Munin MC, Rudy TE, Glynn NW, Crossett LS, Rubash HE. Early Inpatient Rehabilitation After Elective Hip and Knee Arthroplasty. JAMA. 1998;279(11):847-852. doi:10.1001/jama.279.11.847
From the Division of Physical Medicine and Rehabilitation (Drs Munin and Glynn), Department of Orthopaedic Surgery (Drs Munin, Crossett, and Rubash), and the Departments of Anesthesiology and Biostatistics (Dr Rudy), University of Pittsburgh, Pittsburgh, Pa.
Context.— Inpatient rehabilitation after elective hip and knee arthroplasty is
often necessary for patients who cannot function at home soon after surgery,
but how soon after surgery inpatient rehabilitation can be initiated has not
Objective.— To test the hypothesis that high-risk patients undergoing elective hip
and knee arthroplasty would incur less total cost and experience more rapid
functional improvement if inpatient rehabilitation began on postoperative
day 3 rather than day 7, without adverse consequences to the patients.
Design.— Randomized controlled trial conducted from 1994 to 1996.
Setting.— Tertiary care center.
Participants.— A total of 86 patients undergoing elective hip or knee arthroplasty
and who met the following criteria for being high risk: 70 years of age or
older and living alone, 70 years of age or older with 2 or more comorbid conditions,
or any age with 3 or more comorbid conditions. Of the 86 patients, 71 completed
Interventions.— Random assignment to begin inpatient rehabilitation on postoperative
day 3 vs postoperative day 7.
Main Outcome Measures.— Total length of stay and cost from orthopedic and rehabilitation hospital
admissions, functional performance in hospitals using a subset of the functional
independence measure, and 4-month follow-up assessment using the RAND 36-item
health survey I and the functional status index.
Results.— Patients who completed the study and began inpatient rehabilitation
on postoperative day 3 exhibited shorter mean (±SD) total length of
stay (11.7±2.3 days vs 14.5±1.9, P<.001),
lower mean (±SD) total cost ($25891±$3648 vs $27762±$3626, P<.03), more rapid attainment of short-term functional
milestones between days 6 and 10 (36.2±14.4 m ambulated vs 21.4±13.3
m, P<.001; 4.8±0.8 mean transfer functional
independence measure score vs 4.3±0.7, P<.01),
and equivalent functional outcome at 4-month follow-up.
Conclusion.— These data showed that high-risk individuals were able to tolerate early
intensive rehabilitation, and this intervention yielded faster attainment
of short-term functional milestones in fewer days using less total cost.
TOTAL JOINT replacements for the hip and knee are among the most common
surgical procedures in the United States. Rehabilitation is essential to minimize
disability after surgery, yet pressure on clinicians to reduce length of stay
has limited the use of hospital rehabilitation services. Most patients can
be discharged directly home in 5 to 7 days if they are medically stable and
have completed a postoperative rehabilitation program.1
Those who were significantly older, lived alone, and had an increased number
of comorbid conditions were at high risk for requiring further inpatient rehabilitation
services before returning home.2 However, there
are no practice parameters to determine how best to treat high-risk patients
who are unable to go directly home after joint replacement surgery.
Two factors that influence the type of rehabilitation program selected
for high-risk individuals include the timing of when to begin inpatient rehabilitation
and the intensity of therapy services needed to attain good outcomes. Some
rehabilitation facilities accept patients as early as the third postoperative
day to reduce patients' length of stay. However, it is unknown if elderly
patients can benefit from intensive therapy this early after surgery. If patients
are transferred to a rehabilitation facility prematurely, rehabilitation costs
could be increased if patients are unable to participate in intensive physical
and occupational therapies, or if they are transferred out of rehabilitation
because of acute medical or surgical complications. It has not been determined
if more intensive therapy for high-risk joint replacement patients produces
a faster rate of improvement, or if outcome is independent of the amount of
Using an acute inpatient rehabilitation setting, this study randomized
high-risk patients to either begin inpatient rehabilitation on postoperative
day 3 (day 3 treatment group) or to delay inpatient rehabilitation until 7
days after surgery (day 7 treatment group). Since the acute care hospital
provides some rehabilitation services at a lower intensity compared with inpatient
rehabilitation, and most patients who go home can be discharged within 7 days,
this time interval was used as the control. Total length of stay, total cost,
complications, and functional outcomes were measured. We tested the hypothesis
that the day 3 group would demonstrate less total cost and a faster rate of
functional improvement because intensive therapy was delivered earlier after
surgery. Additionally, we hypothesized that beginning rehabilitation efforts
earlier following surgery would not be detrimental to long-term outcomes.
