Reorganizing an Academic Medical Service: Title and subTitle BreakImpact on Cost, Quality, Patient Satisfaction, and Education

AS MANAGED CARE captures an increasing share of the health care market, academic health centers are placed in a tenuous position because of the extra costs of training¹ and the high acuity of their patients.^{2 - 4} Some training programs have placed increased emphasis on cost-effectiveness in conferences, provided feedback on resource use to house staff, or added utilization-review nurses or physicians to the inpatient wards.^{5 - 11} A few have created nonteaching faculty services to deliver efficient, high-quality care to some inpatients.^{12 - 13} But these changes may be insufficient given the major expenditures attributable to inpatient care even in capitated health care systems.^{14 - 15} Moreover, by removing trainees from the care of capitated patients (in the case of nonteaching services) or providing administrative oversight devoid of an educational component (in the case of external utilization review), such changes may insulate residents and students from the lessons of managed care, thereby squandering opportunities to educate them in evidence-based medicine, practice guidelines, collaborative care, and other skills relevant to the future health care environment.^{16 - 17}

We hypothesized that attending physicians who were focused on inpatient medicine, were attuned to cost-effectiveness, and were promptly and intensively involved in the care of each patient would have the greatest chance of decreasing costs while preserving quality, education, and patient satisfaction. We reorganized our academic medical service to test this hypothesis.

Moffitt-Long Hospital at the University of California, San Francisco (UCSF), is a 520-bed referral center and community hospital. The inpatient general medical service consists of 4 teams, each admitting every fourth night and composed of a faculty member, 1 resident, 1 or 2 interns, and 0 to 3 students. There is a separate cardiology service, and separate bone marrow, liver, and renal transplant services. Traditionally, the vast majority of ward attending physicians have served in this role for 1 month each year. The majority of patients admitted to the medical service are cared for as outpatients by a full-time faculty member (independently or with residents) and are seen on the day of admission by this physician, by faculty in the emergency department, or by both. These faculty members generally provide input into the initial management of admitted patients, and may continue to provide some input during the course of the hospitalization. The inpatient attending physician, however, is the physician-of-record on the staff medical service, and the inpatient team makes most clinical decisions regarding hospitalized patients. Most inpatient attending physicians, though available on the day of admission, first heard about admitted patients on the morning after admission. The inpatient teams continue to provide care for their patients in the intensive care units. About 13% of admissions each year are patients who continue to be cared for in the hospital by their private physicians; these admissions were not considered in this analysis.

On July 1, 1995, we divided the medical ward teams into 2 services, each with 2 teams each month: the managed care service (MCS) and the traditional service (TS). The differences were (1) MCS faculty members served as attending physicians more often (57% served 2 or more months vs 4% of TS faculty, P<.01; 3 MCS attending physicians served 3 or 4 months [Table 1]); (2) MCS attending physicians examined (for daytime admissions) or discussed (for nighttime admissions) patients at the time of admission and gave the resident early input into diagnostic and treatment decisions; (3) MCS attending physicians became involved in practice guideline and quality improvement activities related to the inpatient service; and (4) MCS attending physicians were given an explicit mandate to "increase quality and decrease costs." The MCS faculty were members of the existing faculty pool (n=6), members of our integrated practice group who cared for capitated outpatients (n=5, each of whom only attended for 1 month), or new hires (n=3). To recruit faculty members for the MCS, 1994-1995 ward attending physicians were asked if they wanted to participate in the new service. Eleven faculty members were willing; the remainder of the MCS was staffed by 3 new faculty members interested in serving on the MCS. At the time of the MCS formation, no data on individual faculty members were available regarding prior year's efficiency or outcomes to allow for informed "cherry-picking"; MCS attending physicians were chosen based on interest and availability rather than demonstrated ability to manage care effectively. Fully capitated patients (n=77 in 1995-1996) were preferentially admitted to MCS teams and were therefore excluded from all reported analyses (their inclusion in the analysis, however, had no significant impact on any of the results); otherwise, patients were allocated to teams based on alternate days of admission. The MCS and TS teams admitted to the same hospital floor and worked with the same nurses, discharge planners, and information systems. Residents and interns (each of whom serve on the Moffitt-Long general medicine service for only 1 month each year) were assigned to MCS or TS teams based on the same scheduling considerations as in prior years without consideration of prior performance. The MCS faculty members who served as ward attending physicians for more than 1 month received a small stipend for the extra work involved; there was no incentive system linked to resource use, clinical outcomes, or satisfaction. To minimize bias, providers did not receive feedback about their performance or learn the results of the experiment until the end of the academic year; however, all faculty members and house staff on both services knew of the reorganization, the goals of the MCS, and the fact that outcomes, costs, and satisfaction were being measured.

