Long-Term Outcomes of Joint Replacement Rehabilitation Patients Discharged From Skilled Nursing and Inpatient Rehabilitation Facilities

Article Outline

• Abstract
• Methods
  • Participating Facilities
  • Study Group
  • Measures and Instruments
  • Data Collection
    • Facility-level data
    • Patient-level data
  • Data Analysis
    • Descriptive analyses
    • Bivariate analyses
    • Duration of follow-up period
    • Multivariate analyses
• Results
  • Study Facilities
  • Study Group
    • Final study sample
    • Comparing study participants with nonparticipants
    • Study group characteristics across settings
  • Outcomes
    • Bivariate comparisons
    • Multivariate analyses
• Discussion
  • Implications for Practice and Policy
  • Study Limitations
• Conclusions
• Acknowledgments
• References
• Copyright

Abstract 

DeJong G, Tian W, Smout RJ, Horn SD, Putman K, Hsieh C-H, Gassaway J, Smith P. Long-term outcomes of joint replacement rehabilitation patients discharged from skilled nursing and inpatient rehabilitation facilities.

Objective

To examine functional and health status outcomes of patients with joint replacement discharged from a skilled nursing facility (SNF) or an inpatient rehabilitation facility (IRF).

Design

Postdischarge follow-up interview study at 7.5 months after admission.

Setting

Five freestanding SNFs, 1 hospital-based SNF, and 6 IRFs.

Participants

Patients (N=856): 561 with knee replacement and 295 with hip replacement.

Interventions

None.

Main Outcome Measures

FIM and Short-Form 12-Item Health Survey (SF-12).

Results

Among patients with knee and hip replacement, IRF patients made larger motor FIM gains from admission and discharge to follow-up. IRF patients, however, were admitted with lower FIM scores and also had more to gain (especially given the ceiling effects within the FIM at follow-up). When adjusted for case mix, IRF patients made larger motor FIM gains and had higher SF-12–related scores among patients with hip replacement but not among patients with knee replacement. Multivariate regressions found modest setting effects that favored IRFs, and the setting effects explained only a modest portion of the variance in motor FIM outcomes.

Conclusions

At follow-up, patients with joint replacement discharged from IRFs had better motor FIM outcomes than those discharged from freestanding SNFs and the hospital-based SNF. Settings did not differ materially in terms of SF-12 outcomes. Findings do not favor one setting decisively over another. A sole focus on initial postacute placement overlooks the larger trajectory of postacute care that needs to be managed to achieve superior outcomes.

Key Words: Arthroplasty, replacement, hip, Arthroplasty, replacement, knee, Rehabilitation, Rehabilitation centers, Skilled nursing facilities, Treatment outcome

List of Abbreviations: BMI, body mass index, CMG, case-mix group, CMS, Centers for Medicare and Medicaid Services, CSI, Comprehensive Severity Index, ER, emergency room, HMO, Health Maintenance Organization, IRF, inpatient rehabilitation facility, JOINTS I, Joint Replacement Outcomes in Inpatient Rehabilitation Facilities and Nursing Treatment Sites, JOINTS II, follow-up study to Joint Replacement Outcomes in Inpatient Rehabilitation Facilities and Nursing Treatment Sites (JOINTS I), LOS, length of stay, PCS, physical component summary, PPS, prospective payment system, SF-12, Short-Form 12-Item Health Survey, SNF, skilled nursing facility

STUDIES HAVE EXAMINED the functional outcomes of rehabilitation patients with knee or hip replacement on discharge from a postacute facility such as an SNF or an IRF.¹, ², ³, ⁴ Presumably, rehabilitation in an SNF or an IRF should have sustained effects beyond discharge with respect to patient function, health, and well being. Providers and policy makers alike want to know the extent to which outcomes differ among SNF and IRF patients at follow-up.

To date, only 1 study has attempted to examine the postdischarge outcomes of rehabilitation patients with a knee or hip replacement across postacute settings. Using Medicare claims data from January 2002 to June 2003, Buntin⁵ evaluated the postdischarge outcomes of joint replacement patients discharged from SNFs and IRFs. Buntin⁵ examined outcomes within 60-day and 120-day windows beginning with the day of discharge from an acute care hospital. Use of Medicare administrative data, however, did not allow the study to examine health or functional outcomes apart from death and institutionalization rates, both of which are low-frequency events for patients with joint replacement that minimized differences between SNF and IRF outcomes. This article examines a larger array of health and functional outcomes.

This article reports the findings of a 6-month to 9-month follow-up study that examined functional and health status outcomes of joint replacement patients discharged from 3 types of facilities—freestanding SNFs, a hospital-based SNF, and IRFs. The article addresses 2 questions: (1) what are the follow-up outcomes of patients with joint replacement discharged from each of the 3 types of facilities, and (2) how are patient-related and setting-related differences associated with follow-up outcomes? A companion article examines postdischarge use of rehabilitative and other health services and incidence of medical complications in the study's follow-up period.⁶

Back to Article Outline

Methods 

The follow-up study, known as JOINTS II, was a sequel to the JOINTS I study, a prospective observational cohort study that examined functional outcomes at discharge of 2152 patients with joint replacement discharged from SNFs and IRFs. JOINTS I patients were enrolled consecutively from February 2006 through February 2007. JOINTS II was a follow-up study of a subsidiary cohort of about 1000 patients interviewed several months later between August 2006 and October 2007.

Participating Facilities 

Study facilities for the follow-up study were a subset of the 22 facilities that participated in the original JOINTS I study examining outcomes at discharge.⁷, ⁸ Thus, facility selection and recruitment processes that governed the original JOINTS I study also shaped the types of facilities that agreed to participate in JOINTS II, the follow-up study. The study's most important facility recruitment goal was geographic diversity. The original JOINTS I cohort consisted of facilities from each of the nation's 4 major census regions, a mix of freestanding and hospital-based facilities, a mix of for-profit and not-for-profit facilities, and facilities in markets with high and low managed-care penetration. The original JOINTS I study sought facilities that could bring at least 200 or more patients into the study, but only a few SNFs could meet this 200-patient goal. Within these recruitment parameters, facility participation was entirely voluntary, and thus neither JOINTS I nor JOINTS II facilities comprised a national probability sample of facilities, nor were they intended to be.

Study Group 

To be included in both JOINTS I and JOINTS II studies, patients had to be (1) 21 years or older, (2) admitted pursuant to hip or knee replacement of any type, and (3) admitted from any source. The results here exclude patients who (1) had a hip replacement subsequent to a hip fracture, (2) died before the interview with them could be conducted, although a limited interview was conducted with a family member, (3) had an additional joint replacement in the follow-up period for a different joint, or (4) had a postacute care rehabilitation LOS longer than 52 days.

The analyses here exclude patients with hip fracture because their demographic profile was different (older and more female), the etiology of their condition was different, and their joint replacements were not elective. We excluded the few who died because death rates were no greater than population-based death rates and because, in several instances, there was no respondent who could provide adequate information about events during the follow-up period that might influence follow-up outcomes. We excluded 2 patients with an LOS of 52 days or more because their LOS was much longer than the nearest patient. They were both SNF patients and accounted for less than .25% of all patients in the study group.

