Background Recovery of function such as the ability to walk without an assistive device after total hip arthroplasty (THA) is not always automatic.
Objective This study investigated whether predetermined variables could be used to identify patients who might have functional limitations at 6 months following THA.
Design A prospective, observational cohort design was used.
Method Demographics and baseline measures, including age, sex, and preoperative Lower Extremity Functional Scale (LEFS) score, were collected 1 to 3 weeks prior to surgery from 40 participants who were scheduled to undergo THA. Six weeks after surgery, a second LEFS score was recorded along with each participant's body mass index and the THA procedure performed; walking speed and balance also were assessed at this time using the 10-Meter Walk Test, the Timed “Up & Go” Test, and the Functional Reach Test. At 6 months following surgery, each participant's functional outcome was determined from the final LEFS score and the need for an assistive device. Classification and regression tree (CART) analyses and logistic regression were used to establish which of the variables could predict outcome at 6 months.
Results Body mass index, sex, and age were identified by CART analysis as predictors to classify participants who did not reach successful outcome status. Logistic regression revealed that sex (female) was the only individual variable that predicted outcome at 6 months. Walking speed was the only performance variable identified as a predictor for outcome using CART analysis.
Limitations Only a limited number of variables were observed due to the small sample size.
Conclusion It is possible to identify those patients who are at risk for an unsuccessful outcome through the use of variables such as body mass index, age, and sex.
Total hip arthroplasty (THA) is a common procedure and the treatment of choice for patients with intractable pain and limited function arising from arthropathy of the hip joint.1 According to the Centers for Disease Control and Prevention (CDC), 230,000 THAs were performed in 2007 in the United States.2 More than 70% of all THAs are performed because of osteoarthritis in the hip, and THA also is the surgery of choice for complications following hip fractures, rheumatoid arthritis, and congenital dislocation of the hip.3–5
Total hip arthroplasty is customarily performed to reduce pain and improve functional mobility, in particular the ability to walk.3 Regaining the ability to walk after THA requires that the patient achieve an appropriate range of motion in the hip joint, adequate strength of the hip musculature, and sufficient balance during the single-limb stance phase of the walking cycle. Several studies have shown that at 1 year after THA surgery, some patients still have postural instability and strength deficits in the surgical limb, affecting their walking speed and maximum walking distance.3,6–8 In a study by Trudelle-Jackson and Smith,8 strength deficits of 10% to 18% in the musculature surrounding the surgical hip were found 1 year following surgery. This same study showed mean differences of 25.9%, 27.2%, and 31.5% between the involved and uninvolved sides for mediolateral stability, anteroposterior stability, and total stability, respectively. Using dynamic posturography, Nallegowda et al7 identified significant differences in dynamic balance, range of motion, and altered weight bearing that existed between patients who had undergone THA and controls. Vissers et al9 reported in a recent systematic review and meta-analysis that patients who underwent THA recovered their physical function to within 80% of controls within 6 to 8 months postoperatively.
Having less than optimal muscle strength or balance deficits may necessitate the use of an assistive device following THA. Many patients use an assistive device prior to surgery because of the pain experienced in the hip during walking. They also may use an assistive device immediately after surgery, but they do not expect to need it for the long term. However, some patients will continue to need to use an assistive device following surgery for as long as 2 years.10 Furthermore, some patients will walk without an assistive device 6 months postoperatively, but their speed of walking is 15% to 25 % below normal. This decrease in speed can last between 2 and 4 years following surgery.11
Some studies indicate that physical therapy is effective in improving functional mobility such as walking in the acute postoperative phase through 1 year following THA.8,12,13 Although it may seem logical for a course of physical therapy to be a standard intervention to maximize functional mobility after THA, not all patients undergo physical therapy intervention beyond the acute postoperative phase. Many patients do receive hospital-based physical therapy following surgery that addresses transfer ability and gait with an assistive device, but these patients do not receive outpatient physical therapy.3 When patients do not have the opportunity for rehabilitation after THA to maximize recovery of strength, range of motion, and balance in a controlled environment, they run the potential risk of a suboptimal recovery.14
Currently, the literature provides scant information about protocols for identifying patients who will have suboptimal outcomes following THA. Studies that investigated the capabilities of individual variables such as body mass index (BMI),15–17 age,18,19 sex,20,21 preoperative function,22 and surgical approach23 to predict function did not consider the interaction of these variables in the prediction equation. It is reasonable that the combination of variables will provide a better way to identify those patients who are at risk of suboptimal outcomes. Early identification of this subgroup of patients is crucial because when these patients have suboptimal motion patterns such as altered gait kinematics due to weakness of the hip abductors, they run the risk of joint instability, implant loosening, or other complications.14 Using the traditional radiographic, range of motion, and patient-reported outcome measures to show the “success” of a THA may not fully identify patients who have functional mobility limitations.4,24,25
The purpose of this study was to use demographic information and the outcomes of physical performance measures collected at 6 weeks after surgery to identify patients who might be at risk for limited functional mobility 6 months postoperatively. Patients were identified as having limited functional mobility at 6 months after surgery if they continued to need any assistive device for ambulation and had limited functional change, as identified by the change score on the Lower Extremity Functional Scale (LEFS). The LEFS change score was predetermined as at least a 9-point change from before surgery to 6 months after surgery.
A prospective, observational cohort design was used to investigate the ability of 5 predetermined variables and the outcomes of 3 physical performance tests recorded at 6 weeks after surgery to predict the functional outcome at 6 months following THA.
A sample of 40 consecutive patients from 3 orthopedic surgeons in a defined geographical region was used for this study. Potential study participants were screened by the orthopedic surgeons. Once these potential participants were identified, the orthopedic surgeons' offices provided them with information about the study and the contact information for the principal investigator. The participants then contacted the primary investigator for further details about the study. At this time, the inclusion and exclusion criteria were reviewed with each participant to determine his or her eligibility for inclusion in the study (Tab. 1). If the participant was eligible and willing to participate, a preoperative meeting was arranged within 1 to 3 weeks of the scheduled surgery. The participant was fully informed about the study before signing the written consent forms. The written consent forms in this study were approved by 2 independent institutional review boards.
Participants in this study were contacted by telephone following their THA to schedule their 6-week study visit. At 6 months following surgery, the principal investigator called the patient on the telephone to establish his or her need for an assistive device. Also at 6 months, participants were mailed and required to complete a third and final LEFS.
The initial assessment of study participants occurred in a hospital-based orthopedic outpatient clinic 1 to 3 weeks prior to each participant's date for surgery. At this time, the following variables were recorded: age, sex, and the current LEFS score. Six weeks following THA, the surgical approach used in the THA was recorded from the participant's medical file, as was the patient's BMI at the time of surgery. Also at this time, participants underwent assessment of walking speed and balance, and they completed a second LEFS. All participants were instructed in a home exercise program (HEP) of range of motion and strength exercises and were given written instructions for the exercises. To improve adherence in completing the HEP, participants were given an exercise log to complete. At 6 months following each participant's THA, the principal investigator telephoned the participant, who reported if he or she used an assistive device indoors or outdoors because of the weakness or balance deficits in the region of the surgical hip. Also at this time, a third and final LEFS was mailed to each study participant; participants completed the LEFS and returned it to the principal investigator along with the daily exercise log.
Demographic variables (age, sex, and BMI), the LEFS, and the surgical approach used were chosen as the first set of outcome variables in this study. These variables often are available in a patient's medical chart, and they have been used previously as predictors of outcome following THA.15,17,20,23 The LEFS is a measure of activity limitation developed for the lower extremities and has been shown to be a valid tool in the measurement of lower-extremity function in a population of patients with orthopedic problems. A change of 9 points on the LEFS has been shown to represent a minimal clinically important difference.26
The second set of outcome variables were 3 physical performance measures (walking speed, Timed “Up & Go” Test [TUG], and the Functional Reach Test [FRT]). These measures are easily and quickly performed in physical therapy clinics and do not require expensive equipment. The reliability of these measures has already been established.27–29
Sample Size Justification
A minimum number of participants were not required for the classification and regression tree (CART) analysis because it is nonparametric. Specifically, CART analyses do not rely on a normal distribution of the data, which often leads to the need for relatively large samples. Based on the projected number of hip replacements in the study's geographic region over the last several years, 40 was a reasonable number of participants to enroll in a 6-month time frame, allowing for the 6-month postoperative follow-up. In comparison with CART analysis, there are guidelines for sample size relative to the number of independent variables that are being considered in logistic regression. It is normally recommended that there be at least 10 participants for each variable examined in the logistic regression.30
Classification and regression tree analysis was performed using R statistical software (Lucent Laboratories, Paris, France) in order to determine the relationships between 2 groups of “predictor” variables and the functional outcome of the participants 6 months postoperatively. The R statistical software uses the twoing method for the CART analysis. A participant was identified as having a successful functional outcome 6 months postoperatively if he or she met the following criteria: the ability to walk without an assistive device and a minimum change score on the LEFS of at least 9 points when assessing from the preoperative functional level to the functional level 6 months following surgery. A participant was identified as not having a successful outcome if either or both of these criteria were not met.
