|
Hospitalists and the Doctor‐Patient Relationship
David Meltzer*
Source: The Journal of Legal Studies, Vol. 30, No. S2, The Regulation of Managed Care Organizations and the Doctor‐Patient Relationship<break></break>A Conference Sponsored by the University of Chicago Law School, the Division of Biological Sciences, and the MacLean Center for Clinical Medical Ethics (June 2001), pp. 589-606
Published by: The University of Chicago Press for The University of Chicago Law School
* Associate Professor, Department of Medicine, and Associated Faculty Member, Harris School of Public Policy and Department of Economics, University of Chicago. This paper was prepared for a conference on the regulation of managed care organizations and its impact on the physician-patient relationship organized by the Center for Medical Ethics, the Division of Biological Sciences, and the Law School of the University of Chicago held at the University of Chicago, November 2000.
Abstract
Hospitalists—physicians whose practice focuses on the care of hospitalized general medicine patients—are increasingly common in the United States, often displacing primary care physicians from this role. While advocates of hospitalists point to evidence of cost reduction and perhaps improved short-run outcomes, critics question whether costs or long-run outcomes are improved and whether there may be insidious effects on the doctor-patient relationship. I de?ne a framework for addressing these questions, assess the available evidence, and identify a research agenda to better understand the implications of the hospitalist movement for the doctor-patient relationship. Using a framework that emphasizes general and patient-speci?c knowledge as crucial to a successful doctor-patient relationship, I argue that the success of the hospitalist model will depend on its ability to continue to develop its areas of technical expertise while ensuring care that is both continuous and appropriate to the needs and values of individual patients.
I. Introduction
The care of hospitalized patients by “hospitalists”—often de?ned as physicians who dedicate at least 25 percent of their practice to inpatient care1—is a recent, growing, and controversial trend in health care delivery in the United States. Motivating this trend is the belief that physicians who devote a large fraction of their time to inpatient care will develop expertise in caring for inpatients and gain familiarity with hospital organization that will allow them to more effectively and ef?ciently care for their patients.2 Although the term “hospitalist” was ?rst de?ned only in 1996, 3 the idea that certain physicians
1 Robert M. Wachter & Lee Goldman, The Emerging Role of “Hospitalists” in the American
Health Care System, 335 New Eng. J. Med. 514 (1996).
2 Id.
3 Id.
specialize in inpatient care is much older and, in fact, has been the dominant model in most countries other than the United States.4
Although primary care physicians have traditionally been responsible for both inpatient and outpatient care in the United States, the increasing demands of both inpatient and outpatient settings and pressures under managed care to control costs have vastly increased interest in hospitalist models of care. This has been reinforced by the dramatic shift of medical care to ambulatory settings over the past several decades so that hospitalization occurs less frequently, and hospitalized patients are generally much sicker than was the case previously. As a result, even a full-time general internist now often has only one or two patients hospitalized at a time, 5 and those patients that are hospitalized are generally very sick, so of?ce-based clinicians may ?nd the inpatient setting an increasingly dif?cult one in which to work. Indeed, concerns such as these have motivated a substantial number of physician groups and health maintenance organizations to move to hospitalist models of care, in some cases even making transfer of care to a hospitalist mandatory.6
In contrast to this enthusiasm concerning hospitalists on the part of some payers and physicians, there are deep concerns on the part of many primary care physicians and patients about the loss of continuity of care between the inpatient and outpatient settings. 7 This has even led to efforts by some primary care physicians to seek legislation that bans mandatory hospitalist models. 8 Nevertheless, the “hospitalist movement” continues to gain momentum. 9 One sign of the growing recognition and importance of hospitalists is the establishment in 1998 of the National Association of Inpatient Physicians, which is devoted to hospitalists and already has over 1,500 members. 10 Advertisements offering positions for hospitalists abound in major medical journals, and the ?rst formal training programs are being formed. 11
4 Naoki Ikegami & John C. Campbell, Medical Care in Japan, 333 New Eng. J Med. 1295 (1995); John W. Peabody, S. R. Bickel, & James S. Lawson, The Australian Health Care System, 276 JAMA 1944 (1996); Kevin Grumbach & John Fry, Managing Primary Care in the United States and in the United Kingdom, 328 New Eng. J. Med. 940 (1993).
5 American Medical Association, Physician Marketplace Statistics: Pro?les for Detailed Specialties, Selected States and Practice Arrangements (1998).
6 Phyllis Maguire, Use of Mandatory Hospitalists Blasted, 19 ACP-ASIMObserver 1 (1999).
7 Harold C. Sox, The Hospitalist Model: Perspectives of the Patient, the Internist, and Internal Medicine, 130 Annals Internal Med. 368 (1999); Stephen D. Fihn, Specialization by Practice Location, 14 J. Gen. Internal Med. 205 (1999); Bruce Bagley, Hospitalists and the Family Physician, 58 Am. Fam. Physician 336 (1998).
8 Maguire, supra note 6.
9 RobertM.Wachter & Scott Flanders, The HospitalistMovement and the Future of Academic General Internal Medicine, 13 J. Gen. Internal. Med. 783 (1998).
