INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study was undertaken to assess the comparative outcomes and utilization of resources between a university-based team of physicians and a comparably specialized team of private practitioners in a long-term acute care facility (LTAC). Prior investigations comparing teaching and nonteaching hospital settings have concluded that medical care provided by academic services is often of high quality but carries extra expense or is hampered by inefficiency (1). Zimmerman and colleagues (11) found that risk-adjusted hospital death rates and length of stay were similar between teaching and nonteaching hospital intensive care units; however, use of resources was significantly greater in teaching hospitals. The focus of this study was the relationship of the organization of the two services to outcome measures and measures of efficiency of care.

Approximately 3% to 6% of patients admitted to an acute hospital intensive care unit (ICU) require prolonged intensive care because of either complex pulmonary or cardiac disease or other complex multi-system problems (12, 13). These patients consume a disproportionate amount of intensive care resources (14). In recent years, numerous LTAC facilities have been constructed specifically to care for these patients, often at lower costs than at acute hospitals (15, 16). Most LTAC facilities provide a full range of intensive care services, such as mechanical ventilation and weaning, vasopressor support, hemodialysis, and minor surgery. patients requiring more complex diagnostic testing or major surgical intervention require transfer to an acute facility.

We studied the outcomes of critically ill patients in an LTAC, half of whom received their care from community-based physicians (CB), and the other half from university-based clinicians and fellow trainees (UB). Both physician groups were composed entirely of specialists in pulmonary and critical care medicine. Patients were assigned arbitrarily to the care of one of the two physician groups. The UB group provided full-time coverage of their patients at the LTAC with salaried attending physicians and fellow trainees. The CB group provided coverage with physicians typically reimbursed as fee-for-service, and were assisted by fellow trainees only in emergencies. We compared the major outcomes of the patients cared for by these two physician groups, including length of stay (LOS), time to liberation from mechanical ventilation (MV), likelihood of liberation from MV, and mortality during the 1-yr study period. We also examined differences in practice behavior, including utilization of consultation services, utilization of ICU beds, frequency of withdrawal of care, prevalence of do not resuscitate (DNR) orders, frequency of liberation of patients who did not survive to discharge, and simulated physician reimbursement for care provided.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Setting

The LTAC we observed was a free-standing facility that was part of a national corporation. The hospital, located in a large urban community, had no association with any teaching hospital or acute care facility. Patients requiring a prolonged course of mechanical ventilation or other critical modality were accepted, regardless of illness severity, if stable for transfer. Patients requiring more than 10 cm H₂O positive end-expiratory pressure (PEEP) or inspired oxygen of more than 60% were screened prior to admission by the medical director. The hospital had a seven-bed ICU with full monitoring capabilities and 68 ventilator-capable floor beds.

The UB service included seven attending physicians and fellow trainees. The attendings and fellows rotated onto the service in 1-mo blocks. These physicians were salaried by the university and were either board-certified or board-eligible in pulmonary and/or critical care medicine. The physicians in the UB group had been in practice for an average of 5 yr (range, 1-15). The CB service consisted of nine physicians belonging to three private practice groups, all of whom had either direct or once removed fee-for-service reimbursement. All CB physicians were board-certified or board-eligible in pulmonary and critical care medicine, and they had been in practice for an average of 8.7 yr (range, 1-18). One of the three private groups was responsible for covering the CB service each week; 24-h in-house emergency coverage was provided for both services by the fellow trainees.

Upon admission to the LTAC, patients were assigned to either the UB service or the CB service in a nonselective manner on an alternating basis by the admissions coordinator unless the patient had been managed previously by one of the physician groups at another hospital. Those patients previously managed by one of the physician groups were automatically admitted under the same physician group's care. If the census on one service became disproportionately large, the smaller service would receive additional previously unknown patients out of the alternating rotation. During their hospitalization, the patients received the same nursing and therapist care and organizational support from the hospital. All weaning efforts were guided by daily physician orders.

