In academic medical centers, multidisciplinary work rounds usually consist of individuals working in parallel rather than working together in interdependent teams. These working groups usually fail to manage time efficiently, detracting from teaching as well as interfering with completion of patient care tasks.1 In recent years, Toyota Production System (TPS)—a model for improving quality and efficiency in manufacturing—has been applied to health care with great success.2 To improve the teamwork and efficiency at the University of Florida Health, Shands Hospital, we initially attempted to apply TPS to design a system for multidisciplinary work rounds. We experienced resistance from both residents and faculty, however, receiving complaints like “Patients aren’t cars.” We subsequently realized that championship athletic teams succeed because they apply the same principles as TPS,3 and to improve the likelihood of adoption, we recast TPS principles as athletic analogies: (1) Playbooks describe in detail the role of each player; TPS uses protocols. (2) Understanding who is passing and receiving (e.g., the quarterback knows how to supply the ball to the running back to prevent fumbles); TPS emphasizes effective customer–supplier relationships. (3) Game Films, which the coaches and players use to continually review player performance, allows them to examine what went well and what could be improved. Similarly, TPS teaches scientific inquiry to encourage hypothesis-based improvement.
We predicted that adopting a rounding system based on athletic principles would shorten the duration of rounds, shorten hospital length of stay, reduce 30-day readmissions, and improve physician and medical student satisfaction without negatively affecting patient satisfaction.
The pilot study, a two-phase prospective trial, compared medical teams who underwent training and weekly supervision (experimental team, see below) versus teams who received no supplemental training (control team). The first phase of the trial occurred from February to July 2009. From August 2009 to August 2011, our trial was placed on hold while the principal investigator (F.S.) was on sabbatical. The second phase began September 1, 2011, and ended January 31, 2012.
Both teams were assigned to the same medical unit and consisted of a faculty attending, a resident, two or three interns, three medical students, a case manager, a faculty pharmacist, and two pharmacy students. To ensure participation of the bedside nurse, the experimental group scheduled a specific time to be at each patient room. Physicians were assigned by the predetermined call schedules. Medical attendings rotated every two weeks and house staff every 30 days. During the trial period, team members only rarely were assigned to both the experimental and control teams (1 out of 16 attending physicians and 2 out of 29 residents; 0 students, pharmacists, or case managers). The control group was “blinded”; however, informal questioning revealed that many members were aware of the trial. The medical teams admitted every other day, admitting patients from the same locations, ensuring a comparable case mix.
About three or four days before starting, the principal investigator (F.S.) gave a 20-minute orientation to the experimental group, discussing TPS’s principles and the analogous athletic terms. Playbooks were introduced, which included descriptions of each team member’s role, with a player analogy: attending as coach, resident as quarterback, intern as running back, student as redshirt freshman, bedside nurses as offensive line, case manager as assistant coach, pharmacist as assistant coach. Then the key customer–supplier relationships were discussed: caregiver–patient, doctor–nurse, and doctor–case manager. Finally, the concept of having a game-film mentality (i.e., the importance of offering continual feedback and striving to continually improve) was emphasized. Team members were asked to review their individual playbooks and practice the fundamental communication protocols (problem-based presentations of daily progress presenting subjective complaints, objective findings, assessment, and plan [SOAP], and following an unexpected change in the patient’s condition: situation, background, assessment and recommendations [SBAR]) outlined on the “Gatorounds” Web site (http://gatorounds.med.ufl.edu, named after the alligator, our college mascot).
The principal investigator (F.S.) observed the experimental team weekly and made suggestions for improvement. For example, teams often failed to form a circular huddle, inadvertently excluding the bedside nurse. In several teams, the senior physician dominated the management discussion, rather than asking guiding questions, and created a steep hierarchy of authority. These findings were shared either publicly or privately, and appropriate changes were encouraged. He and the case manager (M.L.) graded attending adherence to their roles using a 30-point, 6-category, 5-point grading scale (1 = worst, 5 = best): (1) facilitating horizontal communication, (2) trouble-shooting, (3) involving nurses and case managers on rounds, (4) following the rounding schedule, (5) encouraging work-sharing, and (6) knowing the clinical details of the patients before rounds. In the second phase of our trial, a priori we set a performance threshold for attendings to score above 15 points, excluding teams whose attending scored below this value, as well as the matching control. These teams were excluded because in the first phase of our trial we discovered that attendings with low performance scores had longer lengths of stay and higher readmission rates (see below). Scores were not shared with the attendings.
We administered anonymous surveys to all participating physicians, medical students, and patients. All patients mentally capable of responding were surveyed biweekly using paper questionnaires handed out and later retrieved by one of the investigators. With the exception of the attendings (surveyed only in the second phase), all other participants were surveyed throughout the study period. Participants rated responses on a Likert scale (5 = strongly agree, 4 = agree, 3 = don’t care or didn’t notice, 2 = disagree, 1 = strongly disagree).