Clinical outcomes included in-hospital mortality, postdischarge mortality, hospital readmissions within 10 days¹⁸ (both to our hospital and others), and functional status after discharge. Sources for these outcome data included the Moffitt-Long Hospital database, a telephone survey administered in November 1996 to all English-speaking patients discharged from the service in 1995-1996 (for other hospital readmissions and functional status), and the National Death Index. Cost and resource use outcomes included length of stay (LOS), total hospital costs, and subspecialty consultations. The cost data were obtained through the Transition Systems Inc database (Transition Systems Inc, Boston, Mass), which calculates the expense associated with the use of each product or service (in terms of cost per charge or per current procedural terminology code) within each hospital department. No physician costs or charges are included. All costs are expressed in 1995-1996 dollars. Patient satisfaction was assessed by the November 1996 telephone survey. The patient survey was approved by the UCSF Committee on Human Research. Physician satisfaction and/or education was assessed by a written survey administered to house staff and faculty rotating off the service each month. The surveys drew questions pertaining to overall satisfaction and educational emphasis from a number of previously validated surveys on physician satisfaction and education.^{19 - 22}

Differences in faculty and patient characteristics and responses to the house staff and faculty surveys were tested using t tests for continuous variables and χ² analyses for dichotomous variables. Data that were not normally distributed were analyzed using nonparametric methods. We performed multiple linear regression analyses to examine differences in continuous outcome variables (ie, hospital costs, LOS, and patient ratings of health) and multiple logistic regression analyses to examine differences in dichotomous outcome variables (eg, readmissions and mortality). All regression analyses were initially performed including only the year of admission and the admission service (MCS vs TS) as independent variables. To control for differences in patient characteristics and case mix across the different years and services, subsequent analyses also included patient age and sex, payor, and both the highest-frequency diagnosis related groups (DRGs) (89, 489, 395, 174, 88, 79, 296, 475, 320, and 96) and highest-cost DRGs (483, 76, 416, 144, 20, 202, 468, 415, 488, 14, 423, 240, 207, and 467). Analyses using DRG weights as the case-mix adjuster produced essentially the same results, but did not account for as much of the variation in the dependent variables as the high-frequency and high-cost DRG adjustments. Cost and LOS outliers (more than 3 SDs above the mean) were truncated by reclassifying their data back to 3 SDs above the mean. Sensitivity analyses reclassifying outliers to the 99th percentile, using a log-transformation, or dropping patients whose costs or LOS were more than 3 SDs above the mean produced essentially the same results.¹

Between July 1, 1995, and June 30, 1996, there were 1623 admissions to the medical service at Moffitt-Long Hospital, compared with 1707 admissions in 1994-1995 and 1567 in 1993-1994 (Table 2). There were no significant differences in age, sex, and insurance, but there were a few differences in DRGs between the MCS and the TS, and between the MCS and prior years.

There were no significant differences in major clinical outcomes between the MCS and the TS, or between the MCS and prior years (Table 3). Two hundred seventy-five patients (54% of eligible patients with reliable contact information and 84% of patients actually contacted) agreed to participate in the telephone survey. Postdischarge functional status, assessed during this survey, was no different in MCS vs TS patients.

The average adjusted LOS on the MCS (4.3 days) was significantly lower than on the TS (4.9 days; P=.01), and lower than during the previous year (5.0 days in 1994-1995; P<.001 compared with MCS) (Table 4 and Figure 1). The average adjusted hospital cost was also significantly lower on the MCS vs the TS ($7007 vs $7777; P=.05) and vs 1994-1995 ($8078 in 1994-1995; P=.002 compared with MCS) (Table 4 and Figure 2). The overall adjusted cost of care on the medical service in 1995-1996 was $11448129, a savings of $582890 compared with the average cost of care on the TS extrapolated to all admissions that year. Additional multivariate analyses demonstrated that most of the cost reduction was explained by the shortened LOS (data not shown).

The proportion of patients receiving subspecialty consultations fell from 42.1% in 1993-1994 to 32% in 1995-1996. There were no differences in the frequency of consultation on the MCS and the TS, although there was a trend on the MCS toward fewer total subspecialty consultations for patients who had at least 1 consultation (Table 4).

