Journal of Nursing Education

The articles prior to January 2012 are part of the back file collection and are not available with a current paid subscription. To access the article, you may purchase it or purchase the complete back file collection here

Major Article 

The Relationship Between Nurse Education Level and Patient Safety: An Integrative Review

Renee T. Ridley, RN, MSN, CFNP

Abstract

The relationship between patient safety and nurse education level has implications for current and prospective nurses, hospital administrators, policy makers, and nurse educators. This integrative literature review assesses the current state of science on the topic during a 20-year period, using the Agency for Healthcare Research & Quality’s Patient Safety Indicators to measure outcomes. Twenty-four studies of variable quality were included. Although studies suggest that increasing RN dose (i.e., number of care hours) and skill mix (versus LPN) are associated with improved patient safety, evidence linking RN education level (i.e., BSN, ADN, diploma) is sorely lacking.

Abstract

The relationship between patient safety and nurse education level has implications for current and prospective nurses, hospital administrators, policy makers, and nurse educators. This integrative literature review assesses the current state of science on the topic during a 20-year period, using the Agency for Healthcare Research & Quality’s Patient Safety Indicators to measure outcomes. Twenty-four studies of variable quality were included. Although studies suggest that increasing RN dose (i.e., number of care hours) and skill mix (versus LPN) are associated with improved patient safety, evidence linking RN education level (i.e., BSN, ADN, diploma) is sorely lacking.

Ms. Ridley is Lecturer and Clinical Instructor, Department of Nursing, Murray State University, Murray, Kentucky, and a PhD candidate, Saint Louis University, St. Louis, Missouri.

The author wishes to thank Dr. Helen Lach at Saint Louis University for her assistance and continued support during the development of this article.

Address correspondence to Renee T. Ridley, RN, MSN, CFNP, Lecturer and Clinical Instructor, 120 Mason Hall, Murray, KY 42071; e-mail: renee.ridley@murraystate.edu.

Received: April 21, 2006
Accepted: September 12, 2006

Hospital cost containment often means the restructuring of nursing personnel. When this is the case, concerns regarding patient outcomes are inevitable. Although the effects of nursing preparation (RN versus LPN) have been used to predict patient outcomes in previous studies, most have been conducted indirectly, with the primary purpose being multidimensional, such as determining the influences of environmental and institutional factors on mortality rates. In addition, the inclusion of all levels of RN education, such as bachelor of science (BSN), associate degree (ADN), and diploma, has only recently been used to assess patient outcomes.

Although nursing education levels are clearly defined, patient outcomes are less straightforward. The ultimate negative outcome, mortality, has been used in the majority of early research. Studies focusing on nursing-sensitive outcomes have empirically determined which variables should be used to determine quality care. For example, the American Nurses’ Association (ANA), the National Quality Forum, and the Agency for Healthcare Research and Quality (AHRQ) have all researched patient outcomes that are affected by nursing practice. According to Savitz, Jones, and Bernard (2005), the AHRQ’s Patient Safety Indicators (PSI) dominate due to a “combined evidence-based, expert panel approach” on comparison of these three measures (p. 381). Although there are other integrative reviews that have focused on the relationship between nurse staffing and patient outcomes, none could be located that used the AHRQ PSIs to synthesize studies examining nurse education level and patient outcomes. Therefore, the purpose of this literature review was to integrate empirical findings regarding the relationship of nurse education level and patient safety, using PSIs as defined by the AHRQ (2003) to determine applicable patient outcome measures (Table 1).

Agency for Healthcare Research and Quality Patient Safety Indicators

Table 1: Agency for Healthcare Research and Quality Patient Safety Indicators

Literature Review

The literature search for this review began with a key word search using the PubMed, CINAHL, Cochrane Review, and Dissertation Abstracts International electronic databases. The following key words for the search were used: educational status, hospital mortality, nursing staff, hospital, registered nurses, and patient safety. Published works were initially limited to empirical studies and review articles published within the 20-year period from 1986 to 2006. Review articles were then discarded after the references and bibliographies were searched for additional empirical studies. As suggested by Cooper (1998), references from all empirical studies then were reviewed and potential studies were retrieved based on the titles (ancestry approach). Key researchers were determined by reviewing these articles; thereafter, these author’s names were entered in the Web of Science database to locate citations of their work (descendancy approach) and potential studies for further review (Cooper, 1998).

Government reports were then accessed via http://www.access.gpo.gov/, and the Institute of Medicine’s report on nurse staffing and quality of care was retrieved to review applicable studies (Wunderlich & Davis, 1996). The Computer Retrieval of Scientific Projects database was used to determine whether ongoing research was being conducted. One such study was found to have begun in August 2005 and is set to be completed by January 2009 (Lake, 2006). In addition, according to L. Aiken (personal communication, March 14, 2006), primary investigator of the most applicable study to date (Aiken, Clarke, Cheung, Sloane, & Silber, 2003), her research group is conducting a larger study now in Pennsylvania, California, and New Jersey to replicate earlier work linking nurse education and patient mortality. Epidemiological reviews via http://www.annualreviews.org revealed no further prospects for the current integration.

