nursing

The role of theory in improving patient safety and quality health care

Barbara A. Mark, PhD, RN, FAAN Linda C. Hughes, PhD, RN Cheryl Bland Jones, PhD, RN

ple et of vide ean- led l to Mer- ads igh por- elop g to the ost

lex e in

alth lity cur ela-

ause nd

es in ere. gical are

ry tion- de-

why t of tural nce- est atic

e— of

icted. ing ally

ot be pro-

t of ality

the een

s the rse s

the very, ts ugh

im- cally usly , the ely

od-

ssor

ity of

orth

a at 0.

By examining selected research that investigates the relationship between nurse staffing and adverse events, we demonstrate the problems that result from the absence of a strong theory to guide this research. There is considerable work to be done in explicating the theory underlying empirical studies of the relation- ship between nurse staffing and outcomes. Key con- structs must be placed within a theoretical context, proposed causal mechanisms underlying the empiri- cal or hypothesized relationships must be identified, and critical mediating and moderating variables must be recognized.

T heory can be used to mean anything from a sim guess or mere speculation to articulation of a s interrelated and logical statements that pro

explanatory and predictive power. The diverse m ings that can be attributed to the word theory Merton1 to suggest that theory has the potentia obscure rather than create understanding. Despite ton’s cautionary note, development of theory that le to an understanding of nursing’s contribution to h quality and safe patient care is of paramount im tance. Such theory is essential if we are to dev knowledge that can be generalized from one settin another, recommend ways of organizing to improve delivery of quality patient care and, perhaps m importantly, avoid simplistic solutions to the comp problem of insuring safe and high quality patient car organizational settings.2

Successful theory development in the area of he care quality and safety will be reflected in our abi not only to explain why empirical relationships oc but also to predict conditions under which these r

Barbara A. Mark is the Sarah Frances Russell Distinguished Profe at the University of North Carolina at Chapel Hill, Chapel Hill, NC. Linda C. Hughes is a Research Associate Professor at the Univers North Carolina at Chapel Hill, Chapel Hill, NC. Cheryl Bland Jones is an Associate Professor at the University of N Carolina at Chapel Hill, Chapel Hill, NC. Reprint requests: Dr. Barbara A. Mark, University of North Carolin Chapel Hill, Carrington Hall CB#7460, Chapel Hill, NC 27599-746 E-mail: [email protected].

Nurs Outlook 2004;52:11-6. 0029-6554/$–see front matter © 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.outlook.2003.10.010

J A N U A

tionships are most and least likely to hold true. Bec formulating “good theory” is so difficult, Sutton a Staw3 identified examples of what theory isnot. Theory is not represented by an extensive list of referenc which logical relationships are explained elsewh Models and variables must be accompanied by lo statements that explain not only how variables related but alsowhy they are related. Similarly, theo is not represented by a description of observed rela ships, beta weights, or factor loadings. Data can scribe what relationship exists, but cannot explain the relationship exists. Theory is not a laundry lis variables or a diagram used to depict a struc equation model. Nor is theory rendered in evide based clinical practice guidelines or in so-called “b practices.” Theory results from a process of system and logical reasoning— either deductive or inductiv through which the occurrence or nonoccurrence some phenomenon can be understood and pred Such theory will be useful in building an understand of the practice of nursing as it is enacted in soci constructed organizational settings.4 If organizations exist because they can achieve goals that can n accomplished by individuals alone, then the mere cess of organizing introduces a “systemness”5 to the work of nursing that must be an integral componen any theory that seeks to explain and predict high qu and safe patient care.

In nursing, there are promising beginnings in effort to better understand the relationship betw nurse staffing and patient outcomes. An example i growing body of empirical research relating nu staffing to adverse patient events.6 –13 Yet even a dissemination of these findings begins to influence organization and management of nursing care deli the underlying question ofhow nurse staffing affec adverse events remains unexplained. It is only thro a clear understanding of thewhy and the how that innovative policies and management practices to prove quality and patient safety can be systemati developed, methodically implemented, and rigoro evaluated. In the absence of such understanding introduction of policies to guide nurse staffing are lik to be both haphazard and unsuccessful.