10 Report of John Nelson, President of National Association of Inpatient Physicians, April 1999.
11 Wachter & Flanders, supra note 9.hospitalists 591
But despite the growth of interest in hospitalists, there have been few scienti?c evaluations of the concept, and fundamental issues about how hospitalists may alter the doctor-patient relationship remain unexamined. While advocates of the hospitalist movement point to evidence of cost reductions and perhaps improved acute outcomes, critics question whether costs or longrun outcomes are truly improved and whether there may be insidious effects on the nature of the doctor-patient relationship that may ultimately undermine the quality of health care. In this paper, I attempt to de?ne a framework for addressing these questions, assess the available evidence concerning those questions, and identify a research agenda to better understand the implications of the hospitalist movement as it relates to the doctor-patient relationship. Using a framework that emphasizes the roles of both general and patientspeci?c knowledge as crucial to a successful relationship between doctor and patient, I argue that the hospitalist movement is in fact highly consistent with widely accepted patterns for structuring the doctor-patient relationship elsewhere in medicine. This suggests that the success of the hospitalist model will depend on its ability to continue to develop its areas of technical expertise as a discipline while identifying strategies to work with patients and their primary care physicians to ensure care that is both continuous and appropriate to the needs and values of individual patients.
In Section II, I develop a stylized framework for examining the advantages and disadvantages of medical specialization. The basic conceptual model draws on two classic, yet strikingly different, writings on the merits of specialization from the economics literature and the medical literature: Adam Smith’s The Wealth of Nations 12 and Francis Peabody’s On the Care of the Patient.13 While the former argues strongly for the importance and merits of specialization, the latter argues for the importance of care that is based on a deep personal knowledge of the patient. Yet despite these differences in emphasis, a common model is seen to form the foundation for both analyses: the ef?ciency of specialization is determined by a balance of the advantages of increasing expertise that is gained with increasing specialization versus the increasing costs of coordinating care for an individual patient that must be borne as specialization increases. Seen from this vantage point, the desirability of specialization in any speci?c context depends on the magnitude of these effects in that context; in Section III, I describe the state of existing empirical evidence for such effects for hospitalists. Overall, both the advantages relating to increased expertise and the disadvantages relating to the coordination of care are found to be signi?cant, with the available evidence favoring the use of hospitalists but with clear holes in the information required to make a comprehensive assessment. In Section IV, I conclude and outline an agenda for further research to address this question.
12 Adam Smith, The Wealth of Nations (Modern Library 1965) (1776).
13 Francis W. Peabody, The Care of the Patient, 88 JAMA 877 (1927).592 the journal of legal studies
II. A Theoretical Framework for Specialization
As alluded to above, the theoretical framework I will use here to examine specialization views specialization as having both advantages related to increased expertise and disadvantages related to the challenges in coordinating domains of expertise to accomplish a speci?c objective. In the context of medical specialization, this translates into technical expertise in speci?c clinical domains on the one hand and knowledge of the patient on the other. Although classic scholars of the economics of specialization, such as Smith, and classic scholars of the doctor-patient relationship, such as Peabody, have tended to emphasize different sides of this trade-off, one ?nds in both literatures a striking similarity in describing this trade-off as balancing the value of expertise versus the costs of coordination. As I describe below, such a framework helps to illustrate a variety of important issues in medical specialization that extend well beyond hospitalists.
A. On the Bene?ts of Specialization: Adam Smith
The greatest improvement in the productive powers of labour, and the greater part of the skill, dexterity, and judgement with which it is any where directed, or applied, seem to have been the effects of the division of labor. [Adam Smith, The Wealth of Nations, bk. I, ch. I (1776)]
So begins the Wealth of Nations, with this passage shortly followed by Smith’s now classic description of a pin factory in which the division of labor increases the productivity of workers manyfold over what they could accomplish if they worked on their own. Smith then goes on to describe these advantages of specialization as stemming from (1) improved “dexterity that comes from repetition,” (2) savings of time that would otherwise be spent “switching between tasks,” and (3) “the application of machinery, invented by workmen or others.” It is dif?cult not to be struck by how well these simple advantages, described generically by Smith more the 200 years ago, appear to apply to the debate about hospitalists today. What Smith described as improved dexterity with repetition is know today in the economic literature as a part of “learning by doing”14 and in the medical literature as “the volume-outcomes relationship”15 and has been documented for a wide range of clinical conditions ranging from laparoscopic cholecystectomy16 to
