Improvement in the practice environment of nurses in acute care hospitals has been the focus, challenge, and recommendation of many studies, commissions, and committees. Lack of a productive, healthy work environment has been related to nursing shortage, poor quality of nurses' work lives, nurse job dissatisfaction, low productivity, and poor-quality unsafe patient care (American Hospital Association, 2002; American Organization of Nurse Executives, 2006; Hall, 2005; Institute of Medicine, 2004). An initial attempt to measure the attributes of an excellent staff nurse work environment began with the 1984 development of the 65-item Nursing Work Index (NWI; Kramer & Hafner, 1989). Based on a comprehensive list of environmental factors identified by nurse executives and staff nurses in the original magnet hospital study (McClure, Poulin, Sovie, & Wandelt, 1983), the NWI was designed to measure nurse job satisfaction and productivity, two of the four outcomes upon which the original magnet designation was made. Over a 10-year period, items not selected by staff nurses as impacting either productivity or job satisfaction were removed. In a causal model study, Kramer, Schmalenberg, and Hafner (1989) found that over 80% of nurse job satisfaction, attraction, and retention were due to productivity of quality patient care. For this reason, only productivity ofquality patient care was used as a criterion when staff nurses in 14 magnet hospitals were asked to select, from the remaining 37 NWI items, those that were essential to a productive work environment (Kramer & Schmalenberg, 2002). The following eight care processes or relationships were selected: (a) clinically competent peers, (b) collaborative nurse-physician relationships, (c) clinical autonomy, (d) support for education, (e) perception of adequate staffing, (f) nurse manager support, (g) control of nursing practice, and (h) patient-centered cultural values. A tool, the Essentials of Magnetism (EOM), was developed and tested to measure each element essential to a productive work environment (Kramer & Schmalenberg, 2004).

Environment is the aggregate of the conditions, influences, forces, and cultural values that influence or modify an individual's life in a community such as a clinic or clinical unit. Several psychometrically sound work environment tools are focused on specific elements of the environment such as staffing (Irvine, Sidani, & McGillis-Hall, 1998) and perceived workload (Lacey et al., 2006). Three other tools that purport to measure work environment more broadly by assessing the absence or presence of attributes have been found to lack validity, psychometric precision, and a unifying theoretical focus (Cummings, Hayduk, & Estabrooks, 2006). The EOM measures the aggregate of the environmental characteristics identified by magnet hospital nurses as essential to a productive work environment. Each essential is measured by a subscale.

Background

Essentials of Magnetism Instrument

The EOM differs from other instruments that measure attributes of a work environment in that it is based on the process feature of Donabedian's (1980) paradigm. In this model, 3 components necessary for evaluation of the quality of health care systems are proposed-having the right things (structure), doing the right things right (process), and obtaining desired results (outcomes). The EOM respondents are requested to identify extent of agreement or disagreement with the steps or components of a process or relationship rather than their perception of the degree to which the attributes or conditions are present. These items from the autonomy subscale of the EOM illustrate process measurement: type of decision making in nursing-unique and overlap spheres of practice; degree of risk in making independent decisions; the effect of bureaucratic rules and regulations on decision making; administrative approval and managerial support for independent or interdependent decision making; and augmentation of clinical knowledge through evidence-based practice initiatives (Kramer, Maguire & Schmalenberg, 2006; Kramer et al., 2007a). These items from the National Database of Nursing Quality Indicators (American Nurses' Association, 2007) illustrate measurement of respondent's perception of presence of an attribute: "Nurses need more autonomy in their daily practice" and "Physicians in general cooperate with nursing staff."

Evaluation of the work environment requires assessment of the linkages between structure, process, and outcomes. Using the Magnet Hospital Standards as the structure, the EOM measures the essential processes and makes it possible to establish whether the magnet standards enable the staff nurse-identified care processes and relationships that lead to desired outcomes. Two single-item 10-point indicators, extrapolated from a four-point scale by Aiken, Clark, and Sloane (2002), were used to measure the two outcomes, overall job satisfaction (OJS) and nurse-assessed quality of care (QC).