In this article we discuss conceptual and meth

11R Y / F E B R U A R Y N U R S I N G O U T L O O K

Role of Theory in Improving Patient Safety and Quality Health Care Mark et al.

ological issues that are central to the goal of building strong theory that can explain nursing’ s contribution to patient safety and quality care in acute care settings. This discussion is organized around four questions that reflect the intellectual processes through which causal mechanisms can account for the relationships observed among variables and, in so doing, build theory that has both explanatory and predictive power. These questions include: ● What are the major constructs of interest and how

should they be measured within the context of the theory? This places the constructs within a nomolog- ical network,14 a prerequisite to establishing con- struct validity.

● Why do the constructs relate to each other in the way that is proposed, ie, what are the causal mechanisms underlying the hypothesized relationships? Theory must also include explanatory frameworks

that permit identification of variables that mediate observed relationships as well as contextual factors that moderate the magnitude of those relationships: ● What are the critical mediators in the proposed

relationships? ● What are the contextual factors that moderate the

proposed relationships? This article examines these four questions in the

context of research investigating the relationship be- tween nurse staffing and adverse patient events. Our purpose is to demonstrate the problems that arise when theory is absent: First, research findings contradict each other, conceptual and statistical power is reduced, and potential effects are masked; and, second, the develop- ment of a coherent body of knowledge is hindered and the design of interventions to improve quality and patient safety is made difficult.

WHAT ARE THE MAJOR CONSTRUCTS OF INTEREST AND HOW SHOULD THEY BE MEASURED WITHIN THE CONTEXT OF THE THEORY? In studies investigating the relationship of nurse staffing and patient outcomes, the major constructs seem obvi- ous: Nurse staffing and patient outcomes. While iden- tification of the constructs is, in fact, obvious, their conceptualization and measurement is not. For exam- ple, from a theoretical perspective, what does the term “ nurse staffing” mean?

Since theory ought to drive measurement, consider the current use of multiple definitions of nurse staffing: (1) actual number of hours of care delivered by RNs (also called “ nursing hours” ); (2) RN hours as a percent of all nursing care hours; and (3) percent of total staff who are RNs. The last two operational definitions are frequently referred to as “ skill mix,” despite the fact that one measures skill mix in terms of hours and the other measures skill mix in terms of people.7–9 The fact that “ skill mix” is operationally defined differently in

12 V O L U M E 5 2 ● N U M B E R 1 N U R S I N G O U

different studies makes comparison of findings difficult. Other studies6,11,13 have measured nurse staffing as the number or percent of RN full-time equivalents (FTEs). These represent different conceptual approaches to measuring different phenomena—they are not multiple indicators of the same construct. Although they have not previously been discussed as such, one can consider a staffing measure that focuses on hours of care as a construct reflecting nursing care patients actually re- ceive (assuming that productive hours—not paid hours—is measured). In contrast, staffing as measured by FTEs can be thought of as a unit-level phenomenon reflecting the unit’ s capacity to deliver nursing care. However, neither of these measures provides informa- tion about the actual staffing levels needed to deliver safe care; instead, they report only about staffing levels used in delivering nursing care.

The importance of distinguishing among these con- structs is illustrated in the following two examples. First, imagine the nursing unit with nearly an all RN staff. One would expect that patients on that unit would receive the vast majority of their care from an RN. However, poor management, inadequate support ser- vices, and/or poor medication delivery and supply systems might significantly reduce the amount of nurs- ing care actually received by patients, even though staffing— defined as the proportion of staff or staff FTEs who are RNs—is high and presumed to be excellent. Second, use of such different measures may lead to conflicting results that are difficult to reconcile. For example, one recent study9 found that higher skill mix (ie, a greater proportion of care provided by RNs) was associated with a lower incidence of pneumonia. The study also found, however, that more total licensed hours (ie, the total hours of care delivered by registered and licensed practical nurses) were associated with a higher incidence of pneumonia. The discrepancy may be attributed to the inclusion of licensed practical nurses in one measure and not in the other. The current lack of theoretical development within this particular realm of research makes it difficult to logically integrate these findings.

A theory that proposes clear boundaries for the construct of nurse staffing would help to alleviate the dilemma that results from multiple, overlapping, and conceptually non-distinct measures. Such a theory would also make clear the distinction between “ nurse staffing” and “ nursing care”— which are often implic- itly assumed to be one and the same. Many individual and organizational factors affect the actual delivery of nursing care such that merely having more RNs may not necessarily yield either more care delivered by RNs or higher quality of care. Specifically, more registered nurses may provide better nursing care, which may improve quality and patient safety, but only under certain conditions. For example, more RNs—if they lack the requisite intellectual capacity, skill, motivation,

T L O O K

Role of Theory in Improving Patient Safety and Quality Health Care Mark et al.

and a caring attitude—would not likely improve quality of patient care simply by virtue of there being more of them. Also, if organizational conditions are such that the environment is not conducive to professional nurs- ing practice, care may suffer regardless of the number of RNs available to deliver care.

