What effect does physician "profiling" have on inpatient physician satisfaction and hospital length of stay?

The cost of providing health care to hospitalized patients is enormous. While recent data on national trends in hospitalization show that the average length of a hospital stay in the United States has declined from 7.3 days in 1980 to 4.9 days in 2000 [1], spending on hospital care increased 9.5% in 2002 to $486.5 billion, marking the first time the rate of hospital spending outpaced the overall health care spending rate of growth since 1991 [2]. The Department of Veterans Affairs ("VA") operates one of the nation's largest healthcare systems, providing health care for eligible veterans. Facilities operated by the VA traditionally have had significantly longer lengths of stay than those of the private sector [3–5] and, while there is evidence that the gap in length of stay between VA facilities and the private sector may be narrowing [6], there is still increasing pressure to reduce costs. As hospital spending continues to grow and as reimbursement for hospital expenses has moved towards the prospective payment system, hospitals have a major incentive to decrease resource utilization, including hospital length of stay. Various methods have been proposed to control hospital costs and decrease inpatient length of stay, including utilization review, critical pathways, physician education, and the use of hospitalists [7–16].

Physician profiling is another process that has been used in an attempt to constrain spending and reduce inpatient length of stay [17–19]. Physician profiling is the practice whereby the practice pattern of a single physician or a group is expressed as some measure of the use of resources and/or length of stay during a defined period for the population served. The resulting profile is then compared with a norm that is either based on practice (such as profiles of other physicians) or on standards (such as practice guidelines) [20]. According to results of the 2000–2001 Community Tracking Study Physician Survey from The Center for Studying Health System Change, more than a third of physicians have reported that practice profiles had influenced their practice of medicine [21]. In addition, results indicate that physicians may view profiling in a more positive light than before. Nearly half of physicians surveyed, however, still maintained a mixed or negative view of physician profiling [21]. Nevertheless, some believe that comparing a physician's resource consumption level to a benchmark figure may encourage the physician to conserve resources to avoid classification as a resource consumption outlier.

Physician profiling has had mixed success [17–19, 21–23]. We thus evaluated whether physician profiling affects physician satisfaction and hospital length of stay, while assessing physicians' views concerning hospital cost containment and the quality of care they provide.

Inpatient length of stay is generally the single greatest factor contributing to overall hospital costs [28]. The average length of a hospital stay has declined over the past decade; however, hospitals still face pressure to reduce costs by restricting length of stay for hospitalized patients. While there are significant geographic variations in the use of inpatient hospital services in the Veterans Health Administration [29], lengths of stay at VA hospitals have historically exceeded those of community hospitals [30, 31]. Recent analyses, however, suggest that the disparity in length of stay between VA facilities and non-federal hospitals is lessening [32].

The decrease in inpatient length of stay may reflect, among other factors, the recent use of more aggressive utilization management techniques, the shifting of care to outpatient venues [33], and the hospitalist model of care [7, 9, 11–13]. Since 1995, Veterans Health Administration facilities have undergone extensive restructuring and realignment in order to improve healthcare service delivery and administrative operations [34]. However, as the veteran population ages and new veterans enter the Veterans Health Administration system, the challenge of increasing efficiency will become even more important. All health professionals, however, are challenged to reduce unnecessary stays and services without sacrificing quality of care. As hospitals face pressures to contain costs, it is important to understand the usefulness of various methods of reducing resource utilization, including patient length of stay.

Previous studies evaluating the effects of physician profiling are instructive in placing our results in proper context. Ross et al. [18] undertook a study in which physicians were sent profile reports that included information regarding the number of patients and lengths of stay for each DRG the profiled physician had admitted during two-month periods. Each profiled physician was also sent a letter that included benchmark figures for lengths of stay for the critical pathways that were in use at the study facility. The results suggest that profiling had a strong influence on physicians' behaviours, resulting in a decrease in average length of stay. Balas and colleagues [23] performed a multilevel meta-analysis designed to assess the clinical effect of peer-comparison feedback intervention (profiles) in changing practice patterns. They found physician profiling to have a statistically significant though minimal effect on the utilization of clinical procedures. Evans and colleagues [17] analyzed the effectiveness of one hospital's introduction of physician length of stay profiling. Controlling for physician, DRG, and patient severity level, data for patients treated by 400 physicians over 42 months were analyzed, including both pre-profiling and post-profiling periods. There was a significant increase in the percentage of physicians who achieved the length of stay benchmark after the introduction of profiling.

It is also important when interpreting the findings of the current study to consider several potential limitations. First, since the intervention site consisted of a single VA medical center, some may question the external validity of our findings. However, our intervention was not "VA-specific" in that a similar type of intervention could readily be implemented in virtually any type of hospital. Second, the cause of the significant decrease in length of stay may have less to do with profiling and more to do with other issues. Specifically, as Evans and colleagues [17] have pointed out, results from profiling studies may also reflect the influence of other factors, such as physicians substituting more procedures for the reduction in length of stay in response to concerns such as exposure to legal liability. Finally, since we did not conduct a randomized trial, confounding variables – such as case-mix changes or co-interventions – could have influenced our results. We did, however, adjust for case-mix and temporal trends at other institutions in our statistical analyses.

In this study, we describe an investigation of physician profiling as a method of decreasing inpatient length of stay. While the majority of attending physicians agreed that trying to contain costs is the responsibility of every physician, nearly half expressed negative feelings toward being profiled prior to the intervention. A much smaller proportion of physicians who were profiled, however, viewed the process negatively; being profiled had no apparent effect on satisfaction during profiling. Additionally, profiling of attending physicians appeared to have no significant effect on senior residents' perception of the degree of involvement by their attending physicians between pre- and profiling years. Finally, the majority of both senior resident and attending physicians reported that they provided extremely high quality care. Although survey responses suggested that profiling would not cause physicians to discharge patients early, after taking into account the background trend and relative to the non-profiled sites, length of stay at the profiled site decreased by one-third of a day in the profiling year.

Despite limitations discussed previously, our results raise several important questions. While anticipation of being profiled was viewed as negative by a large number of attending physicians, most of those who were profiled did not view the process negatively. It may be that the unknown (i.e., the possibility of being profiled) is more threatening than the actuality of being profiled. In addition, while it is difficult to determine to what extent the reduction in length of stay was due to the profiling, it is noteworthy that physicians' reported intent that being profiled would not cause them to decrease patient length of stay differed from the results (i.e., a reduction in length of stay while being profiled). Our findings indicate that profiling may be an effective method of decreasing length of stay without affecting physician satisfaction.