The EOM is a four-point Likert scale with weighted responses. In any process measurement, some steps or components are more important than others. For example, two types of RNMD (nurse-physician) relationships described by interviewees, collegial and collaborative, were identified as being more instrumental to quality patient care than were two other types, formal and hostile. To obtain the differential item weights for each subscale, 392 nurses in 7 magnet hospitals not studied previously rank-ordered scale items (steps and components of the process) in terms of how important each item factored in to giving quality patient care. The mean rank given each item by the 392 respondents is the weight used in scoring (Kramer & Schmalenberg, 2004).

Single-Item Outcome Indicators

Global single-item indicators, congruent with nursing's emphasis on wholism and individualism, require that participants consider all aspects of a phenomenon and differentially weight aspects according to their own values, ideals, and situation (Youngblut & Casper, 1993). Two outcomes, QC and OJS, were measured on single-item indicator scales similar to the four-point scales used by Aiken et al. (2002) and by Ulrich, Lavandero, Hart, et al. (2006). In a study of 2,969 nurses in Pennsylvania hospitals, Aiken et al. reported that 20.8% rated QC as poor to fair (lower 2 points on the four-point scale). This percent is virtually the same (20.9%) as that reported by Ulrich et al., using the same scale in their national online survey of 4,034 critical care nurses. In the Aiken study, breakdown by magnet status was not provided, but it was probably very low as there were few Pennsylvania magnet hospitals in 2002. The critical care study reports that the 715 nurses (21% of the sample) in magnet hospitals rated QC significantly higher (p < .05) than did their counterparts in non-magnet hospitals (Ulrich, Woods, Hart, 2007), but the percentage figures on the four-point scale were not provided. With respect to OJS, the critical care study reports that 24% of the 4,034 nurses were very dissatisfied or somewhat dissatisfied, whereas 18% of the magnet hospital subsample indicated that they were somewhat or very dissatisfied.

2003 Psychometric Study of the EOM

Following the generation of grounded theories and item development, the EOM was administered to almost 4,000 staff nurses in 26 regionally dispersed magnet and non-magnet hospitals. Exploratory factor analysis was used to identify subscale dimensions. Validity and reliability studies were conducted, and criterion validity was established by the "known group" method (nurses in magnet hospitals) (Kramer & Schmalenberg, 2004). Analysis of variance (ANOVA) with post hoc multiple comparisons (Tukey honestly significant difference statistic) was used to identify statistically significant (p <= .05) homogenous subsets in the 26-hospital 2003 sample. Usually three, but on a few variables, only two groups were identified. The homogenous subset with the highest scores was labeled Magnet, second highest was Magnet-aspiring, and last group was non-Magnet. The Profile range was established by adding and subtracting 1/2 standard deviation (SD) from the homogenous subset mean. Not all magnet hospitals fit the Magnet Hospital Profile and, on some variables, magnet-aspiring and non-magnet hospitals scored in the magnet profile (Kramer, Schmalenberg, & Maguire, 2004). In 13 (81%) of the 16 magnet hospitals, staff nurses confirmed a satisfying and productive work environment (Kramer et al., 2004) by scoring within the National Magnet Hospital Profile range on at least seven of the eight essentials of magnetism and on the total EOM score.

Purpose of Study

In 2006, substantive changes were made in the Adequacy of Staffing and Nurse Manager Support subscales of the EOM, and the tool was relabeled EOMII (Kramer & Schmalenberg, 2005; Kramer et al., 2007b). The primary purpose of this secondary analysis of the EOMII data aggregated from 10,514 staff nurses in 34 hospitals was to test the hypothesis that nurses in hospitals designated as having excellent work environments (magnet) will report more satisfying (fulfills personal and professional needs); (Kramer & Hafner, 1989) and productive (being able to give quality patient care) work environments than will nurses in hospitals not so designated. Productive work environments were assessed through scores on the EOMII and the two single-item outcome indicators. Additional aims were to (a) evaluate the psychometrics of the EOMII (including stability by comparing data from EOM and EOMII); (b) update the National Magnet Hospital Profile so that hospitals can compare their environments with that of excellent hospitals; (c) investigate linkages between the OJS and QC outcome variables, the Professional Job Satisfaction (Total EOMII) score, and the processes constituting a productive work environment; and (d) describe relationships between and among the nurse attributes of education and experience, work characteristics of shift and clinical unit, and the contextual variable of type of hospital, with the processes of a productive work environment.