Study Patients

We screened all patients (185) transferred to the LTAC hospital between August 1, 1995 and July 31, 1996 from ICUs or step-down units at 37 acute care hospitals in the area, who had not previously been an inpatient at the LTAC hospital. To limit selection bias and to ensure that all physician-patient relationships began on the day of transfer, we eliminated all patients who had received care at the transferring facility from either the CB group (51 patients) or the UB group (16 patients). The final sample was 118 patients (42 CB patients, 76 UB patients). Eighty-eight percent of patients were covered by Medicare, 22% had Medicaid, and 90% were covered by either Medicare or Medicaid.

Clinical Characteristics and Outcome Measures

All clinical information was abstracted by three trained chart reviewers from medical charts housed at the LTAC facility, including age, gender, diagnoses leading to ICU admission, events at transferring hospital, survival and discharge disposition, hospital and intensive care unit LOS, dates of weaning attempts, presence and date of DNR orders, and evidence of either use of cardiopulmonary resuscitation (CPR) at death or withdrawal of MV or dialysis, despite continued dependence on these modalities.

Acute Physiology and Chronic Health Evaluation (APACHE) II scores on LTAC admission were calculated according to the method described by Knaus and coworkers (17). Successful liberation from MV was defined as being independent of MV for seven consecutive days and nights, regardless of outcome after that period. An attempt at breathing without MV support was defined as attempted unassisted ventilation planned for at least 1 h. Functional status upon LTAC admission was determined from physical and occupational therapist records.

Survival at 1 yr was assessed by follow-up telephone interviews of patients or families performed by a nursing case manager or by examination of Medicare records. Follow-up at 1 yr was 100%. Inter-rater reliability was confirmed using kappa statistics and measures of correlation.

Simulated Medicare Reimbursement for Care Provided

We simulated the amount of reimbursement a physician received for care provided using standard Medicare provider reimbursements for Current Procedural Terminology (CPT) codes regarding management of critically ill patients (18, 19). Specifically, any ICU day was designated "critical care, evaluation, and management, first hour" (CPT 99291) and reimbursed at $196.18/d. Any day on the medical floor on MV was designated both "ventilation assist and management; subsequent day" (CPT 95657), reimbursed at $54.12/d, and "subsequent hospital care; straightforward/low complexity" (CPT 99231), reimbursed at $37.65/d. Any day on the medical floor not on MV was designated "subsequent hospital care; straightforward/low complexity" (CPT 99231) and reimbursed at $37.65/d.

Statistical Analysis

All p values are two-sided. Distributions of measures were assessed, and appropriate statistical tests were employed. Specifically, patients' and physicians' characteristics were compared using Student's t test for continuous variables, and Pearson's chi-square statistics for categorical variables. Lengths of stay, time to liberation from MV, and time until first attempt at liberation from MV were compared using the Student's t test and the Wilcoxon test (20). For testing the time to liberation, patients still mechanically ventilated were considered censored at the time of death. Alternative formulations, such as ignoring these patients, or censoring them after all events, did not substantively change the results. Number of consultants, LOS in the ICU, and differences in physician's reimbursement were tested using the Wilcoxon-Mann-Whitney rank sum test (21). Liberation from MV, death at the LTAC, withdrawal of care, prevalence of DNR orders, use of CPR at death, prevalence of liberation prior to death, and patient disposition were compared using Pearson's chi-square statistics, and Fisher exact test for infrequent events. Survival of patients between physician groups was compared using the log-rank test (20), where follow-up was 100%. Version 5.0 of STATA (Stata Corporation, College Station, TX) was used for all analyses.

All analyses were repeated, including data on the 67 patients who met our exclusion criteria. These analyses yielded substantively similar results, many of which had more statistical significance than the results we present here. The study was reviewed and approved by the institutional review board of the University of Chicago.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patient Comparisons

Baseline characteristics (Table 1) for the 118 study patients were comparable between the UB and CB groups. The two groups were similar in age, APACHE II score, serum albumin, hospital days prior to admission, functional level, gender, and alveolar-arterial oxygen gradient. Seventy-four percent of the UB patients and 71% of the CB patients were mechanically ventilated at admission to the LTAC, and APACHE II scores for those subgroups were similar. Percentages also were similar for those who had bacteremia prior to admission, and those who were acutely ill from lung disease, heart disease, surgical disease, or neurologic disease.

