In response to the growing pressures for cost-containment attendant to the growth of managed care, health care organizations are changing the division of labor between physicians and nurses. Many observers believe corporatized and industrialized health care will substitute nurses and other allied health professionals into physicians' traditional roles. This study tests for evidence of doctor-nurse labor substitution by exploring the impact of hospitals' resource environments and organizational characteristics on the amount of patient counseling, education and case-management performed by doctors and nurses, and how this work is divided between doctors, housestaff and nurses.

Three years of patient surveys (N=50,000) in 45 university hospitals are used to create measures of the amount of patient care performed by nurses and doctors at each hospital, as well as a third variable, the percent of patients reporting they had a physician case-manager. Subtracting reported doctor work from reported nurse work creates a fourth "division of labor" variable. Based on these four measures, some hospitals are found to be more "physician-centered," where patient counseling, education and case-management is less often delegated to nurses and housestaff, and more often performed by attending physicians, while the care at other hospitals is more "nurse-centered," where attendings delegate more patient care to nurses and housestaff.

These labor variables correlate to a variety of environmental and hospital features. For instance, staffing mix has a clear correlation to the dependent variables: the more nurses per bed, the more patients perceive that nurses do, and the less they say doctors do; the more doctors per bed, the more doctors do; the more housestaff per bed, the less doctors are perceived to do.

Next I propose several stages of labor substitution that university hospitals will go through in adapting to the growth of managed care, resulting in a curvilinear relationship between market evolution and division of labor:

(a) First, an initial phase of little managed care, in which the hospital is the academic physicians' workshop, and physicians devote themselves to research and delegate patient care to nurses and housestaff. In this phase, university hospital care is relatively nurse-centered.

(b) Next, an intermediate phase of managed care growth, where physicians and administrators begin to respond to competitive markets by substituting licensed practical nurses (LPNs) for registered nurses (RNs), and encouraging physicians to return to patient case-management and direct patient care. In this phase, the lay-offs of RNs and re-incentivization of academic physicians creates a more physician-centered division of labor.

(c) Finally, a latter phase of managed care growth and administrative rationalization, where administrators begin to substitute RNs for physicians in patient care and case management.

The data support the theorized curvilinear relationship of managed care market evolution and the doctor-nurse division of labor.

The history of American hospitals has been cast in three phases (Perrow, 1965; Rosenberg, 1987; Stevens, 1989).

The growth of managed care has increased pressures for administrative rationalization and control. In particular, managed care has encouraged hospitals to collect data on practice patterns, and institute more efficient patterns of inpatient care, with the goal of reduced lengths of stay. Many of these efforts at practice reform have had the effect of expanding nurses' roles as providers and organizers of clinical care.

This paper examines whether there is evidence that hospitals' responses to the growth of managed care has effected the division of labor between doctors and nurses in the provision and organization of care.

Much attention has been given to HMOs' experiments with expanding roles for paraprofessionals, such as nurse practitioners. Hospitals' experiments with redesigning work have also been much studied. But analysts have rarely examined the effects of cost-containment on the labor substitution of non-specialist R.N.s into physician task turf.

Many hospital administration strategies for cost-containment have consequences for the doctor-nurse division of labor. Quality assurance programs, new staff organizations, and informatics systems may shrink or expand nurses' roles, without this ever having been an explicit goal. Since physician resistance to organizational innovation is often a given, and nurses are quite sensitive to changes in their task turf, administrators make every effort to ignore and obfuscate potential effects of administrative initiatives on the balance of power and responsibilities between these two groups. Administrators often use a mantra of the need for "physician buy-in" and physician leadership in their determined effort to ignore the effects of administrative reforms on physician autonomy.

For fifteen years health planners have attempted to address the shortage of primary care physicians by calling on medical schools to train more of them, and by encouraging the use of nurses, midwives, and other non-physician health professionals in these front-line roles (Graduate Medical Education National Advisory Committee, 1980; Weiner et al., 1987; Safriet, 1992). But in the 1980s, instead of changing medical school priorities or challenging physician authority by substituting paraprofessionals into physician roles, American health care filled primary care roles by "substituting down" physician specialists into primary care roles (Weiner, Steinwachs and Williamson, 1986; Steinwachs, et al., 1986).

Nonetheless, the growing cost crisis and the growing pressure for health care reform have renewed debate about the appropriate divisions of labor in medicine. Social workers and nurses are assuming roles in patient case-management, including assessment, coordination, advocacy, referral, teaching, home visiting, crisis intervention and medication monitoring (Etheredge, 1989). Backed by diagnostic and treatment protocols, nurses are being used as primary-care providing gatekeepers, and are making a growing number of diagnostic and treatment decisions.

As hospitals shift from being doctor's cooperatives in low competition markets to administratively dominated firms in competitive markets there are increasing pressures to substitute nurses and other paraprofessionals for physicians. Research on the quality of care shows that expanding nurses' roles in these ways can benefit clinical outcomes, patient satisfaction and cost-effectiveness. The largest review of nursing substitution, the Office of Technology Assessment's (1986) report Nurse Practitioners, Physician Assistants and Certified Nurse-Midwives: A Policy Analysis. reviewed twenty-four major studies comparing paraprofessionals' quality of care to physicians'. OTA found that paraprofessional performance was equivalent to physicians' performance in ten studies, superior to physicians in twelve studies, and inferior in only two studies. Numerous studies have subsequently confirmed that nurses, nurse specialists and nurse practitioners can equal or excel physicians at communication (Taylor, Pickens, and Geden, 1989) and improved clinical outcomes (Kane, et al., 1989; Hall, et al., 1990).