Methods

Sample

This volunteer, unit- and hospital-representative sample consisted of 10,514 staff nurses from 34 hospitals, 18 of which were magnet and 16 were comparison hospitals. The 34 hospitals were from among hospitals requesting evaluation of the quality of their work environment using the EOMII and the two outcome indicators. This evaluation was usually part of self-study in preparation for Magnet, Baldrige, or Employer of Choice applications, or they were hospitals undertaking a general internal self-evaluation, or they were hospitals from a particular region of the country who agreed to be tested to ascertain if they qualified for another research study. Selection of the 34 hospitals was made on the basis of 3 criteria: Staff had to have been tested on the EOMII between June 2005 and November 2006; a representative sample (Verran, Gerber, & Milton, 1995) consisting of staff nurses (at least 35% but preferably 50%) from every unit/clinic with a registered nurse complement of 5 or more registered nurses participated in the testing; and the hospital agreed to have their data added to a National EOM Data base.

Instrumentation

EOMII: Six of the eight subscales of the EOMII are the same as the EOM. Additions and deletions were made in the Adequacy of Staffing and Nurse Manager Support scales on the basis of additional observation, interviews, and extension of the grounded theories. Specific changes are detailed in Kramer & Schmalenberg (2005) and Kramer et al. (2007b).

OJS: Overall job satisfaction was measured using a 10-point single-item indicator. The stem was: "Considering all aspects of your job as well as your own values, ideals and goals, how satisfied are you with your current nursing job? (Circle any number on the scale)." Benchmarks provided were 0 (it's terrible), 5 (I'm satisfied), and 10 (I love it).

QC: This 10-point single-item indicator, modified from a four-point tool developed by Aiken (2002), used the stem: "Select a number that indicates the usual quality of care provided to patients on your unit." Benchmarks provided were: 0 (Dangerously low), 5 (It's safe but not much more), and 10 (Very high quality).

Data Collection Procedure

All hospitals procured their own institutional review board approval, including procedures for consent, anonymity, and confidentiality. The EOM tool was provided by the authors; dissemination and collection of data was done by on-site investigators who returned the data to the authors for scoring, analyses, and report writing. Primary analyses were individualized for each hospital and transmitted to the hospital via confidential reports.

The method used to establish the National Magnet Hospital Profile range and mean on the aggregated 34-hospital data was described in the "Background" section. This method of establishing Profile ranges resulted in a narrower range than using the 95% confidence interval automatically provided by SPSS v.15 programs.

Analysis

ANOVA with post hoc multiple comparisons (Tukey honestly significant difference statistic) was used to test the three hypotheses on the aggregated 34-hospital EOMII data. It was hypothesized that magnet hospital (the main independent variable) staff nurses would report higher scores on all subscales of the EOMII (Hypothesis 1) and on the QC (Hypothesis 2) and OJS (Hypothesis 3) outcome indicators than would their counterparts in comparison hospitals. ANOVA was also used to obtain the statistically significant homogenous subsets needed to update the National Magnet Hospital Profile (Aim b) and to establish linkages between outcome variables and essential work processes. Confirmatory factor analysis was used to test the structural integrity of the EOMII subscales; test-retest and internal consistency reliability coefficients were used to determine aspects of reliability (Aim a). The effects of the demographic, work, and context variables of education, experience, shift, clinical unit, and type of hospital (i.e., academic or community) on the dependent variables were tested by entering the demographic, work, and context variables as independent variables (multivariate analysis of variance) and also by entering them as covariates (multivariate analysis of covariance) (Aims c and d).

Results

Description of Hospital Sample

In this 34-hospital sample, 9 hospitals were located in medium-sized cities, 14 were in large cities, and 11 in large metropolitan areas. All major census tract regions (Northeast, Southeast, North Central, South Central, Midwest, Mountain West, Northwest, and Southwest) were represented by hospitals in both the magnet and comparison subsamples, except there were no Southwest hospitals in the comparison sample. Nine magnet and nine comparison hospitals were academic teaching hospitals. Nine magnet and seven comparison hospitals were community hospitals.