Theory is no less problematic when patient outcomes are considered. A great deal of attention is currently directed at “ nurse-sensitive” patient outcomes, which have been described, in general terms, as outcomes that are sensitive to “ variations” . However, it is not clear how the term “ variations” is being used. Is it variations in the amount of nursing care that is provided? Is it variations in the quality of nursing care that is pro- vided? Is it variations in how nursing work is organized to provide nursing care? Or is it variations in both quality and amount of care? Additionally, it is unclear at what organizational level—the hospital or the nurs- ing unit—we should measure the relationship between nurse staffing and adverse events. Since we know that nurse staffing varies by unit, how might organizational level variables affect the relationship between nurse staffing and adverse events on a nursing unit? Would we expect certain outcomes to be more (or less) sensitive to changes in numbers of RNs in the hospital, or on the unit, than to changes in the number of hours of care provided by RNs?

Another issue to be considered is that understanding the relationship between nurse staffing and different outcomes, indeed different classes of outcomes, may require different theories. Most of the research on nurse staffing, in fact, has examined nurse staffing relative to only one type of patient outcome—adverse patient events. However, different theories might be needed to explain the relationship between nurse staffing and positive patient outcomes, which reflect more than the absence of adverse events. Similarly, different theories might be called upon to explicate the relationship between nurse staffing and the prevention of potential adverse events. New theoretical formulations are begin- ning to address these issues, but undoubtedly will require additional research. An example is Tucker and Edmondson’ s distinction between errors (defined as “ the execution of a task that is either unnecessary or incorrectly carried out and that could have been avoided with appropriate distribution of pre-existing informa- tion” , and problems (defined as “ disruption in a work- er’ s ability to execute a prescribed task because either: something the worker needs is unavailable in the time, location, condition, or quantity desired and, hence, the task cannot be executed as planned; or something is present that should not be, interfering with the desig- nated task” ).15 Thus, gaining a better understanding of the relationship between nurse staffing and patient outcomes, errors and problems, and latent conditions in the practice environment is paramount in demonstrating nursing’ s contribution to quality and safe patient care.

J A N U A

If we hypothesize that such differential effects might exist, then we move to the next of the four questions posed earlier: how and why do the constructs relate to each other in the way that is proposed?

HOW AND WHY ARE THE CONSTRUCTS RELATED TO EACH OTHER? This question addresses the proposed causal relation- ships among the constructs of interest and explanations for the proposed relationships. In other words, this question focuses on the theoretical basis of expecting that changes in nurse staffing will, in fact, result in changes in adverse events. Popular arguments make the case “ more RNs, better outcomes”— but this argument only addresses the issue of what rather than why this relationship holds. That is, why “ more RNs, better outcomes” ? Why are causal linkages between nurse staffing and patient outcomes as hypothesized? Con- sider, for example, the following set of theoretical statements: ● RNs’ communication with physicians is hypothesized

to be more timely, complete and accurate than non-RN staff communication with physicians;

● This “ enhanced” communication between physicians and nurses is hypothesized to result in early recogni- tion and intervention in potentially hazardous patient situations;

● Therefore, nursing units characterized by a greater proportion of RN staff will have fewer adverse events than a nursing unit with a lower level of RN staffing. Given this theoretically derived set of causal state-

ments, the measurement of “ nurse staffing” becomes obvious: “ nurse staffing” is measured by the proportion of RNs to total nursing staff, not the number of hours of care delivered by RNs. An explanation for the use of this nurse staffing measure is that the proportion of RN staff is relevant to communication with physicians and unit-level capacity to deliver nursing care. Conversely, the number of hours of care is not an appropriate measure of nurse staffing because this measure would capture data on the nursing care delivered to patients rather than unit level nursing capacity and operations. These theoretical statements do not, however, clearly address— either conceptually or operationally—the identification of appropriate and theoretically relevant adverse events. Thus, theory should guide the develop- ment of hypothesized relationships and the selection of measures for both independent and dependent variables, nurse staffing and adverse events, respectively.

