The Early Impact of the Inpatient Rehabilitation Facility Prospective Payment System on Stroke Rehabilitation Case Mix, Practice Patterns, and Outcomes

Article Outline

• Abstract
• Background
  • Design of the IRF-PPS
  • Impact of the IRF-PPS
• Methods
  • Methods That Pertain to the Analysis of the Impact of the IRF-PPS
  • Pre- and Post-PPS Periods
  • Medicare and Non-Medicare
  • Two-Way, 3-Way, and Multivariate Analyses
• Results
  • Study Group Characteristics and Impact on Stroke Case Mix
  • Impact on the Utilization and Provision of Stroke Rehabilitation Services
  • Impact on LOS
  • Impact on the Amount, Intensity, and Duration of Physical and Occupational Therapy Services
  • Impact on Outcomes
    • Functional status at discharge
    • Discharge disposition
    • Multivariate analysis
• Discussion
• Conclusions
• Acknowledgments
• References
• Copyright

Abstract 

DeJong G, Horn SD, Smout RJ, Ryser DK. The early impact of the inpatient rehabilitation facility prospective payment system on stroke rehabilitation case mix, practice patterns, and outcomes.

Objective

To determine the early effects of the inpatient rehabilitation facility (IRF) prospective payment system (PPS) on stroke rehabilitation case mix, practice patterns, and outcomes.

Design

Prospective observational cohort study.

Setting

Three IRFs in the United States.

Participants

Consecutively enrolled convenience sample of 539 stroke rehabilitation patients treated between 2001 and 2003 in 3 IRFs.

Interventions

Not applicable.

Main Outcome Measures

Length of stay (LOS), therapy utilization, FIM instrument gain, and discharge destination.

Results

The IRF-PPS had no material short-term effect on stroke rehabilitation case mix and LOS for the study facilities. Facilities shifted physical and occupational therapy resources from those in the most severe case-mix groups (CMGs) to those in the moderate CMGs. Those in the more severe CMGs also experienced a noticeable decline in FIM score gain over the course of the rehabilitation stay. Using multivariate analyses, the authors discerned no major role for the IRF-PPS in explaining pre- and post-PPS differences in utilization and outcome among study facilities.

Conclusions

For the 3 study facilities, IRF-PPS did not materially reshape stroke rehabilitation case mix, utilization, and outcome in the early stages of PPS implementation, apart from the shift in therapy resources from more severely involved stroke patients to moderately involved patients. The study’s findings are limited to 3 facilities, and a longer time horizon is needed to more fully determine the effects of the IRF-PPS.

Key Words:  Prospective payment system , Rehabilitation

INPATIENT REHABILITATION FACILITIES (IRFs) are a major venue for poststroke rehabilitation, and patients with stroke are the second-largest group of people served by IRFs, accounting for nearly 20% of all IRF discharges.¹ For better or for worse, payment systems are major drivers of poststroke rehabilitation care. The largest payer for inpatient rehabilitation care remains the Medicare program. Medicare pays for 65% of all IRF-level stroke care in the United States (Sam Fleming, eRehabData.com, personal communication, August 2, 2005), and its payment systems shape access, utilization, and costs of IRF-level care. In 2002, the Centers for Medicare and Medicaid Services (CMS) implemented a prospective payment system (PPS) for IRFs. Using in-depth stroke rehabilitation data from 3 IRFs across the nation, this article provides a preliminary assessment of the early impact of the IRF-PPS on stroke rehabilitation case mix, practice patterns (ie, length of stay [LOS], service mix, intensity), and short-term outcomes (ie, functional status, discharge disposition).