After all studies were retrieved, they were initially reviewed for study design and variable applicability. For example, studies that took only nurse staffing into account as the independent variable and did not specify level of education (i.e., RN or LPN) were discarded. Although studies from countries other than the United States were included (Estabrooks, Midodzi, Cummings, Ricker, & Giovannetti, 2005; McCloskey & Diers, 2005; Sasichay-Akkadechanunt, Scalzi, & Jawad, 2003; Tourangeau, Giovannetti, Tu, & Wood, 2002), the level of education had to be clearly stated and comparable to the nursing education system in the United States. One study from the United Kingdom was deleted for this reason, as nursing education was described in grades instead of levels. Studies in which the dependent variable(s) did not meet the criteria specified by the AHRQ’s PSI were not used. Only studies that used data from acute care hospitals were included, whereas data from long-term care units were excluded. After all studies were finalized for inclusion, as suggested by Garrard (1999) they were read again and abstracted into a matrix format.

This integrative review followed the methods described by Cooper (1998) and Garrard (1999). First, the topic was abstracted into its basic concepts, or variables. Next, research methods (i.e., design, subjects, data collection instruments and procedures, and data analysis) were described (i.e., quality of data). Then, major findings were synthesized (i.e., what is known). Missing and inadequate topics are discussed (i.e., what needs to be known), along with questions for further research (i.e., what subsequent direction needs to be taken). Critical analyses are integrated throughout to provide an evaluation of how well the literature represents the issue.

Nurse Education

Nurse education has overwhelmingly been dichotomized as either RN or LPN, categorizing all nurses prepared for RN practice into one group, despite being educated at three different levels. On analysis of the current literature, it was noted that nursing education as an independent variable has been defined as either RN/LPN dose or number of RN/LPN care hours. For example, some studies used RN/LPN full-time equivalents per 100 patient days (Farley & Ozminkowski, 1992), per 1,000 patient days (Mark, Harless, McCue, & Xu, 2004), per patient adjusted day (Kovner & Gergen, 1998; Kovner, Jones, Zhan, Gergen, & Basu, 2002), per patient day (Cho, Ketefian, Barkauskas, & Smith, 2003), per patient admission (Manheim, Feinglass, Shortell, & Hughes, 1992), per average patient daily census (Person et al., 2004), or per occupied bed (Bond, Raehl, Pitterle, & Franke, 1999; Silber et al., 2002) to operationalize either RN or LPN dose; whereas Needleman, Buerhaus, Mattke, Stewart, and Zelevinsky (2002) used RN/LPN hours per patient day.

Nursing skill mix, or number of RNs proportionate to other nursing staff, has also been used to define and measure nurse education as an independent variable. For example, researchers used either the percentage of RN employees (Shortell & Hughes, 1988) or percentage of nurses who were RNs (Aiken, Smith, & Lake, 1994; al-Haider & Wan, 1991; ANA, 2000; Blegen, Goode, & Reed, 1998; Blegen & Vaughn, 1998; Cho et al., 2003; Estabrooks et al., 2005; Hartz et al, 1989; Krakauer et al., 1992; Lichtig, Knauf, & Milholland, 1999; McCloskey & Diers, 2005; Sasichay-Akkadechanunt et al., 2003; Tourangeau et al., 2002).

It was not until 2003 that level of nurse education became more specified. For example, Aiken et al. (2003) divided RNs into highest level of nursing education obtained (e.g., BSN or higher, ADN, diploma). Estabrooks et al. (2005) and Sasichay-Akkadechanunt et al. (2003) differentiated RN levels as well. Because these three studies were the first to use these refined definitions of nurse education, they limited their outcome measures to include only failure-to-rescue (Aiken et al., 2003) and mortality (Aiken et al., 2003, Estabrooks et al., 2005; Sasichay-Akkadechanunt et al., 2003).

Patient Safety

AHRQ (2003) defined patient safety as “freedom from accidental injury or avoiding injuries or harm to patients from care that is intended to help them” (p. 8). Although negative in connotation, 30-day postadmission mortality has been the most common patient safety outcome identified in the literature (Aiken et al., 1994, 2003; Estabrooks et al., 2005; Hartz et al., 1989; Krakauer et al., 1992; Manheim et al., 1992; Person et al., 2004; Silber et al., 2002; Tourangeau et al., 2002) followed by in-patient mortality (Al-Haider & Wan, 1991; Blegen et al., 1998; Farley & Ozminkowski, 1992; Mark et al., 2004; Person et al., 2004; Sasichay-Akkadechanunt et al., 2003; Shortell & Hughes, 1988) and mortality in which timing is unspecified (Bond et al., 1999; McCloskey & Diers, 2005; Needleman et al., 2002).

By 1998, patient outcomes other than mortality began to be used in these retrospective analyses. For example, decubitus ulcers and unspecified infections (Blegen et al., 1998), along with cardiopulmonary arrest (Blegen & Vaughn, 1998), postsurgical thrombosis, urinary tract infection (UTI), pneumonia, and pulmonary compromise (Kovner & Gergen, 1998) are thought to be sensitive to nurse staffing. Other researchers followed thereafter, using these same outcomes in the next few years (ANA, 2000; Cho et al., 2003; Kovner et al., 2002; Lichtig et al., 1999; Mark et al., 2004; McCloskey & Diers, 2005; Needleman et al., 2002). Other researchers pioneered the inclusion of failure-to-rescue in the outcomes, defined as deaths in patients with serious complications (Aiken et al., 2003; Needleman et al., 2002; Silber et al., 2002), as well as shock and physical and metabolic derangement (McCloskey & Diers, 2005; Needleman et al., 2002).