In the early stages of theory development, research- ers tend to focus on demonstrating the existence of a relationship— broadly conceptualized— between inde- pendent and dependent variables, in this case, the hypothesized relationship between nurse staffing and adverse patient events. This is a common progression of knowledge development in any science, but is particu-

13R Y / F E B R U A R Y N U R S I N G O U T L O O K

Role of Theory in Improving Patient Safety and Quality Health Care Mark et al.

larly prevalent in studies of nurse staffing and adverse events. The wide availability of administrative and other large databases may actually fuel this research strategy: Because data are widely available and acces- sible, researchers ask the questions that can be an- swered of the data. It is in the later stages of theory development where researchers begin to address prob- lems in conceptualization, methodologies and conflict- ing and/or equivocal results.

In the social and organizational sciences, “ clean” relationships between independent and dependent vari- ables are unlikely—real social and organizational life is just too complicated to be modeled so simple. Drawing on this assumption would be a useful approach in the study of relationships such as nurse staffing and patient outcomes where complex social and organizational factors likely impact observed relationships. Thus, the next stage of theory development turns to the identifi- cation of “ mediators” and “ moderators”— other vari- ables that in specific ways affect the relationship be- tween the independent and dependent variables. First, we turn to the identification of variables that play a potential mediating role in the relationship between nurse staffing and adverse events; then we discuss possible moderators of this relationship.

WHAT FACTORS MEDIATE THE RELATIONSHIPS WE OBSERVE? Mediators are variables that explain the mechanism of action that produces the observed relationships between independent and dependent variables.16 The advantage of testing models in which mediators are specified is that it forces the researcher to consider the theoretical basis upon which relationships can be predicted. Medi- ators thus help explain the relationship between nurse staffing and, for example, variables related to quality or patient safety.

An area of inquiry that has significant potential to inform research in the area of nurse staffing, quality care, and patient safety is being investigated in the field of organizational industrial psychology, where re- searchers are attempting to unravel the work processes that contribute to employee emphasis on safety issues and a reduction in industrial accidents. In particular, the

Figure 1. LMX as a mediator of the relationship between nurse staffing and adverse events.

14 V O L U M E 5 2 ● N U M B E R 1 N U R S I N G O U

theory of leader-member exchange (LMX) has been used to explain the motivational process through which employees engage in voluntary behaviors that have “ add-on” value in terms of being beneficial to the organization.17 The basic premise of LMX theory is that managerial actions that demonstrate positive regard for an employee create a desire on the part of the employee to reciprocate through behaviors that are seen as highly valued by the manager.18

The leader-member exchange (LMX) theory is espe- cially relevant to the area of safety and risk mitigation because safety-related behaviors tend to be seen by employees as voluntary and, therefore, as behaviors where “ corners can be cut” in order to complete work assignments. According to Weick19 this tendency re- sults from the fact that, ideally, nothing happens in response to accident prevention behaviors. As a result, it is easy to assume that these behaviors are of lesser importance and can be omitted when time is limited. LMX theory, however, hypothesizes that high quality exchange relationships will result in the performance of safety and risk mitigation behaviors when employees view these behaviors as highly valued by the manger. Hofmann and Morgeson20 tested and confirmed this hypothesis by demonstrating that, in the context of high quality leader-member exchange relationships, open communication about safety issues was associated with a reduction in work-related injuries.

Based on these findings, one could, for example, hypothesize that the relationship between staffing levels and adverse events is mediated by the quality of the exchange relationship between nurse managers and their staff. In other words, better staffing allows the nurse manager to invest more time and energy in the development of meaningful social exchange relation- ships with members of the nursing staff. As a conse- quence, staff nurses who experience a high quality exchange relationship with their manager would be more likely to respond proactively to managerial com- munication about the importance of safety-related be- haviors that contribute to the prevention of adverse patient events.

Figure 1 illustrates the proposed role of quality of LMX as a mediator. Four conditions must be present for mediation to be demonstrated.16 There must first be a statistically significant relationship between the inde- pendent variable (nurse staffing) and the dependent variable (adverse events). Second, there must also be a statistically significant relationship between the inde- pendent variable (nurse staffing) and the mediator (quality of LMX). Third, after controlling for the effects of the independent variable (nurse staffing), there must be a statistically significant relationship between the mediator (LMX) and the dependent variable (adverse events). Finally, complete mediation is demonstrated when, after controlling for the quality of LMX, the relationship between nurse staffing and adverse events

T L O O K

Role of Theory in Improving Patient Safety and Quality Health Care Mark et al.

is zero. The more likely scenario in this example is one of partial mediation, where, after controlling for LMX quality, the relationship between nurse staffing and adverse events is significantly reduced. In other words, part of the relationship between nurse staffing and adverse events is explained by LMX. This example provides an illustration of how identification of poten- tial mediators can increase theoretical richness, and may enhance both the predictive and explanatory power of the underlying theory.

