Postacute Rehabilitation in Patients With and Without Dementia

One in every three older adults in the US, which is about 6 million people, experience dementia, and this rate is expected to more than double to 13.8 million by 2060.¹ People with dementia admitted to acute care hospitals face many challenges irrespective of the setting to which they are discharged.² About one-fifth of Medicare-insured hospitalized patients receiving acute care are discharged to short-term skilled nursing rehabilitation facilities (SNFs) with the goal of reducing hospital length of stay.3 In 2017, approximately 1.6 million people aged 65 years and older in the US received short-term rehabilitative services in SNFs posthospitalization.³ Patients are admitted to SNFs to receive rehabilitative services intended to restore function so that they may be safely discharged to their community or home.

The research to date exploring the effectiveness of rehabilitative care in SNFs has primarily focused on patients who are otherwise healthy and without complex conditions, including dementia.^4-6 Studies have shown promising potential in the SNF quality measures, such as the rate of discharge to the community, shorter length of stay, improved mobility, decreased mortality, and a decreased rate of readmission after a SNF stay; however, many of these studies have focused on patients who are generally healthier, such as those undergoing joint replacement procedures, rather than people with cognitive decline or other risk factors.^4,7-9

Other studies have reported that despite increases in treatment intensity and fee-for-service (FFS) rehabilitative care, patients who are insured by managed care plans have experienced fewer 30-day readmissions, were more often discharged to the community, and were less likely to require long-term care (LTC).^4,10 However, these positive outcomes have been overshadowed by the rapid and questionable rise in patients who have received increased intensity of therapies to the exclusion of those who may need it most, such as those with cognitive impairment, functional loss, or both, including impairment in activities of daily living (ADLs) and sensory loss.^3,11

The limited research that has examined the impact of rehabilitative services in SNF patients with dementia admitted from acute care settings indicates these people may not benefit from the rehabilitation services as they exist. Further, they may be at risk for transfer as an LTC patient. However, the causes of these poor outcomes are not clear.^2,12-14 The research examining rehabilitative services for SNF patients with comorbid conditions admitted from acute care settings is variable. One study reported that nonmodifiable risk factors, such as baseline function, cognition, and demographics, and modifiable risks, such as sensory impairment (eg, hearing and vision loss), depressive symptoms and pain, were associated with an increased likelihood of transitioning into LTC rather than discharge back to the community.²

Before 2019, the structure for rehabilitative services was prospective payment system, or FFS system, which prompted Medicare reimbursement based on the intensity of services measured in minutes of rehabilitative service (ie, physical, occupational, and speech therapies).¹¹ Revenue utilization groups were used to categorize patients based on higher minutes and intensity of rehabilitative services. However, for at least a decade, concerns have been raised about the effectiveness of the FFS system. First, it has favored intense rehabilitative services for patients who were healthy, cognitively intact, and able to respond to directions and follow-through on prescribed exercises and treatments, often to the exclusion of patients who have greater cognitive impairment.¹¹ Second, the focus on treatment intensity has resulted in increased costs. As an example, Medicare payments to SNFs for ultrahigh therapy rose from $5.7 billion in 2006 to $10.7 billion in 2008, representing a nearly 90% increase in cost.¹¹ Since 2011, the increase in cost amid mixed benefits with regard to the outcomes of rehabilitative therapies in SNFs has raised concerns about the cost of care in relation to its benefits.³

The rapid and unsustainable increases in costs related to rehabilitation intensity in SNFs resulted in changes to FFS Medicare reimbursement in 2020. These changes are focused on assuring equitable reimbursement to a wider variety of patients. This includes more medically complex patients, specifically those who may require care that is not focused on FFS rehabilitative therapies. This restructuring of Medicare has increased payments for medically complex patients for costs such as pharmaceuticals and shifted the focus away from intense rehabilitative therapy. The goal is to establish more equity among SNFs and to encourage SNFs to attract a broader case mix, but it is not yet clear whether this goal has been met.³

While these changes are meant to provide more equitable reimbursement for a greater case mix of patients, it is not clear what impact this will have on overall long-term outcomes, such as length of stay, both in the hospital and the SNF setting, or upon SNF discharge to the community. However, underscoring all the efforts to provide more equity in the Medicare system is the issue that little work had been done before the 2020 Medicare changes to explicitly examine the outcomes of rehabilitative therapies in patients with dementia. This is critically important given that the number of people with dementia who are discharged from acute care to SNFs is expected to increase significantly in ensuing years. As a result, the purpose of this study was to retrospectively examine the effects of each of the rehabilitative therapies—occupational therapy (OT), physical therapy (PT), and a combination of the two (OT+PT) under the FFS structure on the change in physical function from admission to quarterly assessment or discharge in patients with and without dementia (aged >65 years) admitted to SNFs in Massachusetts.