All patients evaluated for total-hip arthroplasty or total-knee arthroplasty
at the University of Pittsburgh orthopedics office were considered for inclusion
between August 1994 and November 1995. Eligible subjects included patients
at high risk for requiring inpatient rehabilitation after elective hip and
knee arthroplasty: (1) 70 years of age or older and living alone, (2) 70 years
of age or older with 2 or more comorbid conditions, or (3) any age with 3
or more comorbid conditions.2 A comorbidity
checklist that assessed 11 major conditions was developed to extract significant
medical history from both the clinical record and patient interviews. The
surgical procedure could be a primary arthroplasty or a revision of a previous
arthroplasty. Individuals were excluded if the indication for surgery included
stabilization for tumor, acute fracture, femoral osteonecrosis, or hemophilic
arthropathy. Additionally, patients were excluded if medical or surgical complications
occurred postoperatively and precluded scheduled rehabilitation transfer.
Each patient who met the criteria for inclusion read and signed the institutional
review board informed consent form before data collection began.
Eighty-six subjects met the inclusion criteria. These patients were
assigned randomly to begin inpatient rehabilitation on postoperative day 3
(day 3 group) or remain on the orthopedic service until postoperative day
7 (day 7 group) and then pursue inpatient rehabilitation. A random listing
of 100 numbers, using 0 to equal day 3 group and 1 to equal day 7 group, was
generated in blocks of 10. Separate lists were used for hip and knee patients
to maintain balance of the randomized design. The codes were administered
by a blinded executor who kept them in a locked filing cabinet that was inaccessible
to the principle investigator and research coordinator. Randomization occurred
preoperatively in the orthopedic office after patient eligibility was established
and the consent form was signed. Based on our previous data,2
a sample size of 40 patients per group was determined to have adequate statistical
power (β=.80) to detect major effect sizes (SD=0.8) for length of stay
and total costs.
Patients were treated at a university medical center for all aspects
of the study, which included an orthopedic surgery floor and a 20-bed inpatient
rehabilitation unit. Patients in the day 3 group had earlier exposure to the
inpatient rehabilitation unit, even though both treatment groups started therapy
at the same time.
While in the acute care orthopedic surgery service, patients were scheduled
to receive two 30-minute physical therapy sessions beginning on postoperative
day 2 and one 30-minute occupational therapy session beginning no later than
postoperative day 3 during weekdays only. The comprehensive inpatient rehabilitation
program included two 60-minute physical therapy and two 60-minute occupational
therapy sessions daily, as well as recreational therapy and clinical psychology
services as needed. One 30-minute physical therapy session was given to all
patients, regardless of location, on Saturday.
The rehabilitation and orthopedic units both used clinical pathways,
and several practical considerations were routinely used. Pain was aggressively
treated with ice packs applied to the incision, and narcotic analgesia was
always given before morning and afternoon therapy sessions. The surgical wound
was monitored for drainage and prophylactic antibiotic coverage with a first-generation
cephalosporin instituted when needed, although therapy was infrequently withheld
because of wound drainage. Discharge from the rehabilitation unit was determined
by the interdisciplinary team of clinicians, some of whom were not blinded
to randomization. However, standardized, objective criteria were used for
discharge. All patients had to be ambulating greater than 45 m, performing
transfers and all self-care independently with adaptive equipment, and demonstrating
the ability to safely return home.
Length of Stay and Cost Analysis. The total hospital length of stay, which included the days in the orthopedic
and rehabilitation units, was compared between treatment groups. All physical
and occupational therapy sessions were counted throughout the orthopedic and
rehabilitation admissions. Data were collected by the Medical Archival System
Charges were retrieved from the individual transaction detail, including
transaction code, department code, quantity, date of service, and charge.