Back to Article Outline

Background 

The IRF-PPS had been a long time in coming. When Congress initiated the Medicare diagnosis-related group (DRG)–based PPS for short-stay acute care hospitals in 1983, it exempted specialty hospitals (ie, rehabilitation centers, children’s hospitals, psychiatric hospitals, long-term care hospitals) and various postacute venues (eg, skilled nursing facilities [SNFs], home health agencies) from a PPS. Congress left these facilities to be paid on a modified cost basis as provided by the Tax Equity and Fiscal Responsibility Act of 1982 (TEFRA). Both Medicare DRGs and cost-based reimbursement for postacute care led to a rapid expansion of postacute facilities of all types from the mid 1980s to the mid 1990s. In 1997, Congress passed the Balanced Budget Act of 1997 to curb this growth by authorizing the Health Care Financing Administration to establish PPSs for IRFs, SNFs, and home health agencies. Congress later authorized a PPS for long-term care hospitals in the Balanced Budget Refinement Act of 1999 (BBRA 1999). When implementing these legislative mandates, CMS instituted different PPS methods for each postacute setting with different start dates and phase-in periods.

This article examines the early impact of the IRF-PPS on stroke rehabilitation patients, practices, and outcomes. More specifically, it examines the impact on stroke case mix (patient severity, case-mix groups [CMGs]); functional status; severity index; utilization (ie, LOS); the mix, duration, and intensity of therapy services; and outcomes at discharge (ie, functional status, discharge destination). It does not attempt to evaluate the indirect effects of other postacute PPSs on IRF-based stroke rehabilitation. Other postacute PPSs shape the willingness of various postacute providers to enter or exit the stroke rehabilitation market and their willingness to accept certain types of patients and thus indirectly shape the case-mix and practice patterns observed among IRFs. At this early stage, we do not have a clear picture of how the IRF-PPS is shaping stroke rehabilitation care and its outcomes. We do know anecdotally that many IRFs have made adjustments in their programs in the wake of the IRF-PPS, but we do not know how they have adjusted their programs and the effects that these adjustments may have had on the nature of stroke care and its outcomes.

Design of the IRF-PPS 

CMS initially sought to implement a per diem PPS known as resource utilization groups for IRFs as it had in the SNF industry. Instead, with the passage of the BBRA 1999, the IRF industry prevailed on Congress to have CMS implement a per-discharge PPS known as function-related groups (FRGs) based on each patient’s functional profile on admission to rehabilitation. In other words, Medicare would pay IRFs a fixed amount per discharge or per case based mainly on each patient’s functional status at admission.

The FRG approach was originally developed by Stineman et al² with industry backing in the early to mid 1990s. Using rehabilitation LOS as a proxy for resource utilization, Stineman attempted to determine for each impairment group, such as stroke, those patient characteristics (eg, functional status [as measured by the FIM instrument], age) that best explained variation in LOS. Based on the results of this work, the initial FRG system classified rehabilitation patients into 1 of 53 distinct groups according to each patient’s impairment (eg, stroke), functional status (eg, FIM motor score), and—in some instances—age. Subsequent refinements undertaken by Stineman et al,³ the Rand Corporation,¹, ⁴, ⁵ and CMS⁶ eventually led to a 95-group classification system now referred to as CMGs. At the time of this study, there were 14 CMGs for stroke rehabilitation based on a patient’s motor or cognitive FIM scores on admission, and in 7 CMGs, the patient’s age is also taken into account (table 1).⁷

Table 1. Stroke CMGs and CMG Groupings by Relative Tier Weights

Source: Centers for Medicare and Medicaid Services.⁶

One of the later additions to the patient classification system for the IRF-PPS was an adjustment for comorbidities. Providers argued that their patients presented a host of comorbidities that affected resource utilization, as did each patient’s functional status. In short, they argued that the function-based classification system overlooked the medical acuity and comorbidities that also drive resource utilization. The Health Care Financing Administration (now CMS) responded and had its principal contractor, Rand, take another look. Rand found that adding comorbidities did help explain additional variance in resource utilization. The final rule implementing the new PPS ranks each comorbidity according to 1 of 3 tiers of severity specific to each patient’s main impairment (eg, stroke). Thus, each CMG comes with 4 weights—3 for different levels of comorbidity severity and a fourth for no comorbidities.⁶ The CMG weight assigned to each patient depends on the most severe comorbidity the patient presents. Although comorbidities are factored into the new PPS, there is uncertainty, if not controversy, about the current approach used to capture this dimension of patient need.