We conducted a secondary data analysis using the 2013 Minimum Data Set (MDS) 3.0 database. These data were relevant to our goal of retrospectively examining the FFS system in people with and without dementia before SNF reimbursement changes took place. This study was an exempt protocol per the university’s Institutional Review Board.

Study Sample

The MDS assessment data included 6396 people age 65 years and older who were admitted to an SNF in 2013 from an acute care hospital in Massachusetts.

Study Measures

Main outcome. The main outcome variable was the difference in total ADL scores between the first quarterly assessment or discharge and admission, reflecting the change in functional status among these patients from admission to quarterly assessment or discharge. The total ADL score is the sum of all available ADL ratings for both self-performance and with support provided: bed mobility, transfer, walk in room, walk in corridor, locomotion off unit, locomotion on unit, dress, eating, toileting, personal hygiene, and bathing. Each ADL outcomes has a rating score ranging from 0 to 9, with higher values indicating less physical function. Our previous analyses showed that these ADL measures have high reliability.² The total ADL score had a very close to a normal distribution.

Primary independent variables. The primary exposures in the analyses were OT, PT, and combined OT and PT. They were measured by the number of minutes that OT and PT were received, respectively, for each patient, reflecting the intensity of each intervention received.

Controlling variables. In our multiple linear regression (MLR) analyses, we also included the following control variables: demographics, final discharge status, and hearing, vision, and delirium.

Demographics. We controlled for age, sex (male vs female), race (per the MDS as either White or non-White), and marital status (never married, married, widowed, separated, divorced). Dementia diagnosis (yes or no) was also extracted from the MDS.

Final discharge status. Our study sample had four different types of discharge status: community, acute hospital, deceased, or SNF. Functional status is associated with discharge status, therefore, we included this variable to control for its effect. We were also interested in testing whether this factor moderated the intervention effects.

Vision. Vision was measured on four levels: adequate, minimal difficulty, moderate difficulty, and highly impaired. This variable was recoded as vision loss: no (adequate) or yes (other categories).

Hearing. Hearing was measured on five levels: adequate, impaired, moderately impaired, highly impaired, and severely impaired. The variable is recoded as hearing loss: no (adequate) or yes (other categories).

Data Analysis Methods

Descriptive statistics. To describe the characteristics of our study sample, we computed descriptive statistics on the study measures using frequency (percentage) for categorical variables and mean (standard deviation) for continuous variables.

MLR analysis for all available patients. To determine changes in function in patients with dementia who were admitted to SNFs, we performed an MLR analysis using the MDS 3.0 assessment data of all 23,677 patients aged 65 and older who were admitted to SNFs in 2013 from acute care hospitals in Massachusetts. We examined the interventions of OT, PT, or OT+PT, respectively, controlling for demographics, dementia, hearing, vision, and delirium. All observations with any variables containing “No assessed” or “No information” or “Skip pattern” were excluded from the analysis. The initial model included the intervention (OT, PT or OT+PT), interaction between intervention and dementia, interaction between intervention and final discharge status, and other controlling variables. We used SAS PROC GENMOD to fit the MLR models using the maximum likelihood estimation. The likelihood ratio test is used for selecting statistically significant independent variables with P<.05. A final model was fitted with the intervention variable and all the significant independent variables and interactions. MLR analysis for patients with dementia. To investigate the intervention effects among patients with dementia, we performed a similar MLR analysis for changes in function of the MDS 3.0 subsample of people with dementia (n=6396).

Demographics

As shown in Table 1, the sample was mostly female (66.3%) and White (92.9%), with a mean age of 82.4 (standard deviation, 8.1).

Abbreviation: SD, standard deviation.

MLR Analysis of Changes in Function for the Whole Study Sample

All results from this MLR analysis are summarized in Table 2. From these final models, all interventions were not statistically significant (OT: β, .02; 95% CI, −0.01 to 0.05; P=.10; PT: β, .05; 95% CI, −0.01 to 0.05; P=.17; OT+PT: β, .01; 95% CI, −0.006 to 0.03; P=0.14), whereas dementia was significant for all OT, PT and OT+PT models (OT, P=.01; PT, P=.05; OT+PT, P=.04), indicating cognitive function did have an impact on the change in physical function in the entire study population.
 