Method

All studies in the review pool were conducted with non-experimental research, using a retrospective design. All studies were cross-sectional except for three that were longitudinal, with data collected over an extended period of time (Farley & Ozminkowski, 1992; Mark et al., 2004; McCloskey & Diers, 2005).

Participants

Hospital units were used to collect data, the majority of which were located in the United States (Aiken et al., 1994, 2003; al-Haider & Wan, 1991; ANA, 2000; Blegen et al., 1998; Blegen & Vaughn, 1998; Bond et al., 1999; Cho et al., 2003; Farley & Ozminkowski, 1992; Hartz et al., 1989; Kovner & Gergen, 1998; Kovner et al., 2002; Krakauer et al., 1992; Lichtig, 1999; Manheim et al., 1992; Mark et al., 2004; Needleman et al., 2002; Person et al., 2004; Shortell & Hughes, 1988; Silber et al., 2002). Countries where other studies were conducted include New Zealand (McCloskey & Diers, 2005), Canada (Estabrooks et al., 2005; Tourangeau et al., 2002), and Thailand (Sasichay-Akkadechanunt et al., 2003).

A few of the studies used only one hospital for data collection (Blegen et al., 1998; Sasichay-Akkadechanunt et al., 2003), whereas the largest study used 6,668 hospitals (Person et al., 2004). In the study by Farley & Ozminkowski (1992), the number of hospitals used is not clearly stated. Hospitals in all reviewed studies were acute care; however, some studies retrieved data from Medicare discharges only (Aiken et al., 1994; al-Haider & Wan, 1991; Bond et al., 1999; Hartz et al., 1989; Manheim et al., 1992; Person et al., 2004; Shortell & Hughes, 1988; Silber et al., 2002), meaning that the older adult population (older than age 65) was well represented.

Many studies had very specific inclusion and exclusion criteria. For example, several included only prespecified ages or medical-surgical conditions (Aiken et al., 2003; Cho et al., 2003; Estabrooks et al., 2005; Farley & Ozminkowski, 1992; Person et al., 2004; Shortell & Hughes, 1988; Sasichay-Akkadechanunt et al., 2003; Tourangeau et al., 2002). Other studies used more general criteria, such as all medical-surgical adults (Bond et al., 1999; Hartz et al., 1989; Mark et al., 2004; McCloskey & Diers, 2005; Needleman et al., 2002; Silber et al., 2002).

A few studies made great efforts to get results that were more generalizable than others. For instance, researchers selected hospitals randomly from predefined substrata for two studies (Kovner & Gergen, 1998; Krakauer et al., 1992); in another study, hospitals were divided into nine regions across the United States, providing varied representation (Manheim et al., 1992). Aiken et al. (1994) used a multivariate matched sampling procedure to decrease sampling bias and confounding variables, which are common in retrospective data analyses. Because Blegen et al. (1998) used one large hospital for their initial study, all inpatient units were used; thereafter, Blegen and Vaughn (1998) used all inpatient units in 11 hospitals. Other researchers used all inpatient units as well (ANA, 2000, Kovner et al., 2002), with some excluding psychiatric and rehabilitation units only (Lichtig et al., 1999).

Data Collection Instruments and Procedures

The principal source of United States mortality information was based on federal claims data from the Health Care Financing Administration’s (now called the Centers for Medicare and Medicaid Services) Medicare Provider Analysis and Review file (Aiken et al., 1994; al-Haider & Wan, 1991; ANA, 2000; Bond et al., 1999; Hartz et al., 1989; Krakauer et al., 1992; Manheim et al., 1992; Mark et al., 2004; Shortell & Hughes, 1988; Silber et al., 2002). Several studies also used the American Hospital Association’s Annual Survey for data such as nurse staffing and other hospital characteristics (Aiken et al., 1994, 2003; Al-Haider & Wan, 1991; Bond et al., 1999; Farley & Ozminkowski, 1992; Hartz et al., 1989; Kovner & Gergen, 1998; Kovner et al., 2002; Manheim et al., 1992; Mark et al., 2004; Needleman et al., 2002; Person et al., 2004; Silber et al., 2002). In addition, the AHRQ’s Healthcare Cost and Utilization Project data were used (Farley & Ozminkowski, 1992; Mark et al., 2004).

Other countries also collected information from national data sets. For example, due to the small size of New Zealand (equivalent to Colorado), McCloskey & Diers (2005) were able to use the National Minimum Data Set and the Nursing Workforce Dataset to collect data on 85 of their country’s hospitals. In addition, the Canadian Institute for Health Information maintains a Hospital In-patient Database that was used for one of the Canadian studies (Estabrooks et al., 2005).