Measures and Instruments 

Table 1 describes the dependent and explanatory variables used in this study. The table lists 3 groups of explanatory variables: (1) those related to the status of the patient on admission, (2) those related to the inpatient setting and the services received, and (3) those related to the patient's postdischarge use of rehabilitative and other health services. We used 2 key patient assessment instruments in JOINTS II to measure dependent variables: the FIM and the SF-12.⁹, ¹⁰, ¹¹, ¹²

Table 1. Study Variables, Their Measurement and Source in JOINTS II Follow-Up Study

Variable	Measures and Coding	Source
Admission status variables
Patient demographics	Age: continuous	Medical record
	Sex: 1=female, 0=male
	Race: black (1=yes 0=no), white (1=yes 0=no), others (1=yes 0=no)
Payer	Medicare: 1=Medicare non-HMO, 0=others
Health and functional status	Bilateral replacement: 1=yes, 0=no
	Revision, measured as whether patient had previous artificial joint replaced: dichotomous, 1=yes, 0=no
	BMI: continuous
	Admission CSI: continuous
	Admission motor FIM: continuous
	Admission cognitive FIM: continuous
	CMG: 801–802 (1=yes 0=no), 803–804 (1=yes 0=no), 805–806 (1=yes 0=no)
	Case-mix relative weight: continuous
Inpatient stay variables
Process of care	Onset days: number of days from surgery to admission to settings	Medical record
	LOS: continuous
	Intensity: PT with OT hours a day: total hours of PT and OT divided by LOS	Point-of-care documentation
Facility characteristics	Facility type: Fs-SNF (1=yes 0=no), IRF (1=yes 0=no), Hb-SNF (1=yes 0=no)	Facility questionnaire
	Facility volume (measured by the number of all patients with joint replacement served by facility in 2006): low volume (1=20–45, 0=other); medium volume (1=100–183, 0=other); high volume (1=272–347, 0=other); very high volume (1=604–606, 0=other)
	Site variables: 12 dummy variables (0/1) for each of 12 sites
Postdischarge follow-up variables
Follow-up period	Follow-up period: number of months from postacute admission to being interviewed, continuous	JOINTS II follow-up survey
Health care use	Hospitalization: 1=yes, 0=no
	ER use: dichotomous, 1=yes, 0=no
	Number of outpatient therapy visits: continuous
	Number of home therapy visits: continuous
	Total number of outpatient and home therapy visits: continuous
Help at home	Help available at home, 1=yes, 0=no
Outcomes	Motor FIM: continuous	Medical record and phone FIM survey conducted as part of JOINTS II follow-up survey
	Follow-up motor FIM gain: follow-up motor FIM minus discharge motor FIM
	Total motor FIM gain: follow-up motor FIM subtracts admission motor FIM
	SF-12 items	JOINTS II follow-up survey.
	SF-12 PCS: continuous
	SF-12 mental component score: continuous
	General health: excellent=5, very good=4.4, good=3.4, fair=2, poor=1
	Physical health: excellent=5, very good=4.4, good=3.4, fair=2, poor=1
	Satisfied with quality of life: excellent=5, very good=4.4, good=3.4, fair=2, poor=1
	Limited moderate activities: limited a lot=1, limited a little=2, not limited at all=3
	Limited stair climbing: limited a lot=1, limited a little=2, not limited at all=3
	Work accomplished less than you would like: 1=yes, 2=no
	Limited in work or activities: 1=yes, 2=no
	Had a lot of energy: scale, all of the time=6, most of the time=5, a good bit of time=4, some of the time=3, a little of the time=2, none of the time=1
	Living at home: 1=yes, 0=no
	Interfered with social activities by pain: scale, all of the time=1, most of the time=2, some of the time=3, a little of the time=4, none of the time=5
	Experienced pain? 1=yes, 0=no
	Pain score: Likert scale, 0–10
	Had a fall: 1=yes, 2=no
	No. of times fallen (only for those who had a fall): continuous

Abbreviations: Fs-SNF, freestanding SNF; Hb-SNF, hospital-based SNF; PT, physical therapy; OT, occupational therapy.

The FIM served as the study's principal measure of patient functional status in both JOINTS I and II. We obtained FIM scores on admission to, and discharge from, postacute care (JOINTS I), and at follow-up (JOINTS II). To ensure that the FIM was administered properly in each study facility, IT HealthTrack, an independent FIM training and follow-up survey organization, assisted in training the clinical staff at each of the study sites. Freestanding SNF clinicians, who do not use the FIM as routinely as IRF clinicians, completed a 3-day training session in which they were taught how to use the instrument. Each clinician was required to score 100% on a pencil-and-paper examination that tested their knowledge of the FIM and its uses. This followed the FIM training procedures required of IRF clinicians. IT HealthTrack's trained nurse interviewers administered the phone version of the FIM at follow-up. The telephone version of the FIM is considered both valid and reliable.¹³, ¹⁴, ¹⁵, ¹⁶

Although the FIM includes both motor and cognitive FIM items, we used only the motor FIM as the main measure of follow-up outcome because joint replacement is mainly a motor function issue and because the cognitive FIM demonstrated significant ceiling effects on discharge and even more so at follow-up. The motor FIM consists of 13 items, and patient scores can range from 13 to 91.

The SF-12, derived from the Medical Outcomes Study 36-Item Short-Form Health Survey,⁹, ¹² is one of the most widely used patient assessment instruments in clinical and health services research today. The SF-12 was administered only at follow-up and was embedded in the follow-up interview protocol administered by IT HealthTrack. The SF-12 consists of 2 subscales: (1) the PCS and (2) the mental component summary. Because physical, not mental, recovery is the main goal of joint replacement, we focused on the PCS to measure the overall physical well being of the patient at follow-up. The PCS includes the following SF-12 items: (1) perception of general health, (2) participation in moderate activities, (3) climbing several flights of stairs, (4) accomplishing less than one would have liked because of physical health, (5) being limited in work or other activities because of physical health, and (6) bodily pain. The scale of each item varies from 1 to 6, and a series of transformation formulas was used to compute the PCS scores.

We also used the CSI as the study's main severity adjuster. The CSI provides a detailed, disease-specific measure of severity of illness and acuity. The CSI takes into account over 2200 potential variables. These include patient demographic characteristics, medical history, physiologic parameters, laboratory findings, and a large variety of patient signs, symptoms, and physical findings. The CSI score reduces the patient's physiologic and psychosocial complexity to a single overall score based on the extent and interaction of all the patient's various health conditions. The CSI, in this instance, is expressed as a continuous score with no upper limit. For purposes of this follow-up analysis, we used the patient's admission CSI score as the measure of patient acuity. The predictive validity of the CSI has been demonstrated in several rehabilitation and long-term care–related studies.¹⁷, ¹⁸, ¹⁹, ²⁰, ²¹, ²²

Data Collection 

Data on facility characteristics came mainly from 2 sources: (1) the provider of service files compiled by CMS for each Medicare-certified provider from CMS's Online Survey and Certification Reporting System to which individual study facilities report key facility characteristics, and (2) a facility characteristics questionnaire sent to each study facility to validate and supplement the data obtained from the CMS provider-of-service files. These 2 sources provided data on a facility's geographic location, profit status, bed size, volume type, number of beds dedicated to orthopedic or joint replacement patients, occupancy rate, availability of physician and pharmacy services, payer mix, staffing ratios, and more (see table 1 in DeJong et al⁶).