Charge data were converted to cost using total-loaded ratios of cost to charge,
which are hospital-specific, government-mandated standards used to estimate
cost in prior joint replacement studies.5,6
The ratios of cost to charge have been shown to have good reliability for
comparing the relative costs of patients with the same diagnosis in different
institutions.7 Each department's unique fiscal
year 1995 ratios of cost to charge were applied to arrive at the total-loaded
cost for each transaction. The total-loaded costs were aggregated to arrive
at the department costs for day-of-surgery costs, hospital postoperative costs,
and rehabilitation costs for physical and occupational therapy. Physician
costs were derived from average Medicare reimbursement for the type of surgical
procedure and physiatric evaluation and management.
Functional and Health Status Assessments. Baseline data were obtained approximately 4 weeks before surgery to
assess general health status and function. General health status was obtained
with the RAND 36-item health survey I (RAND-36) and data were confined to
2 primary health constructs, physical and mental domains.8,9
The functional status index (FSI)10 was used
to assess overall function based on important tasks that are needed to complete
activities of daily living. The FSI has 3 dimensions that measure difficulty,
pain, and assistance. Both of these self-report instruments have been validated
in prior hip and knee arthroplasty studies.11- 13
A subset of the functional independence measure (FIM)14
was used daily in both the orthopedic and rehabilitation units to measure
immediate postoperative function. The therapists in the study were trained
to use the FIM, and we have previously demonstrated excellent interrater reliability
using a FIM-based measurement system in the acute care hospital after joint
replacement.15 To assess the perceived benefit
of the rehabilitation program, follow-up data using the RAND-36 and FSI were
mailed at 16 weeks16 from the date of hospital
Data Analyses. Analysis of variance, using the general linear model because of unequal
sample sizes, was used for length of stay and cost data. To provide better
protection against type I error, a multivariate analysis of variance (MANOVA)
approach was used to analyze RAND-36 and FSI change scores. Repeated-measures
MANOVA was used for the FIM data that were divided into 3 periods. Power or
square-root transformations were applied to 3 variables to achieve adequate
normality for the analysis of variance and MANOVA analyses. Raw means and
SDs for all variables are reported here for clarity. Exact probability χ2 analyses, computed with the StatXact program,17
were used to analyze dichotomous and ordinal measures. A P value of <.05 was used as statistically significant, and Bonferroni
correction was applied to post hoc analyses when the primary analyses were
The intention-to-treat principle was maintained in this study whenever
possible. Because of differences in study dropouts between the day 3 and day
7 groups, efficacy analyses that included only subjects completing the rehabilitation
arm of the study also were computed for length of stay and cost data.
Of the 86 patients randomized to the study, 71 patients (83%) completed
the rehabilitation arm of the study following replacement surgery (Table 1). Fifteen patients were excluded
or dropped out after surgery (Table 2),
with no exclusions occurring before randomization. The number of patient exclusions
was not significantly different between the 2 rehabilitation conditions (χ2 [1 df]=0.47, P=.57).
A MANOVA analysis between those excluded and those completing treatment for
FSI and RAND-36 presurgery scores indicated no significant differences. Analysis
of variance data indicated no significant differences with respect to age
and the number of comorbid conditions, and χ2 analyses indicated
no significant differences with respect to gender.
As displayed in Table 1,
the baseline characteristics for subjects completing the rehabilitation component
were similar between the 2 treatment conditions (day 3 vs day 7) and the 2
types of replacement surgeries (hip vs knee). Analysis of variance and χ2 analyses for the variables listed in Table 1 indicated that for knee patients, a higher proportion of
rheumatoid arthritis patients were randomized to the day 3 group (χ2 [1df]=4.92, P<.05).
For all other pretreatment measures, no significant differences were found
between groups, indicating that the randomization procedure was successful
in producing comparable groups.
