Lai, Jennifer C. MD, MBA; Montero, Alex MD, MPH; Lebwohl, Benjamin MD; Brown, Robert S. Jr MD, MPH
Residency training in internal medicine remains a largely inpatient, ward-based experience, where the trainee is exposed to patients with a variety of diseases—ranging from acute presentations, such as pneumonia and pulmonary embolism, to exacerbations of chronic illnesses including liver disease and chronic obstructive pulmonary disease.1 Traditionally at our institution, patients are admitted to a medicine service and managed by medicine housestaff under the direct guidance of either the ward attending or an admitting private attending, regardless of the acuity or complexity of the patient's condition, often with consultative input from specialist physicians.
Recently, however, external factors have had a significant impact on the educational environment within residency training, necessitating a variety of adjustments to the traditional service model.2–4 Economic pressure to reduce length of stay (LOS) has shifted care to the outpatient setting, resulting in trainees' caring for inpatients with conditions of higher acuity. Additionally, in many academic health centers, patients receiving quaternary-care interventions, including organ transplantation, have become increasingly common and represent a unique subgroup of patients with high-acuity, complex conditions who require management input from a subspecialist outside of the traditional service model. Lastly, implementation of resident work hour restrictions and increased economic pressures to see patients largely in the outpatient setting for attending physicians have potentially decreased opportunities for formalized teaching between attending physicians and residents, thus lessening residents' satisfaction with the training experience.5–7 As residency programs adjust to these competing influences, the traditional service model of patient care may no longer represent the optimal structure for delivering high-quality care while still achieving trainee educational goals.
In response to the changing training environment, we implemented the liver service—a novel ward-based educational model for internal medicine residency training—dedicated to patients with end-stage liver disease. We hypothesized that for these quaternary-care patients, a specialized housestaff team structure directly supervised by a subspecialist would improve the clinical care of a group of patients with similarly high-acuity, complex conditions while concurrently enhancing trainees' educational experience. We carried out the study reported here to determine the impact of this new housestaff service model.
Traditional housestaff service model (preintervention)
This study was carried out within New York Presbyterian Hospital–Columbia Presbyterian Medical Center, a large academic health center with a high-volume liver transplant program performing more than 120 liver transplants per year. Before the intervention, all patients requiring admission for medicine-related diagnoses (e.g., urinary tract infection, pulmonary embolus) were admitted under the traditional housestaff service (THS) model onto one of several internal medicine housestaff teams, consisting of four interns and four second- or third-year residents who rotated every four to five weeks. These teams were supervised by ward attendings, both generalists and specialists, who rotated on a monthly basis. Patients were allocated to these housestaff teams on the basis of their likely admitting diagnosis—patients with congestive heart failure exacerbation were admitted to a general cardiology service; patients with pneumonia were admitted to a general medicine service. Under the THS model, patients presenting with advanced liver disease were admitted to any one of several internal medicine services but received primary management input outside of the THS model from a consulting multidisciplinary hepatology team. The hepatology team rounded and met independently of the internal medicine housestaff team but communicated with the housestaff team via daily progress notes in the patients' charts or via periodic, informal verbal communication as needed.
Specialized housestaff service model (intervention)
In June 2004, our internal medicine residency training program created the liver service, using the specialized housestaff service (SHS) model. This service was dedicated to inpatients presenting with decompensated liver disease requiring a referral for liver transplantation or with advanced liver disease currently undergoing a liver transplant evaluation, regardless of admitting diagnosis (e.g., pneumonia). An attending hepatologist, who rotated every two weeks, and his or her hepatology team—consisting of a gastroenterology fellow, a liver transplant fellow, a hepatobiliary/liver transplant surgeon, and a transplant psychiatrist—were directly incorporated into the supervising structure of the housestaff team. The housestaff team consisted of four interns and four second- or third-year residents who rotated every four to five weeks, consistent with the preintervention period. Under the SHS model, the multidisciplinary hepatology team formally met with the housestaff team on a daily basis for 90 minutes during morning rounds. This time was devoted to communicating issues relevant to patient management, discussing daily plans, and teaching management skills for common problems that arise in the care of patients with decompensated liver disease. In addition, weekly, case-based lectures focused on recently admitted patients with decompensated liver disease, in a format that was similar to general medicine didactics held in the preintervention period.