14 Kenneth J. Arrow, The Economic Implications of Learning by Doing, 29 Rev. Econ. Stud. 155 (1962).
15 Harold S. Luft et al., Hospital Volume, Physician Volume, and Patient Outcomes: Assessing the Evidence (1990).
16 Southern Surgeons Club, Michael J. Moore, & Charles L. Bennett, The Learning Curve for Laparoscopic Cholecystectomy, 170 Am. J. Surgery 55 (1995).hospitalists 593
care for human immunode?ciency virus.17 Below I show data from our own experiment with hospitalists at the University of Chicago that shows that the same relationship between experience and outcomes holds for hospitalists. The savings of time spent switching between tasks is also highly relevant for the debate on hospitalists. Particularly in community settings, it is precisely the time costs associated with moving between the hospital and a community-based of?ce that have probably been the most important drivers of the growth of hospitalists. As hospitalization rates have fallen, primary care physicians have been increasingly reluctant to take time out of a busy
of?ce practice to see what is often only one or two patients in the hospital at a given time. 18 Moreover, from the patient perspective, the primary care physician may be some distance away from the hospital when an emergency arises, and the time required to get to the hospital may often mean that the primary care physician is not physically present when urgent decisions must be made. And ?nally, although perhaps not at ?rst obvious but probably no less important than the others, “the application of machinery, developed by workmen and others,” has an important parallel in the discussion about hospitalists—namely, that hospitalists may both develop and better utilize knowledge and systems of care that may then allow both the hospitalists and other physicians to better care for their patients. Such advantages of hospitalists are manifest both in the growing efforts within hospital medicine to de?ne, develop, and disseminate a knowledge base19 and in the key role that hospitalists play in systems development in many of the institutions in which they operate.
It is not by chance that Smith chose to begin the Wealth of Nations with this discussion of specialization. Indeed, it is the bene?ts of specialization that ultimately produce the incentives for exchange through markets that is the ultimate focus of The Wealth of Nations. A corollary of this, however, is that the costs of accomplishing those exchanges must be considered to be the costs of specialization. Indeed, specialization must surely have some disadvantages, or it would be carried everywhere to its maximum. At a super?cial level, Smith’s answer to this is usually summarized in his statement that the division of labor is “limited by the extent of the market.” For example, a given number of people are eventually able to produce so many pins that, although the addition of more workers and further division of labor could
17 Charles L. Bennett et al., Relation between the Hospital Experience and In-Hospital Mortality for Patients with AIDS-Related Pneumocystis carinii Pneumonia: Experience from 3,126 Cases in New York City in 1987, 5 J. Acquired Immune De?ciency Syndromes 856 (1992); Valerie E. Stone et al., The Relation between Hospital Experience and Mortality for Patients with AIDS, 268 JAMA 2655 (1992); Mari M. Kitahata et al., Physician’s Experience with the Acquired Immunode?ciency Syndrome as a Factor in Patients’ Survival, 334 New Eng. J. Med. 701 (1996).
18 American Medical Association, supra note 5.
19 Wachter & Goldman, supra note 1.
further enhance productivity, there would simply be no market for so many pins. Smith illustrates this with the observation that the division of labor is taken furthest in goods produced for mass consumption and in larger towns and cities, where the extent of the market is large. These are both important observations, but further examination of Smith’s argument shows a more subtle analysis—that these limits on the division of labor can often be overcome by improved transport that can expand the market. Thus, a more subtle restatement of Smith’s argument is that the division of labor is ultimately limited by the costs of coordination. This is in fact the core of modern economic theories of specialization.20 But while most of theWealth of Nations focuses on the social institutions and market functions that create such coordination, the costs of accomplishing such coordination are not Smith’s focus; despite the reality of the social costs of the brokers, traders, and others who maintain markets, Smith’s emphasis is on the “invisible hand” of the market. For a richer understanding of the costs of coordination, especially in a medical context, one has to turn elsewhere.
B. On the Costs of Specialization: Francis Peabody
The treatment of a disease may be entirely impersonal; the care of the patient must be completely personal. The signi?cance of the intimate personal relationship between physician and patient cannot be too strongly emphasized, for in an extraordinarily large number of cases both diagnosis and treatment are entirely dependent on it, and the failure of the young physician to establish this relationship accounts for much of his ineffectiveness in the care of patients. [Francis Peabody, The Care of the Patient, 88 JAMA 877 (1927)]
Peabody’s simple statement of the importance of the personal relationship between the doctor and patient captures the essence of an immense and important literature on the doctor-patient relationship that has followed it. But in reading Peabody, one is struck by how in framing his work he drew on many of the same ideas about expertise that Smith drew upon and then proceeded one step further: To begin with, the fact must be accepted that one cannot expect to become a skillful practitioner of medicine in the four or ?ve years allotted to the medical curriculum. Medicine is not a trade to be learned but a profession to be entered. It is an everwidening ?eld that requires continued study and prolonged experience in close contact with the sick. All that the medical school can hope to do is to supply the foundations on which to build. When one considers the amazing progress of science in relation to medicine during the last thirty years, and the enormous mass of scienti?c material which must be made available to the modern physician, it is not surprising that the
20 Gary S. Becker & Kevin M. Murphy, 107 Q. J. Econ. 1137 (1992).
schools have tended to concern themselves more and more with this phase of the educational problem. And while they have been absorbed in the dif?cult task of digesting and correlating new knowledge, it has been easy to overlook the fact that the application of principles of science to the diagnosis and treatment of disease is only one limited aspect of medical practice. The practice of medicine in its broadest sense includes the whole relationship of the physician with his patient.