Physicians have traditionally delegated many tasks to nurses, but as clearly subordinate workers (Abbott, 1988). Physicians do delegate when they can extend their income-generating capacity by carrying a larger patient panel, such as in physician-owned ambulatory clinics. And physicians are willing to allow extensive delegation of tasks which they no longer wish to perform, as long as the practitioners are under the formal guidance of physicians. Ferraro and Southerland (1989), for instance, found that physicians have the least resistance to expanding the use of "physician assistants" in poor urban areas. On the other hand, Ferraro and Southerland also found that, despite the growing shortage of pediatricians and obstetricians, physicians were most resistant to expanding use of physician assistants in these areas. Obstetric and pediatric work is largely routine, usually not requiring the presence of a physician, and strong pressures from the women's movement, midwives and pediatric nurse practitioners are pressing for independent, as opposed to subordinate, practice. In other words, physicians are willing to delegate work as long as the delegation is in their interest; when the delegation takes on the character of "loss of turf" they begin to resist.

The declining autonomy of physicians, changes in the status of women, and the calls by health economists for changes in practice regulation, have made physicians much more sensitive to protecting the traditional division of labor in medicine (Levitt, 1977). Nursing organizations and researchers are actively challenging the constrained practice autonomy of nursing, and urging expanded, independent roles (Safriet, 1992).

As nurses have pursued their professionalization project and their protest against their subordinate role in the sexual division of labor, they have been led in contradictory directions. On the one hand nursing has aspired to professionalization and clinical specialization, involving the expansion of their task turf. On the other hand, nurses could organize unions and seek control over work along the model of the industrial workers. The union model reifies current job responsibilities and task turf and militates against the professionalization project. The "real medical professional" is always on call, and will do whatever their patient needs, while collective bargaining agreements with hospital administrators specify exactly what can be expected of nurses, in terms of over-time and task turfs. These two paths have led researchers to distinguish between white-collar and blue-collar nurse cultures (Smith, 1985; Coburn, 1988).

Nursing unionization has been portrayed by some as a consequence of the proletarianization of nursing work. Writers have argued that nursing has been proletarianized as nurses have lost autonomy in the work place (Wagner, 1980) and nursing work became increasingly specialized and fragmented (Bellaby and Oribabor, 1977). Gray (1989) points to many problems with this nursing proletarianization thesis as an explanation for nursing unionization. Rather than moving from skilled labor to unskilled labor, as the Braverman (1974) thesis implies, nursing has moved from relatively low levels of technical skill to higher levels, and from little formal education to years of formal education (Melosh, 1982). Rather than losing control over their work, nurses have gone from having no protection from arbitrary dismissal or work regulations, to having many protections and avenues of appeal (Melosh, 1982). Nursing specialization has not led to less autonomy in their work but greater autonomy (Gray, 1989). Rather, Gray (1989) suggests that the women's movement and gender militancy are the chief motivators of labor militancy.

While nursing work has not been proletarianized, the increasing differentiation and specialization of nursing work, and efforts at work redesign in hospitals, have pushed nurses into further contradiction between unionization and professionalization. As hospitals seek to replace RNs with licensed practical nurses (LPNs) and other nurse extenders, professionalism-oriented nurses have embraced the changes and proposed "differentiated practice models" specifying different roles for nurse specialists, staff RNs, and nurse extenders. They welcome these changes as a way of re-assigning "scut" work from B.A. and M.A. prepared nurses to those with six-week to six-month training. These work redesign efforts are often accompanied by role expansion for higher-trained nurses, such as nursing case-management and the cross-training of nurses to do tests and procedures previously done by ancillary workers.

On the other hand, nursing unions have militantly opposed these work redesign efforts as ways of cutting RNs, reducing nursing's autonomy, increasing RN work loads and task turfs without proper compensation, and endangering the quality of patient care by over-using "unskilled workers" (Sherer, 1994).

The nurse extender proponents, mostly advance practice nurses, recognize that labor substitution of non-RNs for RNs is contentious, and have done research to demonstrate that nurse extenders will actually improve RN work satisfaction by delegating "scut work" to the LPNs and aides. A study of implementation of nurse extenders, substituting for RNs through attrition, in a Boston university hospital found that the RNs were neither more nor less satisfied with their work, while the cost per patient day decreased (Garfink, Kirby, Bachman and Starck, 1991).

But the key to the differentiated practice model for nursing managers is that it frees B.A. and graduate nurses to take on more of the higher skill patient tasks, and gives them greater autonomy over their work, without threatening their control over the nursing process.

In other words, the differentiated practice model is the recognition that the further professionalization of RNs and graduate trained nurses requires shifting back to nurse aides the less complex tasks that RNs assumed in the 1980s as part of primary nursing. The hope of graduate trained nurses and nurse managers is that this task shift will enable them to expand their task turf with more professional responsibilities, such as patient education and case-management. Blue-collar staff RNs, however, are much more divided on this approach, believing that shifting tasks to nurse aides, and giving up the primary nursing model, will costs them their jobs. As Robert Brannon (1994) concludes in his study of work re-organization in nursing:

These trends suggest that managed care pressures on hospitals may speed two parallel task-turf shifts: the substitution of nursing assistants for registered nurses in lower skilled tasks, and the substitution of nurses for physicians in tasks at the intersection of their domains.

The physician's model of the normative division of labor assumes that nurses have some role in patient education and counseling, but that the more general case-management tasks, such as the coordination of patients through admission, testing, consultation with specialists, and discharge, remain the responsibility of physicians. Similarly, while physicians may be willing to cede to nurses responsibility for making diagnostic and treatment decisions on minor illnesses, the diagnosis and treatment plan of the principal illness should remain with the physician. In other words, the physician's mental map of task turf involves a series of concentric circles which places patient counseling and education in an outer circle, and case-management closer to the center of the physician's core responsibilities.

Labor substitution, consequently, probably comes in stages, progressing from the patient history-taking and education tasks which physicians have already often delegated to nurses in subordinate jurisdiction, to case-management, and finally to the tasks which are the core of the professional claim to autonomy, i.e. diagnostic and treatment decisions. In this study I will be examining task delegation and substitution in the first two realms, patient counseling and case-management.

Figure 1: Overlapping Task Turfs Between Doctors and Nurses

The hospitals in this study are all academic medical centers. Academic medical centers' base of support has shifted since the 60s, away from federal research monies and philanthropic donations, to income from patient services. Medical service income now accounts for half of the revenues of American medical schools, up from 5% in 1966.

