Obesity and Inpatient Rehabilitation Outcomes Following Knee Arthroplasty: A Multicenter Study

Study design

This was a retrospective, comparative study using data compiled from computerized medical records systems of 15 independent rehabilitation hospitals along the east coast. These facilities represented urban, rural and academic settings. The facilities were chosen to increase the applicability of the findings to the general population. Patient admission dates for rehabilitation were between 1 January 2002 and 31 March 2006. Medical directors of these facilities were provided standardized instructions on data collection procedures by the investigative team; information regarding patient demographics (13,14,15), insurance carriers, surgical revision status (12), post-surgical instructions and specific hemtalogic variables (glucose, hematocrit, hemoglobin, white blood cell count) that could potentially affect main outcome measures (13,14,16,17) were collected. Contact was periodically made with each study site to ensure accurate completion of data collection. This investigation was approved by the Human Investigation Committee at University of Virginia.

Participants

Consecutive obese and non-obese patients with diagnoses of either primary TKA or revision TKA were identified as participants for this analysis with no bias of selection by the data collectors at each facility (N = 5,428). Subjects were stratified into four groups based on BMI values: non-obese (BMI < 25 kg/m²), overweight (25–29.9 kg/m²), moderately obese (30–40 kg/m²), severely obese (BMI > 40 kg/m²) (18).

Study variables

Patient descriptive variables included age, gender and ethnicity. Physiological measures performed at the time of admission to rehabilitation included a metabolic chemistry panel, complete blood count, vital signs, height and weight. The number and type of comorbidities were compiled from medical records using the International Code of Diseases (ICD-9) for each patient. The type of TKA procedure (primary or revision), TKA revision etiology (pain, mechanical and infection related reasons) and weight bearing status were obtained from medical records and confirmed from each medical chart by the surgeons' discharge summary.

Study outcome variables

The criterion measures were the LOS, functional independence measure (FIM) scores and 18 subscores at admission and discharge, and total, pharmacy, occupational and physical therapy rehabilitation hospital charges.

Functional improvement during inpatient rehabilitation is generally measured using the 18-item FIM (19). The FIM estimates performance of tasks that can be broadly categorized as activities of daily living, mobility, and cognitive domains. The FIM tool is an important representative measure of the overall success of the interdisciplinary rehabilitation program. The FIM tool is a validated measure to capture overall functional improvements (19). Cognition was included with the motor scores as the two major total scores included in the overall FIM score. As part of the FIM score were other specific scores for activities of daily living. While all 18 FIM subscores were collected for each patient, categories that related to lower body functioning included weight bearing transfers (bed to chair, tub/shower, toilet), level of independence for walking and stairs (20).

As an estimate of the rate of functional gain made during the rehabilitation intervention, a FIM efficiency score was calculated. FIM efficiency was defined as the change in FIM from admission to discharge divided by the LOS (FIM efficiency = FIM points gained/total days). LOS, total hospital, pharmacy and individual therapy charges were obtained from medical records. Discharge disposition locations were determined from the case management section within the patient charts. Patients were discharged to one of four locations: home, a skilled nursing facility, acute transfer back to the hospital, or assisted living.

Rehabilitation intervention

During the rehabilitation program, each patient completed a comprehensive interdisciplinary inpatient rehabilitation under the supervision of a physiatrist and therapy team. Each patient received ∼3 h of supervised therapy daily from both physical and occupational therapists. This rehabilitation plan is summarized in Table 1. Standardized protocol driven therapy sessions were conducted twice daily, once in the morning and afternoon. During therapy sessions, patients engaged in activities to increase flexibility and range of motion, improve independence in performing activities of daily living, and to improve proprioception, gait and balance.

Statistical analyses

Statistical analyses were performed using the Statistical Package for the Social Sciences software (version 14.0). All data are expressed as mean ± s.d. of measurement. Frequency distribution of diagnosis etiologies, ethnicities, smoking and alcohol use, comorbidity types and disposition locations were analyzed using Chi-Square tests (χ²). Since there were four groups of different sample sizes, non-parametric Kruskal–Wallis tests were used to determine whether differences existed among the groups for outcome variables. Given that our previous work has shown that TKA type (primary vs. revision) influences TKA outcomes, we examined the potential interactions between the BMI group and TKA status on the main outcome variables of admission and discharge FIM scores, FIM efficiency, LOS and total charges by applying a univariate ANOVA with the group factors of BMI group and TKA status in this study. Significant interactions are presented. Hierarchal linear regression models were generated to examine whether BMI was a significant contributor to the models predicting LOS, FIM efficiency and total hospital charges. Models were adjusted for several factors that influence these outcomes (age, gender, race, revision status, race, comorbidity number, admission hematocrit) (13,14,21). The α level was established at 0.05 a priori for all statistical tests.