Alternatively, some researchers used state-level data (Aiken et al., 2003; ANA, 2000; Cho et al., 2003; Estabrooks et al., 2005; Lichtig et al., 1999; Silber et al., 2002; Tourangeau et al., 2002), whereas others used data directly from either the community (i.e., for demographics) or hospitals themselves (al-Haider & Wan, 1991; Blegen et al., 1998; Blegen & Vaughn, 1998; Person et al., 2004; Sasichay-Akkadechanunt et al., 2003). Even more specifically, some nurse staffing and education data came directly from RNs that were surveyed (Aiken et al., 2003; Estabrooks et al., 2005). Although useful data, response rates on nurse questionnaires from the studies by Estabrooks et al. (2005) (52.8%) and Aiken et al. (2003) (52%) may have biased results, whereas well-designed interview data may have achieved a higher response rate of 80% to 90% (Polit & Beck, 2004).

Data Analysis

Descriptive and multivariate statistical procedures (i.e., multiple regression and correlation) were used to analyze all data, with the goals of describing data and answering research questions related to the variables of interest. Some of the studies were not primarily being conducted to answer questions relative to the effect of nurse education on patient safety; however, questions about this topic were answered indirectly with multiple variables being tested at one time. For example, the purpose of the study conducted by Krakaeur et al. (1992) was to compare the results of risk adjustment achieved with claims model data versus clinical model data. Despite this, the researchers conducted a correlational analysis that revealed an inverse relationship between the proportion of RNs and 30-day mortality rate.

One of the advantages of using multivariate statistics is the ability to statistically control for the effects of confounding variables, such as patient acuity, that cannot otherwise be controlled with design methods (Aiken et al., 2003; ANA, 2000; Blegen et al., 1998; Blegen & Vaughn, 1998; Bond et al., 1999; Estabrooks et al., 2005; Farley & Ozminkowski, 1992; Kovner et al., 2002; Mark et al., 2004; Needleman et al., 2002; Person et al., 2004; Sasichay-Akkadechanunt et al., 2003; Silber et al., 2002). On the other hand, each study used a different set of variables to control for patient characteristics. For example, Bond et al. (1999) used the percentages of intensive care unit days (intensive care unit days/total inpatient days), emergency department visits (yearly emergency department visits/average daily census), and Medicaid patients (Medicaid discharges/total discharges) to control for severity of patient illness. In contrast, patient mix was acuity adjusted in the ANA (2000) study using nursing intensity weights to recognize differences in patients’ need for nursing care.

Results

The majority of studies found no significant association between either RN skill mix or RN dose and in-hospital mortality (al-Haider & Wan, 1991; Blegen et al., 1998; Sasichay-Akkadechanunt et al., 2003; Shortell & Hughes, 1988). Mark et al. (2004) found that increasing RN staffing had a diminishing marginal effect on reducing in-hospital mortality rates. In contrast, several studies found a significant association between either RN skill mix or RN dose and in-hospital mortality (Farley & Ozminkowski, 1992; Person et al., 2004). The three studies that did not specify timing for mortality (Bond et al., 1999; McCloskey & Diers, 2005; Needleman et al., 2002) were split with two of the studies finding no significant association between the variables (McCloskey & Diers, 2005; Needleman et al., 2002) and one finding that mortality rates decreased as staffing level per occupied bed increased for RNs, whereas mortality rates increased as staffing level per occupied bed increased for LPNs (Bond et al., 1999).

Overwhelmingly, 30-day mortality was consistently found to have a significant inverse relationship with either RN skill mix or RN dose (Estabrooks et al., 2005; Hartz et al., 1989; Krakauer et al., 1992; Manheim et al., 1992; Person et al., 2004; Silber et al., 2002; Tourangeau et al., 2002). Although in their study, Aiken et al. (1994) noted that RN skill mix was not a predictor of 30-day mortality, in 2003 they found that a 10% increase in the proportion of BSN nurses was associated with a 5% decrease in likelihood of death (30-day mortality). Estabrooks et al. (2005) concurred with the latter findings by suggesting that hospitals with a higher proportion of BSN nurses were associated with lower rates of 30-day patient mortality (95% confidence interval).

Regarding outcomes other than mortality, pneumonia was most consistently found to be inversely related to both RN dose (Kovner & Gergen, 1998; Kovner et al., 2002; Needleman et al., 2002) and RN skill mix (ANA, 2000; Lichtig et al., 1999). In fact, Cho et al. (2003) found an inverse RN dose at a rate of 1 hour RN increase per patient day to be associated with an 8.9% decrease in the odds of pneumonia; in addition, they found a 10% increase in RN proportion (skill mix) to be associated with a 9.5% decrease in the odds of pneumonia. On the other hand, Mark et al. (2004) determined that RN dose had no consistent effect on pneumonia.

Studies using decubiti as an outcome measure found an inverse relationship with RN skill mix (ANA, 2000; Blegen et al., 1998; Lichtig et al., 1999), with the exception of Mark et al. (2004), who found that RN dose had no consistent effect on decubiti. Moreover, McCloskey & Diers (2005) reported that when RN skill mix in New Zealand increased 18% from 1989 to 2000, there was a significant increase in decubiti as well; however, the number of acute patient days increased during this time and a rise in patient acuity therefore occurred. This could explain the difference in findings because it was not controlled for in the statistical analysis. Other influencing factors to consider include the notation that RN education in New Zealand is available through 3-year BSN programs, as opposed to 4-year BSN programs in the United States; in addition, enrolled nurses, who have licensure and scope similar to LPNs, were being phased out during the time of data collection in the New Zealand study, which may have greatly influenced the negative results. For example, although RNs became higher in number and proportion during this period of LPN lay-off, it is likely that short staffing influenced the outcome, which might not have been observed if the study had been completed during a different time period.