In addition to the function-based and comorbidity-modified patient classification system, the IRF-PPS also makes adjustments for (1) transfers—patients who are transferred from an IRF to other settings of care, (2) outliers—patients who have exceptionally long LOSs, and (3) interrupted stays—patients whose stay in an IRF is interrupted because of an acute condition that may require a temporary stay in an acute care hospital.

The IRF-PPS also makes adjustments in the per-case payment amount for market- and facility-level characteristics: (1) local wage rates—the payment system adjusts for the relative cost of labor in a given metropolitan statistical area; (2) rural status—the payment system provides an additional 19.14% payment for IRFs located in rural areas; and (3) low-income patient adjustment—the payment system provides additional payment for IRFs serving a disproportionate number of low-income patients.

Before these adjustments, the base rate for an average case in fiscal year 2004 was $12,525. In CMS parlance, this is known as the “conversion factor”—that is, the factor that is converted to a payment amount based on CMG-comorbidity weight, local wage adjustment, rural status, and low-income adjustment.

Impact of the IRF-PPS 

Although considerable research work has been expended on the design of the IRF-PPS, comparatively little research has been published on the probable or actual impact of the IRF-PPS on access, case mix, utilization, costs, outcomes, and other issues such as provider equity, efficiency, financial performance, and gaming. Before IRF-PPS implementation, considerable work was done by both researchers and providers to estimate the probable financial impact of the IRF-PPS using simulation analyses and other techniques.⁸ The chief limitation of this work is that investigators assumed that provider behavior would remain constant and that case-mix and practice patterns would therefore remain static.

It will be years before the many direct and indirect effects of the IRF-PPS on stroke rehabilitation can be fully observed, as we have learned from the implementation of the DRG-based PPS for short-term acute care hospitals. Providers will continue to make adjustments in stroke rehabilitation as they learn from their experiences during the first several years of implementation. Given the IRF-PPS implementation in 2002, there has been little time to report the new payment’s impacts. This article provides an early window into the ways in which a geographically diverse group of 3 stroke rehabilitation providers have altered their case-mix and practice patterns and how these have affected utilization and short-term outcomes.

Back to Article Outline

Methods 

The methodology governing the full PSROP, provided in this supplement by Gassaway et al,⁹ provides a detailed description of the larger study’s participating facilities, patient selection criteria, data collection instruments including their validity and reliability, and a detailed description of the project’s final study group. The methodology is summarized in by Maulden et al.¹⁰ The institutional review boards at Boston University and at each participating IRF approved the study.

Methods That Pertain to the Analysis of the Impact of the IRF-PPS 

The PSROP offers a rare opportunity to examine the early impacts of the IRF-PPS, because patients were enrolled both before and after the implementation of the IRF-PPS in 2002. Three of the 6 facilities enrolled a substantial number of patients with stroke both before and after the implementation of the IRF-PPS. Hence, this analysis is limited to just these 3 facilities. The other 3 facilities enrolled patients predominately before or after the implementation of the IRF-PPS and we chose to exclude these facilities because they did not provide an adequate before-and-after view of how the case mix, practice patterns, and outcomes changed with the implementation of the IRF-PPS at these facilities.

Table 2 outlines the study group size and enrollment before and after the implementation of the IRF-PPS at each of the 3 facilities. These 3 facilities enrolled 567 stroke rehabilitation patients. Of this number, 28 patients had some missing FIM data and therefore could not be classified into 1 of the 14 stroke CMGs. We excluded these 28 patients, leaving a total of 539 patients included in this analysis (see table 2).

Table 2. Study Group Enrollment Pre- and Post-PPS

The 3 IRF facilities provide some geographic diversity—1 on the East Coast, 1 on the West Coast, and 1 in the middle of the United States. All 3 facilities were rehabilitation units in academic health centers.

Pre- and Post-PPS Periods 

The IRF-PPS sought to bring greater equity among IRFs that previously had received widely varying levels of reimbursement under the old TEFRA system and to foster access for potential rehabilitation patients by tailoring the amount of payment to the functional and medical needs of each patient. In presenting our results below, we compare findings in the post-PPS period with those from the pre-PPS period. All 3 IRFs also had a ramp-up period before the IRF-PPS implementation date. This ramp-up period varied from 1 to 6 months. We speculated that IRF behavior with respect to admissions and processes of care might already have started to change during this ramp-up period. Accordingly, we considered 3 time periods for analyses: (1) a pre-PPS period, (2) a post-PPS period, and (3) a practice period in preparing for the IRF-PPS implementation. On closer examination of the data, we determined that nearly all patient and practice parameters during the practice or ramp-up period were nearly the same as those during the pre-PPS period and that the most marked changes, where they were discernable, occurred with the implementation of the PPS—that is, during the post-PPS period. Thus, we folded the ramp-up or practice period into the pre-PPS period and present our results below for only 2 periods, the pre-PPS period and the post-PPS period.

Medicare and Non-Medicare 

The IRF-PPS applies only to stroke patients with Medicare and not to patients covered by other types of health plans. Medicare is the major driver of rehabilitation practice and its requirements and effects are known to spill over to patients covered by other health plans. We tested for Medicare and non-Medicare differences with respect to practice patterns and did not detect sufficient differences to exclude non-Medicare patients from the analyses presented below.

Two-Way, 3-Way, and Multivariate Analyses 

In the results that follow, we examine the changes—pre- and post-PPS—across study group characteristics, medical and functional status, service utilization, and outcomes. In most instances, we added a third dimension to the analyses when examining changes from pre- to post-PPS—namely, the CMG groupings—to help account for case-mix differences. To simplify matters, we grouped patients into mild, moderate, and severe groupings (table 1).

Even in 2-way and 3-way analyses, there may be differences that can be explained only when all potential independent variables are considered concurrently. Thus, we used both ordinary least-squares (OLS) and logistic regression analyses to help explain differences in utilization and outcomes in the pre- and post-PPS periods. We sought to control for patient differences to determine how much of the variance could be explained by the IRF-PPS. We used a stepwise procedure that ceased when no other variables met the .08 level of significance for entry into the model.

One of the challenges in the regression analyses was how to capture the IRF-PPS in our regression models. We took 2 approaches. First, we used a simple dichotomous pre- and post-PPS variable. Second, we considered each facility’s TEFRA limit before PPS. We hypothesized that the effect of the IRF-PPS on utilization and outcome would, in part, be a function of the IRF’s pre-PPS TEFRA limit—that is, we had to take into account how high or how low the TEFRA limit was relative to the expected payment under PPS. To do so, we adjusted each facility’s TEFRA limit by the CMS wage index to account for differences in labor purchasing power across market areas. We applied the CMS wage index to both the labor share of the TEFRA limit and to the entire TEFRA limit, and in both cases, the results were essentially the same: 1 facility had a high adjusted TEFRA limit and 2 of the facilities had low adjusted TEFRA limits relative to their expected payment under PPS. In our model, we hypothesized that facilities having a high adjusted TEFRA limit, for example, had the financial wherewithal to provide a richer mix of therapy, offer longer LOSs, and with additional inputs, produce better outcomes in the pre-PPS period.

Back to Article Outline

Results 

Study Group Characteristics and Impact on Stroke Case Mix 

Table 3 describes the 3-facility study group’s principal characteristics in terms of age, sex, race, type of stroke, payer mix, and each group’s medical and functional profile. By tailoring the level of payment to the functional status and medical complexity of each patient with stroke on admission, the IRF-PPS was designed to encourage IRFs to admit patients based on the functional and medical needs of each patient with stroke. The old pre-PPS, it was thought, encouraged IRFs to admit less complex patients to maximize financial margins.

Table 3. Study Group Characteristics Pre- and Post-PPS Enrollment

Patient Characteristics	Pre-PPS (n=235)	Post-PPS (n=304)	Total⁎ (N=539)	P
Mean age (y)	66.0	65.7	65.8	.828†
Sex (% male)	50.6	50.7	50.7	1.00‡
Race (%)				.214‡
White	60.1	62.2	61.2
Black	26.8	21.1	23.6
Other (including Hispanic)	13.2	16.8	15.2
Payer (%)				.016‡
Medicare	51.9	55.3	53.8
Other	48.1	42.8	45.1
Unknown or missing		2.0	1.1
Type of stroke (%)				.479‡
Hemorrhagic	22.6	25.3	24.1
Ischemic	77.5	74.7	75.9
Mean admission FIM score	62.4	61.0	61.6	.465†
CMG (%)				.093‡
Mild (CMG 101–103)	8.1	13.8	11.3
Moderate (CMG 104–107)	49.4	43.8	46.2
Severe (CMG 108–114)	42.6	42.4	42.5
CMI	1.39	1.42	1.41	.613†
Mean admission CSI§	20.1	21.9	21.1	.142†
Mean days from stroke onset to rehabilitation	10.9	10.7	10.7	.822†

