Impact of Hospitalist Care Model on Patient Outcomes in Acute Medical Unit


 BACKGROUND: The present study aimed to assess a newly introduced, hospitalist-run, acute medical unit (AMU) model in Korea. The AMU in our institution started in October 2015. Four hospitalists managed patients with acute medical needs that were admitted through the emergency department (ED). STUDY DESIGN: We conducted a retrospective cohort study of all medical inpatients admitted through the ED from June 1, 2016 to May 31, 2017, at a tertiary care hospital. We evaluated 6391 patients whether the hospitalist care in the AMU improved patient outcomes compared to standard non-hospitalist care. METHODS: We created multivariate analysis models to compare the clinical outcomes of patients cared for by hospitalists with the outcomes of patients cared for by non-hospitalists. RESULTS: In the adjusted models, compared to the non-hospitalist group, the AMU hospitalist group had a lower in-hospital mortality (OR: 0.46, P <0.001), a lower intensive care unit (ICU) admission rate (OR: 0.39, P <0 .001), a shorter hospital length of stay (coefficient: -1.349, SE: 0.217; P <0.001), and a shorter ED waiting time (coefficient: -3.021, SE: 0.256; P <0.001). There were no significant differences in the 10-day or 30-day re-admission rates (P = 0.493, P = 0.201; respectively). CONCLUSIONS: The AMU hospitalist care model was associated with reductions in in-hospital mortality, ICU admission rate, length of hospital stay, and ED waiting time. These findings suggested that this AMU hospitalist care model might be adaptable to other healthcare systems to improve care for patients with acute medical needs.

reported the e cacy of AMUs have described reductions in mortality and lengths of hospital stays [8][9][10].
The Seoul National University Bundang Hospital adopted both the hospitalist care model and the acute medicine unit model. Our institution established the rst hospital medicine center in Korea and opened a hospitalist-run AMU in August 2015. The AMU started with a 20-bed ward, where patients with acute medical needs that were admitted through the emergency department (ED) received appropriate and timely medical care. They were discharged within 72 h after the end of treatment in the AMU, or they were transferred to another special ward for additional care, as required [4]. In 2017, the hospitalist team of our hospital reported the effects of the AMU operation. It reduced the ED waiting time by 40%, and it signi cantly shortened the length of hospital stay (LOS), from 10 days to 9.1 days [11]. However, no study has thoroughly evaluated its effects on patient outcomes, including inhospital mortality, the intensive care unit (ICU) admission rate, or the re-admission rate. Therefore, this study aimed to evaluate the patient outcomes of an AMU that was operated according to the Korean hospitalist care model and to provide supporting data that might inform the design of the most e cient Korean hospitalist care model.

AMU setting
The AMU had 28 beds, but started with 20 beds. Medical patients waiting in the ED were randomly admitted to either the AMU or the medical ward, as soon as a bed was available. Patients in the AMU were cared for by four internists that specialized in infection, pulmonary disease, critical care, and endocrinology. Patients admitted to the AMU were rapidly evaluated and treated by internists, who served as hospitalists, with experience in treating patients with acute medical illnesses [11]. This study included all patients with acute medical needs that were admitted to the ED during the daytime, 7 days/week, in the AMU.
On the other hand, non-hospitalist inpatient care was performed by subspecialists and residents in a specialty medical ward. In the specialty medical ward, residents were mainly in charge of inpatient care, under the supervision of an attending physician. The AMU hospitalist care at our institution mainly focused on acute and general care. In contrast, non-hospitalist care in the specialty medical ward was focused on long-term and specialized care. The AMU models of Korea and the UK are compared in Fig. 1 [12].
Study design This retrospective cohort study employed a secondary analysis of data from clinical records and hospital administrative information in Electronic Medical Record (EMR) of our institution. The local Institutional Review Board approved the waiver of consent and other study procedures.

Subjects
All medical inpatients admitted from the ED between June 1, 2016 and May 31, 2017 were included for the case and control groups. We excluded admissions to the ICU and surgical ward via the ED. A ow diagram of the allocation of the study population is presented in Fig. 2.