Nursing shortage issues were also reflected on regarding the results found in the studies conducted in the United States. According to the National Sample Survey of Registered Nurses conducted every 4 years by the Bureau of Health Professions (Health Resources & Services Administration, 2006), RNs employed in nursing increased every 4 years, beginning with 1,277,041 in 1980 and re porting 2,421,461 in 2004. On consideration of the United States’ population growth from 1960 to 2000 (Figure 1), one must question whether the rise in employed RNs was enough to keep up with the census increase (U.S. Census Bureau, 2006). On analysis of the percentage of RN employment trends in the United States (Figure 2), it was determined that there was an RN shortage in the mid to late 1980s, as well as in the mid to late 1990s, with recovery continuing at the turn of the century.

United States Population, 1960 to 2000.Source: U.S. Census Bureau (2006).

Figure 1: United States Population, 1960 to 2000.Source: U.S. Census Bureau (2006).

RN Shortage Trends in the United States, 1980 to 2004.Source: Health Resources & Services Administration (2006).

Figure 2: RN Shortage Trends in the United States, 1980 to 2004.Source: Health Resources & Services Administration (2006).

Four of six studies conducted in the United States during the first shortage (Farley & Ozminkowski, 1992; Hartz et al., 1989; Krakauer et al., 1992; Manheim et al., 1992) determined that RN status (either dose or skill mix) positively affected patient outcomes, whereas two of the six studies (Aiken et al., 1994; al-Haider & Wan, 1991) determined no relationship. In addition, findings from studies conducted in the United States during the second nursing shortage (Aiken et al., 2003; ANA, 2000; Cho et al., 2003; Needleman et al., 2002) showed significant relationships between patient outcomes such as pneumonia, failure-to-rescue, postoperative infection, mortality, UTIs, shock or cardiac arrest, and RN status, especially among nurses with a BSN degree (Aiken et al., 2003). Therefore, neither shortage seemed to negatively affect outcomes; however, it is unknown whether the positive nursing impact would have been even greater during a different time period.

On consideration of all studies, postoperative infections and UTIs were both found to be inversely related to RN skill mix (ANA, 2000; Kovner & Gergen, 1998; Lichtig et al., 1999) and RN dose (Needleman et al., 2002); however, the same exceptions were noted as above with decubitus ulcers with Mark et al. (2004) finding no relationship and McCloskey & Diers (2005) reporting a positive association between UTIs, wound infections, sepsis, and RN skill mix. Blegen et al. (1998) also found no association between un-specified infections and RN dose.

Failure-to-rescue was consistently found to have a significant inverse relationship with RN dose (Needleman et al., 2002; Silber et al., 2002). Aiken et al. (2003) even found a 10% increase in proportion of BSN nurses to be associated with a 5% decrease in the likelihood of failure-to-rescue.

The findings regarding shock or cardiopulmonary arrest were divided in the studies by Blegen & Vaughn (1998), who found no association with RN skill mix, and Needleman et al. (2002), who found an inverse relationship with RN dose. In addition, similar findings were noted regarding postsurgical thrombosis and pulmonary compromise in the study by Kovner and Gergen (1998), who found an inverse relationship between these outcomes and RN dose. On the other hand, McCloskey and Diers (2005) found these variables to be positively associated with RN skill mix along with physical and metabolic derangement. Of special notation, the two studies that examined the relationship between LPN dose and adverse events found none (Kovner et al., 2002; Needleman et al., 2002).

The greatest limitation noted in overall findings was the inconsistency of data retrieved from multiple databases. For example, adverse outcomes (PSIs) derived from data from the Health Care Financing Administration (now called the Centers for Medicare and Medicaid Services) were determined via International Classification of Diseases, Ninth Revision, Clinical Modification codes (Centers for Medicare and Medicaid Services, 2008), which were validated by an expert panel (Cho et al., 2003); however, other databases were maintained at the hospital level with alternative definitions of measures and varying levels of content validity. A standardized set of measures that defines nursing education level and performance, along with evidenced-based patient safety outcomes sensitive to nursing care is needed at the national level to enhance reliability of findings.

A summary of findings (Table 2) revealed that 30-day mortality and pneumonia were found to be inversely related to both RN skill mix and RN dose. In other words, as hospitals staffed more RNs versus LPNs, there were fewer patient deaths noted within 30 days of admission and fewer cases of hospital-acquired pneumonia. Although both RN skill mix and RN dose were well represented across studies, RN dose (17 studies) was more significantly related to total PSIs than RN skill mix (14 studies). Decubitus ulceration was noted to be inversely related to RN skill mix in three studies. Urinary tract infections were inversely related in two studies using both RN skill mix and RN dose (four studies total). Other PSIs found to be inversely related to either RN dose or skill mix include in-hospital mortality, mortality in which timing was unspecified, failure-to-rescue, pulmonary embolus or deep vein thrombosis, pulmonary failure, postoperative infections (unspecified), and shock or cardiac arrest. PSIs that were included in data collection but not noted to be sensitive to nursing care include sepsis, physical or metabolic derangement, and wound infection.