Patient-level data came from 2 main sources: the JOINTS I database and the JOINTS II telephone follow-up survey. The JOINTS I study database included data obtained from the patient's chart and data obtained at the point of care. Using study identification numbers, the study team linked JOINTS I data with the telephone follow-up survey data. Combined, these 2 data sources provided an in-depth profile of (1) the patient's status at 3 points in time (admission, discharge, follow-up) and (2) the patient's use of rehabilitation and other health care services for those participating in both JOINTS I and II.

IT HealthTrack conducted the study's follow-up interviews. The 50-question JOINTS II follow-up telephone interview asked patients about their general health, living situation and social support, functional status, depression and mood, pain, community participation and employment, additional joint replacements, postdischarge medical complications, and all postdischarge health care use including physician visits, hospitalization, ER use, nursing home residential care, home-based rehabilitation, outpatient rehabilitation, and any additional rehabilitation in a freestanding SNF, hospital-based SNF, or IRF. The telephone questionnaire included both the SF-12 and the telephone version of the FIM. All nurse interviewers were certified in FIM administration. Interviews lasted from 15 to 35 minutes depending on the number of skip patterns that applied to a given respondent.

The study was aided by the study's clinical practice team, which consisted of clinical representatives from each of the participating sites. With the clinical practice team's assistance, we repeatedly revised and pretested the follow-up questionnaire prior to implementation.

Data Analysis 

We conducted our data analysis in the classic 3-step fashion: (1) descriptive analysis using measures of central tendency and frequency, (2) bivariate analyses with and without case-mix adjustment, and (3) multivariate analyses to explain study outcomes.

In characterizing our study facilities, we divided facilities into 3 groups: freestanding SNFs, IRFs, and a hospital-based SNF that had features of both an SNF and an IRF as described by DeJong et al.⁷, ⁸

In characterizing our study patients, we divided them into 2 groups of patients with knee or hip replacement. We also considered whether the patient had an original replacement or a revision and whether it was unilateral or bilateral. We further characterized the study group at admission and discharge across facility types in terms of their demographic profile, medical acuity (CSI), functional status (FIM), presence of comorbidities, CMG, payer mix, and LOS. Because JOINTS II patients were a subset of those participating in JOINTS I, we compared respondents and nonrespondents to address potential selection issues as described in an accompanying article.⁶

Bivariate analyses were conducted to compare follow-up outcomes of patients across settings. Analysis of variance and chi-square tests were used to highlight differences in follow-up outcomes among freestanding SNF, hospital-based SNF, and IRF patients.

To adjust for case-mix differences, we used direct standardization, a classic case-mix method. First, we decided what factors needed to be controlled or adjusted. We used the 6 CMGs used in 2006 by Medicare's IRF-PPS for patients with joint replacement (CMGs 801–806). Although 4 comorbidity tiers are used currently to refine the IRF-PPS CMGs to account for other health conditions that the patient may have, we found no patients with tier 1 comorbidities, the most severe tier, and only a few patients with tier 2 comorbidities. We also found that patients from different settings had a vast range of CSI scores even when they were in the same CMG and tier. We therefore based the case mix on 2 factors: CMG group and CSI score.

Second, we took into account patient distribution across various IRF-PPS CMGs. The distribution of patients across CMG groups showed that there were few patients in some cells. To achieve minimum sample size needed for valid statistical power and a relatively even distribution of patients across cells, we collapsed the 6 CMGs into 3 combined groups: CMG 801 and 802, CMG 803 and 804, and CMG 805 and 806.

Third, although CSI produces a continuous score (and this score was the value that was allowed to enter into the regression analyses in table 5), for direct standardization that was used for case-mix adjustment in table 4, we needed to categorize CSI. Thus, we collapsed the continuous CSI scores into 3 categories using different cutoff points for patients with knee replacement and patients with hip replacement by taking into account the distribution of CSI scores among patients with knee and hip replacement, respectively. CSI scores for patients with knee replacement were collapsed into 3 groups (CSI≤28, 29–46, ≥47). For patients with hip replacement, CSI scores were collapsed (CSI≤20, 21–31, ≥32). Thus, with 3 CMGs and 3 CSI groups, there are theoretically 9 CMG-CSI subgroups for patients with knee replacement and 9 subgroups for patients with hip replacement for a total of 18 condition-specific case-mix subgroups.

Using these 9 + 9 groups, we found insufficient numbers of freestanding SNF and hospital-based SNF patients in our more severe subgroups: 3 knee replacement subgroups and 4 hip replacement subgroups had too few patients from freestanding SNFs and the hospital-based SNF, although there were ample numbers of IRF patients. However, we could not justify collapsing these groups into less severe groups. Thus, in our case-mix adjustment, we omitted all but one of the more severe subgroups. In doing so, we lost 8 freestanding SNF patients, 9 hospital-based SNF patients, and 71 IRF patients to case-mix adjustment. In short, the case-mix adjustment allows us to compare mainly patients represented in the less severe subgroups.

We used IRF patient sample as the reference population and thus computed the outcomes for each stratum as if the freestanding SNF or hospital-based SNF patients had the same case mix as the study IRFs.

Our goal was to obtain 6-month follow-up outcomes, but local circumstances often required a longer follow-up interval. Thus, we had to consider how variation in the time from admission to follow-up might affect patient outcomes. We evaluated the duration of the follow-up periods across the 3 settings and, in the multivariate analyses seeking to explain outcomes, we explicitly allowed duration of the follow-up period to enter as an independent variable in the multivariate analyses.

We used 3 steps to create models for knee replacements and hip replacements separately to examine the influence of potential explanatory variables on 2 of the study's outcome variables at follow-up: motor FIM and PCS of the SF-12. In step 1, we allowed only patient admission status variables including age, sex, race, days from joint replacement to rehabilitation admission (otherwise known as onset days), patient acuity as measured by the patient's CSI score, whether the replacement was a revision, whether it was bilateral, patient's admission motor and cognitive FIM scores, and payment source to enter the model.

In stage 2, we allowed patient and process variables associated with the postacute stay including type of facility (ie, freestanding SNF, hospital-based SNF, IRF), annual facility volume, LOS, and intensity of therapy (ie, physical therapy and occupational therapy hours a day) to enter the model. We also ran regressions omitting LOS and therapy intensity because they were strongly associated with setting and could obscure the setting effect.

In step 3, we added predictor variables from the follow-up period including duration of follow-up period, use of outpatient and home-based rehabilitation services, hospitalizations, and ED visits. We included only hospitalizations and ED visits that were not associated with the original joint replacement. We theorized that if the hospitalization or ED visit was associated with the original joint replacement, it would be a reflection, direct or indirect, of the medical management provided in the postacute setting and would therefore diminish potential setting effects that otherwise might be observed. To include hospitalizations and ED visits related to the joint replacement would remove accountability from the postacute setting.