The first group of predictor variables were age, sex, BMI, presurgical LEFS score, and surgical approach, and the second group of variables were the results of the physical performance measures. With the first group, the CART analysis was used to determine whether a combination of any of the 5 predetermined variables could predict outcome following THA. The second group of predictor variables included 3 physical performance measures commonly used by physical therapists: the TUG, the FFT, and walking speed. Scores on these measures were compared with normative values for age and sex.29,31,32 A participant was identified as having met or not met the normative value for each physical performance test when the measures were conducted 6 weeks following surgery.
Logistic regression was performed using the SPSS statistical software package, version 14 (SPSS Inc, Chicago, Illinois). Because the outcome of this study was dichotomous, it was fitting to use logistic regression to identify predictors. A forward step-wise logistic regression was used to assess variables that could predict the outcome. Potential variables were inserted into the regression model one at a time until the residual could no longer be explained by an additional variable. The significance level was set at .05.
Forty-eight patients undergoing THA were screened between December 2008 and July 2009. Forty of these patients met the inclusion criteria. An outline of the intake and attrition of participants during the study is provided in Figure 1. All 40 participants underwent the preoperative testing 1 to 3 weeks prior to their scheduled surgery date. Two of the 40 participants were dropped from the study at the 6-week postoperative testing: 1 participant did not undergo THA because of a diagnosis of leukemia, and the other participant did not undergo THA because of cardiac-related concerns. The remaining 38 participants returned 6 weeks following their THA for the postoperative visit. At the 6-month follow-up, 1 participant did not return the final LEFS. The demographic and functional characteristics of the participants are listed in Table 2.
The visual representation of the CART analysis for the first group of variables is presented in Figure 2. At the root node, the first level of dichotomy occurs where participants fall into either the left node, where their BMI is less than 34 kg/m2, or the right node, where their BMI is 34 kg/m2 or higher. Of those participants who had a BMI of 34 kg/m2 or higher, 66% were unsuccessful in their functional outcome at 6 months after THA. Of the participants who had a BMI of less than 34 kg/m2, only 13% had an unsuccessful functional outcome. At this initial level of dichotomization, 83.78% of the participants were correctly classified into the appropriate branch.
A further dichotomization occurred in those participants with a BMI of less than 34 kg/m2 with the variable of age. All participants who were younger than 68.5 years of age had successful outcomes. Of those participants aged 68.5 years or older, 24% had an unsuccessful outcome. The percentage of correctly classified participants dropped to 53.06% in this second dichotomy.
A third dichotomy occurred for participants who were 68.5 years of age or older. This dichotomy occurred relative to sex. Ten percent of the female participants had an unsuccessful outcome compared with 60% of the male participants. The percentage of correctly classified participants was 76.47% in this third dichotomy.
The final dichotomy occurred again with age. All female participants aged 73.5 years or older had successful outcomes; 20% of the female participants under the age of 73.5 years had an unsuccessful outcome. At this fourth and final level of dichotomization, 33% of the participants were correctly classified.
The results of the logistic regression indicated the overall model of one predictor (sex: female) was statistically significant in distinguishing between a successful and an unsuccessful outcome at 6 months following THA (−log likelihood=34.37, P=.039). The model correctly classified 78% of the participants. Female participants were more likely than male participants to have a successful outcome.
A visual representation of the CART analysis for the second group of variables is presented in Figure 3. Data from 37 participants were fed into the classification system (root). In the first layer of classification in Figure 3, the left branch indicates those participants who did not achieve or surpass the walking speed for their age- and sex-matched equivalent. Twenty-six participants did not meet this equivalent, but 19 (69%) of these 26 participants still had a successful outcome at 6 months following THA. The right side of the first layer of classification represents the 11 participants who did meet the age- and sex-matched comparative walking speeds. Every participant in this branch had a successful outcome. In this single level of dichotomization, only 48.64% of the participants were correctly classified. The TUG and the FRT were not identified by the CART analysis as predictors of success at 6 months.
Forward stepwise logistic regression was conducted to establish which variables (walking speed, TUG, and FRT) were predictors of status for success 6 months after THA. Regression results indicated that no variables were able to distinguish between successful and unsuccessful outcomes at 6 months following THA.
The identification of predictor variables for less-than-optimal functional outcome following THA may allow physical therapists to provide appropriate and timely interventions to improve function such as the ability to walk with optimal motion patterns. The purpose of this study was to establish whether either a selection of predetermined variables or the results from 3 physical performance tests that were completed 6 weeks after THA could predict a patient's outcome at 6 months following THA.