Patient characteristics

Characteristics and comorbidities of the study sample are shown in Table 2. The severely obese group contained a greater proportion of younger, shorter, female patients and African Americans than the non-obese and moderately obese groups (P < 0.05). Similarly, resting glucose levels were also higher in the severely obese than in the non-obese or moderately obese groups (P < 0.0001). The proportions of patients with unilateral/bilateral or primary revision TKAs were not different among groups (P > 0.05); patients in the severely obese group had a higher prevalence of revisions, particularly related to mechanical reasons. Frequencies of patients instructed for specific weight bearing status were also not different among BMI groups. Fasting glucose levels were 5.1–15.5% higher in the severely obese group than the non-obese and moderately obese groups (P < 0.05); admission hematocrit values ranged from 30.2 to 30.5% with no difference among groups (P > 0.05).

Comorbidities of the patient sample were collected. Major comorbidities included diabetes mellitus, rheumatoid arthritis, peripheral arterial disease, hypertension, congestive heart failure, anemia, hypothyroidism, cellulitis, urinary tract infection, neuropathy, depression and anxiety. The prevalence of these comorbidities were not different among the four groups (the lowest P value for the χ² tests of these distributions was 0.189), indicating homogeneity among the study groups.

BMI and functional outcomes

The raw FIM scores from admission to discharge and score changes (Discharge FIM–Admission FIM), LOS, FIM efficiency and itemized hospital charges are presented in Table 3. The FIM change was significantly higher in the BMI <25 kg/m² compared to all the three remaining groups (P < 0.0001). The LOS was 6% shorter in the moderately obese group than in the non-obese group (P < 0.0001). The lowest FIM efficiency value of all groups was found in the severely obese group (P = 0.044); this FIM efficiency value ranged on average from 8.1 to 16.2% lower than in the remaining three groups.

The severely obese group had the highest daily charges among all BMI groups (P < 0.0001). Discharge destinations were similar among all the BMI brackets; similar proportions of patients were discharged home, transferred to a skilled nursing facility or transferred back to acute care in the hospital.

FIM categories and subscores

The motor and cognition FIM scores and FIM scores related to lower body function are presented in Table 4. The changes in motor and cognition FIM scores were higher in the non-obese group than the severely obese group (P < 0.05). Discharge scores for lower body dressing, tub-shower transfers, walking and toilet transfers were higher in the non-obese than the very severely obese group (all P < 0.05). However, the change scores in toilet transfer and walking without assistance change scores were higher only in the non-obese group than the severely obese group of all the FIM subscores (all P < 0.05).

Interactions of BMI and primary/revision TKA status

Several interactions were found for the main clinical outcome variables when analyzed using BMI and TKA status as shown in Table 5. While there was no significant interaction found for the change in total FIM scores from admission to discharge for these two factors (F = 0.894, degrees of freedom = 3, P = 0.443), this may be explained by a low power value for this specific analysis (power = 0.538). There were no significant interactions for BMI–TKA status among percent changes in motor and cognition FIM scores; there were significant interactions for discharge motor and cognition FIM scores (both P < 0.0001). These FIM subscores are presented in Figures 1 and 2.

A significant interaction was found for TKA status and BMI group for LOS (F = 2.771, degrees of freedom = 3, P = 0.042). The interaction for group differences among FIM efficiency values was P = 0.099, with both main effects being significant (P < 0.05). A significant interaction was found for TKA status and BMI group for total charges (F = 2.779, degrees of freedom = 3, P = 0.040). While there were no significant interactions found for physical therapy or pharmacy charges, the interaction between TKA status and BMI group was significant for occupational therapy charges (F = 5.40, degrees of freedom = 3, P = 0.001). Discharge destinations were not different among these interactive factors.

In regression models adjusted for major influential factors, BMI emerged as a significant but very small contributor to the variance of the regression models of LOS, FIM efficiency and total hospital charges, but did not reach statistical significance for independence with walking (Table 6; all P < 0.05).

Obesity and rehabilitation outcomes

While comparative data are scarce regarding the effects of BMI on inpatient rehabilitation outcomes following TKA, we recently completed two studies regarding obesity effects in TKA (13) and total hip arthroplasty (20) populations. Our earlier data indicated that in primary TKA, total FIM scores were 3.9% higher in the obese than in the non-obese group by the time of discharge. Also, a high BMI was associated with a LOS 1.3 days longer, and 21% higher total hospital charges than a low BMI in primary TKA (13). Finally, when BMI was added to a regression model to predict FIM efficiency following TKA, BMI was not a significant predictor (13). While these earlier studies were not able to provide details as to the source of the increased costs due to excessive adiposity, the present study determined that the obesity-related elevation in costs was due to a combination of higher therapy and pharmacy charges.

In this study, the LOS was longest in the non-obese group among all four groups. While this is contrary to our preliminary study, it is likely that the LOS is due to the advanced age of the non-obese group. The average non-obese patient was on an average 11.2 years older than the severely obese patient; previous data indicate that advancing age increases LOS (22) or predicts long LOS (23) in the rehabilitation facility following TKA. Presurgical characteristics such as initially low FIM scores (21) may have contributed to the longer duration of rehabilitation so as to achieve meaningful improvements in the components of the overall FIM score, such as mobility and transfers. Other contributors to longer LOS in the older non-obese group could include physical joint limitations, tolerance to pain or deconditioning as has been suggested in hip arthroplasty (15). Finally, the body composition shift that occurs with aging (muscle mass loss and adipose tissue accumulation) (24), is associated with loss of strength and with perceived and observed functional limitations and poor performance with load bearing activities (stair ascent, 10-m walk, stair descent, and chair stand) (25). It is possible that this body composition shift might have contributed to delayed functional gain, a longer LOS and overall increased facility charges in a fraction of the non-obese patients in that group in the present study.