University hospital administrators struggling to compete in the 1980s market became focused on the question of whether increased federal and state support was the only way for academic medical centers (AMCs) to survive (Schramm, 1983), given their cost and efficiency burdens, or whether it was possible to become viable by pursuing greater independence from their medical schools. These concerns led to the ambitious work of the Consortium for the Study of University Hospitals (CSUH) in the early 80s. The CSUH studied data on 52 university hospitals, and conducted field studies in 15 hospitals (Choi, Allison and Munson, 1986).

The CSUH group applied a resource contingency theory to the hospitals, proposing that their environments consisted of two principal variables: state proactivity and market competitiveness. They assumed that the relevant unit of environmental analysis was the state, and assigned each state a scalar measure of state proactivity and market competitiveness. State proactivity was measured by: degree of political progressivism, average number of months it took the state legislature to pass a bill, and degree of two-party competition. Market competitiveness was measured by: per capita income, urbanization, and number of M.D. specialists per capita. Cluster analysis revealed three groups of states: highly proactive and highly competitive states, moderately proactive and competitive states, and low proactive and competitive states.

Despite this problem of collinearity in the measures they proceeded to attempt to measure the independent effects of state proactivity and market competitiveness on hospital viability and efficiency. Viability was measured by net revenues over total operating expenses. Efficiency was measured by a scale of three variables, occupancy, inventory per occupied bed, and days of revenues in receivables. CSUH also attempted to study structural inertia and other structural features that might mediate or independently effect hospital viability and efficiency.

The CSUH concluded that university hospitals faced two paths, partly determined by their environments and their responses to them. Competitive markets were associated with hospitals becoming more efficient, but less financially viable, as measured by debt. Proactive states, on the other hand, tended not only to provide more generous support for AMCs, but also to provide more generous Medicaid payments and state subsidies, relieving AMCs of some of the burdens the medically indigent.

But while state support was associated with greater financial viability, it also was associated with lower efficiency, and with state expectations that AMCs would assume greater indigent responsibility. In other words, AMCs in proactive states were more dependent on their institutional environment, protected from market competition, and less likely to pursue market-driven efficiency initiatives. Private university hospitals, for instance, needed to debt-finance their capital investment, while public universities could rely on public investment. Since few hospitals saw the pursuit of increasing state dependence and largess as attractive, many hospital began to explore independence from the state and university in order to pursue more aggressive market strategies.

In the wake of the CSUH project, the University Hospital Consortium (UHC), was founded in 1986 to institutionalize collective efforts. The UHC began as a joint-purchasing cooperative, but UHC services have now expanded to include research and strategic planning. Today the UHC is broadly charged by its members to develop strategies to respond to the growth of managed care, medical indigency, and federal cut-backs in funding of research and education.

In the early 1990s, it became obvious to health care economists that the variables used by the CSUH to measure market competition, such as population and physician density by state, were inadequate to describe the new competitive pressures on hospitals. In particular, the rapid growth of managed care was placing university hospitals at a severe disadvantage. The UHC acknowledged these changes by developing a unilinear model of local market stages, and evaluating the strategies and challenges faced by their members at each stage (Iglehart, 1993). The typology saw an inevitable progression towards the dominance of several managed care firms in each local market, with integrated systems, and AMCs at their hub.

Table 1: UHC's Model of Local Market Stage Evolution

This study will use local market stage and managed care penetration at each institution as the two principle variables of market competitiveness, rather than the service or population density measures used by CSUH and others.

Nurses can be expected to assume more physician tasks both when physicians are powerful and delegating, and when physicians are weak and organizations begin assigning physicians' traditional tasks to nurses. In the former situation nurses are under subordinate jurisdiction, and under the latter nurses experience greater autonomy (at least from physicians). But what happens in between the traditional physician-dominated/market-insulated setting and emergent organization-dominated/market-driven settings?

In hospitals with an intermediate balance of power, administrators may not yet have the power to substitute nurses for physicians, and instead seek efficiency within spheres on hospital work that they have under their direct control, i.e. staffing levels of nurses and other ancillary professionals. Consequently, may hospitals in competitive markets, or in anticipation of competitive markets, are replacing RNs with nurse assistants, cutting the overall numbers of nurses, and "cross-training" workers in order to have fewer FTEs do the same numbers of tasks.

Hospitals in the late 80s and early 90s have been shifting away from all registered nurse (RN) staffs to increased use of nurse assistants and licensed practical nurses (LPNs), on the assumption that an all RN staff and the primary and case-management nursing models that require it, are inefficient. Yet there has been little evidence that all-RN staffs are less efficient. For instance, Donovan and Lewis (1987) compared Chicago's Rush-Presbyterian hospital's nursing productivity between 1965 and 1985, controlling for increasing patient acuity. During this period Rush's hospital-cum-university built a leading nursing college, and the staff went from 36% to 94% RNs, yet Rush's nursing personnel doubled their productivity, halving the number of nursing hours devoted per day to patients at each of five levels of acuity. The average census increased, the total number of nursing personnel declined, and the percentage of budget consumed by nursing declined. They conclude that all RN staff models had not contributed to rising hospital costs. In fact, cutting RNs and replacing them with LPNs and nurse aides may reduce overall efficiency and quality since it impairs the ability of the remaining RNs to fill the gaps in inpatient case-management and counseling.

Hospital and health plan administrators in competitive markets do begin to think about the efficiency and quality dilemmas posed by physician practice patterns, however, though they do not yet feel sufficiently empowered to challenge them directly by shifting tasks to nurses. Rather than changing task turf, administrators begin initiatives such as patient satisfaction research and clinical pathways development, which attempt to return physicians to the bed-side and to their traditional case-management responsibilities. This is especially true at academic medical centers, where it is widely understood that patients are poorly case-managed and poorly counseled due to the lack of pre-existing doctor-patient relationships and the rotation of housestaff and attendings.