Table 2. Multiple Linear Regression Analyses of OT, PT and OT+PT for All Patients in Minimum Data Set 3.0 (N=23,671)

Within the entire study population, patients with dementia had a worse deterioration in function compared with patients without dementia. The intervention by dementia interaction was significant for all three interventions (P<.0001 for all three), indicating the rate of deterioration of function was different in the patients with and without dementia. Similar patterns existed for the final discharge status. That is, the change in function as well as the rate of change differed significantly in different discharge groups (all P<.0001). Age and sex were each significantly associated with change in function while patients were under the interventions. Older age and male sex were associated with worse function. Delirium was not significant in all three intervention conditions as shown in Table 1: OT: β, −.13; 95% CI, −1.69 to 1.43); P=.86; PT: β, −.08; 95% CI −1.64 to 1.47; P=.91; OT+PT: β, −.11; 95% CI, −1.68 to 1.45; P=.88).

For the whole sample, final discharge status was significantly associated with change in function (all P<.0001). Final discharge status is also a significant moderator (all P<.0001). White race was not significantly associated with change in function under all three intervention conditions (OT: P=.27; PT: P=.27; OT+PT: P=.27).

MLR Analysis for Changes in Function for Patients With Dementia Only

The main findings from this MLR analysis are summarized in Table 3. These findings indicate that within the subsample of patients with dementia none of these interventions had an impact on changing physical function (OT: β, .01; 95% CI, −0.006 to 0.03; P=.14; PT: β, −.01; 95% CI, −0.06 to 0.03; P=.14; OT+PT: β, −.01; 95% CI, −0.05 to 0.03; P=.59).

Table 3. Multiple Linear Regression Analyses of OT, PT, and OT+PT for Patients With Dementia in Minimum Data Set 3.0 (N=6396)

Abbreviations: OT, occupational therapy; PT, physical therapy; SNF, skilled nursing facility.

For people with dementia, age, sex, and delirium were not significant for change in function across all intervention conditions (OT: P=.53; PT: P=.55; OT+PT: P=.91).

The main finding of our MLR analyses is that the SNF rehabilitative therapies (OT, PT, and OT+PT) did not have any significant effect on changes in physical function among patients with or without dementia in the MDS 3.0 study sample. In the initial MLR of the whole sample, sex and dementia were significantly associated with change in function. However, in the second MLR, including only those with dementia admitted to SNF for rehabilitation, delirium, age, and sex were not significant for changes in function. However, White race was significantly associated with improved function and discharge to community.

These findings suggest that the rehabilitative therapies do not appear to modify function in the general population of those admitted to SNFs or in a subset population of patients with dementia. The findings are similar to the work of others who have described poorer outcomes in the FFS system that focused on intensity of therapies.^4,10 However, it is important to note differences. This work focused on all Massachusetts patients admitted from postacute care to SNFs in 2013 and did not exclude persons based on any diagnoses. In research that focused on patients undergoing total hip replacements, Kumar and colleagues⁴ reported improved outcomes such as length of stay, and discharge to the community as a result of SNF-based rehabilitative services. Because the sample was limited to healthy people undergoing hip replacement, they suggest that future work should include a greater mix of patients. Huckfield and colleagues10 included a more varied sample of patients (eg, poststroke, lower extremity joint replacement, and heart failure). Their findings indicated that SNF-based rehabilitative services resulted in better outcomes in terms of hospital readmission and discharge to the community; however, they did not examine functional outcomes or the impact of treatment intensity.

Treatment intensity and length of stay are all reasonable areas of concern due to rising costs amid variable outcomes, and, appropriately, this has been the focus of critics calling for reform.^3,11 The rising costs are unsustainable. This study reports on a broad sample of patients admitted to SNFs for postacute care following hospitalization. Results indicate that the intensity of rehabilitative therapies did not improve function or final discharge status in the general population of patients admitted to SNFs or in the subset of patients with dementia. These findings suggest there is a need to further examine the root causes that may impact the effect of therapies on patients with dementia, including alignment of the patient’s cognitive abilities and other factors with the rehabilitative treatment plan. While there is a rich body of work supporting several postacute-care-to-home transition models with positive outcomes,^15,16 there is a lack of work focusing on postacute care to SNFs, especially for patients with dementia. The postacute-care-to-home models are individualized and inclusive of family or other supportive resources, which is something the transition from postacute care to SNF lacks.¹⁷ Further, it is important to note that studies to date have not considered the individual as he or she transitions across multiple health care settings or the unique needs of patients with dementia as they transition from home to hospital, and from to SNF to home.

Limitations

There are several limitations to this study. The study focused on modifiable factors, such as the use of OT, PT, and OT+PT, and their potential impact on outcomes related to SNF admission for postacute care. The large MDS provides a wide-lens view that allows us to broadly examine potential issues, such as functional status changes related to therapies. However, at the time we initiated this study, the more recent MDS was not available, and the changes to the payment system had not yet taken place. The data we examined were relevant to our goal of retrospectively examining the impact of rehabilitative services for patients with and without dementia admitted to SNFs for postacute care under the FFS system.