21
From here, Peabody goes on to describe how the physician must understand
the patient’s personal condition to effectively treat a wide range of diseases
in which factors relating to aspects of the patient’s family and/or social
condition or personal psychological factors may play a role. Although Pea-
body’s emphasis is on the importance of integrating the psychological and
social with the physiological, he also makes his point in some concrete ways
that are purely medical and have remarkable resonance with the current
debates about medical specialization: “Now the essence of the practice of
medicine is that it is an intensely personal matter, and one of the chief
differences between private practice and hospital practice is that the latter
always tends to become impersonal....The dif?culty is that in the hospital
one gets in the habit of using the oil immersion lamp instead of the low
power, and focuses too intently in the center of the ?eld....The institu-
tional eye tends to become focused on the lung, and it forgets that the lung
is only one member of the body.”22
Peabody’s description of inpatient medicine in 1927 rings remarkably true
for much of medical care today. As diseases affect multiple organs, at times
one ?nds patients with nearly as many physicians as affected organs, and it
is not uncommon that many of these physicians fail to see the patient as a
whole, often in not only the personal but also the biological sense. It is this
circumstance of the extreme specialization of modern medicine that has in
many ways motivated the resurgence of the primary care physician in recent
years. Surely some of this has been to act as gatekeeper in the sense of
rationing access to specialized care, but the role of primary care physician
as coordinator of care has increasingly emerged as important. For many of
the conditions that primary care physicians once cared for on their own, their
primary role is to act as coordinator of care with one or more subspecialists.
In so doing, the intent is that they not only synthesize the insights and
recommendations of many specialists who may be involved in the patient’s
care but also provide crucial knowledge of the patient as a person, including
the family, social, and economic concerns and personal preferences about
health and medical care. Thus, instead of being eliminated by the growth of
specialized medical knowledge, the role of the primary care physician has
been altered to place greater emphasis on the role of coordinator. Although
he did not focus on the implications for specialization per se, Peabody’s
21
Peabody, supra note 13, at 813–14.
22
Id.596 the journal of legal studies
insight into the care of the patient as requiring a coordinated approach to
understanding both the patient as a person and the interactions among the
diseases that may affect him thus provides an important perspective on
specialization.
C. Toward a Theory of Ef?cient Medical Specialization
Combining Smith’s observations on the advantages of specialization that
operate through expertise and Peabody’s observations on the need for co-
ordination that suggest disadvantages of specialization, it is possible to think
of an ef?cient degree of specialization that balances these two competing
forces. As noted above, such models of optimal specialization have been
described in the modern economics literature.
23
The basic insight of these
models is simply that the optimal degree of specialization increases on the
one hand by the bene?ts of specialization, which tend to increase with the
generation of additional knowledge, and decreases with the costs of coor-
dination on the other.
24
Although such a formulation of medical specialization may be abstract, it
explains a variety of patterns in medical specialization that suggest it is of
more than theoretical interest. For example, the model predicts that special-
ization will rise over time as the total pool of knowledge increases. Seen
from this perspective, the high degree of specialization evident in American
medicine compared to medical practice in other countries can be understood
as re?ecting at least partially its technological sophistication. Yet at the same
time, such a system will generate problems in coordination that will produce
an increasing demand for coordination, especially for persons with a partic-
ular need for coordinated care, such as those with multiple illnesses or im-
paired ability to coordinate their own care. This explains the development
of geriatrics to address the coordination of care needed by older persons who
often have multiple medical problems that require specialized care. Similarly,
23
Becker & Murphy, supra note 20.
24
For example, one might postulate that favorable outcomes of medical care depend on both
technical expertise in executing the speci?c tasks required and coordination of those tasks.
Assume, for example, that technical expertise is given by , where is
aH
f(n, a, H)pnn(1 1)
the degree of specialization (number of specialists), is a parameter re?ecting the bene?ts a(1 0)
of specialization, and is the overall amount of medical knowledge (human capital). H(1 0)
Note that this has the property that expertise is increasing in the degree of specialization, the
bene?ts of specialization, and the total amount of human capital. Note also that the bene?ts
of specialization are increasing in both the bene?ts of specialization and total
aH1
f paHn 1 0 n
amount of human capital. Similarly, one can model coordination costs as a probability (p) that
each of the n specialists’ care would fail to be successfully coordinated, so the overall probably
of successful care is , which is decreasing in both the probability of a failure n
p(n, p)p(1p)
of coordination (p), and the degree of specialization. Combining these two parts of the model,
the expected outcome of care is . To identify the optimal degree n aH
p(n, p)f(n, a, H)p(1p) n
of specialization, we maximize this over n to yield the optimal degree of specialization:
. Thus, optimal specialization increases in the returns to specialization (a) n*paH/(1p)
and total amount of human capital (H) and decreases in the costs of coordination (p).hospitalists 597
if one recognizes that patients, and especially educated ones, may play an
important role in coordinating their own care, one can understand why ad-
vocates of increased access to care for less advantaged persons are often
such strong advocates of physician workforce reform to enhance primary
care. In the section that follows, we apply this same model to try to better
understand the debate about hospitalists.