Even if the administrators are successful in incentivizing physicians to re-assume patient care responsibilities they had turfed out, such as discharge planning, they probably don't do these tasks as well as the nurse in-situ. In other words, if the RN is administering LPNs, she doesn't have as much time to talk to the patient, the LPN can't do it very competently, and the physician won't be able to spend much time doing it.

These reflections lead to an hypothesis of a curvilinear relationship between market evolution and doctor-nurse task turf shifts. I refrain from making an explicit hypothesis about the direction of the relationship to be found in the data, however, since the general curvilinear theory I have elaborated so far is not tied to any particular kind or amount of managed care growth and administrative rationalization. In other words, a negative, positive or curvilinear relationship could all be consistent with the hypothesis, and only a positive asymptotic curvilinear relationship or no relationship at all would be inconsistent with the hypothesis. Consequently I will propose a simpler hypothesis:

Hypothesis 1: The doctor-nurse division of labor and reports of physician case-management will vary with market evolution.

Similarly, the growth of case-management under managed care is expected to increase patient reports of case-management across market stages, until eventually nurses are substituted for physicians. But the survey instrument used in this study includes questions about only physicians' case-management, which is expected to be encouraged by managed care growth, while it is depressed by labor substitution of nurses for physicians. Consequently no explicit hypothesis is made as to the effect of market evolution on physician case-management.

Next I will examine whether the staffing levels of nurses and housestaff are correlated with the patients' reports of their contributions to direct patient care. More nurses per bed are expected to increase nurse-centeredness, while more houseofficers per bed are expected to encourage attendings and nurses to turf patient care to the housestaff. On the one hand, housestaff-heavy hospitals are thought to have less coordination of care, decreasing patients' sense that there is physician case-management. On the other hand, housestaff are physicians, and if patients perceive them as such, housestaff intensity per bed will be correlated with greater reported physician work, and a more physician-centered division of labor.

A second question is whether housestaff are labor complements or labor substitutes for nurses. As labor substitutes, there should be a negative correlation between housestaffing and RN staffing, and, as labor complements, a positive correlation. The correlation between the housestaff and RNs per bed is, in fact, r = .50 (N = 102, sig < .001) supporting a labor complement interpretation. If housestaff are labor complements with nurses, there are stronger reasons to believe that RNs and housestaff together will assume attending task turf, and that their staffing ratios will both be related to a more nurse-centered division of labor, and lower reports of physician case-management.

Hypothesis 2: More housestaff/bed will be correlated with a more nurse-centered division of labor and lower reports of physician case-management.

Hypothesis 3: More nurses/bed, and in particular more RNs/bed, will be associated with a more nurse-centered division of labor and lower reports of physician case-management.

In 1991, the University Hospital Consortium (UHC) proposed that the UHC's 60 members field a common patient satisfaction instrument to their patients. While the survey instrument was being constructed, I consulted for the University of Chicago Hospitals on the project, and expressed an interest in including questions that could be used in my dissertation research. The instrument lent itself to the question of doctor-nurse division of labor, since it asked many exactly parallel questions about whether doctors and/or nurses had done X for the patient. The final version of the questionnaire included ten parallel questions which asked "Did your doctors/nurses do X?" and then allowed for separate answers for doctors and nurses. In addition, I was permitted to insert questions on patients' perception of the degree of intra-staff coordination, whether they were attended by their "regular doctor," and how long they had seen their "regular doctor."

Only 19 of the UHC's 60 members agreed to participate in the first round. Each hospital selected a sample of between 500 and 1500 adult medical or surgical patients, who had been discharged in December 1991 or January of 1992. Approximately 29,000 surveys were mailed to this sample in January of 1992. The response rate was roughly 31% yielding a return of 9,018 responses.

The comparative data provided by this first effort led the UHC to repeat the survey annually, and to expand the survey to ambulatory and others services. The 1993 survey involved 30 hospitals, 16 of whom had participated in the previous survey, and 14 new participants. (Three hospitals from the 1992 sample declined to participate.) Again the sample was drawn from recently discharged adult patients, though some hospitals included obstetrical/gynecology patients in their samples. Roughly 57,000 surveys were mailed out in February and March of 1993, again with a 31% return rate, yielding 17,679 responses. Finally, a third year of the survey was fielded and collected in early 1994. This survey included 43 hospitals, a 31% response rate, and a final sample N of 23,000.

None of the hospitals that participated in the survey had fewer than 200 respondents in any year, and some received more than a 1000 responses in one year. The response rates varied considerably from hospital to hospital, from a low of 19% to a high of 50%. In order to explore the possible response biases introduced by poor response rates, 11 of the hospitals provided data in 1993 to allow comparisons between the demographics of their survey samples and the hospitals' actual patient demographics. This test revealed that there was significant under-representation of younger patients, men, and patients on Medicaid, Medicare and the uninsured. For all hospitals, the higher the percent of patients at each hospital who were on Medicaid or uninsured was also significantly associated with the response rate (r = -.47, sig < .01). Taking this into account, the reports of indigent/Medicaid patients will in some cases be examined separately.

These response rates, and attendant concerns of response bias, are quite common in mailed market survey research, and little methodological work has been done to determine the size of the potential bias. The usual concern with response bias is that those who are at the extremes of the issue, in this case consumer satisfaction, will be those who respond, which could potentially bias the following measures in one direction or another.

Four question were used from the 1992 and 1993 surveys to construct the principal "labor variables" used in this study.

Table 2 : 1992 and 1993 Questions Used to Construct Labor Variables

These four questions are summed into a "work performed by doctors" scale and a "work performed by nurses" scale. Each set was highly intercorrelated (alpha = .82, .83 respectively). To facilitate interpretation of the measures, each question was made equal to 25 points for a potential total of 100 points on the summed measure (No/Never = 0, Rarely = 6.25, Sometimes = 12.5, Usually = 18.75, Yes/Always = 25). If one of the four questions was not answered, the other three responses were rescaled to substitute for the missing response; respondents who answered two or fewer of the questions were given missing values.