Variables
We measured the following outcome variables: in-hospital mortality (IHM), ICU admission, length of hospital stay (LOS), ED waiting time, and unscheduled re-admissions, within 10 days and 30 days. The IHM was de ned as the ratio of inpatient deaths to the total number of inpatients. ICU admission was de ned as an entry into the ICU.
When a patient was admitted more than once during a single hospital stay, only the features of the rst admission were analyzed. The LOS was de ned as the duration of a single episode of hospitalization and was calculated by subtracting the date of admission from the date of discharge. The ED waiting time was de ned as the time spent waiting in the ED before admission to the AMU or a medical ward. Re-admissions were identi ed as an unscheduled admission via the ED, due to any cause, within 10 or 30 days after discharge.
We recorded the following clinical variables of the subjects: gender, age, prior hospitalization history, cardiopulmonary resuscitation (CPR) incidence, cause of ICU admission, referral to a specialty, referral to a specialty unit, consultations, operations (cases performed during the hospitalization, not before), major diagnosis (based on the International Statistical Classi cation of Diseases and Related Health Problems, 10th Revision, Australian Modi cation (ICD-10-AM)), the Korean Triage and Acuity Scale (KTAS), the Age-adjusted Charlson Comorbidity Index (ACCI), and the Acute Physiology and Chronic Health Evaluation (APACHE) II score.
The KTAS consisted of ve stages: resuscitation, emergency, urgent, less urgent, and non-urgent. This was Korea's rst uni ed emergency patient triage system at the national level. It was developed in accordance with the domestic situation, and it was based on the Canadian triage and acuity scale [13]. The KTAS, which is currently applied in emergency medical centers in Korea, is a national standardized classi cation tool for evaluating illness severity. It is the only tool commonly available for assessing patients from the pre-hospital phase to the hospital phase. Previous validity testing showed that the higher the severity level, the higher the ICU and general hospital admission rates, the LOS, the number of clinic consultations, the CT scan rate, the emergency intervention rate, and the ED medical expenses. Thus, the KTAS was con rmed as a valid tool that re ected the severity trend, and it is currently used in national and regional emergency medical centers [13]. Accordingly, in this study, we used the KTAS to compare the severity of conditions among patients.
The comorbidity score was calculated with the Charlson Comorbidity Index (CCI). The CCI score is the sum of 1, 2, 3, and 6 weighted values for 17 disease groups, ranging from 0 to 29; higher scores indicate higher severity [14]. Additionally, because age was determined to be a signi cant factor of survival, it was subsequently incorporated into the Charlson comorbidity score to create a single index that accounted for both age and comorbidity; i.e., the ACCI [15]. The ACCI was calculated with additional points added for age (1point was added for each decade over 40 years of age, ranging from 0 to 4). In addition, the clinical comorbidities were translated into ICD codes, which were used as a risk-adjustment tool based on administrative data [16]. Therefore, in this study, we used ICD-coded data and age to calculate the ACCI score.
The APACHE II scoring system has been widely recognized as an ICU prognostic scoring model [17][18][19]. The APACHE II score utilizes the worst values of 12 physiological variables, including: blood pressure, heart rate, body temperature, oxygenation, Glasgow Coma Score (GCS) during the rst 24 h after ICU admission, an evaluation of the patient's chronic health issues, age, and the type of ICU admission [20]. The APACHE II score was shown to be an accurate measurement of the severity of a patient's disease, and it was strongly correlated with outcome among patients in critical conditions [17]. This score (range 0 to 71) was closely correlated with the risk of hospital death [18]. Consequently, we used the APACHE II score to compare the disease severity among ICU admissions.
Statistical Analysis Categorical variables are reported as percentages, and continuous variables are reported as the mean ± standard deviation (SD). Groups were compared with Pearson's chi-square test or the t-test, as appropriate. ACCI, the LOS, and the ED waiting time are expressed as the median and interquartile range (IQR). For these variables, groups were compared with the Mann-Whitney U test, due to their skewed data distributions. We performed subgroup analyses according to the severity of the patient's condition (based on the KTAS score), the degree of comorbidity (based on the ACCI), and the major disease category (based on the ICD-10).
We identi ed disease codes for the 10 most frequent diseases in the cohort. The remaining diseases were aggregated into a single category designated as "others." We then included these disease codes as dummy variables in the regression models to determine the extent to which they affected the outcome. The main analyses focused on the ve outcome measures for the case and control groups. Logistic regression model (for IHM, ICU admission, and all-cause unscheduled re-admissions as binary outcomes) and linear regression model (for LOS and ED waiting time as continuous variables) were used to adjust for age, gender, prior hospitalization, KTAS, ACCI, CPR incidence, referral to a specialty, operation, consultation, major disease. Using the estimates from the regression models, we presented differences between the hospitalist group and the non-hospitalist groups in the IHM rate, ICU admission rate, all-cause unscheduled re-admission rate, LOS, and ED waiting time.