Number of Studies Showing Inverse Correlation or Positive Predictor Values

Table 2: Number of Studies Showing Inverse Correlation or Positive Predictor Values

Discussion

On the basis of this integrative review, a broad array of patient safety outcomes has been researched using non-experimental methods. Because outcomes thought to be sensitive to nursing care are variables that are not amendable to control or manipulation, retrospective designs using large databases are acceptable for discovering relationships between nurse education and patient safety. For instance, many PSIs were noted to be inversely related to RN care. In other words, acute care hospitals that employ more RNs consistently report fewer adverse events such as 30-day mortality, decubitus ulcers, and pneumonia. On the other hand, there are several variables and concepts that were noted to be missing or inadequately covered in these studies. For example, few studies considered the effects of LPN nursing care on patient safety (Bond et al., 1999; Kovner et al., 2002; Mark et al., 2004; Needleman et al., 2002; Person et al., 2004); whereas all studies included either RN dose or skill mix.

There were also several variables that were notably missing from the reviewed studies. Patient outcomes such as maternal and neonatal birth trauma (i.e., during vaginal or cesarean deliveries) have been identified as PSIs (AHRQ, 2003). For example, birth trauma includes any significant injury to the head, neck, or shoulder, with the exception of preterm infants with a subdural or cerebral hemorrhage. In addition, neonates with a skeletal injury and a diagnosis of osteogenesis imperfecta are excluded as PSIs. It is possible that such outcomes are not easily retrievable because many of them may not be consistently reported or officially coded out of fear of malpractice litigation.

Additional PSIs that were consistently absent from the reviewed studies include complications of anesthesia, foreign bodies left during operative procedures, postoperative hip fracture, hemorrhage or hematoma, wound dehiscence, accidental puncture or laceration, and transfusion reaction. AHRQ (2003) found that these variables are sensitive to patient care when risk adjusted for age, gender, drug-related group, and comorbidity categories; however, it is likely that not all PSIs are sensitive to nursing care (i.e., foreign bodies left during operative procedures, neonatal birth trauma).

Other related patient outcomes pertain to those from a pediatric population. The majority of studies limited patient data to either Medicare (older adults) or adults older than age 18. AHRQ (2006) recently devised a set of pediatric quality indicators that measure pediatric healthcare quality. These indicators will likely show up in future studies focusing on this special population.

Only three studies that included variables more specific than either RN or LPN were reviewed (Aiken et al., 2003; Estabrooks et al., 2005; Sasichay-Akkadechanunt et al., 2003); however, Estabrooks et al. (2005) did not include ADNs as a choice because the study was conducted in Canada. In addition, the study by Sasichay-Akkadechanunt et al. (2003) was conducted in Thailand using only one 2,300-bed university hospital and included information on the percentage of BSN nurses without consideration for the other RN education levels. Although Aiken et al. (2003) collected data on all levels of RN education (i.e., masters, BSN, or ADN degrees, diploma), the ADN and diploma categories were “collapsed into a single category and the educational composition of the hospital staff was characterized in terms of the percentage of nurses holding bachelor’s or master’s degrees” (p. 1618). Therefore, findings were limited to examining whether the proportion of hospital RNs educated at the baccalaureate level or higher was associated with risk-adjusted 30-day mortality and failure-to-rescue, with little emphasis on the original question of what effect a nurse’s education level has on patient outcomes.

In addition, although these three aforementioned studies differentiated RN education into levels, the only outcome variables used were mortality and failure-to-rescue, which are negative outcomes. Because patient safety is well defined, outcome measures that are stated in a positive sense should be developed.

Patient outcomes that are sensitive to nursing process variables such as knowledge, comprehension, application, critical thinking, or clinical judgement must also be explored. More specifically, studies focusing on the relationship between nurse competence and patient outcomes are grossly lacking in the literature, as are studies focusing on patient outcomes related to nursing care provided by nursing students.

Studies that build on previous findings should further explore these inadequacies. Experimental studies using a variety of PSIs empirically known to be sensitive to nursing care should be used to evaluate the differences in patient safety outcomes when care is delivered by LPNs, ADNs, BSNs, MSNs, diploma nurses, and, ideally, nursing students. Integration of computerized documentation of nursing care (e.g., Meditech; Medical Information Technology, Inc., West-wood, MA) could provide data used to analyze these differences. In addition, the following questions could be advanced with well-designed qualitative and quantitative studies: How is nursing competence defined? How is it developed? What concepts validate competence? How are knowledge, comprehension, application, and critical thinking related to competence? What is the association between these concepts and patient safety? What levels of nursing education develop these competencies? What concepts define the variations in nurse education levels? How do various levels of nurse education affect patient safety?

Theoretically, if there is a relationship between level of nursing education and patient safety, and if levels of nursing education are associated with competence, then is there a relationship between nursing competence and patient safety? What is more, if there is a relationship between nursing competence and patient safety, then what teaching methods can best cultivate the development of this competence in nursing students?

