tial patients. These data can also sometimes cloud the findings of research (Seago, 2001). This issue is less significant in nursing homes, where heterogeneous nursing units are much less likely to exist, the resident population is more homogeneous, and variation in patients can be addressed for research studies as needed through case-mix adjustment.

A number of studies of the effect of nurse staffing levels on patient outcomes have attempted to use patient mortality as an outcome measure. However, patient mortality is a problematic nurse-staffing outcome for several reasons. First, patient death is not common; its low frequency makes detecting statistically significant differences difficult (Hartz et al., 1989). Second, while some patients die as a result of injuries related to health care, others die as a result of overwhelming disease. While some studies evaluating the quality of hospital care have used methods to assess the reasons for in-hospital deaths (Brennan et al., 1991; Thomas et al., 2000), studies of nurse staffing that have used patient mortality as an outcome measure have lacked methods for attributing the cause of death to preventable or non-preventable causes. Thus, it is not surprising that these studies do not agree on whether lower nurse-to-patient ratios (i.e., fewer nurses per patient) are associated with higher patient mortality (measured as either in-hospital mortality or death within 30 days of admission). The strongest evidence supporting such a mortality relationship was derived from a study of patients with AIDS (Aiken et al., 1999). This study was conducted in 20 hospitals, aggregated data at the nursing unit level, and had good case-mix controls. Other diagnosis-specific studies have not been able to demonstrate a relationship between nurse staffing levels and patient mortality.

Studies in which patients were not selected by diagnosis also have yielded inconsistent findings about the effect of staffing levels on mortality. Two nationwide studies that aggregated data at the hospital level (Aiken et al., 2002; Bond et al., 1999) found that lower nurse-to-patient ratios were associated with higher patient mortality. This association was not found, however, in other studies examining multiple ICUs (Amaravadi et al., 2000; Shortell et al., 1994) and hospital-level staffing ratios (Hunt and Hagen, 1998; Needleman et al., 2002).

Nonfatal adverse events, such as nosocomial infections and decubitus ulcers, are thought to have a more plausible direct relationship to the availability of hospital nursing staff. A consistent finding across multiple recent studies is that lower nurse-to-patient staffing ratios are associated with higher rates of nonfatal adverse events, including nosocomial infections, pressure ulcers, and cardiac and respiratory failure (Aiken et al., 2002; Cho et al., 2003; Kovner et al., 2002; Needleman et al., 2002). Similarly, a review of evidence pertaining to acute care hospital staffing published in the health professions literature from 1990 to 2001 revealed that of 16 hospital-based studies of the relationship between levels of nursing staff and pa-

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement