Effects of an Acute Care for Elders Unit on Costs and 30-Day Readmissions

Importance Providing high-quality care while containing cost is essential for the economic stability of our health care system. The United States is experiencing a rapidly growing elderly population. The Acute Care for Elders (ACE) unit interdisciplinary team model of care has been shown to improve outcomes in hospitalized older adults. The University of Alabama at Birmingham ACE unit incorporates evidence-based care processes. We hypothesized that the ACE model would also reduce costs.

Objective To examine variable direct costs from an interdisciplinary ACE compared with a multidisciplinary usual care (UC) unit.

Design Retrospective cohort study.

Setting Tertiary care academic medical center.

Participants Hospitalists' patients aged 70 years or older spending the entirety of their hospitalization in either the ACE or UC unit in fiscal year 2010.

Main Outcome Measures Using administrative data, we analyzed variable direct costs for ACE and UC patients. We also conducted a subset analysis restricted to the 25 most common diagnosis related groups (DRGs) shared by ACE and UC patients. Generalized linear regression was used to estimate cost ratios and 95% confidence intervals adjusted for age, sex, comorbidity score, and case mix index (CMI).

Results A total of 818 hospitalists' patients met inclusion criteria: 428 from the ACE and 390 from the UC unit. For this study group (all DRGs), the mean (SD) variable direct cost per patient was $2109 ($1870) for ACE and $2480 ($2113) for UC (P = .009). Adjusted cost ratios revealed significant cost savings for patients with low (0.82; 95% CI, 0.72-0.94) or moderate (0.74; 95% CI, 0.62-0.89) CMI scores; care was cost neutral for patients with high CMI scores (1.13; 95% CI, 0.93-1.37). Significantly fewer ACE patients than UC patients were readmitted within 30 days of discharge (7.9% vs 12.8%; P = .02). Subset analysis of the 25 most common DRGs revealed a significantly reduced mean (SD) variable direct cost per patient for ACE compared with UC patients ($1693 [$1063] vs $2138 [$1431]; P < .001); cost ratios for total (0.78; 95% CI, 0.70-0.87) and daily (0.89; 95% CI, 0.85-0.94) variable direct costs remained significant after adjustment.

Conclusions and Relevance The ACE unit team model reduces costs and 30-day readmissions. In an era when improving care processes while reducing costs is a vital objective for the Medicare program and our nation as a whole, the ACE model meets these goals.

Delivering high-quality care while containing cost is essential for the economic stability of our health care system.¹ By 2030, the proportion of the US population aged 65 years or older will exceed 20%, creating an imminent need for strategies that improve person-centered care while reducing costs. Hospitalization often leads to adverse outcomes such as functional decline,^2-4 delirium,^5-7 undernutrition,⁸ polypharmacy,⁹ and other iatrogenic events^10,11 that lead to higher costs.^12,13 During the past 2 decades, Acute Care for Elders (ACE) units have been developed in hospitals nationwide. The distinguishing feature of ACE units is the use of an interdisciplinary team (IDT) model to provide hospital care as opposed to a multidisciplinary model in which providers from all disciplines deliver care but practice predominantly independently, or in “silos.” An interdisciplinary model integrates disciplines to collaboratively develop the patient-centered care plan. The primary goal of the ACE model is to reduce adverse outcomes in older adults through frequent IDT rounds that recognize and manage geriatric syndromes while initiating transition planning from the day of admission.¹⁴ Previous research of ACE units has shown improved processes of care,^15,16 prescribing practices,^15,17 physical functioning,^14,18,19 and patient and provider satisfaction,¹⁵ as well as reduced rates of restraint use^15,20 and institutionalization.^14,15

We hypothesized that our ACE model would reduce patient care costs for patients whose attending physicians were not geriatricians. The objective of our study was to examine variable direct costs from 2 hospitalist medical units: an interdisciplinary ACE compared with a multidisciplinary usual care (UC) unit. Despite the evidence regarding improved outcomes from ACE units, there is a paucity of data evaluating cost as the primary outcome of an ACE compared with a UC unit that uses the same attending physicians and unit staffing.

