Patient characteristics do not predict the change in physical functioning following arthroplasty measured by the HOOS-PS and KOOS-PS

Abstract

Predicting personalized outcome after arthroplasty improves shared decision-making. The aim of this paper was to determine predictors of functional outcome measured by the Hip disability and Osteoarthritis Outcome Score - Physical function Shortform (HOOS-PS) or Knee injury and Osteoarthritis Outcome Score - Physical function Shortform (KOOS-PS) in patients undergoing total hip (n = 79) or total knee arthroplasty (n = 90) respectively. Patients were assessed at baseline and following arthroplasty. A multiple regression analysis showed that the included variables predicted the change score in HOOS-PS limited (F (8,66) = 3.139, p = 0.005, adjusted R² = 0.188) and the KOOS-PS not significantly (F (8,73) = 0.837, p = 0.573, adjusted R² = −0.016). Concluding, baseline characteristics cannot be used for personalized prediction using the KOOS-PS and HOOS-PS.

1. Introduction

Total hip arthroplasty (THA) and total knee arthroplasty (TKA) are treatments in selected patients to relieve pain and improve physical functioning. However, not every patient experiences these improvements after surgery. Recent studies have demonstrated patient-related predictors at baseline for worse physical functioning following THA or TKA, including older age, higher body mass index (BMI), gender, functional limitation, other comorbidities, high American Society of Anaesthesiology (ASA) classification, current smokers, lower mental health score, lower educational level and severe pain1, 2, 3. The International Consortium for Health Outcomes Measurements (ICHOM) advices to measure several domains in patients undergoing arthroplasty to provide outcome information.⁴ These domains include functional outcome measured by the Hip disability and Osteoarthritis Outcome Score - Physical function Shortform (HOOS-PS)⁵ or Knee injury and Osteoarthritis Outcome Score - Physical function Shortform (KOOS-PS).⁶ Some research has been carried out investigating the influence of patient characteristics on functional outcome measured with the HOOS-PS and KOOS-PS and in patients with hip complaints.³,⁷ However, these previous studies did not analyse patients undergoing arthroplasty. The predicted postoperative functional outcome on the HOOS-PS or KOOS-PS can be personalized, when the patient's characteristics of influence are known. That will improve the information supply, the shared decision-making at the outpatient clinic and can manage expectations.

Therefore, the aim of this paper is to determine variables predicting functional outcome measured by the HOOS-PS or KOOS-PS. These patient characteristics can have prognostication value in total hip or total knee arthroplasty. The variables being explored were based on the literature, including patient characteristics, pain scores and quality of life (QoL).

2. Methods

Patients indicated for TKA or THA at the department of orthopaedic surgery in a district general hospital (St. Antonius Hospital, Utrecht), who were present at the hospital information meeting were prospectively enrolled between February 2017 and November 2017. Exclusion criteria were insufficient understanding of the Dutch written language or mental incompetence. The institutional review board approved this study.

A priori, patient-related characteristics found to be influencing functional outcome in previous studies were determined. These were age, gender, BMI (kg/m2), educational level (range 1–3; 1 = elementary school, 2 = high school graduate, 3 = college graduate), ASA classification (range 1–5; 1 represented best physical status⁸), smoking status (dichotomous; smoker or non-smoker), EQ-5D-5L index score (range 0–1; 1 represented best valued health-related quality of life^9,10 and pain during activity prior to surgery (Numeric Rating Scale (NRS) range 0–10; 0 represented zero pain and 10 was the worst imaginable pain). Al these variables were collected at baseline and included in the analysis.

Patients were asked to complete a paper questionnaire. Patients were assessed at baseline at the outpatient clinic and reassessed after arthroplasty by mail. To ensure full capacity of recovery, hip patients were approached at least three months post-surgery and knee patients at least 6 months post-surgery for post-operative evaluation. The primary outcome physical functioning was assessed with the HOOS-PS or KOOS-PS, consisting respectively of five and seven items to measure a person's disabilities. For each question, four points could be obtained, adding up to a maximum score of 20 for the HOOS-PS and 28 for the KOOS-PS. These scores were converted into a true interval score (0–100) with zero representing no complaints. Furthermore, patients were asked to complete a Numeric Rating Scale for pain during activity and the EQ-5D-5L index score (0–1, 1 represented full health). Postoperative the same questionnaire was sent by mail, including a free of charge return envelope. Patients who returned incomplete questionnaires were contacted by phone to answer the remaining questions.

