KNEE
Patient satisfaction with total knee replacement cannot be predicted from pre-operative variables alone P. N. Baker, S. Rushton, S. S. Jameson, M. Reed, P. Gregg, D. J. Deehan From Institute of Cellular Medicine, Newcastle University, Newcastle upon Tyne, United Kingdom P. N. Baker, MSc, Dip Stat, FRCS(Tr & Orth), Associate Clinical Researcher and Research Fellow, National Joint Registry for England and Wales Newcastle University, Institute of Cellular Medicine, Newcastle upon Tyne NE1 7RU, UK. S. Rushton, BA(Hons Oxon), PhD, Professor of Biological Modelling Newcastle University, School of Biology, Newcastle upon Tyne NE1 7RU, UK. S. S. Jameson, MRCS, National Joint Registry Research Fellow Durham University, School of Medicine, Pharmacy and Health, Queen's Campus, University Boulevard, Stockton-on-Tees TS17 6BH, UK. M. Reed, MD, FRCS(Tr & Orth), Consultant Orthopaedic Surgeon Northumbria Healthcare NHS Foundation Trust, Woodhorn Lane, Ashington, Northumberland NE63 9JJ, UK. P. Gregg, MD, FRCS, FRCS(Ed), Professor of Orthopaedic Surgical Science James Cook University Hospital, South Tees Hospitals NHS Foundation Trust, Marton Road, Middlesbrough TS4 3BW, UK. D. J. Deehan, MD, MSc, FRCS(Tr & Orth), Professor of Orthopaedic Surgery Freeman Hospital, Newcastle upon Tyne NHS Trust, Freeman Road, High Heaton, Newcastle upon Tyne NE7 7DN, UK. Correspondence should be sent to Mr P. N. Baker; e-mail:
[email protected] ©2013 The British Editorial Society of Bone & Joint Surgery doi:10.1302/0301-620X.95B10. 32281 $2.00 Bone Joint J 2013;95-B:1359–65. Received 23 April 2013; Accepted after revision 5 June 2013
A COHORT STUDY FROM THE NATIONAL JOINT REGISTRY FOR ENGLAND AND WALES Pre-operative variables are increasingly being used to determine eligibility for total knee replacement (TKR). This study was undertaken to evaluate the relationships, interactions and predictive capacity of variables available pre- and post-operatively on patient satisfaction following TKR. Using nationally collected patient reported outcome measures and data from the National Joint Registry for England and Wales, we identified 22 798 patients who underwent TKR for osteoarthritis between August 2008 and September 2010. The ability of specific covariates to predict satisfaction was assessed using ordinal logistic regression and structural equational modelling. Only 4959 (22%) of 22 278 patients rated the results of their TKR as ‘excellent’, despite the majority (71%, n = 15 882) perceiving their knee symptoms to be much improved. The strongest predictors of satisfaction were post-operative variables. Satisfaction was significantly and positively related to the perception of symptom improvement (operative success) and the post-operative EuroQol-5D score. While also significant within the models pre-operative variables were less important and had a minimal influence upon post-operative satisfaction. The most robust predictions of satisfaction occurred only when both pre- and post-operative variables were considered together. These findings question the appropriateness of restricting access to care based on arbitrary pre-operative thresholds as these factors have little bearing on postoperative satisfaction. Cite this article: Bone Joint J 2013;95-B:1359–65.