The overall total FIM change, and select specific motor scores (transfers, walking) in the severely obese group were lowest among all groups by the time of discharge. In addition to BMI, the severely obese group was characterized by the highest composition of women and African Americans. Female gender has been identified as a factor associated with low strength and muscle tone (26), a greater prevalence of sedentary behavior (27) and lengthy inpatient rehabilitation LOS in this population (22,23). Both women and African Americans have been shown in previous studies to utilize arthroplasty procedures later in the course of osteoarthritis disease; this delay worsens the presurgical status and pain symptoms of the patient (23). Electing for surgery later in the disease (due to lower willingness) leads to lower expectations of the hospital course, pain, and function following replacement surgery (28); longer recovery periods (23); and potentially less favorable outcomes in African Americans. It is possible that these characteristics, coupled with severe obesity led to a lesser overall functional gain, a greater need for therapies, and subsequently greater hospital and itemized charges during their hospital stay.

Interaction of TKA status and BMI

To our knowledge, this paper is the first to report the significant interactive effects between TKA status (primary vs. revision) and BMI on several main inpatient rehabilitation outcomes including high total and occupational therapy charges, long LOS and low discharge motor and cognitive FIM scores. This suggests that less functional independence is achieved by the time of discharge and prolonged rehabilitation may occur when revision surgery is accompanied by high BMI. This translates into a more costly inpatient rehabilitation stay.

Even though there were relatively few patients in the revision TKA group who were severely obese, this finding is clinically relevant in that obesity is associated with knee arthroplasty failures (29), focal osteolysis (30) and aseptic loosening of the prosthesis (31). In addition, revision TKA rates are higher in the obese than the non-obese, and especially severely obese, patients (32). Given that the prevalence of obesity is increasing in young people and that joint problems are developing earlier in their lifetime, the numbers of revision surgeries in obese patients are likely to increase over the next decades. What is the appropriate clinical venue for these patients? While the costs of inpatient rehabilitation may be higher in obese than in non-obese patients, the excess cost is lower than that of prolonged acute care stays. The excess cost of the therapies in the BMI >40 kg/m² obese group, supports the need for these aggressive therapies. In other discharge environments such as a skilled nursing facility, the provision of rigorous therapies is diminishing due to changes in Medicare reimbursement (7); one speculation is that obese individuals may not derive as much benefit from an environment driven less by mobility and the need for discharge. Also, the techniques of facilitating mobility learned in the inpatient rehabilitation setting may be useful for the patient over a long term in the maintenance of functionality, although this has yet to be shown. Alternatively, the very severely obese population might derive more benefit from an extended stay in rehabilitation to maximize benefits after the arthroplasty procedure before going home. Further research is required to examine whether inpatient rehabilitation has a beneficial functional and economic impact on long-term outcomes compared to prolonged acute care and discharge to the patient's home.

Study limitations

While this retrospective study provides novel data regarding rehabilitation outcomes in a large TKA population, this study has some limitations. Patients were not matched for all possible confounders such as presurgical functional status, patient characteristics, social support, personality traits and motivation levels that can influence recovery processes. Socioeconomic status, acute care complications, and joint pain development during recovery prior to admission may also affect rehabilitation outcomes and should be assessed in future studies. Important information that could assist in interpreting these findings would be body composition, and these assessments were not available for this database.

Finally, while the sample size was very large, the detection of statistical differences need to be interpreted with caution, as they may not necessarily be clinically relevant for all measures. Also, the relatively small contribution of BMI to main rehabilitation outcomes after adjusting for other influential factors is important to note. For example, the differences in total, daily and itemized charges or improvement levels in specific FIM lower body activities were small. This suggests that the average severely obese individual may not find it excessively costly ($10–$36 excess daily charges) or not be very difficult to rehabilitate (with only slightly lower discharge scores of tub-shower, lower body dressing and walking); however, the accumulated costs and lower functional level for some activities at the national level can mean millions of dollars of excessive charges using current estimates of TKA procedures (33) or more physical assistance at discharge for a part of this very obese population.

Conclusions

Excessive BMI does not prevent gains in physical function, but it is associated with lower overall functional change by the time of discharge and higher total and itemized inpatient rehabilitation charges following TKA. The presence of revision TKA and excessive BMI worsens clinical rehabilitation outcomes. While the severely obese population would be permitted to participate in inpatient rehabilitation under the new Centers for Medicare Services eligibility criteria, patients with a range of BMIs derive meaningful functional benefits with rehabilitation. BMI itself may not be a single criterion which defines the appropriateness of rehabilitation following TKA.