Description of the Nurse Sample

Although reports on the attributes and characteristics of the nurse sample in each hospital were presented to each of the 34 hospitals, the characteristics of the aggregated sample of 10,514 nurses has not been presented elsewhere. Almost 50% of this sample of staff nurses from 34 hospitals were prepared at the baccalaureate level with no differences between magnet and comparison subsamples. There were significantly fewer associate degree nurses and more staff nurses with higher degrees in the magnet than in the comparison subsample. Sixteen percent of the sample had 3 years or less experience; the remainder had 3 to 10 years (28%), 10 to 20 years (26%), and more than 20 years of experience (31%). There were no differences between subsamples by experience (Table 1).

TABLE 1. Significance of Differences in Demographic and Work Characteristics of Nurses in Magnet and Comparison Subsamples

Test of the Hypotheses

To the extent that the Magnet Standards are valid criteria of excellent work environments, it was hypothesized that nurses in magnet hospitals would report more productive work environments than would their counterparts in comparison hospitals. This means that they would score higher on all eight essential processes, on total EOMII-labeled Professional Job Satisfaction, and on the OJS and QC outcome indicators. Nurses in magnet hospitals scored significantly (p < .001) higher than did nurses in comparison hospitals on all relationships and processes essential to a productive and satisfying work environment. The F ratios ranged from a low of 9.627 for Nurse Manager Support to a high of 54.340 for Control of Nursing Practice (Table 2). The differences between magnet and comparison subsamples on the outcome variables were also significant. Nurses in magnet hospitals rated OJS as 6.86 (SD = 0.4604) and QC as 8.04 (SD = 0.4376) on the 10-point scale; nurses in comparison hospitals rated OJS as 6.22 (SD = 0.418) and QC as 7.42 (SD = 0.3222). The OJS and QC data are presented in Table 3 as the percent of each subsample selecting the indicated benchmark. All three research hypotheses are accepted.

TABLE 2. Significance of Difference in Eight EOMII Subscales, Total EOMII Score, and Outcome Indicators Between Nurses in Magnet (

TABLE 3. Percent of Respondents in Magnet and Comparison Hospital Selecting Benchmarks on Nurse-Assessed Quality of Care and Overall Job Satisfaction Single-Item Outcome Indicators

Psychometrics

Principal component factor analysis (Table 4) using Varimax rotation with Kaiser normalization (Kim & Mueller, 1978) generated 10 factors. The first seven confirmed the factor analytic structure for seven of the eight essential work processes. All Support for Education and Clinically Competent Peers items loaded on the same factor. Factor 8 contained negative, reverse-scored items from all subscales. Factor 9 consisted of one item from the RNMD scale; Factor 10 consisted of one item from the Nurse Manager Support scale.

TABLE 4. Results of Confirmatory Factor Analysis

TABLE 4. (continued)

The stability aspect of reliability was ascertained through ANOVA comparing subscale scores of the 16 magnet hospitals in the 2003 EOM psychometric study and the 18 magnet hospitals in this 2006 EOMII study on all scales except Adequacy of Staffing, Nurse Manager Support, and Total EOM. Although these were not the same respondents nor the same hospitals at these two time periods, they were all hospitals that adhered to the same structural criteria of excellent work environments (magnet standards) that had not changed during the 3-year period. F ratios ranged from 1.858 on the Support for Education subscale to 9.950 for Clinical Autonomy; none were significant. Clinical Autonomy and Control of Nursing Practice scores were higher in 2003 than in 2006; 2006 RNMD Relationship scores were higher than those in 2003. Internal consistency reliability (Cronbach's alphas) for the 34-hospital sample ranged from .83 to .97. See Table 5 for scale means for the 2003 and 2006 samples and Cronbach's alphas.

TABLE 5. Alpha, Means, and Standard Deviations Between Magnet and Comparison Hospitals on 2003 EOM and 2006 EOMII Data

National Magnet Hospital Profile Update

The percent of the 18 magnet hospitals scoring within the National Magnet Hospital Profile ranged from 78% (n = 14) on Clinically Competent Peers to 100% on Support for Education, Patient-Centered Values, Control of Nursing Practice, and Total EOMII. Seventeen of the 18 hospitals (93%) scored within the National Magnet Hospital Profile range on at least seven of the eight essentials of magnetism and on the total EOM.