III. Expertise, Coordination of Care, and the
Value of Hospitalists
Seen from the perspective of the framework described above, hospitalists
can be seen as offering advantages over primary care physicians to the extent
that they offer greater technical expertise in the provision of inpatient care
but disadvantages relative to primary care physicians to the extent that they
are less likely to know the patient well as an individual, in both medical and
broader personal terms. Whether hospitalists improve or impair outcomes is
ultimately therefore an empirical question, and our analysis suggests that in
examining this question we need to be attuned speci?cally to issues of tech-
nical expertise on the one hand and issues of coordination on the other. What
is the evidence that hospitalists affect outcomes, and, if there are effects,
what do we know about how those effects arise? Are expertise and coor-
dination indeed key mechanisms that affect the effects of hospitalists?
A. What Is Known about the Effects of Hospitalists on
Costs and Outcomes?
It is remarkable that, despite the extraordinary growth in hospitalists, there
have been relatively few systematic studies of the effects of hospitalists on
patient costs and outcomes. In the majority of studies, the major limitation
of the design has been the nonrandom assignment of patients to hospitalist
or nonhospitalist attendings. The results of many of these observational non-
randomized studies have not been reported in the academic literature,
25
but
the studies that have been published illustrate some of the challenges in such
nonrandomized evaluations.
For example, in a study set in a community teaching hospital in western
Pennsylvania, Herbert S. Diamond, Elliot Goldberg, and Janine E. Janosky
found that the implementation of dedicated faculty hospitalists as inpatient
care providers in a teaching hospital for a group of community-based primary
care physicians decreased the median length of stay, median cost of care,
25
F. Michota, T. Lewis, & J. Cash, The Hospitalist: Will Inpatient Specialists Improve Care?
65 Clev. Clinic J. Med. 297 (1998).hospitalists 599
often highly limited from a single institution. Essentially the same concerns
apply to a study of hospitalist versus community physician care for pediatric
asthma and bronchiolitis, which found no effect on either costs or outcomes.
28
A preliminary report of the use of hospitalists at a group of Kaiser Permanente
hospitals found reduced length of stay (but not costs) in hospitals with hos-
pitalists but did not report whether the hospitals with hospitalists also had
shorter length of stay at baseline or control for other possible differences
between facilities or the patient populations they serve.
29
All these studies
point to the important limitations of nonrandomized studies and consequent
need to focus on randomized studies.
To date, there have been only two randomized studies of hospitalists. The
?rst is a study by Robert M. Wachter and colleagues,
30
and the second is a
study that my colleagues and I have in progress at the University of Chicago.
31
Both studies are perhaps best described as “quasi-randomized” since patients
were allocated to hospitalists or nonhospitalists based on the team that hap-
pened to be on call on the patient’s day of admission. While this is not strict
randomization, in neither case is there reason to believe that admission de-
cisions were affected by whether the on-call team was led by a hospitalist
or nonhospitalist, so there is no reason to expect systematic differences in
the patient populations between the services.
In their study, Wachter and colleagues at University of California, San
Francisco (UCSF), compared a group of physicians they called “dedicated
inpatient attendings” to a group of traditional attending physicians.
32
These
dedicated inpatient physicians served as attending physicians more often than
other faculty, provided early input into diagnostic and treatment decisions,
became involved in creating practice guidelines and other activities designed
to shorten length of stay, and were given a mandate to “increase quality and
decrease costs.” Wachter examined outcomes for 1,623 admissions assigned
randomly either to these dedicated inpatient physicians or to traditional ac-
ademic attending physicians and found that patients cared for by the dedicated
inpatient attendings had a .6-day shorter length of stay and $770 lower
average adjusted hospital costs. No statistically signi?cant effect was found
28
M. Seid, K. Quinn, & P. S. Kurtin, Hospital-Based and Community Pediatricians: Com-
paring Outcomes for Asthma and Bronchiolitis, 4 J. Clinical Outcomes Mgmt. 21 (1997).
29
Diane E. Craig et al., Implementation of a Hospitalist System in a Large Health Main-
tenance Organization: The Kaiser Permanente Experience, 130 Annals Internal Med. 355
(1999).
30
Robert M. Wachter et al., Reorganizing an Academic Medical Service: Impact on Cost,
Quality, Patient Satisfaction, and Education, 279 JAMA 1560 (1998).
31
David Meltzer et al., Effects of Hospitalist Physicians on an Academic General Medicine
Service: Results of a Randomized Trial, 14 J. Gen. Internal Med. 112 (Supp. 2, 1999); David
Meltzer et al., Effects of Physician Specialization on Costs and Outcomes on an Academic
General Medicine Service: Results of a Trial of Hospitalists, Annals Internal Med.
(forthcoming).