In the 1994 the question concerning doctors and nurses educating patients about medications was changed to "hospital staff." The response categories of all the yes-no questions were also changed to "yes, completely" - "yes, somewhat" - "no, not at all". Consequently, two new measures were calculated for the 1994 data, based on factor and reliability analysis, using the following four questions:

Table 3 : 1994 Questions Used to Construct the Dependent Labor Variables

Including the questions about satisfaction with the knowledge and skills of the doctor and nurses introduces a somewhat less obvious dimension, potentially measuring the outcomes of the doctor and nurse work than of the work itself, but the inter-correlation of the questions suggested that this would be as good a proxy for "amount of counseling and education provided" as the other questions; the reliability for the doctors and nurses' scales were both .69 . Again, each question was equated to a potential 25 points, making a 0-100 scale for doctors and for nurses, and respondents with one missing response were rescaled to estimate their score out of 100 possible points.

Figure 2 shows the correlation of hospitals' aggregate scores for doctor and nurse work.

Figure 2 : Correlations between Measures for Doctor Work and Nurse Work

The division of labor variable is the aggregate activity (0-100) performed by nurses subtracted from that performed by doctors (0-100). Positive division of labor scores, found in hospitals in the upper-left part of Figure 2, were places where as much counseling and education work was reported for the nurses as for the doctors; these hospitals are referred to as "nurse-centered." Negative division of labor scores, found in the lower right part of the distribution, indicate that more work was reported performed by the doctors than by the nurses; these hospitals are referred to as "doctor-centered."

If the same amount of work is being done in each institution and simply being divided between doctors and nurses there should be a negative correlation, rather than the positive correlation found above, between the "work performed by doctors" scale and the "work performed by nurses" scale. In other words, if patients at a hospital reported that the doctors performed at a high level they also reported that the nurses performed at a high level. This finding suggests that some independent variables will be associated with greater or lesser reports of both doctor and nurse work, and thus have an insignificant effect on the division of labor, while other variables may be associated with one or the other component measure, and thus also with the division of labor. Throughout this study I will therefore examine the component measures alongside the division of labor measure.

A second implication of the positive correlation between doctor and nurse work is that it supports the Shortellian work harmony thesis over the work conflict thesis, at least at the current stage of health care organizational evolution. The work harmony thesis suggests that in well-functioning work units, with good intra-provider communication and well-defined roles, work will be performed at a higher level by all providers. Again, this does not mean that there isn't variation in the division of labor between the providers, but that a categorization of units or hospitals must have a dimension of overall function orthogonal to the division of labor.

It is also interesting to note that the aggregate nurse work scores vary twice as much as the physician work scores. There are three reasons why this might be so. In the first place, in addition to reporting the amount of "work" each provider did in each category, these scales are also picking up a more general assessment of the satisfactoriness of the provider. Patients generally evaluate their providers with two broad categories, "technical competency" and "interpersonal competency" (Hall and Dornan, 1988; Matthews and Feinstein, 1989); patients are generally poor at estimating the former and good at judging the latter. The difference in the spread of evaluations between doctors and nurses may thus arise when patients evaluate physicians with vague assessments of technical competencies, while applying more discriminating standards of interpersonal competencies to nurses. In other words, patients may see all physicians, especially academic physicians, as doing pretty much what doctors are supposed to do, which is act to knowledgeable, while they perceive greater variances among nurses doing what they should do, which is act like they care.

A second explanation relates to the intra-professional variation in the two groups. While physicians range in education from new housestaff officers to senior specialist attendings, this variation is less stark to the patient than the variation in nursing staff from nursing aides, with minimal training, to clinical specialists with advanced graduate training. As I show later, composition of the nursing staff varies dramatically from hospital to hospital and is correlated to the variation in patient assessments of nursing work. While the housestaff composition also varies, and is also correlated to patients' reports of work as I will show, the smaller housestaff variation probably does not produce as much subjective variation in provider quality as the variations in nursing staff.

Finally, physicians are drawn from a national labor pool while nurses are drawn from a local labor pool. Thus it is more likely that hospital-to-hospital variation will reflect area-specific characteristics in the nursing staff, while the physician staff will reflect more national homogeneity. For instance, a Southern AMC may have Southern nurses with a more positive orientation towards service professions than Northern nurses, while the AMC's physicians come from all across the U.S. This would also produce greater variance among nurses than among physicians.

Whatever the cause of this greater variation among nurses, the result is that the division of labor measure is more reflective of the variation in nurse work. To correct for this, I have constructed a standardized division of labor measure, for each year's measures, using the formula:

While the unstandardized scale was strongly correlated with the nurse work scale and not correlated with the physician work scale, the standardized measure is equally correlated to its two components. Throughout most of the rest of this thesis only the standardized division of labor measure will be used, and the two ends of the division of labor scale will be referred to as "nurse-centered" (the positive end) and "physician-centered" (the negative end).

Patients' answers to the following question were used as a measure of the prevalence of physician case-management, and this question was the same across the three years of surveys. ( I attempted to have this question worded as "Who was in charge of your care? A doctor, A nurse, No one" but the survey committee did not see this as a relevant question.):

"Was there one particular doctor in charge of your care in the hospital?"

Yes 1

No 2

Not Sure 3

Physician case-management is distinct from patient education and counseling, and some form of patient case-management could occur of which the patient was not aware. But given that this is a measure of case-management as perceived by the patient, and thus would require some patient interaction, reported physician case-management is correlated to patient counseling, as shown in Figure 3.

Figure 3 : Correlations of Percent of Patients Reporting Physician Case-Management and the Division of Labor at Each Hospital

* Signif. LE .10 ** Signif. LE .05 *** Signif. LE .01

Table 4 below illustrates that for hospitals, as for patients, reports of physician case-management are correlated with a more physician-centered inpatient experience, and vice versa. Interestingly, however, while the relationship of physician case-management to the division of labor variable is consistent across the three years, this relationship is mediated through nurses' education and counseling work in 1992, while in 1993 and 1994 it is mediated through physicians' education and counseling work.