We conducted a retrospective cohort study examining cost of care among hospitalists' older patients who spent the entirety of their hospital stay in either the ACE or UC unit in fiscal year 2010. The institutional review board at the University of Alabama at Birmingham (UAB) approved this study.

The UAB Hospital is a 1146-bed academic medical center. An adult of any age requiring admission for an acute medical illness is assigned to either the teaching or hospitalist service (which includes nurse practitioners without house staff). The hospitalists admit patients primarily to 2 units: the ACE and the UC unit. Patients are assigned to a unit based mainly on bed availability. Patients met the inclusion criteria for this study if they (1) were aged 70 years or older, (2) met inpatient admission criteria, and (3) spent the entirety of their hospitalization as a hospitalists' patient in either the ACE or UC unit in fiscal year 2010. A total of 818 hospitalists' inpatients met inclusion criteria and were included in this analysis: 428 from the ACE and 390 from the UC unit. From this primary study group of patients, we also conducted a subset analysis restricted to the most common diagnosis related groups (DRGs) from both units to reduce outliers; this resulted in 25 DRGs shared by 260 ACE and 214 UC patients (eTable.).

In fiscal year 2010, the 25-bed ACE unit cared for adults of all ages; 39.2% of patients were aged 70 years or older. The ACE unit delivered the following care model for all patients aged 70 years or older: (1) geriatric assessments, including cognitive and functional screens; (2) daily (Monday-Friday) geriatrician-led IDT rounds focusing on geriatric care; (3) discharge planning from the day of admission; and (4) communication of team recommendations to the appropriate caregiver (eg, hospitalist, nurse, family, or patient). The IDT members included the ACE nurse coordinator, geriatricians, nurses, physical and occupational therapists, pharmacists, a registered dietician, and social workers.

In the IDT rounds, the ACE coordinator and social worker reported results of functional, cognitive, and psychosocial assessments. Nurses reported daily assessments regarding geriatric syndromes (eg, pain, mood, cognition, and mobility). This was followed by reports from each discipline, including medication review, and the collaborative development of a daily geriatric care plan. Patients' baseline and current functional dependencies were assessed by screening questionnaires for basic (Katz index)²¹ and instrumental (Lawton index)²² activities of daily living. Cognition was assessed with the Mini-Cognitive Assessment.²³ The ACE unit did not have formalized nurse-driven care protocols or order sets, but it did use evidence-based geriatric care processes known to reduce the risk of incident geriatric syndromes⁷ and address patient or caregiver stress (Table 1). These services were provided by volunteers or supervised trainees at no additional cost. The hospitalists did not participate in the IDT rounds.

The hospitalists also admitted patients of all ages to a UC unit. This unit was staffed with the same disciplines as the ACE unit, but geriatric screens and daily IDT rounds focusing on geriatric syndromes were not conducted, nor were the volunteer-staffed care processes used. Because the same hospitalists care for patients on both the ACE and UC unit, this UC unit served as our comparison treatment area. During the study period, the UC unit also conducted a daily team meeting, or discharge planning rounds, attended by hospitalists and nurse practitioners, the nurse coordinator, the unit charge nurse, a social worker, and care managers. Geriatric assessments were not performed, and other disciplines did not participate.

Hospitalists and collaborating nurse practitioners rotated patient care duties between the ACE and UC units. Nonphysician staffing was similar in the 2 units (Table 1). Both units had identical assigned nursing hours, calculated as worked hours per patient day, reflecting consistency between the units with respect to unit-based nursing staff.²⁴ Both units formally consulted non–unit-based disciplines, such as physical therapy and social services. Both units also used nurse coordinators, neither of whom provided any direct medical patient care. The ACE nurse coordinator administered the geriatric screens; coordinated care between all team members, including volunteers and outpatient providers; and ensured that the daily geriatric care plan was implemented. The UC nurse coordinator assisted with communication to patients' primary care physicians and collaborated with social services in discharge planning.

Using the UAB Hospital administrative database, we conducted a retrospective analysis of costs of care for the ACE and UC patient populations meeting inclusion criteria. Additional analysis was restricted to the 25 most common DRGs shared by ACE and UC patients (eTable). The hospital administrative database, Transition Systems Incorporated (TSI), was used to obtain financial information for each patient. The TSI data includes payer source, hospital charge, cost, allocated bad debt, and net revenue for each patient. The TSI system generates each patient's total costs from the sum of fixed direct costs, variable direct costs, and indirect costs. Fixed direct costs are those associated with patients that cannot be linked to a specific patient (eg, the salary of the unit nurse manager). Variable direct costs are attributed to direct patient care (eg, laboratory tests, medications). Finally, indirect costs cannot be attributed to a specific patient or department (eg, utilities, salaries of hospital accountants).

Given that fixed and variable staffing was identical on the 2 units, our study's primary outcome measure was total variable direct costs. In addition, daily variable costs were calculated by dividing total variable direct costs by length of stay (LOS). Variable direct cost was chosen as the primary outcome because it represents the cost stemming directly from daily patient care and would therefore provide the best measurement for the effect of the ACE model on costs. Other variables examined included patient demographics, the case mix index (CMI), the comorbidity score derived from the High-Risk Diagnoses for the Elderly Scale,²⁵ LOS, readmissions to UAB Hospital within 30 days, and total cost. The CMI is defined as the sum of the relative weights of the DRGs for a patient population divided by the number of patients in that population, and it reflects patient complexity and expected resource consumption.²⁶

Significant differences between groups were assessed by using χ² tests for categorical data, t tests for normally distributed continuous data, and 2-sided Wilcoxon rank sum tests for medians. Because cost and LOS data were not normally distributed, t tests with bootstrap resampling were used to test means.²⁷ Generalized linear models were used to calculate crude and adjusted cost ratios and 95% CIs. Adjusted models included age, sex, comorbidity score, and CMI. Several distribution families (normal, log normal, and gamma) were considered for multivariate models of variable direct costs. For each, we reviewed the Kolmogorov-Smirnov statistic for goodness of fit and determined that the log normal distribution fit the data best. Additional analysis was stratified by CMI group (low, moderate, or high). All analyses were performed using SAS software, version 9.2 (SAS Institute, Inc), and differences were considered statistically significant at P < .05.

There were no significant group differences for demographic variables or in-hospital mortality (Table 2). The ACE and UC groups had similar CMI and comorbidity scores (Table 3). Analysis including all DRGs revealed that the mean (SD) total variable direct cost per patient was $2109 ($1870) for ACE and $2480 ($2113) for UC (P = .009). Adjusted cost ratios stratified by CMI revealed significantly reduced cost ratios for ACE patients with low (0.82; 95% CI, 0.72-0.94) or moderate (0.74; 95% CI, 0.62-0.89) CMI scores and cost-neutral care for patients with high CMI score, all in comparison with UC patients. In addition to cost savings, a significantly smaller proportion of ACE patients than UC patients were readmitted to UAB Hospital for any cause within 30 days (7.9% vs 12.8%; P = .02). There was no difference between the proportions of ACE and UC patients with at least 1 admission to UAB Hospital in the 6 months preceding the study admission date.

Analysis restricted to the 25 most common DRGs revealed a significantly reduced variable direct cost per patient for the ACE compared with the UC unit (mean [SD], $1693 [$1063] vs $2138 [$1431]; P < .001). Mean daily variable direct costs were also significantly less for the ACE unit ($484 [$162] vs $545 [$120]; P < .001). Among patients with the 25 most common DRGs, cost ratios for total variable direct costs (0.78; 95% CI, 0.70-0.87) and daily variable direct costs (0.89; 95% CI, 0.85-0.94) remained significant after adjustment (Table 4). Fewer ACE patients were readmitted to UAB Hospital within 30 days for the 25 most common DRGs, but despite an absolute difference of 4%, this difference was not statistically significant (7.3% vs 11.2%; P = .14).