32
Wachter et al., supra note 30.600 the journal of legal studies
on either in-hospital or postdischargemortality, readmission, emergency room
use, or patient, house staff, or faculty satisfaction, but the con?dence intervals
on these outcomes were wide. For example, the 95 percent con?dence interval
on in-hospital mortality ranged from 75 to [1]37 percent. Also, an additional
important limitation of the study is that only three of the 14 dedicated inpatient
physicians met the de?nition (proposed by Wachter) of “hospitalist”—that
is, attending on an inpatient service at least 3 months per year.
In the other randomized study, reported only in abstract form to date,
33
my colleagues and I at the University of Chicago have analyzed the results
of our own experiment with hospitalists. Beginning in July of 1997, one of
our four general medicine teams was covered by two general internists who
switched from attending on the inpatient wards 1 month per year to attending
6 months per year. Our initial analysis examined the costs and outcomes
from the ?rst 2 years of our study of the 6,511 patients who were cared for
by either these hospitalists or the other attendings who continued to attend
on the inpatient service for only 1 month per year.
Our results suggest that our design effectively randomized patients to
hospitalists and nonhospitalists. Speci?cally, 75.2 percent of patients were
cared for by nonhospitalists and 24.8 percent were cared for by hospitalists,
as predicted when one of four teams is staffed by hospitalists, and no sig-
ni?cant difference were present in the age, gender, race, diagnosis mix, or
comorbidity levels in the two patient groups.
Details of our statistical analysis are presented elsewhere,
34
but our key
?ndings with respect to resource use are that length of stay and costs were
either the same or slightly reduced for the hospitalists in year 1, but the
savings for the hospitalists rose substantially by year 2, to about .5 days and
$780 per admission. Interestingly, these decreases in length of stay and cost
by year 2 are very close to the .6 day and $770 savings reported at UCSF.
As we discuss further below, we believe that these increases in savings over
time provide important evidence about the mechanisms by which hospitalists
may have their effects and, speci?cally, the role of experience in determining
the effectiveness of hospitalists.
While this evidence of cost reductions is surely supportive of one objective
of the growing use of hospitalists, a meaningful assessment of their value
must include an assessment of their effects on patient outcomes. Our initial
analyses of outcomes from the ?rst 2 years found no signi?cant difference
between hospitalists and nonhospitalists in in-hospital or 30-day mortality,
readmission, physical function, or overall patient satisfaction. However, as
with the UCSF results, the con?dence intervals were large. For example, the
95 percent con?dence interval on 30-day mortality in our initial results in-
33
Meltzer et al., supra note 31.
34
David Meltzer et al., Effects of Hospitalist Physicians on an Academic General Medicine
Service: Results of a Randomized Trial (unpublished manuscript, Univ. Chicago 2001).hospitalists 601
cluded a 25 percent increase and 50 percent decrease in mortality rates with
hospitalists.
35
The results were similar in multivariate analyses controlling
for diagnosis-related group (DRG) weight and Charlson comorbidity index
and show the need for substantial sample sizes to further limit these con?-
dence intervals. Interestingly, however, the trends favored the hospitalists in
all measures except overall patient satisfaction, which was identical.
In these initial analyses of outcomes, we were able to assess mortality
only when the patient died in-hospital or we were able to contact the patient’s
family when we called for a follow-up interview. Since then, suf?cient time
has passed for deaths to appear in the Social Security Death Index, and we
have been able to match that by Social Security number, name, and date of
birth to vastly increase our follow-up rate and effective sample size. These
analyses suggest that, as with the more limited mortality data presented above
with smaller sample size, there appears to be a trend over the 2 years for
mortality to be lower for hospitalists at 30 days, with about a 1 percentage
point reduction in mortality (from 5.4 to 4.6 percent). With this larger sample
size, we are also able to separately analyze years 1 and 2, and, strikingly,
this shows no difference in year 1, but a reduction in mortality of almost 2
percentage points in year 2 that reaches statistical signi?cance at . p ! .04
Similar analyses for year 2 at 60 days ?nd essentially identical results. There
was no statistically signi?cant effect on mortality at 1 year, although the
trend again favored the hospitalists.
While we think it is highly credible that hospitalists may reduce mortality,
it is surprising to us that these mortality effects at 30 days are as large as
they appear to be—as much as a quarter to a third lower than mortality for
the nonhospitalists. We note, however, that the 95 percent con?dence interval
on these effects on mortality in year 2 includes effects as low as .1 percent,
so it is very important to be cautious in interpreting these results. This is
reinforced, of course, by the fact that these results re?ect the experience of
only two hospitalists in a single academic medical center. Nevertheless, as
we examine these results, it is dif?cult not be left with the strong impression
that there were substantial improvements in costs and outcomes.
B. What Mechanism Might Explain Effects of Hospitalists on
Costs and Outcomes?
