatively impact outcomes or by “identifying aspects of the work system that can be used to compensate for the negative aspects” (Carayon, 2009, p. 319). For example, the impact of patient-to-RN ratios on patient care quality may be optimized when an RN’s lack of work experience is offset with good coworker support and workplace design that facilitates collaboration. The association between patient care quality indicators and the following factors has been empirically examined: education, autonomy (Estabrooks et al., 2005), quantitative workload, physical work environment (Gurses et al., 2009), nurse–physician relations, nurse manager support, nurse involvement in organizational decision-making, RN perceptions of staffing adequacy (Kim et al., 2009), RN-reported patient-to-RN ratios (Patrician et al., 2010), job satisfaction, coworker support (Kenaszchuk et al., 2010), structural empowerment defined as access to information and resources and support to do the job and opportunity for professional growth (Laschinger, 2008), and Magnet recognition (Stone et al., 2007). On the basis of the model of Carayon (2009), we included three additional person factors (positive affectivity, negative affectivity, and work motivation) and one task factor (variety). We operationalized this model with the variables listed in Figure 1. Conceptual definitions, scale sources, sample items, item ranges, and response options for these variables are presented in Tables 1 and 2. |
| Outcome variable. We used RNs’ ratings of patient care quality as the indicator of quality. It was derived from the question, “How likely is it that a patient will receive high-quality care in your work setting (your unit or organization if not working in hospital setting)?” The question is similar to the single-item perceived quality question used by others (Patrician et al., 2010), measured on a 4-point scale, ranging from 1 (very likely) to 4 (not likely at all). Because of a small number of data points in the not likely at all (n = 28) and not too likely (n = 2) categories, which make analysis unstable, these categories were combined with somewhat likely to create a dichotomous outcome variable with categories coded as 1 (very likely) and 0 (somewhat likely, not too likely, and not likely at all). Predictor variables. There were 18 predictor variables: four person variables (work motivation, negative affectivity, positive affectivity, and education), three task variables (variety, autonomy, and quantitative workload), nine organization variables (nurse–physician relations, workgroup cohesion, supervisory support, organizational constraints, promotional opportunity, procedural justice, reported patient-to-RN ratio, perceptions of staffing adequacy, and Magnet recognition), one physical work environment variable, and one RN job outcome variable (job satisfaction). Fourteen of the predictor variables were measured with multi-item Likert-type scales (see Table 2). To collect information about staffing variables, we asked (a) how many patients the surveyed RN cared for during his or her most recent shift (those who reported not caring for patients or clients or caring for more than 20 patients or clients per shift were excluded from the analysis) and (b) how often were there not enough RNs to adequately care for the patients, with response options ranging from 1 (every shift) to 4 (rarely). Control variables. On the basis of the model used in this study and prior empirical work (Estabrooks et al., 2005; Kenaszchuk et al., 2010; Kim et al., 2009; Poghosyan et al., 2010, Stone et al., 2007), we controlled for the effects of several variables that have been reported to influence RNs’ ratings of patient care quality. The control variables were: gender, race/ethnicity, work status, type of position, shift type, work schedule, unit type, work experience, age, and wage. We aimed to test direct effects of person, task, organization, and physical work environment factors on RN-rated patient care quality to answer the following research question: What are the independent effects of RN work environment factors on RN-rated patient care quality, adjusting for RN staffing? |
Back to Top | Methods Ethical approval for the study was obtained from the institutional review boards at the participating institutions. |
Back to Top | Design We used a cross-sectional, correlational study design. Data were collected by mailed survey between January and May of 2009 in the third wave of a multiyear study designed to follow the career trajectories in a national sample of RNs over 10 years (Kovner et al., 2007). We used a modified Dillman (2007) design, which included five contacts with participants and a $5 cash incentive. The question concerning RNs’ ratings of patient care quality was only included in the Wave 3 survey. |