Study design and data collection
There were two parts to this study: (1) a retrospective pre/post intervention study evaluating the impact of the intervention—the SHS model—on LOS, and (2) an assessment of housestaff satisfaction with the THS model and the SHS model.
For the pre/post comparison study, the preintervention cohort consisted of all 170 patients who were admitted to any of the five transplant hepatologists from July 1, 2003, through May 31, 2004. The postintervention cohort consisted of all 252 patients who were admitted to those same five transplant hepatologists from July 1, 2004, though May 31, 2005. All data on these admissions, including demographics, insurance status, admitting diagnosis, admission time, discharge time, and case-mix index (CMI, a measure of inpatient acuity based on admission diagnosis related group), were collected retrospectively from our hospital's electronic clinical information systems. LOS was calculated using admission and discharge date and time. The Model for End-Stage Liver Disease (MELD) score was calculated using the discharge laboratory data input into the formula:
For the control group, we obtained LOS and CMI information on patients admitted to a nonspecialized ward admitted under the THS model during the same time (pre- and postintervention) period.
Equation (Uncited)Image Tools
To assess housestaff satisfaction, a survey, available on paper and online, was administered in October 2006. All 118 current internal medicine residents (postgraduate years one, two, and three) who had rotated through both the SHS model (liver service) and a separate service with the THS model (general cardiology) at some point during their residency were invited to retrospectively rate their experiences with each service on a voluntary and anonymous basis. The general cardiology service was selected as the comparison rotation for this survey because it employs the THS model and routinely includes a subgroup of patients with advanced cardiac disease requiring primary management input from a multidisciplinary cardiac transplant team—analogous to the manner in which patients with advanced liver disease were cared for in the preintervention study period. Using this approach was necessary because we were unable to survey for satisfaction on the preintervention liver service due to changing residents over time as well as the possible bias due to different eras. The survey asked questions pertaining to the educational experience, multidisciplinary team support, confidence about patient management, service design, and overall satisfaction with each service. Although they were not asked explicitly to compare the two services, respondents were asked to rate their satisfaction with each service using the same questions within a single survey. Housestaff were not informed of the purpose of this survey.
The survey and the retrospective review received approval from the Columbia institutional review board.
We used SAS software version 9.1.3 (SAS Institute Inc., Cary, North Carolina) for all statistical analyses. The primary outcome was LOS. Our independent variable was the patient-care model (THS versus SHS models). Covariates were age, insurance status (dichotomized as Medicaid versus non-Medicaid), MELD score, and CMI. To minimize the effect of outliers, LOS was normalized using logarithmic transformation. Analysis was performed on both the nontransformed and the log-transformed data. Univariate analysis was carried out employing a Student t test for continuous covariates and chi-square for categorical covariates. All variables associated with LOS (P < .2) in univariate analysis were included in our multivariate analysis. A multivariable logistic regression model was then constructed to examine the independent effect of implementation of the liver service on LOS. To examine secular trends in the primary outcome, we calculated LOS for a nonspecialized, medicine housestaff ward on the same floor during the same study periods. Housestaff satisfaction survey responses were analyzed using a Student t test comparing the mean scores between the two groups.
Patient and staff demographics
As stated earlier, of the 422 patients who were admitted under an attending hepatologist and followed by an internal medicine housestaff team, 170 patients were admitted in the preintervention period (July 1, 2003, through May 31, 2004), and 252 patients were admitted in the postintervention period (July 1, 2004, through May 31, 2005). Baseline characteristics were similar between the two groups, including age, gender, MELD score, CMI, and insurance status (Table 1).