That the hospitalist model has signi?cant promise is further reinforced by
the fact that the improvements in both costs and outcomes appear to increase
over time, which suggests that perhaps the experience hospitalists are ac-
cumulating over time (experience that in a sense de?nes the difference be-
tween the hospitalist job and the 1-month-per-year attending job) is indeed
the operative factor in the differences we are seeing between hospitalists and
35
Id.602 the journal of legal studies
nonhospitalists over time. This is of course not the only hypothesis to explain
why hospitalists may have lower resource use or better outcomes than non-
hospitalists. These include, in addition, (1) greater availability for inpatient
care such as being more likely to see patients on the day of admission or
multiple times per day owing to greater presence on the wards, (2) increased
initiation or utilization of speci?c system-level strategies, such as practice
guidelines, (3) increased teaching of cost-effective practices to house staff
that may diffuse throughout the general medicine service, and (4) selection
of low-cost or high-quality providers as hospitalists. While all of these mech-
anisms may play some role, none of them seems likely to be able to explain
the absence of effects initially and subsequent growth of cost savings for
hospitalists compared to nonhospitalists over time. This points again to the
role of the accumulation of experience in determining the effects of hospi-
talists. To examine this experience hypothesis, we constructed measures of
overall and disease-speci?c experience using the total volume of patients
seen by the physician up to the date the patient was admitted and the volume
of patients with the same diagnosis seen by the physician up to that date (as
de?ned by the three-digit ICD-9 code36
of primary diagnosis). As expected,
analysis of the means of the experience variables showed that the hospitalists
had more overall and disease-speci?c experience on average than the non-
hospitalists. For example, considering a measure of disease-speci?c experi-
ence, the average number of cases with the same primary diagnosis that a
hospitalist had seen in the study period to date by the time a patient was
admitted was 10, compared with only 2.5 prior cases for a nonhospitalist.
To assess the role of the greater availability of hospitalists to see patients on
the day of admission, we also controlled for whether the patient was admitted
on a weekday or weekend, since only hospitalists had no clinic on weekday
afternoons. We also constructed a measure of whether the hospitalist saw the
patient on the day of admission by matching physician billing records to the
admission dates. This showed that the hospitalists were more likely to see
patients on the day of admission than were the nonhospitalists, but only on
weekdays, when they presumably have greater availability because they do
not have clinic.
In regression analyses using these variables, we found strong evidence
that the effect of hospitalists is largely explained by disease-speci?c volume.
This suggests that increases in resource savings that occurred over time were
concentrated in the more common diagnoses with which the hospitalists were
able to acquire additional disease-speci?c experience and supports the hy-
pothesis that disease-speci?c experience may be an important mechanism by
which hospitalists are able to reduce resource use. Parallel analyses for mor-
tality were not as striking in statistical signi?cance but did show a trend
36
National Center for Health Statisitics, International Classi?cation of Diseases, 9th Revision
(1998).hospitalists 603
toward lower mortality for hospitalists with greater disease-speci?c experi-
ence. This again reinforces the conclusion that the apparent advantages of
hospitalists may consist of increased experience and, in particular, increased
disease-speci?c experience. Seen from this perspective, hospitalists, like the
workers in the pin factory Smith describes, seem to provide yet another
example of the merits of specialization.
C. What Do We Know about the Disadvantages of Hospitalists?
No one would dispute the need for further studies to better understand the
effects of hospitalists on costs and the broad set of outcomes described above,
including mortality, readmission, and patient satisfaction. Moreover, reason-
able persons could easily disagree about the ability of such measures to
provide a comprehensive assessment of outcomes. Nevertheless, probably
the most important limitation of the “randomized” studies described above
is that they have all been performed in academic medical centers in which
the comparison group is not patients cared for by their primary care physician
but patients cared for by another doctor whom the patient also does not
already know, only one with less current inpatient experience than hospi-
talists. Thus, while the data from the existing randomized trials suggest that
experienced hospitalists may offer some advantages over traditional academic
attendings who also do not know the patient, they do not address the fun-
damental concern relevant in community settings that hospitalists may un-
dermine the traditional relationship between the primary care doctor and his
or her patient. It is surely not inconceivable that a primary care physician
familiar with a patient could accomplish better outcomes than even a highly
experienced hospitalist not familiar with the patient.
Although ideally one would perform experiments in community settings
by randomizing patients to hospital care provided either by their primary
care physician or by a hospitalist, such experiments would require community
based primary care physicians to agree to have their patients randomized out
of their care. It is perhaps not surprising that, to date, no such opportunities
have been identi?ed except perhaps within the context of a fully integrated
health system (Kaiser).
37
However, while the results of such a trial would
surely be of interest, it would likely leave open the possibility that the bene?ts
of having one’s primary care physician provide care in the hospital would
be greater in a traditional fee-for-service environment.
So in the absence of an ideal randomized trial, what can we say about the
potential value of the doctor-patient relationship? One piece of evidence
comes from our hospitalist experiment itself, in which we originally gave
the hospitalists certain weekend days off with coverage provided by tradi-
37
Andrew D. Auerbach et al., Implementation of a Voluntary Hospitalist Service at a Com-
munity Teaching Hospital: Improved Clinical Ef?ciency and Patient Outcomes (unpublished
manuscript, Univ. California, San Francisco, December 2000).604 the journal of legal studies
tional general medical attendings in order to prevent the hospitalists from
burning out from 6 months per year of attending on the wards. This primarily
affected patients admitted during the week, who therefore experienced a
discontinuity of care if they were in the hospital when the weekend arrived.