2.1. Statistical analysis

Statistical analyses were performed with SPSS (IBM SPSS version 24.0.). Descriptive statistics were used to describe the study population. Baseline and postoperative outcome means were compared using the paired samples t-test. Multiple regression analysis was conducted to analyse the prediction of the variables on physical functioning change score (change/delta score; score following arthroplasty minus baseline score on the HOOS-PS and KOOS-PS. With the delta change scores used as outcome variables, multiple regression analysis was conducted with age, gender, BMI, ASA classification, educational level, smoking status, pre-operative pain and the EQ-5D index as predictive variables. A p value of <0.05 was considered statistically significant.¹¹ To meet the rule of thumb that on average 10 patients per independent variable are required for analysis,¹¹ the goal was to include at least 80 patients in each group.

3. Results

In total 189 patients were included, 86 in the THA group and 103 in the TKA group. Of the 86 THA patients, 7 (8.1%) patients were later on excluded; 2 because of missing data and 5 were lost to follow-up. This resulted in 79 THA patients with a mean age of 70.0 (±9.5) years and of which 43 (54.4%) were female. In the TKA group, 13 (12.6%) patients were excluded; 6 because of missing items and 7 were lost to follow-up. This resulted in 90 patients in the THA group with a mean age of 68.4 (±8.5) years and of which 61 patients (67.8%) were female (Table 1). The mean follow-up time for the THA group was 6 (±3) months and 8 (±2) months for the TKA group. Both groups improved significantly postoperatively based on HOOS-PS or KOOS-PS and pain scores (Table 2, Table 3).

Table 1.

Baseline characteristics of the THA and TKA patients.

Variable		Total hip (N = 79)	Total knee (N = 90)
		Mean ± SD	Mean ± SD
Age (years)		70.0 ± 9.5	68.4 ± 8.5
BMI (kg/m2)		27.9 ± 4.5	29.4 ± 4.4
		N (%)	N (%)
Gender (female)		43 (54.4%)	61 (67.8%)
Current smoker		8 (10.1%)	8 (8.9%)
Diabetes mellitus		7 (8.9%)	13 (14.4%)
Cardiac disease		21 (26.6%)	21 (23.3%)
Pulmonary disease		10 (12.7%)	13 (14.4%)
ASA classification	I	12 (15.2%)	14 (15.6%)
	II	61 (77.2%)	72 (80.0%)
	III	6 (7.6%)	4 (4.4%)

Abbreviations: BMI = body mass index, ASA = American Society of Anaesthesiology.

Table 2.

Pre- and postoperative outcome scores of the total hip arthroplasty group (N = 79).

		Baseline score	Postoperative score	P-value
HOOS-PS (0–100)		47.5 ± 18.3	23.4 ± 16.8	<0.001
Pain (0–10)	Frequency	7.9 ± 2.3	2.0 ± 2.9	<0.001
	During activity	7.4 ± 2.4	1.8 ± 2.4	<0.001
	During rest	5.5 ± 2.4	1.2 ± 2.2	<0.001
EQ-5D index		0.52 ± 0.24	0.81 ± 0.20	<0.001

Abbreviations: HOOS-PS = Hip Disability and Osteoarthritis Outcome Score - Physical Function Shortform; EQ-5D = EuroQol 5 Dimensions.

Table 3.

Pre- and postoperative outcome scores of the total knee arthroplasty group (N = 90).

		Baseline score	Postoperative score	P-value
KOOS-PS (0–100)		51.5 ± 14.4	32.0 ± 14.5	<0.001
Pain (0–10)	Frequency	8.2 ± 2.3	2.8 ± 3.0	<0.001
	During activity	7.8 ± 2.1	2.9 ± 2.8	<0.001
	During rest	5.4 ± 2.5	1.8 ± 2.4	<0.001
EQ-5D index		0.58 ± 0.16	0.79 ± 0.19	<0.001

Abbreviations: KOOS-PS = Knee Injury and Osteoarthritis Outcome Score - Physical Function Shortform; EQ-5D = EuroQol 5 Dimensions.

3.1. Predictors for functional change

A multiple regression analysis showed that the independent variables statistically significantly predicted the change score on the HOOS-PS F (8,66) = 3.139, p = 0.005, adjusted R² = 0.188. Only preoperative pain added individual statistically significantly to the prediction (patients having more pain preoperative is predictive for a higher change score) (Table 4).

Table 4.