Abbreviation: CMI, case-mix index.

There are 2 ways to examine whether the IRF-PPS encouraged the 3 IRFs to admit patients with stroke with greater functional needs. The first is to examine pre- and post-PPS functional status, as measured by the FIM score at admission. We found that the 3 IRFs combined admitted only marginally more functionally dependent patients with stroke as measured by FIM score at admission. The second is to consider the pre- and post-PPS case-mix distributions by CMG. In this case, we grouped the stroke CMGs into mild (CMG 101–103), moderate (CMG 104–107), and severe (CMG 108–114) groups. Among the 3 facilities represented here, there was a modest shift from the moderate CMG group to the mild CMG group, and the percentage of those in the severe group remained about the same at about 42.5%.

One way to examine whether the IRF-PPS encouraged the 3 IRFs to admit those patients with stroke who had more complex medical needs is to evaluate the pre- and post-PPS patient scores on the Comprehensive Severity Index (the continuous version) at admission. We found that the 3 facilities served a slightly more (but statistically insignificant) medically complex group of patients with stroke in the post-PPS period than they did in the pre-PPS period.

A facility’s case-mix index (CMI) captures, to some degree, both the functional and medical needs of its patients by considering each patient’s CMG assignment (CMG 101–114) and each patient’s tier level assignment within each CMG that takes into account the presumed severity of that patient’s comorbidities. Both a patient’s CMG assignment and tier assignment determine that patient’s case weight (see table 1). Averaging all patient case weights determines a facility’s or group’s CMI, with a higher CMI indicating a more severe case mix. For the study group representing all 3 facilities, we found the CMI relatively unchanged from the pre-PPS period (CMI=1.39) to the post-PPS period (CMI=1.42). We also found little change within each of the 3 facilities represented in the study.

Impact on the Utilization and Provision of Stroke Rehabilitation Services 

The IRF-PPS provides incentives for IRFs to review their practice patterns and processes of care relative to the resources that will be available for each stroke patient given their CMGs and tier assignments. To ascertain the steps taken at each facility with respect to the process of care, we queried each facility’s lead stroke physician. They reported that they made no changes in admission criteria nor did they attempt to achieve a particular case mix. This is confirmed by the results noted above. The stroke physicians reported that they did not establish target LOSs based on PPS apart from the way in which they had always estimated expected LOS in the pre-PPS period. A couple reported that the projected LOS for each CMG gave them an additional benchmark by which to estimate an expected LOS. The 3 facilities already had formal or informal clinical pathways for stroke rehabilitation and did not revisit them in the wake of the IRF-PPS implementation. As a result, they reported no deliberate attempt to alter therapy frequency or intensity.

This said, the facilities reported that they did take steps to evaluate certain care processes, particularly at the front and back ends of a patient’s stay. If they had not already done so, facilities sought to shorten the evaluation and assessment processes to make sure that therapy commenced more quickly and by day 2 whenever possible. At the back end, a patient’s discharge can sometimes be delayed, for example, because of lack of equipment or arrangements at the target destination. Facilities reported that they made attempts to regularly review potential barriers to discharge and to address these barriers well in advance of the projected discharge date. In short, our respondents indicated that the changes were more administrative than clinical.

Impact on LOS 

For the 3 facilities combined, there was virtually no change in overall LOS from the pre- to the post-PPS period. The only marked change was at 1 facility that saw a 3-day decline in LOS, in part due to an increase in patients in the mild CMG group (101–103). Table 4 provides a breakdown in the changes in LOS by CMG. The largest increase, although not statistically significant, in LOS was among patients in the moderate CMG group. As observed in the next section, this group also had the largest increase in therapy time.