WHAT CONTEXTUAL FACTORS MODERATE THE RELATIONSHIPS WE OBSERVE? While mediators illuminate the causal mechanisms that explain why relationships are observed, moderator vari- ables are contextual factors that define the conditions under which such relationships can be predicted.16 In contrast to mediating variables that explain why inde- pendent and dependent variables are related, moderator variables contribute to theory development by specify- ing factors that affect either the direction or magnitude of the relationship between two variables. Most studies examining the relationships between nurse staffing and patient outcomes acknowledge, at least implicitly, that moderating effects may exist by statistically controlling for some or all of the following variables: Hospital size, urban/rural location, and teaching status. Statistical control affords results that can be interpreted as “ all other things being equal.” Acknowledgement that sta- tistical control is necessary indicates that results might be different in the absence of such controls.

Variables measured at the nursing unit level may function as moderators in the relationship between nurse staffing and adverse events. Figure 2 illustrates such a relationship that was examined in a recently completed study which hypothesized that, because hos- pitals experience differential financial pressures in mar- kets with varying levels of managed care penetration, managed care penetration might moderate the relation- ship between nurse staffing and adverse events.13 Using data from a longitudinal panel of 360 hospitals from 1990 –1995, Mark, Harless and McCue13 found a sta- tistically significant relationship between increasing nurse staffing and decreasing mortality ratio (the ratio of actual to expected deaths) for hospitals located in markets where HMO penetration was greater than 7.5%; no such relationship existed in markets with a lower level of HMO penetration. This demonstration of a moderating effect provides additional information about the conditions under which the staffing-adverse events relationship exists.

Two other studies report better patient outcomes on specialized nursing units where patient admissions are restricted to specific diagnoses.21,22 The authors spec- ulate that their findings might be explained by the expertise that nurses develop over time when they care

J A N U A

for patients with the same or a similar diagnosis. So, it is possible that the relationship between nurse staffing and adverse events may be altered when care is pro- vided on specialized nursing units. Studies to explore the potential moderating effect of a contextual variable like this will require demonstration of a significant statistical interaction effect in which the relationship between nurse staffing and adverse events is condi- tioned upon whether patients received nursing care on a dedicated or general nursing unit.

The preceding sections demonstrate why the identi- fication of possible mediators and moderators of the relationship between nurse staffing and adverse events is so important. First, it seems reasonable that hospitals and nursing units that differ from each other in impor- tant ways will also differ in how they deploy RNs, and that these differences may ultimately have effects on quality and patient safety. Second, understanding these differences from a theoretical perspective is critical to designing interventions to improve quality and patient safety. It is critical to know what interventions work, why, for whom, and under what conditions. Without this kind of theoretical specificity, interventions to improve quality and patient safety risk being so diffuse that detection of an intervention effect would be un- likely and where an intervention effect is evident, the challenge would be to explain which aspects of the intervention were responsible for its effects.

SUMMARY In this article, we focused on the research examining the relationship between nurse staffing and adverse events because this area of research has received a great deal of attention, and it allowed us to address four key questions that are critical in developing theory and in advancing knowledge in the area. We have demon- strated the importance of clearly defining the constructs of interest, explicating hypothesized causal relation- ships among the variables, and identifying and testing

Figure 2. Managed care penetration as a moderator of the relationship between nurse staffing and adverse events.

15R Y / F E B R U A R Y N U R S I N G O U T L O O K

Role of Theory in Improving Patient Safety and Quality Health Care Mark et al.

the existence of mediators and moderators of the relationship between nurse staffing and adverse events.

While better documenting and understanding the relationship between nurse staffing and adverse events is critical to patient safety, narrowly focusing on this relationship ignores the intrinsic “ systemness” 5 or con- text in which nursing care is delivered. For example, West23 suggests that contextual characteristics such as the changing division of labor of health care team members, as well as social barriers to communication may influence knowledge sharing across disciplines, thus affecting the level of risk in the organization. Enlarging the focus of inquiry to take into account this “ systemness” increases the likelihood that research findings on the relationship between nurse staffing and adverse events will be successfully integrated into practice.