Discharge and Liberation from Ventilation at the LTAC

While death during hospitalization (47% of the UB group patients versus 43% of the CB group patients) and the percentage of patients dying in the first month (24% versus 21%, respectively) were similar, the average LOS was marginally (19%) less for the UB group (57 versus 70 d, p = 0.19) (Table 2). Those patients who survived until discharge stayed 23% fewer days if cared for by the UB group (68 versus 88 d, p = 0.12). For patients who were mechanically ventilated, the median time to liberation from the ventilator of 33 d for the UB group was substantially less than 51 d for the CB group (p = 0.02); 46% of the UB group's patients were ultimately liberated from MV during their LTAC stay, while only 30% of the CB group's (p = 0.14) patients were liberated from MV. After their LTAC stay, patients were discharged to skilled nursing facilities (50%), home (20%), rehabilitation hospitals (17%), and acute care hospitals (13%) with equal frequencies by each group (p = 0.74).

Resource Utilization and Practice Variations between the CB and UB Groups

We found marked variations in both resource utilization and therapeutic strategy (Table 3). The CB group utilized approximately one more consultant physician per patient than the UB group (3.5 versus 2.75, p = 0.06). Patients treated by the CB group also had marginally longer stays in the ICU (8.7 versus 7.7 d, p = 0.28). The CB physicians also waited a substantially longer time prior to the first trial of breathing without any MV assistance. The median time to the first trial was 9 d for the UB group versus 18 d for the CB group (p = 0.05).

The groups also varied markedly in their approaches to end of life care. Only 33% of the patients of the CB group ever had DNR orders written to prevent resuscitation, while 59% of the UB group had DNR orders (p < 0.01). Perhaps as a consequence of this policy, 50% of the CB patients received CPR at the time of death compared with 14% of the UB patients (p < 0.01). There was also a trend toward the UB group withdrawing life-sustaining therapy more frequently, doing so with 12% of their patients, while the CB group only withdraw life-sustaining therapy in 2% (p = 0.09). The UB group liberated 17% of their patients from MV who were never well enough to be discharged, while the CB group liberated none of their patients before their death at the LTAC (p = 0.15). The UB patients who were liberated from MV but ultimately expired at the LTAC averaged 42 d of ventilator-free life before expiration.

Physician Billing and Bed Usage at the LTAC

Figure 1 illustrates the comparative use of bed days by the two physician groups. The average ventilated patient cared for by the CB group spent 8.7 d in the ICU, 51.7 d on the medical floor on MV, and 8.9 d on the medical floor off MV. In contrast, the average ventilated patient cared for by the UB group spent 8.3 d in the ICU, 32.7 d on the medical floor on MV, and 18.1 d on the medical floor off MV. Similar differences were seen in the care of nonventilated patients: CB patients spent 7.3 d in the ICU and 65.7 d on the medical floor; UB patients spent 4.25 d in the ICU and 50.6 d on the medical floor. Based on these differences in bed and ventilator utilization, we estimated (Table 3) that physicians from the CB group were reimbursed 29% more per patient stay than the UB physicians ($5,961 per patient versus $4,633 per patient, p = 0.12). For patients who survived to discharge, we estimated that the CB group was reimbursed 46% more per patient stay than the UB group ($6,797 versus $4,651, p = 0.03).

View larger version (30K): [in this window] [in a new window]   Figure 1.   Average utilization of bed days: intensive care unit, medical floor while on mechanical ventilation, medical floor while breathing spontaneously, stratified by need for mechanical ventilation upon admission to the long-term acute care hospital and by care provider. UB = university-based; CB = community-based.

Patient Survival between the UB and CB Groups

Despite differences in ICU and MB utilization, and notably different approaches to the use of life-sustaining therapies, we found no evidence that survival was different between the two patient groups (p = 0.79 by the log-rank test). As already presented, we found no difference in survival to 30 d or to discharge from the LTAC. At 6 mo, 37% of the UB patients and 38% of the CB patients were alive. At 1 yr, survival (95% confidence interval) was 29% (19-39%) for the patients of the UB group, and 18% (8-31%) for the patients of the CB group. The similar survival over the first year between these two patient groups is displayed in Figure 2, a Kaplan-Meier plot stratified by care provider.