Table 4. Mean LOS by Stroke CMG Pre- and Post-PPS

Impact on the Amount, Intensity, and Duration of Physical and Occupational Therapy Services 

Table 5 characterizes the amount, duration, and intensity of therapy rendered. We could not report utilization of speech and language therapy because of incomplete data at one of the 3 facilities. We found that the 3 facilities provided somewhat less physical therapy (PT) and occupational therapy (OT) from the pre- to the post-PPS periods but not to any statistically significant degree.

Table 5. Amount of Rehabilitation Therapy Received by Stroke CMG Pre- and Post-PPS

Therapy by Stroke CMG	Pre-PPS	Post-PPS	Change	P⁎
PT
Mild (CMG 101–103)
Mean min of PT	289.7	267.9	−21.8	.716
Mean LOS (d)	7.1	7.8	0.7	.665
Mean days of PT	5.0	4.8	0.2	.870
Mean min of PT/d	45.2	35.1	−10.1	.084
Moderate (CMG 104–107)
Mean min of PT	547.7	645.0	97.3	.072
Mean LOS (d)	13.5	14.8	1.3	.158
Mean days of PT	8.3	10.0	1.7	.031
Mean min of PT/d	39.8	43.4	3.6	.160
Severe (CMG 108–114)
Mean min of PT	1086.0	894.2	−191.8	.017
Mean LOS (d)	25.2	24.1	−1.1	.496
Mean days of PT	17.2	15.6	−1.6	.214
Mean min of PT/d	42.2	37.5	−4.7	.027
Total
Mean min of PT	755.8	698.7	−57.1	.240
Mean LOS (d)	17.9	17.8	−0.1	.909
Mean days of PT	11.8	11.7	−0.1	.882
Mean min of PT/d	41.3	39.7	−1.6	.336
OT
Mild (CMG 101–103)
Mean min of OT	207.4	259.9	52.5	.428
Mean LOS (d)	7.1	7.8	0.7	.665
Mean days of OT	3.7	4.2	0.5	.672
Mean min of OT/d	28.8	34.8	6.0	.240
Moderate (CMG 104–107)
Mean min of OT	474.3	568.1	93.8	.067
Mean LOS (d)	13.5	14.8	1.3	.158
Mean days of OT	7.1	8.7	1.6	.033
Mean min of OT/d	34.7	36.3	1.6	.498
Severe (CMG 108–114)
Mean min of OT	986.2	782.1	−204.1	.009
Mean LOS (d)	25.2	24.1	−1.1	.496
Mean days of OT	15.2	13.6	−1.6	.198
Mean min of OT/d	38.3	33.3	−5.0	.019
Total
Mean min of OT	670.5	616.3	−54.2	.243
Mean LOS (d)	17.9	17.8	−0.1	.909
Mean days of OT	10.3	10.2	−0.1	.918
Mean min of OT/d	35.8	34.8	−1.0	.531

More important is the very noticeable shift in both PT and OT services from those in the more severe CMGs (108–114) to those in the moderate CMGs (104–107). This is the opposite of what was intended under the IRF-PPS, which seeks to provide a more level playing field across all patients by tailoring the amount of payment to the medical and functional needs of each individual patient. Clearly, this is not case here, where we see a shift in resources from more severely impaired patients to more moderately impaired patients. This finding is what one would have expected under the old TEFRA payment system, where a fixed payment ceiling for all patients accompanied by a bonus payment for staying under the ceiling would clearly favor less-impaired patients. This shift in resources is also evident in the decreased LOSs for those in severe CMGs and the increased LOSs among those in the moderate CMGs.

We also tested the observation made by facility representatives that facilities achieved efficiencies by reducing the number of days spent on assessment and evaluation activities or days resulting from administrative barriers to discharge. The mean days spent in therapy declined directly with reduced LOS, and the number of non-PT or non-OT days did not decline as anticipated (see table 5).