Results of multiple regression models, predictors after multivariate analysis on change scores in KOOS-PS or HOOS-PS

Predictor	HOOS-PS		KOOS-PS
	P-value	Adjusted R2	P-value	Adjusted R2
Total	0.005	.188	0.573	−0.016
Age	0.232		0.339
Gender	0.670		0.239
BMI	1.000		0.140
Smoking	0.330		0.337
ASA classification	0.738		0.336
Preoperative pain during activity	0.020*		0.444
EQ-5D index preoperatively	0.106		0.695
Education	0.380		0.525

* = significant.

Abbreviations: KOOS-PS = Knee Injury and Osteoarthritis Outcome Score - Physical Function Shortform; HOOS-PS = Hip Disability and Osteoarthritis Outcome Score - Physical Function Shortform; BMI = body mass index; EQ-5D = EuroQol 5 Dimensions.

A multiple regression was performed to predict the change score of the KOOS-PS using the same predicting variables. These variables did not statistically significantly predicted the improvement in KOOS-PS F (8,73) = 0.837, p = 0.573, adjusted R² = −0.016. None of the individual variables added statistically significantly to the prediction (Table 4).

4. Discussion

This is the first study analysing the predictive value of baseline patient characteristics on the change score of the HOOS-PS or KOOS-PS following arthroplasty. This study selected possible predictive variables from literature and clinical practice. These variables could not explain the improvement of the KOOS-PS and only 19% of the HOOS-PS. Concluding, the variance of the HOOS-PS model explained by eight variables was limited and the KOOS-PS model's variance could be due to error. Therefore, the baseline variables could not be used predicting outcome on these measures for personalized predictions or shared decision making at the outpatient clinic in patients undergoing arthroplasty. There are several likely explanations for the limited predictive value of the baseline characteristic on the HOOS-PS and KOOS-PS.

First, although the most important variables were included, other important variables may be missing. The present study included eight possible predictive variables, based on the literature (1, 2, 3) and clinical practice. It was a priori decided to include a select amount of variables, as including more predictive variables would have consequences for the power of the study.

The second possible explanation for the limited predictive value of the model might be the HOOS-PS and KOOS-PS itself. It might be that the variables are predictive of functional improvement, but the HOOS-PS and KOOS-PS are not capable measuring physical functioning. Previous studies included the same possible predictive variables, however they measured physical functioning with other measurement instruments.^1,2

Third, this study was a single centre study. A multicentre study would provide a wider population to facilitate the subsequent generalisation of the findings. However, by approaching all patients consecutively in a certain period, an attempt was made to obtain a good sample of the standard population.

The literature reported the baseline physical functioning score as a predictive variable to predict post-operative physical functioning scores. Purposively this variable was not included in the current study. Since the objective of arthroplasty is improvement of physical functioning, the change score of physical functioning was used as dependent variable. Therefore, the baseline score is already part of the equation and should not be included as independent predictive variable.

Patient's characteristics could be helpful at the outpatient clinic to predict functional outcome after arthroplasty. Relevant patient characteristics could be used to predict personalized postoperative functional outcome, to improve the information supply, the shared decision-making at the outpatient clinic and manage expectations. However, this study shows that the improvement in functional outcome after hip arthroplasty can only be predicted to a limited extent and the improvement after knee arthroplasty cannot be predicted. This finding has several implications. First, the most common variables (age, gender, BMI, ASA score, etc.) cannot be used for patients counselling or selection of patients for joint arthroplasty to predict change scores on these measurement instruments following surgery. Second, these measurement instruments possibly measure a different construct than other instruments used in previous research. In other words, physical functioning cannot be measured using the HOOS-PS or KOOS-PS. Furthermore, this study shows that case mix variables have no effect on the measurement instrument scores, case mix variables are used to correct variability of outcome scores in e.g. arthroplasty registries and benchmarking through patient characteristics. Although commonly applied, the findings of this study suggest that correcting for case mix variables should not be done.

The results of this study show that the known patient characteristics of influence on physical functioning, cannot predict change scores of the KOOS-PS following total knee arthroplasty. Physical functioning following total hip arthroplasty measured by the HOOS-PS can only be predicted for a small part using eight baseline patient characteristics. Since the case-mix variables have only a slight effect on the scores of the HOOS-PS and KOOS-PS, weighing the results using case-mix adjustments should not be done when judging hospital performances. Furthermore, personalized predicting of the physical outcome is not possible using the most widespread patient characteristics. Therefore, these measurement instruments have no benefit for shared decision making regarding arthroplasty.