The proportion of patients discharged to a skilled nursing facility rose from 5.9% in 1993-1994 to 9.1% in 1995-1996. Despite their earlier hospital discharges, MCS patients were no more likely than TS patients to be discharged to a skilled nursing facility (Table 4).

There were no significant differences in patient satisfaction between MCS and TS patients. Ninety-eight percent of patients discharged from the MCS would "definitely or probably recommend" hospitalization at the UCSF, compared with 93% of TS patients (P=.08). Ninety-six percent of MCS patients (vs 91% of TS patients) were "extremely or somewhat satisfied" with "their main physician's care" in the hospital (P=.13). Ninety-one percent of MCS patients (vs 86% of TS patients) felt they were ready to leave at the time of hospital discharge (P=.25).

Fifty-five (56%) of the interns and residents completed surveys. Sixty percent (15/25) of the house staff on the MCS were "very satisfied" with their experience on the wards vs 33% (10/30) of TS house staff (P=.06), while virtually all house staff on both services were either "satisfied" or "very satisfied" (MCS 96% vs TS 100%; P=.46). Faculty surveys, completed after two thirds of ward months, showed faculty members to be equally satisfied whether on the MCS or the TS (73% [11/15] of the MCS faculty "very satisfied" vs 59% [10/17] of the TS faculty; P=.47); all faculty members on both services were either "satisfied" or "very satisfied."

Forty-seven percent (7/15) of MCS faculty strongly agreed that they emphasized cost-effectiveness (vs 12% [2/17] of TS faculty; P=.05). Both groups felt that they emphasized pathophysiology to the same degree. Fifty-three percent (8/15) of MCS faculty strongly agreed that their presence made an important difference in the quality of care (vs 29% [5/17] of TS faculty; P=.28), while 27% (4/15) of MCS faculty (vs none [0/17] of the TS faculty) said the same about the cost of care (P=.04).

Similarly, MCS house staff stated that cost-effectiveness "was emphasized very much" (68% [17/25] vs 13% [4/30] of TS house staff; P<.001) and that they "learned very much" about cost-effectiveness (36% [9/25] vs 3% [1/30], P=.02). There were no significant differences between MCS and TS house staff on similar questions regarding pathophysiology, practice guidelines, evidence-based medicine, or preparation for practice.

There were no significant differences in costs or LOS based on the number of months worked as attending physicians. For example, mean costs per case for attending physicians who had worked 2 or more months were $7128 ± $7463, as compared with $7355 ± $7786 for attending physicians who had worked for 1 month (P=.82, by Wilcoxon rank-sum test). Similarly, mean LOS was 4.49 ± 3.74 days for 2 or more months worked as attending physicians, as compared with 4.64 ± 4.14 for 1 month worked as attending physicians (P=.64). Attending physician characteristics such as prior year's hospital costs and LOS (for those attending physicians with pre–1995-1996 data), generalist vs specialist, years as a ward attending physician, or sex had no impact on the adjusted cost model.

Patients cared for on our MCS had significantly shorter LOS, lower hospital costs, and equivalent major clinical outcomes when compared with both historical controls and a concurrent control group cared for on our TS. In addition, overall patient, house staff, and faculty satisfaction on the MCS and TS were equivalent. The MCS house staff felt that they learned more about cost-effectiveness, but no less about pathophysiology. Of note, the MCS improvement in efficiency occurred on a medical service that was already relatively efficient prior to the reorganization (in 1994-1995, our medical service's LOS was the fifth lowest in a national sample of 57 academic medical centers [Linda Ono, BA, written communication, 1997]).

Patients were assigned to the MCS or TS based on alternating days of admission. Because there were 7 days in the week, MCS and TS admitting days changed each week. Other than the small number of fully capitated patients who were preferentially admitted to the MCS (and were therefore excluded from the analysis), all other patients were allocated in a fashion that, we believe, approximated randomness. Further support for this assertion can be seen in the strikingly similar demographic characteristics and case mix of patients on the 2 services.

If anything, we believe that our study design served to narrow differences between the MCS and TS rather than to accentuate them. First, although the TS might be considered a "usual care" arm, TS house staff and faculty were aware of the nature of the reorganization and the fact that costs, outcomes, and satisfaction were being measured. Second, although MCS and TS attending physicians' rounds were conducted separately, there were ample other opportunities for information to be exchanged.^{23 - 24} For example, an MCS resident presenting a case at residents' report might share some insights into cost-effectiveness (learned in attending physician's rounds) with his or her TS colleagues. Finally, the formation of the MCS itself probably improved the quality of the TS by decreasing the number of available slots for attending physicians who worked 1 month per year.