Total knee replacement (TKR) has now surpassed hip replacement as the most common joint arthroplasty procedure in the world.1 Approximately 80 000 TKRs are performed in England and Wales annually, at an estimated annual cost of £500 million.2,3 The demand for TKR is predicted to increase six-fold between 2005 and 2030 to reflect an increasingly elderly yet functionally demanding population.1 TKR reduces pain, improves functional capacity, and enhances quality of life for patients with symptomatic end-stage osteoarthritis (OA).4-8 However, despite these benefits, rates of patient satisfaction are less than 85%9-12 and up to 20% of patients fail to demonstrate improvements in health scores post-operatively.13 The Department of Health (DoH) in the United Kingdom routinely collects Patient Reported Outcome Measures (PROMs) for National Health Service (NHS) patients undergoing elective TKR.14 PROMs are designed to assess a surgical episode from the patient’s perspective, but are being increasingly used as tools
VOL. 95-B, No. 10, OCTOBER 2013
for limiting and/or prioritising access to surgery.14,15 Recent publications suggest that these policies create arbitrary thresholds that, when implemented, have no bearing upon post-operative functional improvements and levels of patient satisfaction.14,16 The impact of preoperative PROMs upon post-operative satisfaction is poorly understood and their discriminative capacity in predicting which patients are likely to be satisfied following TKR has yet to be established. A recent examination of the Oxford knee score (OKS)17 demonstrated that it had no predictive accuracy in relation to post-operative satisfaction.14 We aimed to extend this work by modelling the relationships, interactions and predictive capacity of variables available pre-operatively on post-operative patient satisfaction and symptomatic improvement following TKR performed for OA. The additional impact of post-operative variables upon satisfaction and symptomatic improvement were also examined. 1359
1360
P. N. BAKER, S. RUSHTON, S. S. JAMESON, M. REED, P. GREGG, D. J. DEEHAN
Patients and Methods This study was performed as a retrospective cohort analysis using prospectively collected PROMs and data from the National Joint Registry for England and Wales (NJR). Simultaneous applications were made to both the NHS Information Centre (IC) and the NJR research board for access to PROMs and their corresponding NJR records for patients who had undergone primary TKR since August 2008. By linking these two datasets the PROMs data could be combined with the corresponding demographic and operative details held within the NJR. PROMS data were collected by means of questionnaires sent out pre-operatively and at six months post-operatively. Questionnaires consisted of the OKS (scored 0 (worst) to 48 (best))17 and general measure of well-being through the EuroQoL-5D (EQ-5D))18 in addition to measures of patient reported satisfaction and the perception of symptomatic improvement (success) (the latter collected post-operative only). From within the NJR and PROMs datasets a total of 25 011 linked records of TKRs carried out between August 2008 and September 2010 and with a minimum six-month follow-up were identified. A total of 1618 records were excluded as they related to procedures other than primary TKR and a further 595 were excluded as the TKR was performed for an indication other than OA. Therefore 22 798 TKRs were available for analysis. This included 12 785 women (56%) and 10 011 men (44%) with a mean age of 69.7 (SD 8.9; 23 to 100). The mean pre-operative OKS and EQ-5D index were 19.0 (SD 7.7; 0 to 47) and 0.407 (SD 0.310; -0.594 to 1.000), respectively. Post-operative PROMs were collected at a median of 199 days following surgery (interquartile range (IQR) 193 to 213; range 180 to 365). The primary outcome was the patients’ rating of their surgery (satisfaction). This outcome was collected postoperatively using five-point ordinal scales19 by asking patients “How would you describe the results of your operation?”, with the following possible responses: ‘Excellent’, ‘Very Good’, ‘Good’, ‘Fair’ and ‘Poor’. Within this scale the first category was the best response and the fifth category the worst response, which was inversely related to the OKS and EQ-5D scores. A satisfaction rating was available for 22 373 patients (98%) analysed. Explanatory/predictor variables. A structured literature review was undertaken to determine which factors had previously been shown to influence satisfaction following TKR. PubMed, Medline and Embase were searched using the terms “satisfaction” in combination with either “knee arthroplasty” or “knee replacement”. From this search all English language publications between 1990 and 2012 were screened to identify the associations with satisfaction relevant to this analysis and the methods of statistical analysis employed. All of the identified reports focused solely on the direct relationships between a variety of explanatory factors and satisfaction without consideration of any interactions between these factors. Reported associations with
satisfaction included both demographic characteristics or other pre-operative factors including the patient’s age,9,12 gender,9,20 underlying diagnosis,9,21 a lower level of education,11 increasing body mass index (BMI),11 mental health status/depression,22-24 and general health status.12,22 Factors measured post-operatively included any need for revision surgery,10,11,25 post-operative PROMS/symptom improvement9,11,25 and fulfilment of patient expectations.11 Descriptions of the factors available for this analysis are shown in Table I. Statistical analysis. Initial tabular and graphical summaries of the relationship between satisfaction and each of the explanatory variables were supplemented by ordinal logistic regression and structural equation modelling analyses. Ordinal logistic regression with proportional odds was used to analyse the effects of the explanatory variables individually and in combination as satisfaction was measured using a five-point ordinal scale. Structural equation modelling was then used to investigate the direct and indirect effects of the different factors upon patient satisfaction. Ordinal regression models were constructed to assess the relationship between variables available both pre-operatively (age, gender, depression, general health