As discussed in further detail,
38
this resulted in almost eliminating the savings
from hospitalists for patient admitted on weekdays. Thus, the effects of
discontinuity of care in the inpatient setting appear to have the potential to
be very large, even when the doctor-patient relationship has been con?ned
only to that hospital stay.
Another approach is to ask patients how they themselves would value
having their own physician care for them during a hospital stay as opposed
to a hospitalist. While this approach might depend a great deal on patients’
subjective assessment of how hospitalists might affect outcomes versus their
own physician, it has the advantage of also potentially capturing the value
to patients of their personal relationship with their physician. This might
perhaps be re?ected in patient satisfaction, but it is unlikely to be re?ected
in any of the other measures used in studies of hospitalists. We have recently
applied this approach using a national random-digit dial sample to assess
people’s preferences and willingness to pay for care by their physician.
39
Our
results show that about two-thirds of people with a primary care physician
say they would prefer care by their own physician to that by a hospitalist if
they were hospitalized for a general medical condition, while 25 percent
report no preference, and only 9 percent prefer a hospitalist. When asked
about their willingness to pay, however, the average willingness to pay for
care by their primary care physician is less than $200, and about two-thirds
of that average comes from less than 10 percent of the patients whose will-
ingness to pay exceeds the $750 average savings from hospitalists in the two
randomized trials. This argues strongly against a mandatory hospitalist system
and suggests that a system that requires a copayment to have care by one’s
own primary care doctor when hospitalized might be worth considering.With
a copayment set to the ef?cient level of the average savings from hospitalists
($750), those who valued care by their own physician above that amount
would presumably pay the copayment to receive care by their own doctor.
In contrast, those who valued such care by less that amount would not, saving
an average of $750 in costs while having an average willingness to pay for
their own doctor of only $68. This again suggests that the value of the doctor-
patient relationship is not trivial to patients, although it may not be as large
as some might believe. Although there are excellent reasons to question the
validity of willingness-to-pay questions, it is somewhat reassuring that we
38
David Meltzer, Julie Herthko, & Lei Jin, Patients’ Willingness to Pay for Hospital Care
by Their Primary Care Physician versus Hospitalists: Results of a National Survey (unpublished
manuscript, Univ. Chicago, December 2000).
39
Id.hospitalists 605
found that respondents’ stated willingness to pay correlated well with rea-
sonable measures of the quality of the doctor-patient relationship, including
its duration, the patient’s satisfaction with the quality of care, and the patient’s
satisfaction with his or her personal relationship with the doctor.
However, regardless of how one quanti?es the value of the doctor-patient
relationship, it is, as Peabody suggested, not enough to simply treat a disease.
Instead, a successful physician must treat the whole patient, and that involves
some knowledge of how the patient came to be in the condition he or she
is at any speci?c time, even if the physician’s knowledge of the patient’s
past history is con?ned only to the hospitalization. Indeed, the fact that
patients in the University of Chicago hospitalist experiment experienced in-
creases in length of stay and costs when their hospitalists had weekend days
off may suggest that some signi?cant relationship between the patient and
his or her physician had indeed developed in the course of the hospitalization,
so the covering physician was not a perfect substitute. To the extent that
such a relationship can be nurtured in the inpatient setting, as indeed Peabody
himself urged was possible, it is likely that hospitalists may be able to
minimize the effect of their lack of previous contact with the patient.
IV. Conclusions
If optimal medical specialization as it relates to hospitalists must be un-
derstood as re?ecting a balance between the bene?ts of increasing expertise
and the costs of coordination, the national debate over the proper role of
hospitalists is not likely to be fully resolved by evidence any time soon. On
the expertise side, many questions remain, but the ultimate message seems
likely to be that the greater experience of hospitalists in the inpatient setting
translates directly into measurable reductions in resource use and improve-
ment in outcomes. In settings such as many academic medical centers where
hospitalists are replacing physicians who are not patients’ primary care phy-
sicians, resolution of this issue may be all that is needed to make a case for
hospitalists from the patient’s perspective. Understanding the value of hos-
pitalists in a community setting will surely be more dif?cult, but what does
seem clear is that there is the potential for signi?cant advantages to be gained
through greater inpatient expertise and for those advantages to be undone if
care is insuf?ciently coordinated. A reasonable solution, then, is to try to
better understand and strengthen the mechanism by which expertise translates
into improved outcomes and to enhance the potential for hospitalists to relate
successfully to the patient and his or her usual physician. For example, to
the extent that hospitalists can create systems (such as practice guidelines)
that can improve the care provided by all physicians, hospitalistsmay improve
the care provided by primary care physicians. Similarly, to the extent that
hospitalists can improve the quality of their relationships with patients in the
hospital, and smooth the functioning of the handoff, there seems great po-606 the journal of legal studies
tential to realize the advantages of hospitalists without sacri?cing the co-
ordination of care. As hospital medicine continues to develop as a discipline,
its success will likely depend on its ability to address these important needs
for expertise and coordination that together de?ne the appropriate roles for
specialization.
|