The CART analysis showed that BMI, age, sex, and walking speed were variables that could identify the outcomes of the study participants. Body mass index was the variable that led to the initial split in the outcome in the selection of predetermined variables such as age, sex, BMI, surgical approach, and preoperative functional levels. Six participants had a BMI of 34 kg/m2 or higher. Of these 6 participants, two thirds (4 out of 6) had an unsuccessful outcome. Previous prediction studies using BMI as an independent variable were inconclusive concerning whether BMI influences the outcome after THA.15–17,33 The results of the current study support the premise that BMI does influence the functional outcome after THA.
Dowsey and Choong15 and Namba et al16 examined BMI as a risk factor in the occurrence of prosthetic infection after THA. Each of these studies demonstrated a higher infection rate in patients with obesity than in those without obesity, and the authors attributed this heightened infection rate to the increased duration of surgery. Dowsey and Choong15 cited the CDC guidelines, which state that an operative time longer than 2 hours is considered a risk factor for surgical site infection after a THA. It is possible that the higher rate of infection affects the patient's ability to follow a normal recovery pattern after a THA. Moran et al17 and Jackson et al,33 however, contended that the outcome following THA is not influenced by a patient's BMI. Jackson et al33 reviewed 2,026 consecutive cases of THA and dichotomized the patients into 2 groups: those who had a BMI of less than 30 kg/m2 and those with a BMI of greater than 30 kg/m2. They reported that although patients whose BMIs were below 30 kg/m2 had a significantly higher score on the Harris Hip Questionnaire and better hip flexion range of motion (P<.001), there was no difference in patient satisfaction between the 2 groups. Furthermore, there was no significant difference in the rate of revisions between the 2 groups. Basing their conclusion on the lack of difference between revision rates, the authors concluded that a patient's BMI did not affect the outcome.
Another split of the CART analysis occurred with the variable of age. The outcome of those participants who had a BMI of less than 34 kg/m2 was further divided on age. All participants under the age of 68.5 years had a successful outcome. The role of age in determining the postoperative outcome after THA was highlighted by Nilsdotter and Lohmander,18 who found patients under the age of 72 years had higher Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) and 36-Item Short-Form Health Survey (SF-36) scores following THA than those aged 72 years and older, despite both groups having equivalent presurgery scores. These findings were supported in this current study, where all participants who were under 68.5 years of age and had a BMI lower than 34 kg/m2 had a successful outcome.
A third division in the CART analysis occurred around sex. Sex has been shown in previous research to influence outcomes following surgery.20,21 Hawker et al20 found that women had a higher prevalence of osteoarthritis of the hip and knee, had more pronounced symptoms, and had greater disability than men. However, the Hawker et al study did not identify whether women had an improved functional outcome compared with men. In the current study, CART analysis showed that men were more likely than women to have difficulty, and the logistic regression highlighted that women would have a more successful functional outcome.
The outcome variables of surgical approach and change in LEFS score did not provide prediction according to the CART analysis or the logistic regression. However, the inability of these variables to determine prediction could be related to the small sample size in the CART analysis and potential overfitting of the data in the logistic regression. Although preoperative function was not identified as a variable that could predict outcome in the current study, other studies support its predictive role.22,34 Kennedy et al22 found that that preoperative function was significant (P≤.001) in predicting the outcome of 152 participants who had had a THA or a total knee replacement. The results of their study may differ from the findings of the current study because Kennedy et al used an outcome measure (ie, WOMAC) that assessed alternate domains of recovery. The WOMAC captures changes in function and changes in pain levels after THA; the LEFS captures only changes in function. Therefore, the WOMAC and LEFS do not measure the same constructs. The findings of Kennedy et al are supported by Holtzman et al,34 who reported that patients who needed assistance with walking at baseline were more likely to need assistance at 1 year than those who did not need assistance prior to surgery (38% compared with 15%; P<.01).
In the current study, the surgical approach also was not identified as a factor in the prediction of outcome after THA in either the CART analysis or the logistic regression. Klausmeier et al23 examined whether short-term recovery of hip strength and motion differed between the anterior and anterolateral approaches. The results of the kinematic and kinetic studies revealed that although the anterior approach was associated with improved gait speed and peak flexor moments at 6 weeks compared with before surgery, no differences between the 2 approaches was found for most of the isometric strength and dynamic gait measures at 6 and 16 weeks.
Previous predictive research for postoperative outcomes following THA focused mainly on the individualized variables of age, sex, BMI, and preoperative function. Few studies in the literature provide evidence that a physical performance measure such as the TUG or walking speed might predict outcome. The results of the current study did suggest that those participants who walked at a speed at or above their age- and sex-matched comparative values had a successful functional outcome. However, those participants who did not attain the optimal walking speed had a mixed outcome. Of the 26 participants who did not meet their expected values, 69% still had a successful outcome. The assessment of walking speed should allow a physical therapist to clearly identify those participants who will be expected to have a successful outcome. When a participant does not meet these values, further assessment of the variables of BMI, sex, and age should be considered.