Compared to the non-hospitalist group, the hospitalist group had lower rates of IHM (4.8% vs. 9.1%, P < 0.001) and ICU admissions (3.9% vs. 8.7%, P < 0.001). Among the patients admitted to the ICU, the hospitalist group displayed greater disease severity than the non-hospitalist group ( [6.7-19] h, P < 0.001) than the non-hospitalist group. However, there were no signi cant differences between the two groups in the re-admission rates within 10 or 30 days (P = 0.507 and P = 0.248, respectively). We performed subgroup analyses of patients strati ed by KTAS and ACCI scores to determine differences between the two groups (Tables 2 and 3, respectively). Among the more urgent cases, the hospitalist group showed lower rates of IHM (4.8% vs. 9.8%, P < 0.001) and ICU admission (4.0% vs. 9.1%, P < 0.001), compared to the non-hospitalist group.
We performed another analysis of subgroups strati ed by the major disease to determine whether the two groups showed differences in IHM (data not shown; additional le 1). Among patients with malignant neoplasms, infectious diseases, and diseases involving the respiratory system, digestive system, musculoskeletal system, and connective tissue, IHM was signi cantly lower in the hospitalist group than in the non-hospitalist group.        We performed logistic regression analyses to adjust clinical variables potentially associated with the four major outcomes: IHM, ICU admission, re-admission within 10 days, and re-admission within 30 days ( We also performed linear regression analyses to adjust clinical factors associated with LOS and the ED waiting time (Table 5). Both the LOS (coe cient: -1.349, standard error [SE]: 0.217; p < 0.001) and ED waiting times (coe cient: -3.021, SE: 0.256; p < 0.001) were signi cantly shorter in the hospitalist group than in the nonhospitalist group.    [11]. The AMU at our institution was different from typical emergency short-stay units operated by emergency physicians, mainly because the AMU follows these essential principles: intensive, active care, multi-disciplinary teamwork, and rapid diagnostics and therapy [12].
Although those results were con icting, in our institute, AMU hospitalist care was more effective than nonhospitalist care in reducing patient mortality. Our ndings provided evidence that a hospitalist-run AMU was an  [32,48,49]. Despite those con icting results, in our institute, AMU hospitalist care reduced the total LOS, compared to care from non-hospitalists. Recently, another study reported that hospitalist care within the integrated medical model of Korea showed a reduction in LOS, particularly in patients with multiple comorbidities [34]. According to that study, patients that received hospitalist care had shorter LOS, for several reasons. One reason was that the hospitalists were better trained than residents; consequently, they could manage diseases from more perspectives, which increased the likelihood of resolving the condition [34]. That nding suggested that Korean hospitalist care might signi cantly shorten LOS, regardless of the type of care model applied.
Some studies showed that AMU hospitalist care signi cantly shortened the ED waiting time, compared to nonhospitalist care [7][8][9]23]. Consistent with previous studies, in our institution, AMU care also reduced the ED waiting time. This result might be explained by an increase in the bed turnover rate and an alleviation of delays caused by waiting inpatients.
In our study, AMU hospitalist care was not associated with a signi cant difference in unscheduled re-admissions. Previous studies showed that hospitalist care led to signi cantly lower re-admission rates, compared to nonhospitalist care [10,21,22,29,39,48,50]. Others found no signi cant difference in the re-admission rate [30,[32][33][34][40][41][42][43][44]. Those results suggested that the re-admission rate might depend on the type of hospitalist care model applied, disease-related factors, and the hospitalist's roles.  reported that, when a hospitalist provided post-discharge transitional care by telephone to discharged patients, the re-admission rate was signi cantly reduced [50]. Some studies have found that an AMU signi cantly reduced the re-admission rate [51], but most studies on the effects of acute medicine found no effect of hospitalist care on the rate of unscheduled re-admissions [42,43]. In our institution, AMU hospitalist care was focused on the acute treatment of inpatients admitted through the ED; therefore, we did not expect a signi cant impact on the post-discharge readmission rate.
In summary, we provided evidence that AMU hospitalist care in a Korean tertiary care hospital reduced IHM, ICU admission, LOS, and ED waiting time.
Our study had some limitations: (1) it had a retrospective design, and it was limited to a single institution; (2) it was di cult to distinguish whether the effects on outcomes were due to the AMU setting or the care provided by the hospitalists; (3) we did not evaluate patient or staff satisfaction; and (4) we did not perform an economic evaluation, including the medical costs. Future studies are necessary to determine whether the introduction of the AMU will improve patient health in the long term, increase patient or staff satisfaction, and improve the costeffectiveness of patient care.

Conclusions
South Korea adopted both the UK's AMU and the US's hospitalist model and applied a combined care model. This is the answer to the increase in the complexity and severity of inpatient care in internal medicine due to the increase in patient with complex diseases, various chronic, tumor, and immune diseases [4]. Our study showed that the AMU Hospitalist care model of South Korea improved patient outcomes. This care model may be applied in some hospitals. However, it is necessary to apply a different care model depending on the situation (number of beds, human resources, needs) and culture of individual hospitals [4]. In addition, it is necessary to evaluate the effectiveness of various hospitalist care models.