Table Four: Percent of Patients Reporting Physician Case-Managers

1992 N = 19, 1993 N = 30, 1994 N = 43

The hospitals' aggregate measures are highly inter-correlated across the three years of surveys, which adds some confidence in the reliability and validity of these measures.

I use two measures of managed care growth in this study. The first is the 1992-1993 local market stage of the hospital, using the model developed by the University Hospital Consortium in consultation with its members, summarized in Table 1 above. All of the 113 principal Council on Teaching Hospitals (COTH) hospitals were categorizable into one of the market stages.

The principal hypothesis to be tested in this chapter is that managed care market stage will be related to variation in the four principal dependent labor measures. Figure 4 shows the variation in hospitals' aggregate reports of doctor and nurse labor across the four market stages. The significance of the variation across the market stages is measured by the ANOVA technique, and significant variations are marked with asterisks.

Figure 4 : 1992-1994 Doctor and Nurse Work Variables by Market Stage

Market Stage

1992 N = (5) (7) (5) (2)

1993 N = (9) (13) (6) (2)

1994 N = (13) (20) (8) (2)

ANOVA = * Sig2.10 ** Sig2 .05 *** Sig2 .01 (2-tailed)

Figure 4 demonstrates that reports of physician work generally decline as managed care grows, while reports of nurse work are curvilinear with managed care. Figure 5 shows that the division of labor measure (standardized) also varies curvilinearly with managed care growth.

Figure 5 : Division of Labor (1992-1994) by Market Stage

1992 1993 1994

N = (5) (7) (5) (2) (9) (13) (6) (2) (13) (20) (8) (2)

Market Stage Market Stage Market Stage

ANOVA Not Sig. Sig. 2 .05 Not Sig.

Ý The box plots demarcate one standard deviation around the mean, which is indicated by the center line, and the vertical lines indicate the complete range of the distribution.

The relationship of managed care market share to "work for nurses" and "division of labor" is curvilinear, and negative and linear for "work for doctors." "Physician case-management" is a more mixed picture, but it is clear that reports of physician case-management increase from the hospitals in market stages 1 to 3, as shown in Figure 6.

Figure 6 : Reported Rates of Physician Case-Management by Market Stage

92 N = (5) (7) (5) (2)

93 N = (9) (13) (6) (2)

94 N = (13) (20) (8) (2)

* Sig2.10 ** Sig2 .05 *** Sig2 .01 (2-tailed)

In sum, the data support principal hypothesis: managed care penetration does appear to have a strong relationship to the reported amount and division of work in hospitals in the different markets. While the work reported for doctors declines significantly across the four stages, there is a significant curvilinear trend in the work reported for nurses and in the division of labor. Paradoxically, while physician work declines, reports of physician case-management increase with managed care growth.

If we know that nurses save lives and save money in hospitals, why are nurses being replaced? Health care in now a business. Business managers are controlling health care decisions from the top down...The short-term savings seen in replacing RNs is business' short-sighted means of securing funds for new buildings and technology (the real cost drivers in health care) to gain "competitive advantage in the health care marketplace. (Gleeson, 1993).

Staffing ratios are assumed in this study to be a principal intermediate variable between managed care growth and changes in the division of labor. Hospital administrators are adapting to managed care by cutting RNs and ancillary personnel, re-examining the utility of large housestaffs, and attempting to build larger staffs or feeder networks of primary care physicians. At some point in the organizational evolution of health care, I also expect them to engage in hospital management innovations which increase RN staffing.

Three staffing measures are tested below: houseofficers, RNs, and all nurses per hospital bed, as reported in the 1992 COTH survey. Tables 5 and 6 illustrate that staffing ratios differ significantly as managed care increases. (The Ns vary in the comparisons, since not all COTH hospitals provided these staffing data, and some provided data on one kind of staff and not on others.)

Table 5 : Analysis of Variance of Staffing Ratios by Market Stage

Table 6 : Correlations of Staffing Ratios to Percent of Market in HMOs

The data show both significant curvilinear and linear trends in the staffing ratios across market stages, as illustrated in Figures 7 and 8.

Figure 7 : RNs and Nurses per Bed by Market Stage Ý

Market Stage

Nurses/Bed N = (28) (50) (24) (6)

RNs/Bed N = (26) (47) (23) (6)

Figure 8 : Housestaff per Bed by Market StageÝ

N = (28) (51) (25) (6)

Ý The box plots demarcate one standard deviation around the mean, which is indicated by the center line, and the vertical lines indicate the complete range of the distribution.

In short, there is overall a positive correlation of nurse, RN and housestaff staffing with managed care, after an initial decline from market stages 1 to 2.

The amount of managed care is related to AMC's staffing ratios and the reports of labor by AMC's patients. Table 7 shows that the AMCs' staffing ratios are in turn related to reports of labor, suggesting that staffing decisions are an intermediate variable between managed care growth and observed labor performance.

Table 7 : Correlations of Staffing Ratios with Labor Variables

* Sig2.10 ** Sig2 .05 *** Sig2 .01 (2-tailed)

The strongest pattern of relations are found for the 1993 survey data, but the overall patterns are generally maintained across the years:

1. More RNs and nurses per bed are related to a more nurse-centered division of labor, and less physician case-management. RN staffing, however, is not a stronger predictor of patient labor reports than overall nurse staffing, however, suggesting that patients are not distinguishing RNs from licensed practical nurses.

2. More housestaff per bed are related to less reported work for both nurses and physicians. Though housestaff intensity is generally related to a more physician-centered division of labor, it is also related to less reported physician case-management.

The relationship between nurses per bed, division of labor and physician case-management are shown in Figure 9 below. The hospital with the highest nurse-centeredness also has the lowest reported physician case-management and the highest nurse-to-bed ratio, while the most physician-centered hospital has one of the lowest nurse-to-bed ratios and the highest percent of physician case-management.