A two-tailed, unpaired, nonparametric Wilcoxon analysis was used for sample sizes of less than 30. For larger data sets, equal to or above 30, we used a two-tailed, unpaired t test. The relative length of stay (LOS) was calculated by dividing the actual LOS by the expected LOS. We calculated the expected LOS by using standardized multiple linear regression models. Models were developed for each base Medicare Severity Diagnosis Related Group and included variables such as patient demographics (age, race, socioeconomic status) and health status (comorbidities and complications), and arrival means as predictors. Cronbach alpha coefficients were calculated for all survey questionnaires to assess internal consistency, and in all instances coefficients were greater than or equal to 0.7.
Duration of rounds
The duration of work rounds was measured in the second phase of the trial, and this measure reflects the efficiency of the team. As can be seen in Figures 1A (control team) and 1B (experimental team), the experimental team performed more efficiently than the control team. While a typical control team’s rounds periodically extended to 180 minutes and the duration failed to decrease over time, the experimental team was usually able to complete rounds within the 120-minute goal and progressively improved. The mean duration of rounds ± standard error of the mean (SEM) for the experimental group (110.8 ± 2.8, n = 75) was 16 minutes shorter than for the control group (126.1 ± 3.8, n = 64) (P = .0049). The time saved by the experimental group was devoted to didactic teaching sessions for students after rounds.
Adherence to the rounding system
Adherence to the rounding system, determined by a 30-point, 6-category, 5-point grading scale, demonstrated variability during both trial periods. Despite coaching, 19% (3/16) of physicians earned scores below 15 points. They failed to follow the recommended protocols, instead using their personal systems to conduct rounds and interact with team members.
Length of hospital stay
A shorter relative LOS can be achieved by reducing the time required to complete the management plans for each patient. A value of 1 indicates the duration of hospitalization is comparable to that of other patients with the same diagnosis and demographics. During the initial trial period (February to July 2009), the mean LOS ± SEM for the experimental group (0.90 ± 0.04, n = 363 cases) and the control group (0.87 ± 0.05, n = 417 cases) were not statistically different (P = .51). However, the LOS for both groups was 25% lower than the combined internal medicine services during this time period (mean LOS = 1.20, n = 4,683 cases) (Figure 2A).
During the second phase of the trial (September 2011 to January 2012), the experimental group’s mean LOS ± SEM was significantly lower than that of the control group (0.76 ± 0.05, n = 213 cases versus 0.93 ± 0.07, n = 242 cases, respectively; P = .0105) (Figure 2A). Both the control and experimental groups had significantly lower lengths of stay than the other medical services during the same time period (1.11, n = 7,009 cases). If we had included in the analysis the lower performers (attendings with adherence scores below 15: 12.7 versus 28.7), we would have seen a smaller reduction in the mean LOS (experimental group: 0.87 ± 0.08, n = 312 cases versus control group: 1.013 ± 0.11, n = 337 cases; P = .19). The low-scoring physicians had significantly longer mean LOS as compared with high-scoring physicians in the experimental group (1.18 ± 0.17 versus 0.76 ± 0.05, respectively; P = .024).
30-day readmission rates
We considered the percentage of 30-day readmissions in part a reflection of the quality of the care provided. Compared with the control group, the experimental group readmitted 30% fewer discharges within 30 days in both phases of the trial (mean readmission percentage ± SEM: experimental group: 6.95 ± 1.29, n = 576 discharges versus control group: 9.95 ± 1.02, n = 659 discharges; P = .039) (Figure 2B). Both groups demonstrated markedly lower readmission rates (by more than 50%) compared with the mean readmission rate for all the internal medicine services in the hospital during the same periods (21.7%, n = 10,824 discharges) (Figure 2B). Inclusion of the low performers did not affect the readmission rate for the experimental group.
Faculty (attendings) and house staff (residents) in the experimental group expressed greater overall satisfaction (mean rating ± SEM) with the rounding system (attending, control group: 3.9 ± 0.6, n = 10 versus experimental group: 4.4 ± 0.4, n = 11; P = .045) (resident, control group: 3.7 ± 0.2, n = 24 versus experimental group: 4.1 ± 0.2, n = 29; P = .032). They also felt that rounds were more efficient (attending, control group: 3.5 ± 0.3 versus experimental group: 4.4 ± 0.2; P = .005) (resident, control group: 3.5 ± 0.2 versus experimental group: 4.1 ± 0.2; P = .011). Attendings in the experimental group strongly favored the adoption of a standardized rounding system compared with the control group (4.1 ± 0.3 versus 3.0 ± 0.3, respectively; P = .028). Interestingly, both the control and experimental resident groups favored the adoption of a standardized rounding system (3.9 ± 0.2 versus 4.0 ± 0.2, respectively; P = .57). Faculty in the experimental group also felt that they had more time for meaningful teaching than those in the control group (3.7 ± 0.3 versus 2.6 ± 0.4, respectively; P = .018) (Figures 3A and 3B). When asked about the importance of being able to use their own personal system for rounding, and the fear that a systems approach would interfere with their sense of autonomy, physicians in the experimental group expressed less concern than the control group (control attendings: 3.5 ± 0.2, n = 10 versus experimental attendings: 2.0 ± 0.3, n = 11; P = .0031) (control residents: 3.4 ± 0.2, n = 24 versus experimental residents: 2.6 ± 0.2, n = 30; P = .019). These findings represent the desired outcome for our experimental group.