It is noteworthy that the MDS is limited in what it does capture, including potentially important nonmodifiable risk factors, such as demographics. The homogeneity of the sample (ie, predominantly White race) raises questions as to whether structural racism has a role in postacute care disposition. In addition, the MDS does not capture details on culture or gender identity or preference, which is an additional oversight from a health equity perspective. Gender identity is captured only as male or female. Other important individual characteristics that may influence the transfer to an SNF are not captured in the MDS 3.0; these include medical acuity, a calculation of disease burden, prefracture functional status, and the nature of the rehabilitative treatment, and whether it was individualized to the patient’s needs and preferences.

Another limitation of this study was the use of the dichotomous answers (ie, yes or no) that were used to capture dementia status rather than using the continuous Brief Interview for Mental Status (BIMS) scale. Whereas BIMS may not capture dementia status per se and may include such conditions as traumatic brain injury, it does capture the severity of cognitive changes, an important variable not captured in this study and thereby limiting our findings.

It is also not clear from these findings what impact community and family support may have upon the outcomes related to SNF transition for rehabilitative care, including discharge disposition. The sample was limited to Massachusetts, which also limits the generalizability of the findings.

Despite these limitations, the results of our study do support the need to further understand whether more tailored, innovative approaches to therapy would produce different outcomes, especially in light of the new reimbursement structure that is now in place.

This study used the 2013 MDS 3.0 to examine the effects of each of the rehabilitative therapies (OT, PT, and OT+PT) on the change in physical function from admission to quarterly assessment or discharge in patients with and without dementia aged 65 years and older who were admitted to a Massachusetts SNF in 2013 for acute care posthospitalization. Results suggest a need for rehabilitation approaches that are adapted to the complex needs of older adults admitted to SNFs, including those with dementia. These findings indicate a need for more innovative and effective interventions that must be developed to support more successful postacute rehabilitation in SNF care settings and across diverse populations.

Future work should examine the impact of rehabilitation services on the change in physical function from admission to quarterly assessment or discharge in patients with and without dementia aged 65 and older for acute care posthospitalization after changes in reimbursement policies for rehabilitation that were made in 2020. Research is also warranted to understand the differences in care received by those differed from our study sample by race, culture, ethnicity, sex, and gender preference. While the MDS provides a broad picture and allows for a large sample, it is important that it be updated to capture more diverse sample characteristics to understand what may be needed for a more personalized approach to rehabilitation. There is also a need to understand the role that formal and informal community and the involvement of family and friends may have in successful SNF outcomes or discharge disposition. The use of the dichotomous variables to capture the presence of dementia (ie, yes or no) did not allow us to accurate assess the severity of cognitive changes, and so future work should include a measure of dementia severity. Further, given that 86.3% of older adults with cognitive decline have at least one chronic condition, and that 70% of Medicare beneficiaries have two or more chronic conditions,¹⁸ there is a need to determine the impact of rehabilitative services that patients receive in SNFs considering their disease burden.

The findings from this study suggest that rehabilitative therapies did not improve function or discharge status in the larger population of patients admitted to SNFs or the subset population limited to patients with dementia. These findings do not suggest that rehabilitation plans are ineffective, but do suggest the need for further study of current treatment approaches, their effectiness, and possibilities to tailor them to patients’ individual characteristics, needs, and preferences. There is a need to control costs in the health care system, and the findings of this study suggest the need to better align reimbursement with performance, as opposed to a FFS approach. Findings suggest the advisability of aligning reimbursement with rehabilitative approaches that are individualized to patient needs and unique situations, while also considering preferences, to optimize performance outcomes. Furthermore, controlling cost containment measures must consider support in the home or community as well as a host of personal factors that may impact a person’s ability to receive family or community support to return home.

Jane Flanagan, PhD, RN, ANP-BC, AHN-BC, FNAP, FAAN¹ • Marie Boltz, Marie Boltz, PhD, GNP-BC, FGSA, FAAN² • Ming Ji, PhD ³

Affiliations:
¹Department of Nursing, Boston College Wm. F. Connell School of Nursing, Chestnut Hill, MA 02467
²Department of Nursing,Penn State College of Nursing. University Park, PA 16802
³College of Population Health, University of New Mexico. Albuquerque, NM 87131

Disclosures:
The authors report no relevant financial relationships.

Funding: 
J.F. and M.B.M. acknowledge partial support from the Donaghue Foundation. The contents of the article are solely the responsibility of the authors and do not necessarily represent the official views of the Donaghue Foundation.

Address correspondence to:
Jane Flanagan
140 Commonwealth Avenuse, Chestnut Hill, MA 02467
Phone: 617-552-8949
Email: flanagjg@bc.edu