Figure 9 : Division of Labor by Physician Case-Management and Nurses/Bed

It may be that patients do not fully distinguish housestaff from physicians, and that patient perceptions of physician case-management or work done by physicians partly reflects work done by housestaff. Of all 50,000+ respondents, 73% reported being seen by a resident during their stay, 10% said they had not see one, and 17% didn't know or didn't answer. The likelihood of patients reporting that they had been seen by a resident varied considerably between hospitals, from a low of 44% to a high of 83%. As one would expect, the number of housestaff per bed is correlated very highly (N = 43, r =.56, p<.001) with the likelihood that patients report having seen an intern or resident.

Since the number of housestaff per bed are correlated with the number of RNs and nurses per bed at r = .50 (N = 102, sig < .001), it is important to try to disentangle their independent effects if possible. Table 8 below shows OLS regression models fitting housestaff and nurses per bed on the labor variables of the larger 1993 and 1994 surveys. (Models including RNs per bed had very similar results.)

Table 8 : Regression of Housestaff- and Nurse Staffing on Labor Reports

* Sig2.10 ** Sig2 .05 *** Sig2 .01 (2-tailed)

While only one of the variables remains significant in each model, both housestaffing and nurse staffing ratios are significantly related to labor reports in some of the models, and their relationships are relatively consistent. Nurse staffing is related to more reported nurse work, less physician case-management, and a more nurse-centered division of labor. More housestaff per bed are related to less reported nurse work and physician work, and consequently their effect on the doctor-nurse division of labor is not significant. While housestaffing is not significantly related to patients' reports of physician case-management, all four models show a negative relationship. In sum, the relationships generally confirm the "hydraulic" model of staffing as a determinant of the doctor-nurse division of labor proposed in Hypotheses 2 and 3. The relationships also support the theory on which the housestaff hypothesis is based, that housestaff are labor complements for nursing in academic medical centers, and that housestaff and nurses work together to assume patient care tasks from attending physicians.

Staffing ratios are also significantly related to managed care penetration, in parallel to managed care's relationship to the dependent labor variables. Consequently, the data support the general model that staffing ratios are an intermediate variable between managed care and the patient's perception of the inpatient doctor-nurse division of labor.

Just a few years ago, a widespread nursing shortage meant that nurses were able to pick and choose among multiple job offers. Today, nurses find they're competing for choice jobs as they enter a tightening job market. And when they do get jobs, they're likely to find that hospitals are still understaffed...

Waldman, the spokesman for the New York State Nurses Association, speculates that hospitals are downsizing because they're not sure what to expect from healthcare reform. "It's reform panic. Hospitals are looking into their crystal ball and, fearing the worst, they're slashing positions." But as soon as the health-reform picture clears, those changes will be back, he predicted. (Anonymous, 1994: 6)

Health system reform and market forces promise to increase the role of nonphysician providers, especially in primary care...More patients are covered under capitation, which forces providers to focus on cutting costs. "Groups are discovering that it doesn't make sense to hire a $50-per-hour person to do the work of a $20-per-hour person," said Michael J. Parshall, a consultant. (Borzo, 1994: 5)

The nursing vacancy rate has fallen from 13% in 1989, to 8.1% in 1991, to 5.3% in 1992. After more than a decade of concern about the nursing shortage, health care journalism is now replete with articles about hospitals' declining demand for RNs. Nursing unions are arguing vociferously that nursing layoffs are harming the quality of patient care by substituting LPNs for RNs (Sullivan and Brown, 1989; Voelker, 1993). Yet this is happening at the same time that managed care firms are desperately searching for primary care providers, and nursing organizations are making great headway with the argument that nurses should be used in these roles (Hennenberger, 1994).

The analysis of this chapter suggests that both things may be happening at the same time, but in different parts of the country. The data generally support the hypothesis that the division of labor between doctors and nurses varies across market stages. Nurse work and nurse-centeredness vary curvilinearly across market stages, as do the RN and nurse-to-bed ratios to which they are related. In other words, the early stages of managed care growth appear to encourage hospital administrators to cut nursing, while nurses appear to get rehired in latter stages.

Like nurse staffing, housestaff staffing is negatively related to reports of doctor work and physician case-management. This supports the observation that counseling and continuity of care suffer in housestaff-heavy hospitals, and these responsibilities may then default to nurses. There is also a correlation of housestaffing to nurse staffing, and a curvilinear relationship of housestaff per bed with market stage parallel to the nurse staffing curve. This supports the labor complement theory, that housestaff and nurses together act to assume attendings' task turf. The mechanism that needs to be explored here is an informal feedback model between the attendings and the residency program directors as to how many housestaff are required to staff their units, given their interest in and ability to practice medicine as opposed to pursue research, or handle a broader panel of patients. In any case, whatever forces are at work convincing residency program directors to expand or contract housestaff are correlated with the forces convincing hospital administrators to expand or contract nursing staffs.

On the other hand, more housestaff are related to less nursing work. This also supports some degree of substitutability within the hydraulic model of task turf: more housestaff reduce attending and nurse work, more nurses reduce attending and housestaff work.

Abbott, Andrew. 1991. "The future of professions: occupation and expertise in the age of organization." Sociology of Organizations 8:17-42.

______. 1988. The System of Professions: An Essay on the Division of Expert Labor. Chicago: University of Chicago Press.

Anonymous. 1994. "Even nurses are feeling the effects of downsizing." Chicago Sun-Times, January 16: 52.

Bellaby, Paul and Paul Oribabor. 1980. "Determinants of the occupational strategies adopted by British hospital nurses." International Journal of Health Services 10:291-309.

______. 1977. "The growth of trade union consciousness among general hospital nurses viewed as a response to proletarianization." Sociological Review 25:801-822.

Borzo, Greg. 1994. "As market grows, recruiters eye physician extenders." American Medical News, (June 14):5.

Braverman, Harry. 1974. Labour and Monopoly Capital. New York: Monthly Review Press.