We found that an ACE model with an IDT supporting hospitalists who do not participate in ACE team rounds significantly reduces variable direct costs and 30-day readmission rates for patients aged 70 years or older. Extrapolating an average of $371 in variable direct cost savings per patient would result in a $148 400 savings for every 400 patients admitted to the ACE unit. Our results suggest that an ACE team collaboratively supporting attending physicians who are not geriatricians can deliver geriatric care processes cost efficiently. Our study comparing 2 units using different care models eliminates several confounding variables present in previous ACE studies. First, patients with comparable comorbidity and CMI scores admitted to the ACE and UC units were cared for by the same physicians. Second, the units had identical nursing staff to patient ratios and access to non–unit-based disciplines. However, the ACE model redesigned how staff was used. For example, the units had the same pharmacist support, but the role of the ACE pharmacist included IDT rounds and medication reviews. This adjustment of roles does not add costs. The ACE unit did have slightly more physical therapist full-time equivalents available to see patients (Table 1); this additional cost is well below the observed cost savings and was accounted for in the total costs, which were also significantly reduced in the ACE model.

Finally, both units conducted daily team meetings; a geriatrician attended the ACE meeting and 2 or 3 hospitalists attended the UC meeting. The cost for physician time for daily team meetings is much less in the ACE model and represents additional cost savings not included in variable direct costs. We speculate that the savings we observed are explained by improved care processes and coordination that stem from an IDT model, thereby enhancing geriatric assessment and appropriateness of care.

Our findings strengthen previously published studies on ACE units. In a randomized controlled trial of the nation's first ACE unit, Landefeld et al¹⁴ demonstrated improved functional performance and reduced long-term care placement for ACE patients compared with UC patients. In a follow-up cost analysis of this new ACE intervention,²⁸ the start-up costs, including environmental redesign and staff time to develop care protocols for the new unit, added $38 per bed-day for ACE patients, but because the ACE unit reduced LOS (7.5 vs 8.4 days for UC; P = .449), total costs to the hospital were also reduced ($6608 for ACE vs $7240 for UC; P = .93). We incurred no additional costs for care process development, and our ACE unit had not undergone any environmental modifications.

In 2012, the authors who first wrote about the ACE model published a second retrospective cost analysis²⁹; this study included patients aged 70 years or older who were randomized to the ACE or UC unit from 1993 through 1997. This analysis demonstrated a significantly shorter mean LOS for the ACE unit (6.7 vs 7.3 days for UC; P = .004) and therefore lower total costs for ACE patients. However, a Consumer Price Index inflation calculator was used to convert costs from the year of discharge to equivalent costs for 2011.²⁹

Since these original publications on ACE units from the mid-1990s, 2 new trends have emerged in the care of hospitalized elderly patients: (1) many older adults are now admitted to the care of hospitalists and (2) the average LOS for older adults has decreased to 5.6 days.³⁰ Our ACE unit study occurred in this new health care environment and demonstrated reduced costs and readmissions for the ACE model, even in patients whose LOS was less than 5 days, signaling a role for the ACE model in hospitalists' care for older adults.

Finally, one other study demonstrated significant total cost savings for the ACE model in 3 common DRGs (pneumonia, congestive heart failure, and urinary tract infection).³¹ However, the authors do not state whether patients had the same attending physicians, some ACE patients were admitted to a geriatrician, and cost was estimated by using cost to charge ratios. Our study strengthens these findings by including all DRGs, ensuring that the patients in the ACE and UC units were treated by the same physicians in units with the same staffing, and by using variable direct cost data.

In addition to treatment by the same attending physicians, both of our study units conducted a team meeting, thus eliminating the conduct of a daily meeting as a confounder. The UC team rounds followed the conventional discharge planning format and focused on the discharge day and destination but did not address geriatric syndromes or include other disciplines that participated in the ACE team meetings. In 2011, O’Leary et al³² published a controlled trial evaluating the addition of IDT rounds to UC for hospitalists' patients of all ages (mean age, 64 years). Although nurses noted improved communication with the team intervention, the effect on costs was inconsistent. The ACE model focus on geriatric syndromes may explain the cost savings and reduced readmissions noted in our study.