We ran regressions by comparing facilities in 3 different ways: (1) treating each setting (ie, freestanding SNF, hospital-based SNF, IRF) as distinct, (2) combining the hospital-based SNF with the freestanding SNFs and comparing the combination with IRFs, and (3) omitting the hospital-based SNF and simply comparing freestanding SNFs and IRFs.

At follow-up, there is always the possibility that some patients will reach a maximum score on 1 or more outcome measures. Applying a linear regression would be biased by the fact that it is not possible, for example, for patients to obtain motor FIM score higher than 91. Thus, we used censored regression models to account for potential ceiling values. Censored regression models are commonly used in econometrics in cases in which the variable of interest is only observable under certain conditions.

For all these models, we used a stepwise selection process that allowed variables to enter and leave the model, and evaluated the relative strength of each variable by examining its F value. Only those significant variables (P<.05) were allowed to remain in the model.

Back to Article Outline

Results 

Study Facilities 

Twelve of the original 22 JOINTS I facilities participated in the JOINTS II follow-up study. The 12 JOINTS II facilities included 5 freestanding SNFs, 1 hospital-based SNF, and 6 IRFs. Ten facilities elected not to participate in the JOINTS II follow-up study because of local institutional review board limits on recontacting patients, low patient volumes, and demands on facility staff. One or more freestanding SNFs from each of the 4 major census regions participated in the follow-up study. We were unable to recruit an IRF from the midwest region.

Facilities included in the study represent larger freestanding SNFs and IRFs compared with the average size of facilities nationwide. Two of the 5 freestanding SNFs and 1 of the 6 IRFs had a dedicated orthopedic or joint replacement unit. The proportion of nonprofit facilities in this study was similar to the percentage of nonprofit facilities nationally. However nearly all IRFs and 83.3% of SNFs in this study were freestanding, while only 17.9% of IRFs are freestanding nationally. Hence, freestanding IRFs were overrepresented. None of the JOINTS II freestanding SNFs offered 24-hour onsite physician or pharmacist coverage, while all IRFs and the lone hospital-based SNF did provide these services in 2006 (see table 1 in DeJong et al⁶).

Study Group 

The 12 JOINTS II facilities identified 1326 patients for the follow-up study. The survey research team contacted and interviewed 1016 patients or family members, yielding a response rate of 76.6%. For purposes of this article, we excluded (1) 85 patients who had a hip replacement subsequent to a hip fracture, (2) 15 patients who died (although a limited interview was conducted with a family member), (3) 35 patients who had an additional joint replacement in the follow-up period for a different joint, and (4) 2 patients whose LOS was longer than 52 days. The study team also excluded 20 patients who had been interviewed for the wrong rehabilitation stay.

The final sample consisted of 856 patients: 561 patients with knee replacement and 295 patients with hip replacement (table 2). In the overall study group, 65.5% of the patients had a knee replacement and 34.5% had a hip replacement. Moreover, 22.2% of the entire study group came from freestanding SNFs, 23.3% from the hospital-based SNF, and the remaining 54.5% of patients from IRFs. Compared with freestanding SNFs, IRFs served a larger percentage of patients with knee revisions (5.4% and 2.4%, respectively) and hip revisions (15.2% and 7.7%, respectively). IRFs also served a higher percentage of patients with bilateral knee replacements than did freestanding SNFs and the study's hospital-based SNF (18.4% vs 7.1% and 1.4%, respectively). Only 2.9% of IRF patients had bilateral hip replacements, but none of the patients served in freestanding SNFs or the hospital-based SNF had such a procedure.

Table 2. Characteristics of Patients With Knee and Hip Replacements^⁎

Characteristic or Outcome	Knee Replacement				Hip Replacement
	Fs-SNF (n=127)	Hb-SNF (n=140)	IRF (n=294)	P	Fs-SNF (n=65)	Hb-SNF (n=59)	IRF (n=171)	P
Demographics
Age (y)	71.1±9.8	72.5±7.1	71.5±9.5	.399	72.4±11.7	74.3±6.1	70.0±11.7	.022
Sex (% female)	78.7	71.4	72.5	.322	73.9	67.8	68.4	.686
Race (% white)	58.3¶	98.6	74.2	<.001	40.0¶	100.0	84.8	<.001
Lived alone (%)	37.8	33.6	30.3	.312	41.5	28.8	33.9	.318
Payer mix (% non-HMO Medicare)	78.7	79.3	81.0	.845	89.2	93.2	74.9	.002
Admission status
Bilateral replacement (%)	7.1	1.4	18.4	<.001	0.0	0.0	2.9	.158
Revision (%)	2.4	4.3	5.4	.370	7.7	8.5	15.2	.179
Adm CSI†	27.2±11.0	28.9±7.8	31.4±14.6	.005	20.8±7.7	15.8±9.2	20.4±12.6	.015
BMI	34.0±8.0	34.5±7.8	31.5±6.6	<.001	31.0±5.7	30.6±6.6	30.2±6.2	.635
Adm motor FIM	50.3±8.7	46.9±8.3	40.9±7.6	<.001	48.4±7.4	46.3±8.6	38.4±7.6	<.001
Adm cognitive FIM	34.4±2.9	30.5±4.4	30.1±5.2	<.001	34.6±1.5	29.9±3.2	29.1±5.3	<.001
Adm high pain score	3.8±2.5	6.3±2.5	7.2±2.2	<.001	2.9±2.3	5.3±2.8	6.3±2.8	<.001
Onset days‡	3.7±2.3	3.6±1.7	4.2±3.6	.073	3.2±2.0	3.6±1.5	3.9±1.5	.015
CMG (%)
801–802	71.7	70.7	39.8		63.1	67.8	28.1
803–804	22.1	22.1	40.5	<.001	30.8	18.6	40.9	<.001
805–806	6.3	7.1	19.7		6.2	13.6	31.0
Comorbidities (%)
Morbid obesity (BMI≥40)	19.7	19.3	10.9	.017	9.2	6.8	4.7	.414
Hypertension	74.8	81.4	74.8	.280	76.9	72.9	63.7	.108
Diabetes	26.8	27.1	23.5	.633	16.9	20.3	18.1	.883
Ischemic heart disease	11.8	21.4	16.7	.110	10.8	23.7	18.1	.161
Tier 1 (most severe)∥	0.0	0.0	0.0		0.0	0.0	0.0
Tier 2 (moderately severe)	0.0	2.9	2.4	<.001	0.0	0.0	0.6	.463
Tier 3 (mild)	7.1	15.7	28.9		7.7	11.9	15.8
Tier 0 (none on list of comorbidities)	92.9	81.4	68.7		92.3	88.1	83.6
Discharge status
Discharge to community %§	97.6	99.3	96.3	.183	100.0	96.6	94.7	.161
Discharge motor FIM	70.6±7.5	69.8±6.7	68.0±6.5	<.001	70.2±7.5	69.3±6.5	67.3±6.2	.006
Change in motor FIM	20.2±7.3	22.9±8.0	27.2±6.9	<.001	21.7±6.7	23.0±7.8	29.0±6.5	<.001
Follow-up health care use
Home therapy visits	6.8±6.5	2.9±4.9	4.5±6.4	.000	8.1±6.8	3.9±6.6	5.5±7.2	.003
Outpatient therapy visits	12.2±12.8	16.0±11.7	15.0±15.6	.072	6.4±9.4	6.3±13.8	9.4±12.5	.107
Total therapy visits	19.0±14.0	19.0±11.6	19.5±15.2	.916	14.6±11.6	10.2±14.1	14.9±12.6	.045
LOS	13.7±6.5	8.6±3.1	9.7±3.3	<.001	13.3±8.6	9.0±3.6	10.1±4.1	<.001

Source: JOINTS II study.