Medical students in the experimental group (n = 23) were more satisfied with the rounding system than the control group (n = 19) (4.5 ± 0.1 versus 3.4 ± 0.3, respectively; P = .0002), rating their rounds as highly efficient (4.3 ± 0.1 versus 2.8 ± 0.3, respectively; P < .0001) and indicating that they felt more integrated into work rounds (4.6 ± 0.1 versus 4.2 ± 0.2, respectively, P = .05). Teaching evaluations were markedly higher among students in the experimental group compared with the control group (4.2 ± 0.1 versus 3.0 ± 0.3, respectively; P < .0001). The experimental group felt the teaching sessions were at the appropriate level for their knowledge base, while the control students reported that teaching was more appropriate for interns and residents (4.5 ± 0.1 versus 3.7 ± 0.2, respectively; P = .02) (Figure 3C).
There were not any significant differences in patient satisfaction between the two groups (overall satisfaction: experimental group: 4.3 ± 0.1, n = 44 patients versus control group: 4.0 ± 0.1, n = 46 patients; P = .076).
In this study, we piloted a two-phase, 11-month prospective trial (February to July 2009, and September 2011 to January 2012) to test whether a rounding system based on athletic principles would improve LOS, 30-day readmission rates, and physician, student, and patient satisfaction.
Our pilot might encourage larger, more robust investigations that include multi ple medical units and larger numbers of participants at other academic institu tions. We are presently beginning a more comprehensive two-year study that includes four general medical services in our hospital, as well as two general medical services in a smaller university-affiliated private hospital in Montgomery, Alabama.
Our study had a number of limitations. The trial took place at a single institution and included a small number of parti cipants. To ensure a comparable case mix, both the control and experimental group were selected from the same medical unit. This arrangement limited our ability to blind participants, and as a consequence we experienced a large Hawthorne effect. The control group performance with regard to 30-day readmissions (9.95%) was less than half that of the other internal medicine units during the same time period (21.7%), and LOS for both the experimental and control groups was reduced by 16% to 32% as compared with the other medical units. The improved performance of the control group may be explained by the control case manager’s active participation on work rounds. Before the trial, case managers never rounded with the teams, and case managers continued to be absent from rounds on the medical teams not participating in the trial. These findings suggest the central role of the case manager in improving patient flow and reducing hospital readmissions.
Innovators beware: Changing the way things are done is difficult, and immunity to change is a well-known barrier to improvements in all fields, including medicine.4 To learn how to more effectively lead change, the principal investigator (F.S.) suspended the trial after 2009 to participate in a one-year leadership fellowship. On resuming the study in September 2011, he was able to more effectively manage the emotional disequilibrium associated with change. By practicing empathy and focusing on positive behavior, he was able to reduce anxiety among the physicians. He also maintained a low profile and recruited other faculty and residents to champion the new rounding system. Despite these efforts, 19% of the faculty would not adhere to the rounding system. Rogers5 has observed a similar percentage of laggards in every population. For studies of microsystems change, we recommend the selective recruitment of the innovators, early adaptors, and early majority, because this strategy will increase the likelihood of full implementation and significant performance gains. Alternatively, administrators could mandate that all physicians adhere to the rounding system, a common strategy for managing laggards who do not voluntarily embrace change.
Can this system be implemented in other academic medical centers? We have found that the use of athletic analogies creates a new joy at work and quickly creates a sense of camaraderie among students, house staff, faculty, and other caregivers. We recommend naming the new rounding system after the mascot of the school (e.g., Yale University—Bulldog Rounds) or a professional team (New England—Patriot Rounds) and using the mascot as the symbol for a local Web site to promote the rounding system, as we did with Gatorounds.
We believe that the broad implemen tation of our rounding system has the potential not only to reduce costs and improve teaching but also to create positive cultural changes by emphasizing teamwork and the value of establishing protocols for conducting work—two key elements for achieving a culture of safety. Don’t our patients deserve the same efficient and high-quality systems as our championship athletic teams?
Acknowledgments: The authors would like to thank Dr. Dennis Deurelle for his helpful suggestions; Gloria P. Lipori and Paul J. Lipori of UF-Health System for assistance in obtaining length of stay and readmission data; Mia Belleville, RN, and Kelly Jacobitz, RN, for their leadership on the UF Shands 64 Medical Unit; and all the nurses on 64, as well as the many faculty, medical residents, and students who enthusiastically embraced our rounding system. The authors would also like to thank Dr. Amit Sharma of the University of Alabama, Birmingham, for being the first physician outside of our institution to implement our rounding system. Finally, the authors wish to emphasize the important role the Harvard University Advanced Leadership Fellowship (http://advancedleadership.harvard.edu/) played in providing the strategies that allowed us to complete our study. The authors have no conflicts of interest. This study was not funded by any outside sources. Dr. F. Southwick has reviewed and verified all the data and statistical analysis for this paper.