Choi, Thomas, Robert Allison, Fred Munson. 1986. Governing University Hospitals in a Changing Environment. Ann Arbor: Health Administration Press.

Coburn, David. 1988. "The Development of Canadian Nursing: Professionalization and Proletarianization." International Journal of Health Services 18:437-456.

Donovan, Marilee I. and Gwendolyn Lewis. 1987 1. "Costs of nursing services: are the assumptions valid?" Nursing Administration Quarterly 12:1-6.

______. 1987 2. "Increasing productivity and decreasing costs: the value of RNs." Journal of Nursing Administration 17:16-8.

Etheredge, Mary Lou Salome. 1989. Collaborative Care: Nursing Case Management. AHA Publications.

Ferraro, Kenneth and Tammy Southerland. 1989. "Domains of Medical Practice: Physicians' Assessment of the Role of Physician Extenders." Journal of Health and Social Behavior 30:192-205.

Garfink, Christine, Katherine K. Kirby, Sara S. Bachman and Paulette Starck. 1991. "The University Hospital Nurse Extender Model: Part III, Program Evaluation." Journal of Nursing Administration 21:21-27.

Georgopolous, Basil S., F. C. Mann. 1962. The Community General Hospital. NY: MacMillan.

Graduate Medical Education National Advisory Committee. 1980. GMENAC:Volume 6- Nonphysician Health Care Providers Technical . U.S. Department of Health and Human Services.

Gray, David E. 1989. "Militancy, unionism, and gender ideology: A study of hospital nurses." Work and Occupations 16:137-152.

Hall, Judith A., R. Heather Palmer, E. John Orav, J. L. Hargraves, Elixzabeth A. Wright and Thomas A. Louis. 1990. "Performance Quality, Gender and Professional Role: A Study of Physicians and Nonphysicians in 16 Ambulatory Care Practices." Medical Care 28:489-501.

Hennenberger, Melinda. 1994. "For Nurses, Fewer Openings, Specialized Needs." New York Times, (Aug 1):

Iglehart, John K. 1993. "The American Health Care System: Teaching Hospitals." New England Journal of Medicine 329:1052-1056.

Kane, Robert L., Judity Garrard, Carol L. Skay, David M. Radosevich, Joan L. Buchanan, Susan M. McDermott, Sharon B. Arnold and Loyd Kepferle. 1989. "Effects of a geriatric nurse practitioner on process and outcome of nursing home care." American Journal of Public Health 79:1271-1277.

Levitt, Jane. 1977. "Men and Women as Providers of Health Care." Social Science and Medicine 11:395-398.

Melosh, Barbara. 1982. The Physician's Hand. Philadelphia: Temple University Press.

Morlock, Laura L. and Constance A. Nathanson. 1979. "Organizational factors associated with quality of care in 17 general acute care hospitals." Association of University Programs in Health Administration.

Morlock, Laura and Constance Nathanson. 1983. "Power and Influence." Pp. 250-284 in Health Care Management: A Text in Organization Theory and Behavior, eds. Stephen Shortell and Arnold Kaluzny. NY: Wiley.

Office of Technology Assessment. 1986. Nurse Practitioners, Physician Assistants and Certified Nurse-Midwives: A Policy Analysis. Washington D.C.: GPO.

Perrow, Charles. 1965. "Hospitals: Technology, Structure and Goals." Pp. 910-71 in Handbook of Organizations, ed. J. March. Chicago: Rand McNally.

Pfeffer, Jeffrey. 1973. "Size, composition and function of hospital boards of directors: A study of organizational-environment linkage." Administrative Science Quarterly 18:349-364.

Rosenberg, Charles E. 1987. The Care of Strangers: The Rise of America's Health Care System. Basic Books.

Safriet, Barbara J. 1992. "Health care dollars and regulatory sense: the role of advanced practice nursing." Yale Journal of on Regulation 9:417-88.

Schramm, Carl J. 1983. "The teaching hospital and the future of state government." New England Journal of Medicine 308:41-45.

Sherer, Jill. 1994. "Can hospitals and organized labor be partners in redesign?" Hospitals and Health Networks 10: 56-57.

Smith, Elizabeth D. 1985. "Career- versus job-orientation: a power dilemma in nursing and feminism." Pp. 83-105 in Power, Politics and Policy in Nursing, ed. Rita R. Wieczorek. NY: Springer Pubs.

Smith, Howard L. 1958. "Two lines of authority: The hospital's dilemma." Modern Hospital 84:59-64.

Steinwachs, Donald M., Jonathan P. Weiner, Sam Shapiro, Paul Batalden, Kathy Coltin and Fred Wasserman. 1986. "A Comparison of the Requirements for Primary Care Physicians in HMOs with Projections made by the GMENAC." New England Journal of Medicine 314:217-222.

Stevens, Rosemary. 1989. In Sickness and in Wealth: American Hospitals in the Twentieth Century. NY: Basic.

Sullivan, Patricia and Timothy Brown. 1989. "Unlicensed persons in patient care settings. Administrative, policy, and ethical issues." Nursing Clinics in North America 24:557-69.

Taylor, Susan G., J. M. Pickens and E. A. Geden. 1989. "Interactional styles of nurse practitioners and physicians regarding patient decision making." Nursing Research 38:50-55.

Voelker, Rebecca. 1993. "Health care hiring hasn't beat the nursing shortage." American Medical News, February 1:3.

Wagner, David. 1980. "The Proletarianization of Nursing in the United States 1932-1946." International Journal of Health Services 10:271-90.

Weiner, Jonathan P. 1987. "Primary Care Delivery in the United States and Four Northwest European Countries: Comparing the "Corporatized" with the "Socialized"." Milbank Quarterly 65:426-461.

Weiner, Jonathan P., Donald M. Steinwachs, Sam Shapiro, Kathryn L. Coltin, Dan Ershoff and Paul O'Connors. 1987. "Assessing a Methodology for Physician Requirement Forecasting: Replication of GMENAC's Need-based Model for the Pediatric Specialty." Medical Care 25:426-436.