NOTE. Values are mean ± SD or as otherwise indicated.

Abbreviations: Adm, admission; Fs-SNF, freestanding SNF; Hb-SNF, hospital-based SNF.

The JOINTS II follow-up cohort was similar to those who only participated in the JOINTS I study, with a few notable exceptions. For example, JOINTS II freestanding SNF patients presented more medical acuity on admission than their JOINTS I counterparts (CSI=27.2 vs 21.7; P<.001). JOINTS II freestanding SNF patients with knee replacement also had more bilateral replacements (6.8%) than their JOINTS I–only peers (2.3%). JOINTS II IRF patients with knee replacement, compared with their JOINTS I–only counterparts, had lower FIM scores on admission and achieved greater FIM gains on discharge.

Table 2 partitions the study group by type of replacement (hip or knee) and by facility type (freestanding SNF, hospital-based SNF, IRF), and compares the patients' demographic, medical, and functional profile across the 3 settings.

Patients with knee replacement in all 3 settings were typically women in their early 70s. About one third of patients lived alone prior to surgery, and about 80% were paid by non-HMO or fee-for-service Medicare. The percentage of white patients in freestanding SNFs was lower than that in IRFs and the hospital-based SNF. In general, IRF patients had a more severe medical and functional profile on admission than did freestanding SNF and hospital-based SNF patients. IRF patients had higher admission CSI scores, lower admission motor and cognitive FIM scores, higher pain scores at admission, and more onset days, and were less likely to fall into CMG 801 and 802. IRF patients with knee replacement were more likely to have had a bilateral replacement and more likely to have a tier 1 or tier 2 comorbidity (as defined under the IRF-PPS patient classification system) compared with freestanding SNF and hospital-based SNF patients. However, freestanding SNF and hospital-based SNF patients had higher BMI scores and a higher percentage were morbidly obese (BMI≥40) than IRF patients. IRF patients were discharged with lower motor FIM discharge scores but achieved greater gains in motor FIM than did freestanding SNF and hospital-based SNF patients.

Patients with hip replacement were also typically women in their early 70s. However, hospital-based SNF patients were somewhat older, more likely to be white, and more likely to be paid by non-HMO Medicare than were their freestanding SNF and IRF counterparts. Across the 3 settings, patients with hip replacement presented significant differences in severity and functional status on admission and discharge. IRF patients with hip replacement had lower admission motor FIM and cognitive FIM scores, higher maximum pain scores at admission, and more onset days, and were more likely to be classified in CMG 805 and 806 than were freestanding SNF and hospital-based SNF patients. Compared with freestanding SNF and hospital-based SNF patients, IRF patients achieved the lowest discharge motor FIM scores but achieved the highest gain in motor FIM scores. For both knee and hip replacement, hospital-based SNF patients had the shortest average LOS, while freestanding SNF patients had the longest.

Outcomes 

Unadjusted for case mix and duration of follow-up period, all 3 settings had similar follow-up outcomes; there were few statistically significant differences (table 3). Among patients with knee replacement, hospital-based SNF patients had the highest follow-up motor FIM scores, while freestanding SNF patients had the lowest and IRF patients were in the middle. IRF patients made the largest motor FIM gains from admission or discharge to follow-up.

Table 3. Follow-Up Outcomes of Patients With Joint Replacements Prior to Case-Mix Adjustment^⁎

Outcomes	Patients With Knee Replacement				Patients With Hip Replacement
	Fs-SNF (n=127)	Hb-SNF (n=140)	IRF (n=294)	P	Fs-SNF (n=65)	Hb-SNF (n=59)	IRF (n=171)	P
Months from admission to follow-up interview	7.8±1.2	7.4±1.5	8.0±1.7	<.001	7.6±0.8	7.3±1.2	8.2±1.9	.003
Physical health now (%)				.392				.121
Much-somewhat better	81.9	74.3	79.6		83.1	71.2	84.2
About the same	13.4	20	13.6		15.4	20.3	12.9
Somewhat-much worse	4.7	5.7	6.8		1.5	8.5	2.9
Pain score (%)				.107				.366
0	59.1	67.1	57.1		72.3	67.8	66.1
1–5	35.4	25.7	31.6		26.2	23.7	24.6
6–10	5.5	7.1	11.2		1.5	8.5	9.4
SF-12 scores
PCS	48.1±13.2	51.1±14	48.9±14.4	.176	49.6±12.1	48.9±15.8	49.2±15.0	.961
MCS	44.3±8.4	44.8±7.6	45.0±7.5	.691	46.5±6.3	45.3±6.7	45.1±8.4	.406
Motor FIM	85.5±6.0	87.2±4.2	86.6±4.7	.014	86.2±3.8	85.7±5.7	85.5±5.6	.662
Motor FIM gain since admission	35.2±10.1	40.3±7.8	45.8±8.0	.000	37.8±7.5	39.4±8.4	47.1±7.2	.000
Motor FIM gain since discharge	14.9±9.0	17.5±7.3	18.6±6.8	<.001	16.0±7.4	16.4±5.2	18.2±5.7	.018
Motor FIM score=91 (max) (%)	4.7	12.1	11.2	.023	6.1	10.2	8.8	.406
Living at home (%)	100.0	98.6	98.6	.411	100.0	100.0	98.8	.481
Interfered with social activities (%)				.221				.165
All/most/some of the time	11.9	12.1	16.3		10.8	11.9	13.5
A little of the time	15.1	7.9	10.5		20.0	6.8	9.9
None of the time	73.0	80.0	73.1		69.2	81.4	76.6
Had a fall (%)	29.1	12.9	21.8	.005	13.9	15.3	22.2	.245

Source: JOINTS II study.

NOTE. Values are mean ± SD or as otherwise indicated.

Abbreviations: Fs-SNF, freestanding SNF; Hb-SNF, hospital-based SNF; MCS, mental component score.

At follow-up, a noticeable number of study patients attained a motor FIM score of 91, the maximum motor FIM score. Table 3 provides a percentage distribution of those who obtained a motor FIM score of 91 by setting and type of replacement. Prior to case-mix adjustment, hospital-based SNF knee replacement patients were the most likely to obtain this score (12.1%).

There were no statistically or clinically significant differences in the study's other main outcome variable, the SF-12's PCS score.

Hospital-based SNF patients with knee replacement reported the lowest rates of falls, and IRF patients with hip replacement reported the highest rates of falls.