Back to Top | Sample Selection A mailed survey was administered in 2006 (Wave 1) to a sample of RNs licensed for the first time by exam between August 1, 2004 and July 31, 2005, in 1 of 60 sites (51 metropolitan statistical areas and 9 rural counties) in 34 states and the District of Columbia. We replicated the sampling design used for the Community Tracking Study done by the Center for Studying Health System Change (CSHSC; 2003). The Center for Studying Health System Change chose the 60 sites at random with certain known probabilities to represent the U.S. population. We used these same probabilities along with estimates of the number of newly licensed RNs in each site and estimates of each site’s eligibility rate in allocating the sample. The 3,370 respondents to the 2006 survey (response rate of 58%) were surveyed again in 2007 (Wave 2) and 2009 (Wave 3). The sample for this study was selected from the RNs who responded to the Wave 3 survey and who reported working in a hospital (n = 2,007; response rate = 68%). |
Back to Top | Survey Instrument The 98-question survey, described in detail elsewhere (Kovner et al., 2007), was designed to collect data about RN personal (e.g., education), work environment (e.g., nurse-physician relations), and market (e.g., unemployment rate) factors. An expert RN workforce panel, composed of six individuals selected in conjunction with our funding agency, reviewed the survey prior to Wave 3. Minor modifications to the survey were made following the panel’s review. All of the questions were pilot tested with 15 recent RN graduates and undergraduate nursing students in their senior year. The expert panel established content-related validity for all questions. Evidence of construct validity for all multi-item scales presented in Table 2 was established with confirmatory factor analysis on the basis of the incremental and nonincremental fit indices in a sample of 365 RNs from a metropolitan, academic medical center in New York (Model [chi]2 (196) = 437.63, p < .001, comparative fit index = .97, Tucker–Lewis index = .98, root mean square error of approximation = .06; Djukic, Kovner, Budin, & Norman, 2010) and in the Wave 1 sample of our multiyear study (Model [chi]2 (3338) = 13,633.37, p < .001, comparative fit index = .92, Tucker–Lewis index = .92, root mean square error of approximation = .03), with the exception of the physical work environment scale, which was added to the Wave 3 survey. Internal consistency reliability of all the scales using data from this study was acceptable with a Cronbach’s alpha of >.70. |
Back to Top | Data Analysis The data were collected and analyzed at the nurse level. Descriptive statistics were used to summarize the data and to screen for out-of-range values. We also examined frequency histograms to determine if sufficient heterogeneity of the univariate distributions was present. Bivariate correlations for all continuous predictor variables were assessed to check for the presence of colinearity (Pearson’s r > .7, p < .5). For the multi-item Likert-type scales, a mean value imputation was done for the scales in which less than 50% of the values were missing. Regression substitution, using age, age squared, and highest nursing degree was used for estimating missing wage values. Following these substitutions, a listwise deletion was used in the final regression analyses to handle missing data. Logistic multivariate regression was used to answer the research question. Statistical Package for Social Sciences version 18.0 software was used for these analyses. |
Back to Top Back to Top | Nurse and Work Environment Characteristics Of the 2,007 RNs who responded to the Wave 3 survey, 445 were excluded because they did not work in a hospital and 123 were excluded because they did not indicate their work setting. This resulted in a sample of 1,439. In the regression analysis, 213 respondents were excluded after the listwise deletion. This resulted in a regression sample of 1,226. There were no differences (p > .05) in gender, age, marital status, first basic nursing degree, race/ethnicity, or RNs’ ratings of patient care quality between the respondents included and those excluded from the regression analysis. On the basis of bivariate correlations, no predictor variables were correlated above .6. The respondents’ personal characteristics, task, organizational and physical work environment factors, job satisfaction, and ratings of patient care quality are presented in Tables 1 and 2. As shown in Table 1, most worked in direct care positions (79.7%) and non-Magnet-recognized hospitals (70.9%). On the average, RNs had worked for 44 months (SD = 11.2 months) in RN jobs since graduation. Almost three quarters (74.5%) of the RNs reported that there were “enough RNs to adequately care for patients” on some shifts or rarely as opposed to on every shift or most shifts (25.5%). Lastly, 69.5% of RNs reported that patients were very likely as opposed to somewhat likely (28.4%), not too likely (2.0%), or not at all likely (0.1%) to receive high-quality care in their work settings. As shown in Table 2, mean scores for work motivation and negative affectivity were below the midpoint on a 5-point scale (lower numbers denote a lesser degree of the trait), whereas for positive affectivity the mean score was above the midpoint. The RNs reported, on average, a moderate amount of variety and autonomy in their jobs. For quantitative workload, the RNs, on average, reported that they were working very hard and very fast, with little time and more work than can be done well in 1–2 days/week. Furthermore, the RNs, on average, scored above the midpoint of the scales, where higher scores represented more positive ratings for the following work environment factors: nurse–physician relations, workgroup cohesion, supervisory support, promotional opportunities, procedural justice, physical work environment, and job satisfaction. |