Weick24 has argued that people in organizations become so entrenched in pursuit of a particular idea that they either selectively ignore or rationalize evidence that is inconsistent with their perception of the idea—a process he calls “ cultural entrapment.” In our efforts to understand high quality and safe patient care, nursing as a professional group runs the risk of cultural entrapment through continued emphasis on a narrow set of con- structs that adhere to preconceived assumptions about nursing and the way nurses provide care. Our efforts to build theory that is useful for both research and practice will depend on the extent to which we are able to think creatively in terms of identifying constructs that capture the complex interface between the clinical and organi- zational domains within which nurses practice.

REFERENCES 1. Merton R. On Theoretical Sociology. New York, NY: Free

Press; 1967. 2. West E. Management matters: The link between hospital

organization and quality of patient care. Quality in Health Care 2001;10:40-8.

3. Sutton R, Staw B. What theory is not. Admin Sci Quarterly 1995;40:371-84.

4. West E. Organizational sources of safety and danger: Socio- logical contributions to the study of adverse events. Qual Health Care 2000;9:120-26.

5. Berwick DM. Seeking systemness. Healthc Forum J 1992; 25:22-8.

6. Kovner C, Gergen P. Nurse staffing levels and adverse events following surgery in US hospitals. Image: J Nurs Schol 1998;30:315-21.

7. American Nurses Association. Implementing nursing’ s re- port card: A study of RN staffing, length of stay, and patient

16 V O L U M E 5 2 ● N U M B E R 1 N U R S I N G O U

outcomes. Washington, DC: American Nurses Publishing; 1997.

8. Lichtig L, Knauf R, Milholland D. Some impacts of nursing on acute care hospital outcomes. J Nurs Admin 1999;29:25- 33.

9. American Nurses’ Association 2000. Nurse Staffing and Patient Outcomes in the Inpatient Hospital Setting. Wash- ington DC: American Nurses Publishing. 2000.

10. Needleman J, Buerhaus P, Mattke S, Stewart M, Zelevinski K. Nurse-staffing levels and the quality of care in hospitals. New Engl J Med 2002;346:1715-22.

11. Kovner C, Jones C, Zhan C, Gergen P. Nurse staffing and post-surgical adverse events. An analysis of administrative data from a sample of US hospitals, 1990-1996. Health Services Research 2002;37:611-30.

12. Cho S. Nurse staffing and adverse patient outcomes: A systems approach. Nursing Outlook 2001;49:78-85.

13. Mark B, Harless D, McCue M. The influence of managed care penetration on the relationship between nurse staffing and quality of care. Health Econ Submitted manuscript. 2003.

14. Cronbach L, Meehl P. Construct validity in psychological tests. Psych Bull 1955;52:281-302.

15. Tucker AL, Edmondson AC. Why hospitals don’ t learn from failures: Organizational and psychological dynamics that inhibit system change. California Management Review 2003;45:55-72.

16. Baron R, Kenny D. The moderator-mediator variable dis- tinction in social psychological research: Conceptual, stra- tegic and statistical considerations. Journal of Personality and Social Psychology 1986;51:1173-82.

17. Sparrowe R, Liden R. Process and structure in leader- member exchange. Academy of Management Review 1997; 22:522-52.

18. Setton R, Bennet N, Liden R. Social exchange in organizations: Perceived organizational support, leader- member exchange, and employee reciprocity. J Applied Psych 1996;81:219-27.

19. Weick K. Organizational culture as a source of high reliabil- ity. California Management Review 1987;29:112-27.

20. Hofmann D, Morgeson F. Safety-related behavior as a social exchange: The role of perceived organizational support and leader-member exchange. J Applied Psych 1999;84:286-96.

21. Aiken LH, Sloane DM, Lake ET. Satisfaction with inpatient acquired immunodeficiency syndrome care: A national com- parison of dedicated and scattered-bed units. Medical Care 1997;35:948-62.

22. Czaplinski C, Diers D. The effect of staff nursing on length of stay and mortality. Med Care 1998;36:1626-38.

23. West E. Organizational sources of safety and danger: Socio- logical contributions to the study of adverse events. Qual Health Care 2000;9:120-26.

24. Weick KE, Sutcliffe KM. Hospitals as cultures of entrapment: A re-analysis of the Bristol Royal Infirmary. California Management Review 2003;45:73-84.

T L O O K