status, preoperative OKS/EQ-5D) and post-operatively (post-operative OKS/EQ-5D, ‘operative success’) with satisfaction. The ordinal package in the R statistical package26 was used to undertake preliminary model development and assessment of the impacts of patient covariates on satisfaction. Model identifiability was assessed on the basis of the condition number of Hessian of < 104 and when the condition number of the Hessian was in excess of this value the models were rejected on the basis of being ill-conditioned.26 Initial models including both the OKS and EQ-5D in addition to the other patient characteristics were rejected as they did not converge. Models with the OKS alone were also inadequate for the same reason. A resampling exercise using 200 random subsets of 94% of the data showed that models with OKS were not identifiable as the condition number for all replicates was in excess of 105 (mean 6.5 × 105; SD 4332). However as the OKS and EQ-5D were correlated both pre(r = 0.72, Spearman’s rank) and post-operatively (r = 0.80, Spearman’s rank), and prior analysis had demonstrated the pre-operative OKS and satisfaction were poorly correlated (Spearman’s rank correlation 0.04 (95% confidence intervals (CI) -0.01 to 0.08)),14 we proceeded using only the EQ5D in addition to the other patient characteristics. Variables were removed using a two-stage stepwise reduction until only significant variables were left in the model. The assumption of proportionality required in ordinal regression was assessed using the graphical method following Bender and Grouwen28 and Gould.29 This was undertaken by comparing the pairwise separation of thresholds of each ordinal scale for each covariate on the logit scale. The best fit models were then used to predict the likely extent of satisfaction in two distinct situations. THE BONE & JOINT JOURNAL
PATIENT SATISFACTION WITH TOTAL KNEE REPLACEMENT CANNOT BE PREDICTED FROM PRE-OPERATIVE VARIABLES ALONE
1361
Table I. Explanatory variables identified from the structured literature review and their availability within the National Joint Registry-Patient Reported Outcome Measures (NJR-PROMs) dataset Predictors
Details
Age (yrs) Gender Diagnosis
1) Female; 2) Male Analysis restricted to only those procedures performed for osteoarthritis as this group comprises > 98% of all knee replacements performed. Other indications excluded PROMs recorded at six months post-operatively. The number of revisions at this time-point is small limiting the ability to analyse this variable Not recorded in the NJR-PROMs dataset Data only available for 9874 patients (42%). Patients with missing BMI data differed from those with recorded BMI data in relation to reported levels of satisfaction and success. BMI data were therefore excluded from the analysis Patient reported. Indicates whether the patient has previously been given a diagnosis of depression: 1) No; 2) Yes Patient reported. Indicates the patients perception of their own general health with 5 ordered options: 1) Excellent; 2) Very good; 3) Good; 4) Fair; 5) Poor
Need for revision surgery Level of education Body mass index (BMI) (kg/m2)
Mental health status/depression Pre-operative general health Patient reported outcome scores (preand post-operative and change in score) EuroQol index (EQ-5D)
Oxford knee score (OKS)
Symptom improvement (patient perception of ‘operative success’)
Patient expectation
A generic measure of health used for clinical and economic appraisal. It provides a simple descriptive profile of 5 health domains (Mobility, Self-Care, Ability to perform usual activities, Pain/ Discomfort, Anxiety/Depression), which are combined using population weightings to produce a single index value for health status A knee specific measure consisting of 12 questions assessing the level of pain and functional capacity of the affected knee. Scores for each question are rated 0 to 4 and combined to give an overall score out of 48 (0 worst to 48 best) Collected post-operatively alongside satisfaction using 5 point ordinal scales. Assessed by asking patients “Overall, how are the problems now in the knee on which you had surgery, compared to before your operation?” with possible responses ‘Much better’, ‘A little better’, ‘About the same’, ‘A little worse’, ‘Much worse’. Data available for 22 429 (98%) of 22 798 patients Not recorded in the NJR-PROMs dataset
Structural equation modelling (SEM). As the explanatory variables have potential to influence each other we used SEM to investigate their direct and indirect effects on the satisfaction outcome. SEM is a method for investigating the impacts of variables in systems where there are multiple pathways to the final effect. The approach is used to challenge a hypothetical representation of the system pathways using data. In the present context we hypothesised that patient satisfaction is driven by patient characteristics, preand post-operative function and the perception of improvement of symptoms (‘operative success’). For example, this model is able to recognise that post-operative function is dependent upon pre-operative function which in turn is dependent on patient age and gender. Thus there are a set of driving variables that have both direct and indirect effects on satisfaction that are in themselves related. Male and female patients were modelled separately in the analysis. SEM tests whether the variables in the path are interrelated by analysing their variances and co-variances. Goodness-of-fit criteria for each model were then used to identify the simplest model and best explanation for the available data. The data contained both non-normal and categorical variables. Models were therefore fitted using distribution free estimation by weighted least squares which makes no distributional assumptions about the data. Empirical studies have shown this robust in assessing models with more than 1000 to 2000 samples.29 VOL. 95-B, No. 10, OCTOBER 2013
Model fit was assessed using Bollen–Stine bootstrapping of the chi-squared statistic30 and the Root Mean Square Error of Association. Models were fitted in the Lavaan package in R. All modelling was undertaken using libraries in the R statistical package.31 A p-value of p < 0.01 was considered to be statistically significant. As no patient sensitive data were used as part of this analysis it was performed as a service evaluation without need for formal ethical approval.