Back to Top | Predictors of RNs’ Ratings of Patient Care Quality Table 3 shows the results of multivariate logistic regression, in which the effects of predictor variables were analyzed jointly and adjusted for the previously specified controls. At the p < .05 level, none of the person and task variables were related to RNs’ ratings of patient care quality. For a single incremental increase in RNs’ ratings of physical work environment (4.99), workgroup cohesion (1.69), nurse–physician relations (1.40), procedural justice (1.34), and job satisfaction (1.26), the odds of RNs reporting their patients were very likely to receive high-quality care increased by the respective parenthetical factors. In other words, a single incremental increase in RNs’ ratings of the physical work environment corresponded to an almost five times greater likelihood of RNs rating patient care as being high quality. Similarly, one incremental increase in RNs’ rating of organizational constraints decreased the odds of RNs reporting their patients received high-quality care by 44%. Those working in a Magnet hospital were 1.69 times more likely to rate their patients as very likely to receive high-quality care when compared with other quality options. The odds of RNs reporting their patients were very likely to receive high-quality care decreased by 8% for each additional patient they were required to care for; however, RNs’ ratings of staffing adequacy were not related to their ratings of patient care quality. With the exception of race/ethnicity category “other,” none of the previously specified control variables were significant at p < .05. | |
Back to Top | Discussion We used a well-specified measurement model and a national sample of RNs to extend empirical evidence on key work environment factors that influence RNs’ ratings of patient care quality over and above staffing. Health care managers can use the results to inform and focus work environment redesign to support the delivery of high-quality patient care. We found that several factors, which are amenable to organizational change, contribute to RNs’ ratings of quality, adjusting for staffing: nurse–physician relations, procedural justice, organizational constraints, and Magnet recognition status. This is consistent with the conceptual model (Carayon et al., 2006) and with existing studies in which staffing and other factors were examined jointly (Aiken et al., 2002, 2009; Friese, 2005; Kim et al., 2009; Patrician et al., 2010). Moreover, we extend the available evidence by empirically explicating the relationship between RN-rated quality and three variables that have not been previously studied jointly with staffing: workgroup cohesion, physical work environment, and job satisfaction. Contrary to the relationships hypothesized in our conceptual model, two organization variables (supervisory support and promotional opportunity) were not related to RN-rated quality. Similarly, nurse manager support was not related to RN-rated quality in the study of Kim et al. (2009), whereas Friese (2005) reported a positive relationship between the two variables. Staff RNs may not always perceive supervisor support as directly contributing to patient care quality, because supervisors are likely to provide the support indirectly through securing staff RNs with access to resources and information to do their jobs (Laschinger, 2008). Education was not related to RN-rated quality in our study. This is congruent with the findings of Kim et al. (2009) and Patrician et al. (2010) and suggests that, although a higher proportion of RNs with a bachelor’s degree in an institution can contribute to improved patient outcomes, such as 30-day mortality (Aiken et al., 2009), educational background of an RN does not play a role in shaping that RN’s ratings of quality. In contrast to the propositions by Carayon et al. (2006), personality traits and work tasks (autonomy, variety, and quantitative workload) were not related to RN-rated quality. Similarly, Estabrooks et al. (2005) did not find a relationship between autonomy and RNs’ ratings of quality. In our study, quantitative workload was not related to RN-rated quality, whereas Gurses et al. (2009) found a negative relationship between the two variables. The difference in findings between the study of Gurses et al and our study is likely because of differences in measurement model specifications. For example, Gurses et al. did not measure staffing, nurse-physician relations, and other factors that were related to RN-rated quality in our study. |