The pressures of managed care and an increasingly competitive health care marketplace are driving a new organization for inpatient care, one in which primary care physicians (PCPs) refer their hospitalized patients to separate inpatient physicians, a group we have called hospitalists .²⁵ Anecdotal reports indicate that hospitalists may provide more efficient inpatient care than the PCPs themselves.^{26 - 29} Although 1 part of our reorganization involved the use of attending physicians who served in this role more often, our study provides no direct evidence that the MCS increased efficiency was mediated through this additional time spent. However, our power to detect such a relationship was limited; relatively few of the MCS attending physicians met our definition of hospitalist (at least 25% of time as a dedicated inpatient physician),³⁰ and these individuals did not gain their increased experience until late in the year (their third and fourth months of work as attending physicians). Moreover, our staff medical service has always accepted handoffs from PCPs, so our study cannot evaluate the impact of a change from a system in which PCPs remain the inpatient physicians-of-record to a hospitalist-based system.

The attending physicians on the MCS were not chosen randomly. Although it is reasonable to question whether the improved efficiency on the MCS was due to cherry-picking the most efficient attending physicians, we do not believe this to be the case. The MCS and TS faculty members were similar in terms of sex and generalist vs specialist mix, although the MCS faculty members had been on the faculty for fewer years. Moreover, when MCS faculty members with experience in the prior years were analyzed, they were not found to be more efficient during these prior years than were their colleagues who ended up on the TS.

It is possible, perhaps likely, that the key portion of the intervention was earlier faculty involvement and increased commitment to improving inpatient care. Although the vast majority of patients were seen by their outpatient or emergency department physician on the day of admission, most inpatient attending physicians previously were not involved until the morning following admission. It may be that the average half-day shortening of LOS was a result of this attending physician's involvement in care about a half-day earlier on the MCS. These earlier encounters may have led to earlier testing, consultation, and initiation of treatment.

We also gave MCS teams a mandate to "improve quality and decrease costs." It is difficult to quantify the relative contribution of this particular intervention, but experience in other quality improvement activities has demonstrated that statements of purpose are important.³¹ Moreover, although there were no direct incentives to save money or improve quality, MCS faculty may have been particularly motivated to do so since the departmental leadership was behind the initiative and the ultimate future of the MCS was predicated, in part, on its effectiveness.

In the end, we cannot identify with certainty which part of the MCS intervention led to our results. Future research should consider whether the key factor in improving efficiency is increased faculty experience (eg, multiple months of work as an attending physician per year), earlier and more intensive faculty involvement and commitment to inpatient care, greater use of clinical guidelines, or a mandate for change.

The 2 constituencies most concerned about the MCS when we began were our house staff, who worried that faculty members who attended more often and became involved earlier would threaten their autonomy, and our subspecialty consultants, who worried that MCS teams would consult less often in an effort to cut costs. Our results demonstrate that the residents were at least as satisfied on the MCS as on the TS, if not more so. We believe that these results are a tribute to the skill of the MCS faculty in providing meaningful input and education to house staff while respecting residents' needs to formulate their own tentative treatment plans.

Subspecialty consultants worried that the MCS teams, striving for cost-consciousness, would not consult as often, thereby compromising both the quality of care and our educational mission for house staff, students, and subspecialty fellows. In fact, the number of patients receiving consultations on the MCS and TS teams was roughly equivalent, which reassured our specialists.

Work as a medical service attending physician was previously an obligation for the majority of faculty, many of whom spent the bulk of their time engaged in research activities.¹ As serving as a ward attending physician increasingly becomes the domain of highly skilled and clinically experienced faculty members interacting closely with patients and house staff from the time of admission, the opportunities for triple threats to participate in the clinical and educational activities of the medical service may diminish. It will be critical for academic departments to find other ways for talented research faculty to interact with house staff and students as opportunities to work as inpatient ward attending physicians decrease.

Perhaps the most important lesson we learned was related to the process of change in an academic medical center. In our case, changing the culture of our medical service was facilitated by our commitment to measure all of the important variables, including cost, quality, and patient and resident satisfaction, and to base further decisions on the results of these measurements. The continued evolution (and, we think, improvement) of our service has been made much easier by everyone's knowledge that decisions about change will be evidence based.