Impact on Outcomes 

Table 6 provides a glimpse of the changes in functional outcomes at discharge. We found that overall, FIM scores at discharge and FIM score increase from admission to discharge declined somewhat from the pre- to post-PPS periods. However, these overall changes mask some of the changes among CMG groupings. Patients in the mild CMG grouping showed greater functional gains from the pre- to the post-PPS periods, whereas those in the moderate and severe CMG groupings showed a decline in functional gains from the pre- to post-PPS periods. A closer examination from facility to facility also showed some noticeable changes in the amount of functional gain across the 3 CMG groupings. The CMG subgroups become too small at the individual facility level to make any authoritative observations, except to note that observations in 1 facility tend to be cancelled out by another; thus, generalizations are difficult to make apart from these broader observations.

Table 6. Change in Functional Status from Admission to Discharge, Pre- and Post-PPS

Stroke CMG	Pre-PPS	Post-PPS	Change	P⁎
Mild (CMG 101–103)
Admission FIM	94.7	89.3	−5.4	.039
Discharge FIM	108.7	109.1	0.4	.880
Increase in FIM	14.0	19.8	5.8	.015
Moderate (CMG 104–107)
Admission FIM	72.4	71.5	−0.9	.503
Discharge FIM	99.8	96.9	−2.9	.090
Increase in FIM	27.3	25.4	−1.9	.231
Severe (CMG 108–114)
Admission FIM	44.6	41.1	−3.5	.033
Discharge FIM	78.7	69.3	−9.4	.033
Increase in FIM	33.3	28.2	−5.1	.049
Total
Admission FIM	62.4	61.0	−1.4	.465
Discharge FIM	91.8	87.0	−4.8	.015
Increase in FIM	28.7	25.8	−2.9	.034

Overall, the percentage of patients with stroke discharged to home or to another community setting declined from 82.1% in the pre-PPS period to 79.3% in the post-PPS period, although the decline is not statistically significant. Table 7 indicates that the decline in discharge to a community setting occurred across all 3 CMG groupings. A closer facility-by-facility examination of the data uncovered no facilities that might have had a disproportionate impact on discharge disposition.

Table 7. Discharge Destination by Stroke CMG Pre- and Post-PPS

Table 8 presents the results from the OLS regression analysis for 3 measures of stroke rehabilitation utilization—LOS, total amount of PT and OT (measured in minutes), and intensity of therapy as measured by the minutes of therapy per day. We found that we could explain at best up to 38.5% of the variance in utilization. None of the PPS variables entered any of the 3 regression models. Table 9 presents the results for the OLS regression analysis for the change in FIM and the logistic regression analysis for discharge disposition. Again, the PPS-related variables were not major variables that explain utilization and played a secondary role in explaining the increase in FIM score.

Table 8. OLS Regressions for Stroke Rehabilitation Utilization (n=534)

Abbreviation: Coeff, coefficient.

Table 9. OLS and Logistic Regressions for Stroke Rehabilitation Outcomes (n=534)

Back to Article Outline

Discussion 

This study provides an in-depth view of how 3 IRFs responded to the IRF-PPS in the short term. We conclude that, for these 3 facilities, the IRF-PPS did not materially reshape stroke rehabilitation case mix, utilization, and outcome in the early stages of PPS implementation, apart from the shift in therapy resources from more severely impaired patients with stroke to moderately impaired patients. The study’s 3 IRFs reported that they made several administrative adjustments in how patients were processed to achieve greater efficiencies and reduce the number of nontherapy days, but this observation is not borne out in the utilization and outcome data reported here. The IRFs did, of course, have to train staff to comply with the new payment system’s reporting requirements. We did observe some facility-to-facility variations that pretty much cancelled each other out when examining the effects on the entire study group or within similar CMGs. The results of this study do not support the notion that providers would reduce services overall, although we did detect some shifts in resources among patient groups.

This study’s chief advantage is that it provides detailed information about the amount of therapy services received relative to the medical and functional status of each patient with stroke during an important juncture in postacute payment policy. The study’s chief limitation, however, is that it examines the effects of the IRF-PPS among only 3 facilities. Although geographically diverse, we can make no claims as to the representativeness of these facilities relative to the 1200 IRFs in the United States.