Our study has several limitations. Given the predominance of studies demonstrating improved outcomes in elderly patients admitted to ACE units, we did not believe it was appropriate to knowingly prevent admission of a patient to the ACE unit if a bed was available. Furthermore, because of the complexity of our academic medical center, holding beds open for a randomized study is not feasible. Thus, our study was not a randomized controlled trial. However, given the large number of older adults admitted to UAB Hospital each year (11 756 inpatients in fiscal year 2010), bed availability predominantly determined patient assignment to the ACE or UC unit, thus “pseudo-randomizing” patients to the 2 care models.

The mean LOS for the UC patients was markedly below the national average for older adults.³⁰ The UAB hospitalists and geriatricians had been closely collaborating in patient care for 3 years before our study period. In fiscal year 2010, 52% of all inpatient geriatric consultations were in hospitalists' patients (including ACE and UC units). Thus, the hospitalists used the geriatric consultation service for UC patients. The crossover contamination may contribute to the short LOS in the UC patients and underestimation of the cost savings potential of the ACE model for clinical services less familiar with geriatric care principles than our hospitalist physicians. The ACE unit geriatrician plays a consultative role and is not the primary attending physician. Given the declining number of geriatricians nationwide, this ACE model is more sustainable than one requiring an attending geriatrician. The approximate geriatrician time for the daily ACE meeting was 30 minutes. Given the cost savings for a short time commitment, it is feasible that a single geriatrician could facilitate multiple team rounds per day on multiple units, using identical interfacing techniques with attending physicians, which would multiply the cost savings dramatically throughout a hospital.

Other limitations are related to scarcity of geriatric syndrome information available in the hospital database or medical record. The database does not record where patients were living before hospitalization and does not distinguish between subacute rehabilitation and long-term care. This limits our ability to perform in-depth analysis of where patients were admitted from or their discharge destinations. We also were not able to determine the number of 30-day readmissions to other hospitals. Measurement of specific quality outcomes are beyond the scope of this study, but the UAB ACE unit implements the same care processes that have been shown to improve outcomes for other ACE units.^6,16,17,33

In conclusion, our results strengthen the findings of previous studies of ACE units by demonstrating that the ACE model can seamlessly interface with nongeriatrician attending physicians and significantly reduce costs and 30-day readmissions. Reducing 30-day readmissions further lowers hospital risk for withheld reimbursement. In an era when improving outcomes while reducing cost is a vital objective for both the Medicare program and our nation as a whole, the ACE model meets these goals.

Correspondence: Kellie L. Flood, MD, Division of Gerontology, Geriatrics, and Palliative Care, University of Alabama at Birmingham, 1720 Second Ave S, CH-19, Room 219, Birmingham, AL 35294 (kflood@uabmc.edu).

Accepted for Publication: January 25, 2013.

Published Online: April 22, 2013. doi:10.1001/jamainternmed.2013.524

Author Contributions: Drs Flood, MacLennan, and Brown had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Flood, MacLennan, and McGrew. Acquisition of data: McGrew, Green, and Dodd. Analysis and interpretation of data: Flood, MacLennan, McGrew, and Brown. Drafting of the manuscript: Flood, MacLennan, McGrew, and Brown. Critical revision of the manuscript for important intellectual content: All authors. Statistical analysis: MacLennan, Green, and Dodd. Administrative, technical, and material support: McGrew. Study supervision: Flood and Brown.

Previous Presentation: Preliminary analyses of these data were presented at the Plenary Paper Session at the annual American Geriatrics Society Scientific Meeting; May 12, 2011; Washington, DC.

Additional Contributions: We thank geriatricians Donna Bearden, Heather Herrington, and Caroline Harada, and all ACE team members for their dedication to providing the highest quality of care to the patients we serve. We also thank Harry Vickers for his generous donations for ACE unit staff training.