Conclusion

Determining the best nursing staff to provide the safest patient care possible is a daunting task for health care systems. To focus on the most basic characteristic of these professionals, this integrative review has provided an overview of the current state of science regarding how their educational level affects patient safety using AHRQ’s PSIs to evaluate outcomes. Although there is evidence indicating that an increased RN dose and skill mix decreases adverse patient outcomes sensitive to nursing care, the breakdown of nurse education level is virtually nonexistent. Further research to explore the relationship of nursing education levels is needed to strengthen what is currently known so that recommendations for nurse staffing will target safer patient outcomes. Studies also need to investigate the role of competence so implications for nurse educators can determine the most efficient ways of cultivating this necessary ability.

References

  • Agency for Healthcare Research & Quality. 2003. AHRQ quality indicators. Retrieved February 20, 2006, from http://www.qualityindicators.ahrq.gov
  • Agency for Healthcare Research & Quality. 2006. AHRQ quality indicators: Pediatric quality indicators overview. Retrieved March 21, 2006, from http://www.qualityindicators.ahrq.gov/pdi_overview.htm
  • Aiken, LH, Clarke, SP, Cheung, RB, Sloane, DM & Silber, JH2003. Educational levels of hospital nurses and surgical patient mortality. Journal of the American Medical Association, 290, 1617–1623. doi:10.1001/jama.290.12.1617 [CrossRef]
  • Aiken, LH, Smith, HL & Lake, ET1994. Lower Medicare mortality among a set of hospitals known for good nursing care. Medical Care, 32, 771–787. doi:10.1097/00005650-199408000-00002 [CrossRef]
  • al-Haider, A & Wan, TT1991. Modeling organizational determinants of hospital mortality. Health Services Research, 26, 303–323.
  • American Nurses Association. 2000. Executive summary: Nurse staffing and patient outcomes in the inpatient hospital setting. Washington, DC: Author.
  • Blegen, MA, Goode, CJ & Reed, L1998. Nurse staffing and patient outcomes. Nursing Research, 47, 43–50. doi:10.1097/00006199-199801000-00008 [CrossRef]
  • Blegen, MA & Vaughn, T1998. A multisite study of nurse staffing and patient occurrences. Nursing Economic$, 16, 196–203.
  • Bond, CA, Raehl, CL, Pitterle, ME & Franke, T1999. Health care professional staffing, hospital characteristics, and mortality rates. Pharmacotherapy, 19, 130–138. doi:10.1592/phco.19.3.130.30915 [CrossRef]
  • Centers for Medicare and Medicaid Services. 2008. Overview: ICD-9 provider & diagnostic codes. Retrieved February 27, 2008, from http://www.cms.hhs.gov/ICD9ProviderDiagnosticCodes/
  • Cho, SH, Ketefian, S, Barkauskas, VH & Smith, DG2003. The effects of nurse staffing on adverse events, morbidity, mortality, and medical costs. Nursing Research, 52, 71–79. doi:10.1097/00006199-200303000-00003 [CrossRef]
  • Cooper, HM1998. Synthesizing research: A guide for literature reviews (3rd ed) Newbury Park, CA: Sage.
  • Estabrooks, CA, Midodzi, WK, Cummings, GG, Ricker, KL & Giovannetti, P2005. The impact of hospital nursing characteristics on 30-day mortality. Nursing Research, 54, 74–84. doi:10.1097/00006199-200503000-00002 [CrossRef]
  • Farley, DE & Ozminkowski, RJ1992. Volume-outcome relationships and in-hospital mortality: The effect of changes in volume over time. Medical Care, 30, 77–94. doi:10.1097/00005650-199201000-00009 [CrossRef]
  • Garrard, J. 1999. Health sciences literature review made easy Gaithersburg, MD: Aspen.
  • Hartz, AJ, Krakauer, H, Kuhn, EM, Young, M, Jacobsen, SJ & Gay, G et al. . 1989. Hospital characteristics and mortality rates. The New England Journal of Medicine, 321, 1720–1725.
  • Health Resources & Services Administration. 2006. The registered nurse population: Findings from the 2004 national sample survey of registered nurses. Retrieved April 18, 2006, from http://bhpr.hrsa.gov/healthworkforce/rnsurvey04/
  • Kovner, C & Gergen, PJ1998. Nurse staffing levels and adverse events following surgery in U.S. hospitals. Image, 30, 315–321.
  • Kovner, C, Jones, C, Zhan, C, Gergen, PJ & Basu, J2002. Nurse staffing and postsurgical adverse events: An analysis of administrative data from a sample of U.S. hospitals, 1990–1996. Health Services Research, 37, 611–629. doi:10.1111/1475-6773.00040 [CrossRef]
  • Krakauer, H, Bailey, RC, Skellan, KJ, Stewart, JD, Hartz, AJ & Kuhn, EM et al. . 1992. Evaluation of the HCFA model for the analysis of mortality following hospitalization. Health Services Research, 27, 317–335.
  • Lake, ET2006. Nurse staffing and adverse events on inpatient units Ongoing research project at the University of Pennsylvania [Data file]. Retrieved February 27, 2008, from the CRISP database.
  • Lichtig, LK, Knauf, RA & Milholland, DK1999. Some impacts of nursing on acute care hospital outcomes. Journal of Nursing Administration, 292, 25–33. doi:10.1097/00005110-199902000-00008 [CrossRef]
  • Manheim, LM, Feinglass, J, Shortell, SM & Hughes, EF1992. Regional variation in Medicare hospital mortality. Inquiry, 29, 55–66.
  • Mark, BA, Harless, DW, McCue, M & Xu, Y2004. A longitudinal examination of hospital registered nursing staffing and quality of care. Health Services Research, 39, 279–300. doi:10.1111/j.1475-6773.2004.00228.x [CrossRef]
  • McCloskey, BA & Diers, DK2005. Effects of New Zealand’s health reengineering on nursing and patient outcomes. Medical Care, 43, 1140–1146. doi:10.1097/01.mlr.0000182549.85761.cd [CrossRef]
  • Needleman, J, Buerhaus, P, Mattke, S, Stewart, M & Zelevinsky, K2002. Nurse-staffing levels and the quality of care in hospitals. The New England Journal of Medicine, 346, 1715–1722. doi:10.1056/NEJMsa012247 [CrossRef]
  • Person, SD, Allison, JJ, Kiefe, CI, Weaver, MT, Williams, OD & Centor, RM et al. 2004. Nurse staffing and mortality for medicare patients with acute myocardial infarction. Medical Care, 42, 4–12. doi:10.1097/01.mlr.0000102369.67404.b0 [CrossRef]
  • Polit, DF & Beck, CT2004. Nursing research: Principles and methods (7th ed) Philadelphia: Lippincott Williams & Wilkins.
  • Sasichay-Akkadechanunt, T, Scalzi, CC & Jawad, AF2003. The relationship between nurse staffing and patient outcomes. Journal of Nursing Administration, 33, 478–485. doi:10.1097/00005110-200309000-00008 [CrossRef]
  • Savitz, LA, Jones, CB & Bernard, S. 2005. Quality indicators sensitive to nurse staffing in acute care settings. In Advances in patient safety: From research to implementation (Vol 4, AHRQ Publication No 05-0021-4) Rockville, MD: Author Retrieved March 18, 2006, from http://www.ahrq.gov/downloads/pub/advances/vol4/Savitz.pdf
  • Shortell, SM & Hughes, EF1988. The effects of regulation, competition, and ownership on mortality rates among hospital inpatients. The New England Journal of Medicine, 318, 1100–1107.
  • Silber, JH, Kennedy, SK, Even-Shoshan, O, Chen, W, Mosher, RE & Showan, AM et al. 2002. Anesthesiologist board certification and patient outcomes. Anesthesiology, 96, 1044–1052. doi:10.1097/00000542-200205000-00004 [CrossRef]
  • Tourangeau, AE, Giovannetti, P, Tu, JV & Wood, M2002. Nursing-related determinants of 30-day mortality for hospitalized patients. Canadian Journal of Nursing Research, 334, 71–88.
  • U.S. Census Bureau. 2006. Statistical abstract of the United States, 2006. Retrieved April 19, 2006, from http://www.census.gov/
  • Wunderlich, GS & Davis, CK1996. Nursing staff in hospitals and nursing homes: Is it adequate? Washington, DC: National Academies Press.