The attending hepatologists remained the same during the two-year study period. In addition, the nursing-to-patient ratio remained constant in the pre- and postintervention periods on all medicine floors throughout our hospital. In the SHS model, the nurses who cared for the patients with advanced liver disease did not receive any subspecialty training, and they continued to care for general medicine patients as well as the patients with liver disease on the ward throughout the study periods.
Length of stay
There was an absolute reduction in mean LOS (−29.5%; 18.3 versus 12.9 days) and median LOS (−8.8%; 8.0 versus 7.3 days) with the implementation of the liver service. This difference was statistically significant in both univariate (P = .04) and multivariate analysis (P = .0025). Repeat analyses were performed using log-transformed LOS to partially normalize the distribution. Univariate analysis of the log-transformed data revealed similar results, though this did not meet statistical significance (P = .1). Age (P = .15), MELD score (P < .0001), and CMI (P < .0001) were associated with LOS; Medicaid status (P = .5) was not. In multivariate analysis, the liver service was associated with a statistically significant reduction in LOS (P = .05). Analysis of the average LOS on a comparable nonspecialized internal medicine housestaff ward revealed an increase in LOS from the pre- to the postintervention study periods (+19.0%; 7.9 versus 9.4 days). This met statistical significance both before (P < .0001) and after (P < .0001) adjustment for CMI. When controlling for outliers using logarithmic transformation for both groups, mean LOS did not differ significantly between the liver service and the nonspecialized housestaff ward despite an increased CMI in the liver service patients (Table 2).
Of the 118 medical residents who had rotated through both the liver service (SHS model) and the general cardiology service (THS model), 83 residents (70%) completed the survey. Demographic characteristics of the responding residents are reported in Table 3, with a greater percentage of respondents indicating cardiology than gastroenterology as their future specialty (P = .048).
Residents reported greater satisfaction with the SHS model when compared with their experience caring for patients under the THS model in a variety of domains. Residents were more satisfied with the teaching by their supervising attending, multidisciplinary team support for routine matters, confidence about patient management as a result of interaction with their supervising attending, and service design. These parameters all met statistical significance, despite greater dissatisfaction with the ancillary and support staff caring for the liver service patients. Overall, 76 (91%) of the respondents were satisfied with their experience with the liver service compared with only 18 (22%) reporting overall satisfaction with caring for patients with end-stage cardiac disease under the THS model (P < .0001), with a mean score of 4.1 on a 5-point scale, where 5 = strongly agree (see Table 4).
Research in medical education has traditionally focused on the impact of educational interventions on trainee performance, knowledge acquisition, and satisfaction. There has been little investigation into the impact of the structure of ward-based medical education models on patient outcomes.9 Recently, leaders in the field have called for greater accountability in medical education with respect to its impact on patient-centered quality outcomes.10,11 Given the continuing controversy surrounding the impact of resident work hour limitations and ongoing pressure to reduce LOS in the face of rising acuity of inpatients' conditions, such studies are becoming increasingly important.
The present study demonstrates that the implementation of the SHS model was associated with a decreased LOS without sacrificing educational goals. Instead, residents rated their educational experience under the SHS model more favorably than their experience of caring for similarly complex patients with advanced cardiac disease under the THS model. This occurred despite decreased satisfaction with the ancillary and support staff associated with the liver service and an increased level of interest in cardiology as a career compared with gastroenterology.
There are several possible explanations for these findings. Most importantly, the implementation of the liver service—using the SHS model—allowed for the direct incorporation of the hepatology attending, a “quaternary-care attending,” into daily rounds. This increased the direct interaction between the hepatology attendings who were initiating the patient-care plan and the housestaff who were delivering the care. Daily face-to-face communication with all team members present early in the workday became the norm rather than an infrequent event, allowing earlier interventions, a finding that has been previously demonstrated by Wachter et al.12 Additionally, clustering of patients with advanced liver disease onto one housestaff team likely facilitated the direct transfer of clinical management knowledge from one patient to another by the residents themselves. Examples might include knowledge of an appropriate liver transplant workup or the indications for a prompt diagnostic paracentesis. Quick acquisition of such knowledge early in the rotation would allow trainees to subsequently initiate an appropriate clinical management plan at the time of admission of similar patients, even before presenting cases to the hepatology attending.