Links Between and Among Process, Outcome, and Attribute Variables

All correlations between the eight essential processes and the OJS and QC outcome indicators were significant. Adequacy of Staffing, Cultural Values, and Nurse Manager Support were correlated most highly. OJS and QC variables are highly intercorrelated (r = .62; Table 6).

TABLE 6. Correlation Between Essential Processes, Total EOM Score, Overall Job Satisfaction, and Nurse-Assessed Quality of Care

Five percent of the 7,163 nurses in the 18 magnet hospitals in this study rated QC below safety (4 or less on the 10-point scale); 9.4% of the comparison hospital nurses did so (Table 3). With respect to OJS, almost 17% of the magnet hospital nurses reported being somewhat or very dissatisfied with their job as compared to 25.5 % of the nurses in comparison hospitals.

Based on the OJS and QC outcome indicators, three combinations of respondents are possible. Respondents can rate QC on their unit as equal to OJS (QC = OJS), QC as higher than OJS (QC > OJS), or OJS as lower than QC (OJS < QC). Almost 60% of the magnet and comparison subsamples rated QC > OJS; 28% rated QC = OJS, and 12% rated QC < OJS. Chi-square differences among these three groups for the magnet and comparison hospital samples indicated significant (p <= .002) differences; 68% of the magnet sample and 64% of the comparison sample rated QC > OJS. When the QC = OJS and QC > OJS groups are combined, 83% of 10,514 nurses in the 34 hospitals rated the QC given on their units as equal to or greater than their OJS.

There were few significant differences between the QC > OJS or OJS < QC groups when data were analyzed by nurse attributes. Chi-square analysis indicated a significantly larger percentage (p <= .000) of general medical-surgical nurses (12.6% compared to 8.6%) in the QC < OJS group and a significantly larger percent of outpatient nurses (19% compared to 14%) in the QC > OJS group than in the converse groups. A significantly (p <= .001) larger percentage of 8-hr day shift nurses were in the QC > OJS than in the QC < OJS group. More community hospital nurses were in the QC > OJS group, but significance of chi-square (59.371) did not reach the criterion level of .01.

Relationships Between and Among Demographic, Work, and Context Variables

The effects of two nurse attributes, education and experience; two work variables, shift usually worked and clinical unit; and two contextual variables, magnet status and type of hospital (i.e., Veteran's Administration, community, regional teaching, academic, or municipal) were studied. A one-way ANOVA indicated highly significant differences (p <= .000) for all dependent variables (essentials and outcomes). A multivariate analysis of covariance with demographic, work, and contextual variables entered as covariates indicated that the variance due to shift usually worked and type of hospital (community, Veteran's Administration, academic) was not significant. The significant differences in the dependent variables when education, experience, and clinical unit are entered as covariates can be found in Table 7. Magnet status continued to be the dominant source of significant difference in variance. Master's and doctorally prepared nurses scored significantly higher on almost all variables except for Nurse-Physician Relationships where diploma nurses excelled. The general pattern was that nurses with 3 years or less or with more than 30 years of experience reported more productive and satisfying work environments than did nurses with 5 to 10 years and 10 to 15 years of experience. The only exceptions to the most inexperienced and most experienced nurses perceiving the most productive environment was with respect to RNMD Relationships, Staffing, and CNP. The group with 3 years or less of experience reported poorer relationships and less adequate staffing; the 30+ years' group reported lower CNP scores. The only exceptions in the lower scoring groups were with respect to Adequacy of Staffing, Values, and Quality Care where they were replaced by the 3- to 5-year group. Staff working on medical and surgical specialty units (primarily oncology, orthopedics, and telemetry) or in outpatient clinics reported more satisfying and productive work environments than did nurses in psychiatry and in operating and recovery rooms.

