The distribution of the percentage of Black patients across our study hospitals is unimodal and positively skewed, with a median of 6.0%, mean of 10.5%, and a standard deviation of 13.0. These numbers are consistent with the average percentage of Black patients admitted to hospitals each year in the United States (National Center for Health Statistics, 2010
shows the nursing and hospital characteristics of our study hospitals, first across all hospitals and then across the three categories of hospitals defined by the percentage of Black patients. The average workload across all hospitals was roughly five patients per nurse. Approximately equal percentages of hospitals were classified as having poor (26%) and better (27%) work environments. There were no statistically significant differences in nurse staffing ratios and nurse work environments across the hospitals with differing percentages of Black patients. Hospitals with the highest percentages of Black patients were significantly more likely to be large, urban, teaching hospitals (p < .001 in all cases).
Distribution of Hospital Characteristics by Concentration of Black Patients in Hospitals
We also descriptively examined differences in nurse-assessed and HCAHPS outcomes across hospitals with differing percentages of Black patients (see supplemental Table
available online, see directions at end of article). There were significant differences across hospitals with higher and lower concentrations of Black patients in every outcome we considered.
presents the results of using ordinary least squares regression models to estimate the association between the concentration of Blacks, nurse staffing, work environment, and nurse-assessed quality outcomes. In the unadjusted models, the percentage of Black patients in the hospital was a statistically significant predictor of each outcome. For example, the percentage of nurses who were not confident that patients could manage their care upon discharge was over 7 points higher in hospitals with the highest percentage of Blacks (>23%) as compared to those hospitals with the lowest percentage of Black patients (<11%).
Effects of Concentration of Black Patients in Hospitals, Nurse Staffing and Nurse Work Environment on Nurse-Reported Quality Outcomes (n = 568)
In the first set of partially adjusted models that controlled only for nursing characteristics, nurses working in hospitals with the highest percentage of Black patients continued to report poorer confidence in patients’ ability to manage their care upon discharge and more frequent adverse events. Controlling for nurse staffing and the nurse work environment, however, appeared to slightly attenuate the coefficients associated with high concentrations of Black patients. The second set of partially adjusted models accounted for structural hospital characteristics. After controlling for structural hospital characteristics, the effect associated with the percentage of Black patients in the hospital was markedly attenuated in size for three of the four outcomes. The coefficient associated with the high concentration of Black patients became statistically insignificant in the case of frequent patient or family complaints.
In the fully adjusted model, the concentration of Blacks remained a significant and strong predictor of the nurses’ confidence in patients’ management of care after discharge and frequent adverse events. In the HAI models, we observed that the coefficients associated with the concentration of Blacks were lessened the most by controlling for nursing staffing, work environment, and hospital characteristics simultaneously. Nurse staffing was significant (p < .001) for three of the four outcomes (poor confidence in ability to manage care after discharge, frequent patient/family complaints, and ventilator-associated pneumonia), and in all cases, higher workloads were associated with more nurses reporting poorer quality. Nurses working in better work environments reported more confidence in patients’ readiness for discharge and less frequent complaints and adverse events, net of the effects of the concentration of Blacks and nurse staffing.
displays the regression models for the HCAHPS measures. In the unadjusted models for all four outcomes, the concentration of Black patients in the hospital was a significant predictor of lower satisfaction scores. In the models adjusting for nursing characteristics, the concentration of Black patients remained a significant predictor, but the effect was attenuated in three of the four outcomes. In the model that accounted only for hospital characteristics, the effect of being in a hospital with a higher concentration of Black patients was again lessened in three of the four HCAHPS outcomes and became statistically insignificant for “staff gave patients discharge information.” In the fully adjusted model that took nursing and hospital characteristics into account, as well as the hospital’s state and patient response rate, the racial concentration effect became insignificant in all HCAHPS outcomes except the global “definitely recommend” measure. The work environment was a strong predictor of the other three HCAHPS measures in the fully adjusted models, while nurse staffing was predictive of staff giving patients discharge information.
Effects of Percentage of Black Patients in Hospitals, Nurse Staffing and Nurse Work Environment on HCAHPS Outcomes (n = 429)
In the fully adjusted model, the proportion of patients who would definitely recommend the hospital was over 2% lower in hospitals where 11% to 23% of the patient population was Black, as compared to hospitals with less than 11% Black patients. In this same model, as a nurse’s average workload in a hospital increased by one patient, the proportion of patients who would definitely recommend the hospital decreased by approximately 1.5%. Finally, the proportion of patients who would definitely recommend the hospital increased by approximately 4% in hospitals that had mixed environments as compared to poor (or better as compared to mixed). This finding would translate into an increase of 8% in the proportion of patients who would definitely recommend the hospital if a “poor” work environment was improved to be “better.” We have transformed the estimates for this outcome into percentiles to demonstrate the relative magnitude of the effects across the range of values taken by the factors we are investigating. Consider, for example, the percentage of patients who would definitely recommend their hospital, which has a standard deviation in this sample of 9.8. If hospitals were to improve their work environments from poor to good and reduce nurses’ workloads by an average of one patient per nurse, their patients’ satisfaction would shift from the 50th to the 84th percentile of all hospitals.