View larger version (17K): [in this window] [in a new window]   Figure 2.   Comparative 1-yr total survival for long-term acute care patients (p = 0.79). The gray line indicates patients cared for by university-based (UB) providers. The black line indicates patients cared for by community-based (CB) providers.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We examined outcomes for patients managed by UB and CB physicians in a setting where we had strong evidence that patients were allocated to different providers' care by an admissions coordinator in a nonbiased way, and where we had statistical evidence that the patients were similar between the two groups at admission (Table 1). We found that patients cared for by a UB group assisted by fellow trainees in pulmonary and critical care medicine were more likely to have shorter LOS, more likely to be quickly and successfully liberated from MV, more likely to have life-sustaining therapy withdrawn, and less likely to receive CPR at the time of death. In contrast, we found that the patients cared for by a CB group of practitioners were more likely to have long LOS in both the ICU and in the hospital (Figure 1), to be seen by more consultants during their stay, and to accrue greater service charges. Despite these differences in practice, we did not detect a difference in short- or long-term survival between the patient groups (Figure 2). Given these observations and our knowledge that the patients are cared for in the same facility and by the same nurses and ancillary staff, we conclude that providers substantially influence patient outcomes at this LTAC, including liberation from MV, duration of MV, and LOS in the hospital.

The differences between groups in success and time to liberation from MV that were documented in this study have important implications for patients. Shortening time on MV decreases risk of ventilator-associated complications (22), lowers costs, and undoubtedly improves patients' quality of life.

The differences in patient outcomes between the UB and CB services are also important with regard to health care costs. We found that the patients cared for by the CB group spent 13% more days in the ICU, 46% more days on MV, and 23% more total days at the LTAC hospital. Each of these represents major cost issues for the hospital and for Medicare reimbursement. LTAC hospital reimbursement by Medicare is determined according to guidelines from the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA) (23), but a given hospital's TEFRA rate is determined by hospital costs. We estimated that the CB group would have received 29% more in fees from Medicare ($5,961 versus $4,633) for providing care to each patient, and used one more consultant on each case than UB physicians.

Patterns of comfort care by the two groups also warrant special attention, as care provided at LTACs is often end of life care. Notably, the UB physicians were more likely to write DNR orders and withdraw support in terminally ill patients. They were also more likely to liberate patients from MV, even if these patients' conditions never warranted discharge from the LTAC. These findings confirm recent reports that demonstrate that some physicians in critical care settings are more likely to write DNR orders than in previous years (24), as well as a study that suggests that physicians at teaching hospitals may withdraw vasopressor and MV support more frequently than CB providers (25). There remains significant practice variation, even when formal systems of physician-patient communication exist (26, 27).

The design of this study did not provide clear explanations for the differences in patient outcome between patients receiving the teaching and nonteaching services. There may have been differences in technical ability, but we found no evidence of differences in overall survival between the two groups. All of the physicians in our study were clinically active subspecialists in pulmonary and critical care medicine, and they did not differ significantly in years of experience.

An alternative explanation rests in the observation that the UB and CB groups are different in the availability and continuity that they provide to patients at the LTAC. The UB group rotated monthly, while the CB group rotated weekly. In addition, the UB group had a full-time dedicated attending physician and a fellow trainee who spent the majority of his or her day at the LTAC. Many of the CB physicians care for patients at multiple facilities simultaneously. Increased physician presence and continuity might have facilitated the more aggressive weaning trails and use of fewer consultants, and generated more communication between physicians and patients about end of life care (26, 27). This improved communication may account for the greater number of UB patients having DNR orders and withdrawal of life-sustaining therapy.

It is important to emphasize some limitations of this study. First, this is a study of a single LTAC and of two specific groups of physicians during a single year. As such, it may not be indicative of other acute care settings or of all physicians caring for patients at LTAC hospitals. Second, although patients were admitted to both services in an apparently nonbiased manner, the randomization was not formal. Third, when we simulated physician billing, we could not account for the extra cost of the pulmonary fellow trainees.

Our results provide evidence that patients suffering from prolonged critical illness experience significant differences in time to liberation from MV, withdrawal from life support, and cost of care when cared for by different intensivist physicians within the same hospital. We have also demonstrated that a teaching service staffed by academic physicians in a community LTAC hospital can provide health care that is cost-efficient as well as high quality.