Back to Top | Practice Implications For managers, the question is how to best redesign RN work environments to promote high-quality patient care. Health care managers often use information from multiple sources to guide their decision-making. To inform these decisions, we provide evidence on what impacts ratings of patient care quality by the largest group of direct care providers– RNs. Our findings show that making changes in individual work environment factors can lead to improvements in RNs’ ratings of patient care quality. We also illustrate the impact of improving various factors simultaneously while holding RN staffing constant. We used regression coefficients to construct two hypothetical RN cases to estimate predicted probabilities of RN-rated quality. The predicted probability of reporting that patients on his or her unit are very likely to receive high quality care is .83 for the RN with a positive covariate structure (Case 1), defined as an RN who works in a Magnet hospital and scores favorably (+SD) on all of the person, task, organization, and physical work environment factors. Whereas the predicted probability of reporting that patients on her or his unit are very likely to receive high-quality care is only .56 for the RN with a negative covariate structure (Case 2)—defined as an RN who also works in a Magnet hospital but scores unfavorably (-SD) on all of the previously cited factors. In other words, the Case 1 RN perceives that 83 of 100 patients are very likely to receive high-quality care, whereas the Case 2 RN perceives that 56 of 100 patients are very likely to receive high-quality care. Not all work environment factors have equal impacts on RNs’ ratings of quality or require equal resources. For example, the impact of obtaining Magnet recognition is nearly the same as the impact of improving workgroup cohesion. However, investment in workgroup cohesion likely requires fewer resources than applying for and earning Magnet recognition. Furthermore, RNs’ ratings of quality are similarly affected by improvements in procedural justice and nurse–physician relations. Here, organizational costs for improving RN involvement in decision-making (procedural justice) by forming unit-based councils are likely to be less than the cost of interprofessional team building. Particularly novel, we found that a single incremental increase in physical work environment ratings corresponded to the largest improvement in ratings of patient care quality. This is not surprising because our physical work environment scale measured elements across several dimensions: ambient, architectural, and interior design. Improving these elements might be costly, but if an organization is undergoing physical environment redesign, involving staff in design decisions could yield improved perceptions of patient care quality. |
Back to Top | Limitations Although our study used a national RN sample and a well-specified measurement model, it has several limitations that caution generalizing the findings. First, the use of cross-sectional data precludes inference to causal relationships between the predictors and the outcome variable in this study. Next, we relied on RNs’ ratings of patient care quality and did not measure actual patient outcomes. Furthermore, the data are from a group of RNs who are early in their nursing careers. These RNs might have different perceptions of work environment and the quality of patient care from more experienced RNs. Such a small number of the respondents reported that high-quality care is not too likely or not likely at all in their work unit, which may reflect that the RNs lack the experience to critically evaluate the quality of patient care. Alternatively, the RNs might have had difficulty reporting poor patient care quality in their own work unit because this could reflect that the care they are providing is of poor quality. Although, Carayon et al. (2006) suggest that work environment factors interact to impact patient care quality through care or other processes, in this study, only the independent direct effects of factors on RN-rated patient care quality were tested. |
Back to Top | Conclusion Our evidence demonstrates the importance of considering RN work environment factors other than RN staffing when planning improvements that may affect patient care. In summary, our findings show that personality and task factors are not directly associated with RNs’ ratings of patient care quality, while several organization factors, physical environment, and job satisfaction are. This is consequential for health care managers, because these are the factors they have the ability to improve on. One way to integrate the evidence from our study into organizational decision-making is by making these key RN work factors part of health care managers’ dashboards, so that they can adjust staffing needs based on the underlying quality of the other work environment factors. Lastly, RNs are the most numerous health care providers in hospitals, and without ensuring optimally designed work environments to support their delivery of care, making progress in improving the overall quality of care will be quite difficult. |
Bottom of Form |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