TABLE 7. Location of Significant Differences on Eight Essentials and Outcome Indicators by Education, Experience, and Clinical Unit: Partial Result of MANCOVA

Discussion

Results of this study provide evidence that the EOMII is a valid and reliable measure of the work processes and relationships that staff nurses identify as essential to a productive and satisfying work environment. The high degree of intercorrelation among the eight essentials and the OJS and QC outcomes indicates that a productive and satisfying work environment is a multidimensional, integrated phenomenon. An excellent work environment does not evolve from the presence of only a few desired processes. None is optional; all are required.

The EOM consistently identifies excellent work environments and differentiates magnet and comparison hospitals (Kramer & Schmalenberg, 2004). The percent of magnet hospitals in which staff nurses confirm excellent work environments increased from 81% in 2003 (Kramer et al., 2004) to 93% in this study despite the fact that the sample selection in this study skewed the results toward greater excellence and higher standards. All of the 34 hospitals had achieved or were aspiring to some designation of excellence; almost half of the magnet hospitals in the sample were among the highest or second highest EOMII scoring by region of the country.

The updated National Magnet Hospital Profiles also reflects an upward bias. Electing to establish the Profile range at 1/2 SD from the subset mean rather than at the 95% confidence interval resulted in a narrower Profile range. It provides a more accurate picture of the homogenous subset, however, as it includes all high-scoring hospitals (magnet and comparison) and excludes those magnet hospitals not scoring high enough to be included in the highest subset. The Magnet Profile mean will be higher than the mean for all magnet hospitals, unless all of them scored in the highest subset, in which case the two means will be identical.

The National Profiles are useful in assisting excellence-aspiring hospitals to achieve their goals. They provide a great deal of comparative, evaluative information on the overall work environment as well as on the specific processes and relationships that make up that environment. However, to obtain specific information on avenues for improvement, item analysis of the EOMII subscales is often helpful. One of the items on the EOMII Clinical Autonomy subscale concerns the type of decision making needed in the nursing-unique sphere of practice in contrast to that sphere of practice where nursing overlaps with medicine and other disciplines. A word of caution is in order. Instruments are developed to measure constructs, and multiple items are needed to capture a construct. Focusing on a single item will not change the work environment with respect to autonomy, but it is a place to start.

The simultaneous assessment of process and outcome variables from respondents is not ideal and yields a dependent rather than independent measure. The psychometrics of the OJS and QC outcome indicators have not been established, and they are single-item indicators of very complex constructs, subject to the same cautions as noted above. Nonetheless, these indicators have value in assessing the effectiveness of structures designed to enable the processes staff say they need to do a good job, to give quality patient care. The indicators can be completed rapidly, meet little resistance, and are very familiar to both physicians and staff nurses because of their similarity to pain-rating scales. Nurses, managers, and physicians have not objected in any way to completing them, either verbally or by pencil, in any of the studies in which we have used them. Moreover, there is some evidence that they are accurate. Comparing the performance of the Aiken et al. (2002) and Ulrich et al. (2007) samples on QC and JS with that of the 18 magnet and 16 comparison hospitals in this sample, the same relationships are evident as was seen in the statistical determination between the two groups on the multi-item EOMII subscales. Magnet hospital nurses indicate higher quality care and less dissatisfaction than do nurses in comparison hospitals.

Increasingly, single-item measures are being used in large population surveys and in clinical research to measure symptom intensity. They are efficient tools, but only if their effectiveness can be established. Further research is needed to establish the psychometric properties of the OJS and QC and the relationship to patient outcomes such as mortality rates, complications, medication errors, and patient satisfaction. Youngblut & Casper's (1993) review of the psychometric performance of global single-item indicators yielded promising results. Baggs et al. (1999), using a single-item indicator, explored the accuracy of physicians' and nurses' ratings of degree of collaboration with specific patient outcomes and found that nurses were more accurate than physicians. Accurate meant that a low collaboration rating was related to poor patient outcomes and the reverse. This is the kind of research needed on the outcome indicators used in this study. If consistent and positive relationships can be established, then continued and increased use of global single-item indicators is warranted.

Magnet designation is the contextual variable that accounts for most of the variance in the essentials and outcome indicators (dependent variables). This is sustained whether the other attribute and contextual variables are treated as independent variables or as covariates. Some significant difference by education, experience, and unit do prevail.