Both intention-to-treat and efficacy analyses of total length-of-stay
data indicated that patients assigned to the day 3 protocol stayed in the
hospital significantly fewer days than day 7 patients (both with P<.04). This conclusion was not significantly influenced or modified
by the type of replacement received. The length-of-stay data for patients
who completed rehabilitation are presented in Table 3. When dropouts were included, patients assigned to day 3
had a mean length of stay of 11.9 days (SD, 2.9) and day 7 had 13.8 days (SD,
3.1), which remained significantly different (P<.004).
The slight reduction in length-of-stay differences between day 3 and day 7
patients when dropouts were included is because of the longer stay of excluded
day 3 patients compared with those day 3 patients completing rehabilitation,
coupled with the 4 patients in the day 7 dropout group (Table 2) who stayed an average of only 8.5 days because they elected
to go home rather than complete the rehabilitation component of the study.
Intention-to-treat grand total cost analyses indicated no significant
differences between patients assigned to the day 3 and day 7 treatment conditions.
Day 3 patients had a mean cost of $26582 (SD, $4370) and day 7 had a mean
cost of $26880 (SD, $4194). An efficacy cost analysis of those completing
rehabilitation showed the grand total costs were significantly lower for day
3 compared with day 7 (Table 3).
To further delineate these differences, Bonferroni-adjusted post hoc analyses
were conducted for major components of the grand total cost. These analyses
indicated that hospital postoperative costs were significantly less for day
3 patients and that hip replacements were more expensive than knee replacements
for day-of-surgery costs. Other analyses stratified by type of surgery were
not significant. For rehabilitation costs (Table 3), physician costs were significantly higher for inpatient
rehabilitation that was begun sooner for day 3 patients compared with day
7 patients and the daily average for physical therapist plus occupational
therapist costs indicated a higher daily fee for day 3 vs day 7 patients.
This finding demonstrates that the day 3 group actually received more therapy
per day than the day 7 group.
The daily FIM scores were analyzed to determine the rate of functional
change during hospitalization. Because of the differing lengths of stay among
patients, the distributional characteristics of patients' lengths of stay
were evaluated and lower-, middle-, and upper-quartile values were computed.18 This statistical approach created 3 postoperative
periods: days 1 through 5, days 6 through 10, and days 11 or higher. Mean
FIM scores were computed for each of these periods for each patient.19 Data for days 11 or higher were not available for
all patients because of the earlier discharge of some patients, as noted in Table 4 and Table 5.
As displayed in Table 4,
significant differences in physical therapist–assessed FIM scores were
found between the day 3 vs day 7 groups, although no differences were noted
when the data were stratified by surgery type. Specifically, transfers, ambulation,
distance walked, and stair climbing scores were significantly higher (better)
for the day 3 patients during days 6 to 10 of hospitalization. Additionally,
day 3 patients compared with day 7 patients displayed significantly better
scores on ambulation and distance walked during days 1 to 5 of hospitalization.
Results for the occupational therapist–assessed scores were similar
to those of the physical therapist–assessed scores (Table 5), with the day 3 patients demonstrating significantly higher
scores during days 6 to 10 of hospitalization for bathing and lower-extremity
Of the 86 patients randomized to the study, 71 patients (83%, 62 who
completed rehabilitation and 9 dropouts) could be reached and returned the
standardized questionnaires at the time of the 4-month follow-up. Separate
MANOVA analyses for pretreatment FSI and RAND-36 scores between those completing
the follow-up questionnaires and those lost to follow-up indicated no significant
Change scores for the FSI between the 4-month follow-up and pretreatment
values are presented in Table 6.
A MANOVA analysis indicated no significant differences in the magnitude of
these FSI changes between rehabilitation groups or between type of replacement
surgery. A separate MANOVA indicated patients displayed significant changes
from pretreatment to the 4-month follow-up for the FSI pain and difficulty
scales (P<.001), but not for the assistance scale.
Change scores for the 5 RAND-36 scales used in this study are displayed
in Table 7.4
All 3 physical domain scales displayed significant improvements from pretreatment
values at the time of the 4-month follow-up (both P<.001),
but significant differences in the change scores were not found between rehabilitation
groups or replacement type. Similarly, change scores for the 2 mental domain
scales were not significantly different between groups or replacements.