Table 4 presents follow-up outcomes controlling for case-mix differences between settings. IRF patients, patients with knee and hip replacement, made the largest motor FIM gains from admission and discharge to follow-up. IRF patients with hip replacement had the highest scores on nearly all outcomes related to the physical component of the SF-12, although differences among the 3 settings were not material. Among patients with knee replacement, no setting had materially better or worse case-mix–adjusted follow-up outcomes. Summary scores, such as the SF-12's PCS, are perhaps more meaningful because they incorporate several variables concurrently. Differences in PCS across the 3 settings, however, were not large: for knee patients, the difference was 2.5 points; for hip patients, the difference was 3.9 points.

Table 4. Case-Mix–Adjusted Follow-Up Outcomes

Abbreviations: Fs-SNF, freestanding SNF; Hb-SNF, hospital-based SNF; max, maximum; MCS, mental component score.

After case-mix adjustment, 12% of IRF patients achieved maximum motor FIM scores (91); lower percentages of freestanding SNF and hospital-based SNF patients achieved maximum FIM scores (see table 4).

Table 5 displays the results from 2 of 3 sets of ordinary least squares regressions where we included hospital-based SNF as (1) a distinct setting and (2) as an SNF combined with freestanding SNFs in explaining follow-up motor FIM scores. In both analyses, we were able to explain only modest to moderate amounts of variance. More of the explained variance came from patient characteristics and less from setting variables. When the hospital-based SNF was considered a setting apart from other settings, freestanding SNFs were negatively associated with outcomes in the case of patients with knee replacement, and IRFs were positively associated with outcomes in the case of patients with hip replacement. When the hospital-based SNF was combined with freestanding SNFs, IRFs were positively associated with outcomes in the case of knee replacement but otherwise found no association in the case of patients with hip replacement. We found association between the use of follow-up therapy and follow-up outcomes only for patients with hip replacement, but we did not test for all the different combinations of bed-service and postdischarge therapy. After discharge from their initial postacute setting, study patients from all 3 settings go on to use considerable amounts of home and outpatient therapy that is nearly equivalent to what they had originally received during their initial postacute stay.⁶ We examined the residuals to determine how well the models fit the data both numerically and graphically. Our histograms suggested that the errors were random and drawn from a normal distribution with mean 0. We also examined potential interaction effects but found that interaction effects made our models less stable and diminished the amount of explained variance.

Table 5. Results of Regression Models for Follow-Up Motor FIM

Variable	Patients With Knee Replacement						Patients With Hip Replacement
	Model 1⁎			Model 2†			Model 1⁎			Model 2†
	Coefficient	F	P	Coefficient	F	P	Coefficient	F	P	Coefficient	F	P
At admission
Age	–0.07	10.66	.001	–0.08	11.48	.001	NS‡			–0.06	6.81	.009
Female	–1.26	7.94	.005	–1.25	7.64	.006	–1.90	10.34	.001	–1.85	9.10	.003
Race white	NS‡			1.29	7.29	.007	NS‡			NS‡
Onset days	–0.20	8.65	.003	–0.19	8.28	.004	–0.33	4.10	.044	–0.39	4.36	.038
Revision	–2.64	7.56	.006	–2.66	7.58	.006	NS‡			NS‡
Admission motor FIM	0.12	26.77	<.001	0.11	20.52	<.001	0.25	51.79	<.001	0.21	43.11	<.001
Admission CSI	NS‡			NS‡			–0.10	16.17	<.001	–0.09	11.26	<.001
Setting
Fs-SNF	–2.35	21.63	<.001	NS‡			NS‡			NS‡
IRF	NS‡			1.26	7.97	<.001	1.95	9.27	.003	NS‡
High volume	NS‡			NS‡			NS‡			–1.55	7.61	.007
Postdischarge
Number of outpatient therapy visits	NS‡						0.06	6.52	.011	0.06	5.99	.015
R2	0.12			0.11			0.25			0.26
F	12.11			9.26			15.79			14.16
p	<0.001			<0.001			<0.001			<0.001

Abbreviation: Fs-SNF, freestanding SNF.

When running our censored regression models that took into account possible ceiling effects in the motor FIM at follow-up, we found that it yielded the same significant independent variables with somewhat different coefficients.

When we ran the regression models using PCS as the outcome variable, we were able to explain only 4.6% to 10.0% of the variance and thus do not display the results.

In each of our regression models, we included duration from postacute admission to follow-up, but it never was a significant predictor.

Back to Article Outline

Discussion 

The study uncovered few direct setting effects on follow-up outcomes after taking into account patient differences between settings of care. The bivariate analysis, prior to case-mix adjustment, found that IRFs produce larger FIM gains. With case-mix adjustment, differences in FIM gains between settings narrow somewhat. The 3 settings did not differ materially in terms of their case-mix adjusted scores on the SF-12 PCS or its individual components.

There are several possible explanations for the lack of more noticeable differences between settings at follow-up, especially on the SF-12 PCS. First, as we noted in our previous analysis of discharge outcomes,⁷ there is significant practice variation within settings of care that blurs the differences observed between types of settings.

Second, as also noted in previous findings,⁶ patients with joint replacement from all 3 settings go on to use considerable amounts of either home or outpatient rehabilitation services or both—amounts that in many instances equal or surpass the amounts received during the original postacute setting. Moreover, patients receive different mixes of home and outpatient rehabilitation services. The variety of postdischarge rehabilitation service patterns makes it difficult to determine which combinations of home and outpatient services result in better outcomes and obscures the effects of the original postacute placement that might otherwise be observed at follow-up. In short, the comparisons between SNFs and IRFs at follow-up are to some extent artificial when there are really multiple trajectories of rehabilitative care over the course of the entire postacute episode. This does not totally absolve the initial setting of care for being accountable for longer-term outcomes. The initial setting of care bears some responsibility for making sure that the patient receives appropriate postdischarge follow-up care, although the final decision may not be in the hands of the initial postacute setting.

Third, the lack of more noticeable setting effects and the convergence of follow-up outcomes also may be a function of natural recovery processes and not solely a function of the original setting or the additional rehabilitation services obtained after discharge from an SNF or IRF. The passage of time has, perhaps, a way of narrowing differences.

And fourth, the case-mix adjustment allowed us to compare mainly patients represented in the less severe subgroups because the more severe subgroups did not have enough patients across settings. There could be a setting effect for more severely involved patients with joint replacement.

Implications for Practice and Policy 

The findings from the follow-up study do not provide an adequate basis for determining whether patients with joint replacement who may need bed-service rehabilitation should go to a freestanding SNF, a hospital-based SNF, or an IRF for their initial postacute placement. None of the 3 settings presents a clear-cut advantage over the others in terms of follow-up outcomes. The path from acute hospital discharge to follow-up outcome is not a straight line but one with up to several stops along the way. Clinicians and policy makers alike need to consider the entire postacute pathway and the costs associated with each pathway. Current payment systems do not encourage decision makers to consider the entire postacute episode and its attendant costs.

Findings from the JOINTS I study on discharge outcomes suggest that patients with joint replacement with shorter LOSs and more intensive rehabilitation did somewhat better.⁷ Hence, it may be possible to move patients along more quickly from bed-service facilities into nonbed service settings. Trying to fit patients into multiple postacute silos, each with their own admission and discharge processes and their own payment systems, appears inefficient and blurs whatever accountability we may wish to place on the original setting of care.