We acknowledge several limitations to our study. First, the retrospective pre/postintervention design leaves open the possible confounding effect of a decreasing secular trend in LOS during the study period to explain our main finding. However, the finding of an increase in LOS on a comparable medical ward for general medicine patients during the same pre- and postintervention periods argues against this trend as a confounder.
Second, we did not control for the effect of the nursing and allied health professional staff, such as social workers and physical therapists, on the patient outcome of interest. Because the staff did not receive any subspecialty training, and because the staff-to-patient ratio did not change from the pre- to the postintervention periods, it is unlikely that they significantly affected the LOS of the patients within the SHS model. Nevertheless, the volume of patients with advanced liver disease increased in the postintervention period, and we cannot exclude the effect of the staff's increased experience with these patients over time.
Third, the anonymous nature of the housestaff satisfaction survey precludes an analysis of survey responders versus nonresponders, leaving open the possibility of selection bias. However, the distribution of the stated future specialties of the responders closely mirrors the actual distribution of fellowship matches of our graduates during the last three years (gastroenterology 7–9%; cardiology 26–30%), suggesting that, at least in this important respect, our survey responders were representative of our housestaff as a whole.
Lastly, because the residents who had cared for patients with advanced liver disease under the THS model had graduated from the institution at the time of the survey, another service (general cardiology) that was analogous to the preintervention THS model for patients with advanced liver disease was selected as the comparison group for our survey. This raises the possibility that the survey results reflect differences between the hepatology and cardiology services independent of differences between the THS and SHS models. Given the large proportion of survey respondents who were interested in cardiology as a future specialty, however, we feel that this comparison group may have biased the survey in favor of the THS model, potentially strengthening the merits of our survey findings regarding the SHS model of patient care. However, the other possibility remains.
Implications and Conclusions
In light of the recent consensus report by the Alliance for Academic Internal Medicine on residency education reform that aims to foster change that “simultaneously improve[s] the quality of residency training, the satisfaction of residents as they go through training, and the patient care delivered in training settings,”13 the implications of our study are timely and important. In many academic health centers, residents are already responsible for a large volume of these complex, quaternary-care patients with advanced organ disease. Thus, implementation of the SHS model primarily represents a reorganization of an existing educational structure into something better—rather than a fundamentally new, superspecialized education experience for the housestaff. That said, the manner in which this new structure is implemented into an existing curriculum could vary considerably and be tailored to very different educational goals. For example, if the SHS model were to be incorporated into an existing ward rotation through which all residents rotated, it would represent simply a enhanced experience for trainees and patients rather than a more specialized training experience for the housestaff. Alternatively, the SHS rotation could be reserved only for housestaff with a stated interest in a particular subspecialty as a career path.
Quaternary-care interventions such as organ transplantation are increasingly common in academic health centers where many residency programs are based. Care of this subgroup of patients is often multidisciplinary and requires considerable expertise. Our study demonstrates that the implementation of a novel SHS model for patients with decompensated liver disease decreases LOS and increases residents' satisfaction, suggesting that this model may be a more optimal health care delivery structure for the care of these patients. Future research should focus on the impact of this model on other patient outcomes such as mortality or hospital readmission rates as well as additional objective measures of housestaff education such as scores on in-training exams or on rotation evaluations by residents. Additionally, future studies should explore the generalizability of this model to other quaternary-care interventions.
The authors wish to acknowledge Nicholas Fiebach, MD, vice chair of graduate and continuing medical education at New York Presbyterian Hospital–Columbia Presbyterian Medical Center, for his work in developing and implementing the SHS model as well as his input in the housestaff satisfaction survey. The authors would also like to acknowledge James Curty, manager of performance analysis/financial systems in the department of surgery, as well as David Caplan, Christine Valentin, and Olabisi Jegede with the information services/business solutions group at New York Presbyterian Hospital, for their assistance with data acquisition. None of these individuals received any financial compensation for their contributions.