Agency for Healthcare Research and Quality Patient Safety Indicators

Complications of anesthesia
Death in low-mortality drug-related groups
Decubitus ulcer
Failure to rescue
Foreign body left during procedure
Iatrogenic pneumothorax
Selected infections due to medical care
Postoperative hip fracture
Postoperative hemorrhage or hematoma
Postoperative physiologic and metabolic derangements
Postoperative respiratory failure
Postoperative pulmonary embolism or deep vein thrombosis
Postoperative sepsis
Postoperative wound dehiscence
Accidental puncture or laceration
Transfusion reaction
Birth trauma, injury to neonate
Obstetric trauma, vaginal with instrument
Obstetric trauma, vaginal without instrument
Obstetric trauma, cesarean delivery

Number of Studies Showing Inverse Correlation or Positive Predictor Values

Predictor ValueRN Skill MixRN Dose
Mortality (in-hospital)*2
Mortality (30-day post-admit)43
Mortality (timing unstated)*1
Failure-to-rescue*2
Decubitus ulcer3*
Pneumonia34
Sepsis**
Urinary tract infections22
Pulmonary embolus/deep vein thrombosis*1
Physical/metabolic derangement**
Pulmonary failure*1
Wound infection**
Postoperative infections (unspecified)2*
Shock or cardiac arrest*1

Because patient safety is well defined, outcome measures that are stated in a positive sense should be developed.

Authors

Ms. Ridley is Lecturer and Clinical Instructor, Department of Nursing, Murray State University, Murray, Kentucky, and a PhD candidate, Saint Louis University, St. Louis, Missouri.

Address correspondence to Renee T. Ridley, RN, MSN, CFNP, Lecturer and Clinical Instructor, 120 Mason Hall, Murray, KY 42071; e-mail: .renee.ridley@murraystate.edu

10.3928/01484834-20080401-06

Sign up to receive

Journal E-contents