Study Limitations 

This study has several limitations. The study carries forward the limitations of the larger JOINTS I study on which this follow-up study builds. One limitation of both this study and its antecedent study is the representativeness of the study facilities. While the study captured geographic diversity, facility participation was voluntary and thus presents potential selection effects. Moreover, study investigators, by design, sought to enroll a limited number of higher-volume facilities in order to minimize study costs. Both smaller SNFs and IRFs, particularly small IRF units, were underrepresented. Given that patient volume appears to play a role in shaping outcomes, as noted in earlier findings,⁷ this is not immaterial. For example, the lowest number of patients with joint replacement served a year by an SNF in this study is 50, which we defined as a low-volume facility. Yet many SNFs nationally serve fewer than 10 patients with joint replacement a year.²³

Nonetheless, this analysis drew on the experiences of 856 patients, which makes this one of the largest joint replacement rehabilitation studies using primary data collection. Even this sample size was not large enough to represent every subgroup adequately. For example, when we tried to adjust outcomes for case mix by using CMG groups and tiers, some cells had no patients or only a few.

The study's hospital-based SNF was only 1 facility, and caution is warranted in generalizing to all hospital-based SNFs, although there is ample evidence that hospital-based SNFs are noticeably different from freestanding SNFs.²⁴, ²⁵ Likewise, among the study's freestanding SNFs, 1 SNF had a considerably higher volume of patients than other freestanding SNFs. In our analyses, we did consider potential individual facility effects as well as setting effects.

Finally, we need to reconsider the validity of the outcome measures used to examine follow-up outcomes. The motor FIM proved to be a more robust measure of outcome and explained more of the variance in outcome than did the SF-12 and its 2 subsidiary component scales, in part because the admission FIM was also used as a predictor. That said, however, the motor FIM presents ceiling effects that can limit the amounts of functional gains observed and thus flattens the differences between settings that might otherwise be observed. IRF and hospital-based SNF patients were more likely to reach the ceiling than were freestanding SNF patients. On discharge from the initial rehabilitation care settings, however, the highest motor FIM score was 80, and most patients scored between 58 and 78. In conversations with P. Smith (July 2008), we learned that an ongoing study by Smith examining 3-month follow-up motor FIM scores indicates that by the third month, the average motor FIM was about 84 for patients with knee replacement and 83 for patients with hip replacement after gaining 12.4 and 13.2 points, respectively, after discharge. Although we cannot compare the results of the 2 studies directly, these observations suggest that patients' outcomes approach a ceiling within the first several months. Other more sensitive outcome measures should be considered when evaluating follow-up outcomes. The concept of functional independence should be broadened to include instrumental activities of daily living (eg, household activities, leisure activities, active participation in community life).

Back to Article Outline

Conclusions 

Patients with joint replacement discharged from SNFs and IRFs do not differ materially on follow-up in terms of functional status and other outcomes. Some outcomes tilt in favor of IRFs, but the magnitude of the differences in most instances is not large. Differences between SNFs and IRFs are obscured, in part, by the array of postdischarge rehabilitation services that patients go on to receive. Some of the ceiling effects in the motor FIM suggest that more robust follow-up outcome measures are needed to evaluate longer-term outcomes. Although the study used the SF-12 and also evaluated patient performance on individual components of the SF-12, we found few meaningful differences using the SF-12.

Most important, clinicians and policy makers need to examine the entire postacute episode. The sole focus on the initial placement overlooks the larger trajectory of postacute care that needs to be managed to achieve superior outcomes.

Back to Article Outline

Acknowledgments 

We acknowledge the role and contributions of our collaborators at each of the 12 clinical sites represented in the JOINTS II study: Charles Schauer, PhD, and Flo Singletary, MS (Brooks Rehabilitation Hospital, Jacksonville, FL); Hilary Siebens, MD, and Harriet Aronow, PhD (Casa Colina Center for Rehabilitation, Pomona, CA); Jacalyn Lichtenstein, RN, and Lynne Wright, RN, CRRN (HealthSouth Plano Rehabilitation Hospital, Plano, TX); Julie Barth, OTR/L (HealthSouth Scottsdale Rehabilitation Hospital, Scottsdale, AZ); Andrea Curry, RN (JFK Medical Center, Johnson Rehabilitation, Edison, NJ); Barbara Higgins, PT (Brentwood Subacute Healthcare, Burbank, IL); Gina Harris, PT, Natalie Russo, OT, and Marcy Howard (The Cedars at The JCA, Chesterfield, MO); Steve Christensen and Lyle Black (Crosslands Rehabilitation, Sandy, UT); Ann Cottrell, Dawn Haskell, PT, and MaryAnn Morrison, RN (Greenbriar Terrace Healthcare, Nashua, NH); Ellen Logsdon (Harrison Health and Rehabilitation, Corydon, IN); and Karen L. George, OT, Suzanne Besecker, Audrey Hartz, RN, Geraldine Essick, and Pragna Doshi, PT (TCU Reading Hospital and Medical Center, Reading, PA).

We acknowledge members of our staffs who also contributed significantly to the success of this study: Ching-Hui Hsieh, PhD, Michael R. Brown, BA, Elizabeth Newman, OTD, OTR/L, Mary Foley, RN, Cathy Ellis, PT, Naomi Greenberg, PT, Deborah Hutton, OTR/L, and Josephine Kuofie, RN (National Rehabilitation Hospital, Washington, DC); Julie Gassaway, RN, MPH, and Roberta James, MStat (Institute for Clinical Outcomes Research, Salt Lake City, UT); Al Dobson, PhD, Joan DaVanzo, PhD, Namrata Sen, and Kristina Ko, MS (Lewin Group, Washington, DC); and Pam Smith, PhD, Erin Smith, Dave Smith, and Todd Smith (IT HealthTrack, Williamsville, NY).

We also acknowledge the input and contributions of all members of the study's Policy Advisory Panel and the organizations they represent: Rosaly Correa-de-Arauio, MD, PhD (Agency for Healthcare Research and Quality, Rockville, MD); Barbara Manard (American Association of Homes and Services for the Aging, Washington, DC); Robyn Stone, PhD (American Association of Homes and Services for the Aging, Washington, DC); Rochelle Archuleta, MPH (American Hospital Association, Washington, DC); Carolyn Zollar, JD (American Medical Rehabilitation Providers Association, Washington, DC); Mary Fran Delaune (American Physical Therapy Association, Alexandria, VA); Trudy Mallinson, PhD and Anne Deutsch, PhD (Center for Rehabilitation Outcomes Research, Chicago, IL); Christine MacDonell (Commission on Accreditation of Rehabilitation Facilities, Washington, DC); Barbara Braun, PhD (Joint Commission on Accreditation of Healthcare Organizations, Oakbrook Terrace, IL); Elizabeth Sandel, MD (Kaiser Foundation Rehabilitation Center, Vallejo, CA); Susan Klanecky (Madonna Rehabilitation Hospital, Lincoln, NE); Michael Weinrich, MD (National Center for Medical Rehabilitation Research, Rockville, MD); Melinda Buntin, PhD (The RAND Corp, Santa Monica, CA); Reginald Warren, PhD (SeniorMetrix, Hingham, MA); Leigh Callahan, PhD (Thurston Arthritis Research Center, University of North Carolina, Chapel Hill, NC); Carl Granger, MD (Uniform Data System, Amherst, NY); and Michael Munin, MD (University of Pittsburgh Medical Center, Pittsburgh, PA). We acknowledge those nonmembers of the Policy Advisory Panel who participated as observers because of their employment with policy-making agencies and funding sources: James Bowman, MD (Centers for Medicare and Medicaid Services, Baltimore, MD); Ruth Brannon and Philip Beatty (National Institute on Disability and Rehabilitation Research, Washington, DC); and Dexanne Clohan, MD, Justin Hunter, JD, and John Markus, JD (HealthSouth Corp).