Back to Article Outline

Conclusions 

In examining the effects of the IRF-PPS, we need to observe changes over a much longer period of time. There is a learning curve associated with every new payment and policy change, and it could be argued that the 3 IRFs represented in this study were still at the beginning stages of the learning curve. Nonetheless, providers generally are acutely aware of how payment systems affect their fiscal well-being. During the ramp-up period for the IRF-PPS, both the IRF industry and individual facilities conducted simulation analyses to determine how they would fare under the new payment system, given industry-wide and facility case mixes. Although individual facilities made numerous preparations for the implementation of the IRF-PPS, these preparations do not appear to have materially reshaped clinical practice in the short-run apart from the shifts observed in this analysis.

Back to Article Outline

Acknowledgments 

We acknowledge the role and contributions of the collaborators at each of the clinical sites represented in the Post-Stroke Rehabilitation Outcomes Project: Brendan Conroy, MD (Stroke Recovery Program, National Rehabilitation Hospital, Washington, DC); Richard Zorowitz, MD (Department of Rehabilitation Medicine, University of Pennsylvania Medical Center, Philadelphia, PA); Jeffrey Teraoka, MD (Division of Physical Medicine and Rehabilitation, Stanford University, Palo Alto, CA); Frank Wong, MD, and LeeAnn Sims, RN (Rehabilitation Institute of Oregon, Legacy Health Systems, Portland, OR); Murray Brandstater, MD (Loma Linda University Medical Center, Loma Linda, CA); and Harry McNaughton, MD (Wellington and Kenepuru Hospitals, Wellington, NZ).

Back to Article Outline

References 

1. Carter GM , Beeuwkes-Buntin M , Hayden O , et al.   Analysis for the initial implementation of the inpatient rehabilitation facility prospective payment system . Santa Monica: Rand, Centers for Medicare and Medicaid Services; 2002;
   • View In Article
2. Stineman MG , Escarce JJ , Goin JE , Hamilton BB , Granger CV , Williams S . A case-mix classification system for medical rehabilitation . Med Care . 1994;32:366–379
   • View In Article
   • MEDLINE
   • CrossRef
3. Stineman MG , Tassoni CJ , Escarce JJ , et al.   Development of FRGs 2.0 (a classification system for medical rehabilitation) . Health Serv Res . 1997;32:529–548
   • View In Article
   • MEDLINE
4. Carter G , Relles D , Buchanan J , Donyo T , Inkelas M , Spritzer K . A prospective payment system for inpatient rehabilitation. Project memorandum prepared for the Health Care Financing Administration . Santa Monica: Rand/UCLA/Harvard Center for Health Care Financing Policy Research; 1997; Publication no. PM-683-HCFA.
   • View In Article
5. Carter GM , Relles DA , Wynn BO . Interim report on an inpatient rehabilitation facility prospective payment system . Santa Monica: Rand; 2000; Publication no. DRU-2309.
   • View In Article
6. Centers for Medicare and Medicaid Services. Medicare program; prospective payment system for inpatient rehabilitation facilities; final rule. 66 Federal Register 2001 (41315).
   • View In Article
7. Centers for Medicare and Medicaid Services. Medicare program; prospective payment system for inpatient rehabilitation facilities; final rule. 68 Federal Register 2001 (45704).
   • View In Article
8. Sutton JP , DeJong G , Song H , Wilkerson D . The impact of a function-based payment model on the financial performance of acute inpatient medical rehabilitation providers (a simulation analysis) . Arch Phys Med Rehabil . 1997;78:1290–1297
   • View In Article
   • Abstract
   • Full-Text PDF (1180 KB)
   • CrossRef
9. Gassaway J , Horn SD , DeJong G , Smout RJ , Clark C , James R . Applying the clinical practice improvement approach to stroke rehabilitation (methods used and baseline results) . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S16–S33
   • View In Article
   • MEDLINE
10. Maulden SA , Gassaway J , Horn SD , Smout RJ , DeJong G . Timing of initiation of rehabilitation after stroke . Arch Phys Med Rehabil . 2005;86(12 Suppl 2):S34–S40
    • View In Article
    • MEDLINE