Within the follow-up interval, 2 patients had hip dislocation (1 day
3 and 1 day 7), while another hip replacement patient (day 3) was ruled out
for deep vein thrombosis at a local emergency department. Three knee arthroplasty
patients (all day 7) were readmitted to the hospital with diagnoses of deep
vein thrombosis in 2 individuals and congestive heart failure was ruled out
in another. The χ2 analyses indicated no significant differences
between day 3 and day 7 groups with respect to follow-up complications (χ2 [1df]=1.44, P=.30).
The results from this randomized trial indicate that high-risk individuals
undergoing elective hip and knee arthroplasty had shorter total length of
stay, faster attainment of short-term functional milestones, and equivalent
functional outcome at 4-month follow-up if they completed an inpatient rehabilitation
program that began on the third rather than the seventh postoperative day.
Patients in this cohort were primarily female, lived alone, had increased
comorbidities, and were typical of elderly patients receiving total joint
replacement.20 More important, the data showed
that this group could tolerate and benefit from early, intensive rehabilitation,
which allowed a more rapid emphasis on physical independence and comprehensive
assessment of functional problems.11 Patients
in the day 3 group did not have a higher rate of complications requiring transfer
from the rehabilitation unit or increased hospital readmissions once sent
home. In addition, the day 3 group demonstrated longer ambulation distance
walked and superior FIM scores for mobility and self-care measures during
postoperative days 6 to 11. The increased therapy attended per day by the
day 3 group appears to be responsible for shorter total length of stay, since
functional milestones were attained sooner than for the day 7 patients.
Few studies have examined cost in relation to functional outcome after
hip and knee arthroplasty. Healy and Finn5
and Barber and Healy6 examined the differences
in acute care costs pertaining to total-hip arthroplasty and total-knee arthroplasty
during a 10-year follow-up period and found significant price increases related
mostly to the cost of the surgical implant. Liang et al13
reported an average cost of $22730 per patient with significant improvements
in global health and functional status at 6 months postoperatively. Chang
et al21 found total-hip arthroplasty to be
cost-effective in improving quality-adjusted life years for both short- and
long-term outcomes. While these studies support the benefits of joint replacement,
all patients were included rather than focusing solely on high-risk individuals
who have higher hospitalization costs. In our earlier work, treatment algorithms
differed significantly depending on whether patients were at low or high risk
for requiring prolonged inpatient rehabilitation services after total-hip
arthroplasty and total-knee arthroplasty.2
Similar to our results, Cameron and colleagues16
randomized hip fracture survivors to receive either accelerated rehabilitation
or standard care and found significant total cost reductions for the accelerated
group. In this study, decreased length of stay accounted for differences in
cost between groups even though the therapy cost per day was higher in the
accelerated treatment group.
Although 15 patients (17%) randomized to the study did not complete
the rehabilitation arm because of dropouts or complications, there were no
significant differences between day 3 vs day 7 exclusions for basic demographic
and health status measures or in the composition of those excluded as compared
with the remaining high-risk patients whose postoperative health status did
not interfere with rehabilitation transfer. An intention-to-treat analysis
indicated that day 3 patients had a significantly shorter total length of
hospital stay compared with day 7 patients. However, an intention-to-treat
total cost analysis indicated no significant cost differences, even though
an efficacy cost analysis of patients who completed rehabilitation demonstrated
a mean cost savings of $1871 in favor of early rehabilitation. The reason
for the different conclusions is that day 7–excluded patients had significantly
lower total costs than day 3 exclusions, primarily because 4 of the 8 patients
in the day 7 exclusion group elected not to participate in the rehabilitation
component of the study even though this was deemed appropriate. Thus, if the
early inpatient rehabilitation program was generalized, it is likely that
overall costs would, in the worst case, not increase and probably would decrease.
In conclusion, this study supports acute inpatient rehabilitation services
beginning on postoperative day 3 for high-risk patients unable to make transitions
to home after total joint replacement. These data also support the notion
that the rate of recovery can be hastened by settings that provide intensive
therapy services early after surgery.