Back to Article Outline

References 

1. Mahomed N, Koo Seen Lin M, Levesque J, Lan S, Bogoch E. Determinants and outcomes of inpatient versus home based rehabilitation following elective hip and knee replacement. J Rheumatol. 2000;27:1753–1758
   • View In Article
   • MEDLINE
2. Munin M, Rudy T, Glynn N, Crossett L, Rubash H. Early inpatient rehabilitation after elective hip and knee arthroplasty. JAMA. 1998;279:847–852
   • View In Article
   • MEDLINE
   • CrossRef
3. Vincent HK, Weng JP, Vincent KR. Effect of obesity on inpatient rehabilitation outcomes after total hip arthroplasty. Obesity (Silver Spring). 2007;15:522–530
   • View In Article
   • MEDLINE
   • CrossRef
4. Walsh M, Herbold J. Outcome following rehabilitation for total joint replacement at IRF and SNF. Am J Phys Med Rehabil. 2006;85:1–5
   • View In Article
   • MEDLINE
   • CrossRef
5. Buntin MB. Comparisons of Medicare spending and outcomes for beneficiaries with lower extremity joint replacements. Santa Monica: RAND Health; 2005;
   • View In Article
6. DeJong G, Tian W, Smout RJ, et al. Use of rehabilitation and other health care services by patients with joint replacement after discharge from skilled nursing and inpatient rehabilitation facilities. Arch Phys Med Rehabil. 2009;90:1297–1305
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (247 KB)
   • CrossRef
7. DeJong G, Horn SD, Smout RJ, Tian W, Putman K, Gassaway J. Joint replacement rehabilitation outcomes on discharge from skilled nursing facilities and inpatient rehabilitation facilities. Arch Phys Med Rehabil. 2009;90:1284–1296
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (632 KB)
   • CrossRef
8. DeJong G, Hsieh C-H, Gassaway J, et al. Characterizing rehabilitation services for patients with knee and hip replacement in skilled nursing facilities and inpatient rehabilitation facilities. Arch Phys Med Rehabil. 2009;90:1269–1283
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (3689 KB)
   • CrossRef
9. Gandek B, Ware J, Aaronson N, et al. Cross-validation of item selection and scoring for the SF-12 health survey in nine countries: results from the IQOLA project. J Clin Epidemiol. 1998;51:1171–1178
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (252 KB)
   • CrossRef
10. Hamilton B, Granger C, Sherwin F. A uniform data system for medical rehabilitation. Baltimore: Brookes; 1987;
    • View In Article
11. Hamilton B, Laughlin J, Fiedler R. Interrater reliability of the 7-level functional independence measurement (FIM). Scand J Rehabil Med. 1994;26:115–119
    • View In Article
    • MEDLINE
12. Ware J, Kosinski M, Keller S. A 12-item short-form health survey: construction of scales and preliminary tests of reliability and validity. Med Care. 1996;34:220–233
    • View In Article
    • MEDLINE
    • CrossRef
13. Chang W, Chan C, Slaughter S, Cartwright D. Evaluating the FONE FIM: part II: concurrent validity & influencing factors. J Outcome Meas. 1997;1:259–285
    • View In Article
    • MEDLINE
14. Petrella R, Overend T, Chesworth B. FIM after hip fracture: is telephone administration valid and sensitive to change?. Am J Phys Med Rehabil. 2002;81:639–644
    • View In Article
    • MEDLINE
    • CrossRef
15. Smith P, Hamilton B, Granger C. Functional independence measure decision tree: the FONE FIM. Buffalo: State University of New York, Center for Functional Assessment Research; 1990;
    • View In Article
16. Smith P, Illig S, Fiedler R, Hamilton B, Ottenbacher K. Intermodal agreement of follow-up telephone functional assessment using the functional independence measure in patients with stroke. Arch Phys Med Rehabil. 1996;77:413–415
    • View In Article
17. Bergstrom N, Horn SD, Smout RJ, et al. The national pressure ulcer long-term care study (NPULS): outcomes of pressure ulcer treatments in long-term care. J Am Geriatr Soc. 2005;53:1271–1279
    • View In Article
18. Horn SD, Bender S, Ferguson M, et al. The national pressure ulcer long-term care study (NPULS): pressure ulcer development in long-term care residents. J Am Geriatr Soc. 2004;52:359–367
    • View In Article
    • MEDLINE
    • CrossRef
19. Horn SD, DeJong G, Smout RJ, Gassaway J, James R, Conroy B. Stroke rehabilitation patients, practice, and outcomes: is earlier and more aggressive therapy better?. Arch Phys Med Rehabil. 2005;86(12 Suppl):S101–S114
    • View In Article
    • MEDLINE
20. James R, Gines D, Menlove A, Horn SD, Gassaway J, Smout RJ. Nutrition support (tube feeding) as a rehabilitation intervention. Arch Phys Med Rehabil. 2005;86(12 Suppl):S82–S92
    • View In Article
    • MEDLINE
21. Maulden S, Gassaway J, Horn SD, Smout RJ, DeJong G. Timing of initiation of rehabilitation after stroke. Arch Phys Med Rehabil. 2005;86(12 Suppl):S34–S40
    • View In Article
    • MEDLINE
22. Ryser D, Egger M, Horn SD, Handrahan D, Gandhi P, Bigler E. Measuring medical complexity during inpatient rehabilitation following traumatic brain injury. Arch Phys Med Rehabil. 2005;86:1108–1117
    • View In Article
    • Abstract
    • Full Text
    • Full-Text PDF (212 KB)
    • CrossRef
23. RehabCare The impact of the 75% rule on patient outcomes and Medicare expenditures in IRFs and SNFs. 2007 http://www.rehabcare.com/75percentstudy/docs/TheImpactofthe75Rule.pdfAccessed January 14, 2008
    • View In Article
24. Medicare Payment Advisory Commission. Report to Congress: Medicare payment policy. Washington (DC): Medicare Payment Advisory Commission; 2008;
    • View In Article
25. Liu K, Black KJ. Hospital-based and freestanding skilled nursing facilities: any cause for differential Medicare payments?. Inquiry. 2003;40:94–104
    • View In Article
    • MEDLINE
    • CrossRef