The educational level of the nurses in this sample is probably an accurate reflection of the educational level of nurses in these 34 hospitals because, in many instances, the assistance of Research Councils aided in obtaining representative samples. A workforce of 60% baccalaureate or higher degree staff nurses is very high and approaches the goal set by nurse executives in teaching (70% bachelor of science in nursing [BSN]) hospitals and exceeds that set for community (50% BSN) hospitals (Goode et al., 2001). Because, nationally, only 43% of hospital staff nurses hold a baccalaureate degree (Spratley, Johnson, Sochalski, Fritz, & Spencer, 2000), this high percentage could be the result of the high emphasis on Support for Education, including many onsite, well-funded licensed vocational nursing to associate degree in nursing, associate degree in nursing to BSN, and BSN to master of science in nursing (MSN)programs reported by nurses in both the magnet and comparison subsamples in this study. In one hospital that financially supported nurses for onsite MSN education, more than 70% of nurses on one critical care unit had earned a master's degree. The positive correlation between BSN education and lower patient mortality on surgical units (Aiken, Clarke, Chueng, Sloane, & Silber, 2003) provides considerable support for continued funding and availability of onsite degree and other programs that nurses perceive as supportive of education. The high percentage of staff nurses prepared at the BSN, MSN, and even at the doctoral level and the high emphasis on education undoubtedly account for the significant differences by education that overrode the contextual variable of magnet designation.

That the most inexperienced (<3 years) and the most experienced (>30 years) report the most productive, satisfying work environments has not been previously reported and, in many ways, is counterintuitive. It may be that it was not so much a factor of these groups scoring high on the essential work processes, but rather that the comparison groups, particularly the 5 to 10 year group, scored low. Anecdotal data from Chief Nursing Executives indicates that they perceive this group as the most disenfranchised and disillusioned. The vulnerability of the 5- to 10-year and 10- to 15-year experience groups needs to be studied in detail, particularly to ascertain if perception of a poor work environment is related to job retention and performance.

That nurses on general medical-surgical units reported less productive work environments with respect to Staffing, RNMD Relationships, and Quality of Patient Care is not surprising. These units are frequently less well staffed and have more float and inexperienced nurses, and nurses must relate to a wide array of physicians making establishment of collegial or collaborative work relationships much more difficult (Schmalenberg et al., 2005a, 2005b). Because nurses on these units rate both job satisfaction and quality of patient care lower than do nurses on specialized units, research efforts need to be focused on what can be done to improve the work environment on these general units. Study of the environment and working conditions of the high-scoring "job satisfied," "high quality care" outpatient, medical, and surgical specialty units would enable identification of structures and practices that result in excellent working conditions that could then be duplicated on other units.

The high percentage of nurses in both magnet (84%) and comparison (82%) hospitals who rate the quality of care given on their clinical units as equal to or higher than their job satisfaction is far higher than would be expected on the basis of chance alone and confirms a finding reported almost 20 years ago (Kramer et al., 1989). The hygienic aspects of a job are important-salary, fringe benefits, safe parking, clean environment-but, for nurses, being able to do a good job, making a difference, and giving quality patient care are the most important and most satisfying aspects of nursing. That both job retention and retention in nursing are more closely related to being able to give quality care than they are to job satisfaction was recently confirmed by Smith, Hood, Waldman, and Smith (2005), who found that noneconomic factors in the environment such as a service quality orientation were key factors associated with job satisfaction and a positive practice milieu.

Items from the Clinically Competent Peers and Support for Education subscales continue to load on the same factor, indicating that further study, observation, and analysis of interview data are needed to capture the components, similarities, and differences between these processes. This work has begun.

In an address to the National Education Association, Booker T. Washington (2007) said: "Excellence is to do a common thing in an uncommon way." An excellent hospital nurse work environment is one in which nurse leadership provides the right structures, practices, and people. This enables clinical nurses to do the right things correctly, thus producing desired outcomes for patients, staff, and the organization. The results of this study provide evidence that the EOMII has acceptable reliability and validity to assess the quality of the practice environment and the effectiveness of interventions designed to improve that work environment.

References