This study had several limitations. One limitation of the study was the choice of variables that were examined as potential predictors of success after THA. Five predetermined variables were used to answer the first question. These variables may not have been the optimal choices, or there may have been additional variables, such as preoperative pain,34 waiting time to surgery,18 or presurgical health status19 of the patient, that should have been considered. A further limitation of the study with reference to sample size was the greater-than-expected attrition rate.
The generalizability of the results also was a limitation of this study. Participants were recruited from an orthopedic practice in a specific geographical region of the United States. The characteristics of these patients may not have been representative of patients in other parts of the United States or elsewhere in the world. Although the sex and age of the participants in this study are similar to those of the participants of other studies examining prediction of outcome after THA,18–20 the lifestyles of participants in this study may have affected their preoperative functional levels. Adherence to exercising after surgery also may have been a factor that affected outcome. Only 40% of the study participants returned the exercise log when the third LEFS was mailed back to the principal investigator. Further study using a multisite model would increase the generalizability of the results from studies such as this one.
The results may be limited due to the possibility of overfitting the logistic regression analysis with 5 variables rather than the suggested minimum of 10 participants per independent variable.35 If there had been 50 participants in the study, it is possible that the role of preoperative function as measured with the pre-LEFS score might have been identified as a predictor variable of the outcome. Future studies using CART analysis and logistic regression should cast a wider net to capture more demographics and measurable variables that are easily accessible by the clinician. Future studies also should use cross-validation to evaluate the quality of the prediction of the tree. Cross-validation was not completed in this study due to limited access to sufficient data to undertake this process. Therefore, the interpretation of the results of this study should reflect this limitation.
Classification and regression tree analysis provides an innovative way to statistically examine the role of variable interaction in the prediction of functional outcome. It is possible that the participants' BMI, age, and sex were variables that, when considered together, could have influenced the functional outcome at 6 months following THA. These variables were identifiable at 6 weeks following THA. The information from the study can be used on 2 levels. At the first level, physical therapists can influence the BMI of a patient before surgery is performed through education and altered lifestyle. The limitation of this model is that a component of weight loss relates to increased activity, which may or may not be possible when advanced osteoarthritis of the hip is present. The second way that physical therapists can be involved is in the early detection of those patients who might be most at risk for a less-than-favorable outcome. Identification of these patients allows for the timely provision of physical therapy services with a goal of improving their long-term functional outcome.
The Bottom Line
What do we already know about this topic?
Patients typically undergo a total hip arthoplasty (THA) for pain relief and increased functional mobility. However, not all patients regain optimal functional mobility, such as the ability to walk without the need for an assistive device, after their THA. Early identification of patients who are at risk of not regaining optimal mobility would allow for appropriate interventions, including physical therapy interventions, to improve long-term outcomes.
What new information does this study offer?
This study identified factors, such as body mass index (BMI) and sex, that can affect outcomes. For example, male patients with a BMI over 34.1 kg/m2 had a less favorable outcome at 6 months following THA than women with any BMI and men with a lower BMI.
If you're a patient, what might these findings mean for you?
Clinicians may use this information to help identify patients at risk for a less-than-favorable functional outcome following THA, allowing for prompt provision of physical therapy. Weight loss strategies for patients with obesity prior to surgery may assist in optimizing functional outcome at 6 months following THA.
The author thanks her doctoral dissertation committee of Samuel Cheng, PT, ScD, Madeline Hellman, PT, MHM, EdD, and Eric Dugan, PhD, for their support in this study. Thanks also are extended to Holmes Finch, PhD, for his statistical support and to Cheryll Adams, PhD, for providing assistance with the manuscript.
This study was completed in partial fulfillment of the requirements for Dr Slaven's doctoral degree in physical therapy from Nova Southeastern University.
This study was approved by the institutional review boards of Ball Memorial Hospital, Muncie, Indiana, and Nova Southeastern University, Fort Lauderdale, Florida.
A platform presentation of this study was given at the Combined Sections Meeting of the American Physical Therapy Association; February 9–12, 2011; New Orleans, Louisiana. A subsection of the data was presented as a poster presentation in the Orthopaedic Section at the Combined Sections Meeting of the American Physical Therapy Association; February 8–11, 2012; Chicago, Illinois.
- Received December 25, 2011.
- Accepted July 31, 2012.
- © 2012 American Physical Therapy Association