Methods
We studied the nurse workforce and patient outcomes in 665 hospitals in four large states (California, Florida, New Jersey and Pennsylvania), three of which are among the top five states for employment of non-US-educated nurses.
Data Sources
Data were linked from three sources: (i) a 2006–2007 survey of registered nurses who were employed by hospitals in the four states; (ii) patient discharge data that was available from state agencies and; (iii) the American Hospital Association Annual (AHA) Survey of hospitals. Three institutional review boards approved the study: the University of Pennsylvania, University of Florida and Rutgers University.
Sample and Measures
Hospitals. The 665 non-federal adult acute care hospitals studied were from California (n = 271), Florida (n = 168), New Jersey (n = 73) and Pennsylvania (n = 153). Hospitals were characterized by size, teaching status and technology using AHA Annual Survey data. Hospital size was defined as small (<100 beds), medium (100–250 beds) and large (> 250 beds). Hospitals with no medical residents were classified as non-teaching, those with trainee to bed ratios of 1:4 or less were classified as minor teaching hospitals, and those with trainee to bed ratios greater than 1:4 were classified as major teaching hospitals. High technology hospitals were those that provided open-heart surgery, organ transplantation or both.
Nurses. Using licensure lists available from boards of nursing in each state, we surveyed by mail a 35–50% random sample of RNs by state and obtained responses from over 95 000 nurses. Of these, 38 657 were staff nurses who provided direct patient care on inpatient units in hospitals; the rest worked in other settings or were not employed at all. The overall response rate was 39% yielding an average of 60 nurse respondents per hospital (a range of 10–282 nurses per hospital). We conducted an intensive follow-up with a random sample of 1300 non-respondents that included monetary incentives that would not have been feasible in the large initial survey, and were successful in obtaining responses from over 90% of non-respondents. No differences were found between nurses that responded initially and those that did not in terms of their reports about their hospitals.
Non-US-educated nurses were self-identified from a survey item that asked nurses about the country where they received their basic nursing education. We estimated the percentage of non-US-educated RNs per hospital by dividing the number of non-US-educated nurses in each hospital that responded to the survey by the total number of RNs that responded in each hospital. The percentage of nurses that responded to the survey that were non-US educated was 12% overall (19% in California, 17% in New Jersey, 13% in Florida and just under 2% in Pennsylvania), which is only slightly higher than the 9% estimate of non-US-educated nurses that can be derived for hospital nurses in the four states using data from the 2008 National Sample Survey of Registered Nurses. Hospital nurse staffing was calculated for each hospital from nurse survey data by dividing the average number of patients reported by nurses on their units on the last shift by the average number of nurses on the unit. Educational composition was computed as the percent of nurses in each hospital whose highest degree was a bachelor's degree in nursing or higher.
Nurse job satisfaction was measured with a single item: 'On the whole, how satisfied are you with your present job?' with response categories as very dissatisfied, a little dissatisfied, moderately satisfied or very satisfied. Nurse-assessed quality of care was measured with a single item: 'On the whole, how would you describe the quality of care delivered to patients on your unit?' with the response categories as excellent, good, fair or poor.
The nurse practice environment in the various hospitals was measured in two ways. The general difference between US and non-US-educated nurses in describing their work environments was assessed using a single item which asked 'How would you rate the work environment in your job (e.g. adequacy of resources, relations with coworkers, and support from supervisors)?' with four response categories: excellent, good, fair or poor. The nurse practice environment measure included in predictive models was assessed using the Practice Environment Scale of the Nursing Work Index (PES-NWI), which is an extensively validated instrument that is endorsed by the National Quality Forum. In earlier analyses, we used five subscales: nurse participation in hospital affairs (9 items), nursing foundations for quality care (10 items), nurse manager ability, leadership and support of nurses (5 items), staffing/resource adequacy (4 items) and nurse–physician relations (3 items). The staffing/resource adequacy subscale was excluded in the analyses here because it empirically overlaps our direct measure of nurse staffing. Subscale measures were calculated for each hospital by averaging the values of all items on each of the subscales for all nurses in the hospital. These four aggregated subscales were then averaged to produce a single composite measure of the practice environment. PES-NWI subscales and the composite scale range in value from 1 to 4, and in the regression models were standardized to have a mean of 0 and standard deviation of 1.
Nurse burnout was measured with the Emotional Exhaustion subscale of the Maslach Burnout Inventory, an instrument with well-established reliability and validity. High emotional exhaustion, scores of 27 or higher, was designated following norms established by the developers' research.
Patients. Patient discharge data were obtained for the study hospitals from the California Office of Statewide Healthcare Planning and Development (OSHPD), the Florida Agency for Health Care Administration (AHCA), the New Jersey Department of Health and Senior Services (NJDHSS) and the Pennsylvania Health Care Cost Containment Council (PHC4). Our sample included patients aged 19–90 years who were admitted to study hospitals between 2005 and 2006 for general, orthopedic and vascular surgical procedures. Surgical patient discharges were chosen for this study because of the availability of a well validated risk adjustment model which was used in our prior research. Patient outcomes included 30-day mortality and failure-to-rescue (or death from complications). Failure-to-rescue was defined as death within 30 days of admission for the subset of hospitalized patients who experienced complications. Risk adjustment procedures included age, sex, transfer status, surgery type and comorbidities identified by Elixhauser et al..
Statistical Analysis
In the descriptive tables below we show characteristics of the surgical patients in the study, compare characteristics of non-US and US-educated nurses in the study hospitals and compare characteristics of hospitals with low, moderate and high proportions of non-US-educated nurses. Logistic regression models were used to estimate the effects of nurse staffing and percent of non-US-educated nurses on mortality and failure-to-rescue before and after taking account of patient characteristics and hospital characteristics, including the nurse work environment. In these models, we included interactions with non-US-educated nurse utilization and nurse staffing, technology status, teaching status and hospital size. We also included an interaction between nurse staffing and the nurse practice environment that was shown to be significant in an earlier paper. The interaction between nurse staffing and the proportion of non-US-educated nurses is explained by estimating the different effects of non-US-educated nurses in hospitals with different levels of nurse staffing. All regression models were estimated using Huber–White (robust) procedures to account for clustering within hospitals. Analyses were conducted using SAS V9.2 (SAS Institute, Cary, NC, USA) and STATA V10 (STATA Corp, College Station, TX, USA).