Results The relationship between satisfaction and ‘operative success’. We found that 15 882 patients (71%) perceived their
knee symptoms to be ‘much better’ following surgery. However despite symptomatic improvement only 4959 (22%) rated their outcome as ‘excellent’ and the majority of patients rated their surgery as either ‘very good’ or ‘good’ (Table II). In total there were 1592 patients (7%) who rated their surgery as ‘fair’ or ‘poor’ despite reporting symptomatic improvements. This group were generally a year younger (mean age 68.8 years (40 to 92) vs 69.7 years (23 to 100)), had higher levels of depression (176 of 1592 (11%) vs 1642 of 22 798 (7%)) and ‘Fair/Poor’ general health (687 of 1592 (43%) vs 5502 of 22 798 (24%)), and had lower pre-operative knee (mean OKS 16.9 (1 to 41) vs 19.0 (0 to 47)) and general health (mean EQ-5D 0.329 (-0.594 to 1.000) vs 0.407 (-0.594 to 1.000) scores than the
1362
P. N. BAKER, S. RUSHTON, S. S. JAMESON, M. REED, P. GREGG, D. J. DEEHAN
Table II. Distribution of satisfaction and success following total knee replacement (TKR). Data on both satisfaction and success were available for 22 278 (98%) of 22 798 patients Success (n) Satisfaction
Much better
A little better
Much the same
A little worse
Much worse
Total (n, %)
Excellent Very good Good Fair Poor
4886 7344 3374 272 6
59 474 2002 1251 63
3 36 313 646 111
7 18 119 476 199
4 11 23 122 459
4959 (22) 7883 (35) 5831 (26) 2767 (12) 838 (4)
Total (n, %)
15 882 (71)
3849 (17)
1109 (5)
819 (4)
619 (3)
22 278 (100)
Table III. Results of the ordinal regression relating patient satisfaction (5-point scale) to patient demographic data and knee function preoperatively and post-operatively. Regression estimates, standard errors, p-values for the Wald test and odds ratios (ORs) with associated 95% confidence intervals (CI) given for each covariate. An OR > 1 indicates an increasing risk of dissatisfaction relative to the reference group Covariate Pre-operative variables Gender (Reference: female) Male Pre-operative EuroQol 5D (EQ-5D) Depression (Reference: Depression) No diagnosis of depression General health (Reference: Excellent) Very good Good Fair Poor Post-operative variables Success (Reference: Much better) A little better About the same A little worse Much worse Post-operative EQ-5D
Estimate (SE)
Z-statistic
OR (95% CI)
p-value
-0.15 (0.028) 0.42 (0.049)
-5.41 8.46
0.86 (0.81 to 0.91) 1.52 (1.38 to 1.68)
< 0.001 < 0.001
-0.19 (0.056)
3.37
0.83 (0.74 to 0.92)
< 0.001
0.67 (0.088) 1.07 (0.078) 1.28 (0.082) 1.2 (0.116)
8.46 13.7 15.5 10.37
1.95 (1.64 to 2.32) 2.92 (2.50 to 3.40) 3.60 (3.06 to 4.22) 3.32 (2.64 to 4.17)
< 0.001 < 0.001 < 0.001 < 0.001
2.79 (0.048) 4.1 (0.078) 4.98 (0.096)
58.6 52.27 51.7
< 0.001 < 0.001 < 0.001
6.74 (0.133)
50.66
-2.39 (0.075)
-31.85
16.28 (14.82 to 17.89) 60.34 (51.79 to 70.31) 145.47 (120.52 to 175.59) 845.56 (651.53 to 1097.38) 0.09 (0.08 to 0.11)
total population undergoing TKR. There was also a smaller group of 40 patients (0.2%) who reported high levels of satisfaction despite a worsening of their symptoms. Ordinal regression. Ordinal regression with proportional odds assumes that the relationship between each pair of outcome groups is proportional. In effect this means that the relationship between the levels of the satisfaction response and the individual patient covariates are assumed to be the same for each of the levels in the response, but that these have different intercepts (thresholds). Interpretation of the regression coefficients as odds ratios (OR) is complex because the OR for the effect of a covariate are common across all classes of the ordinal outcome, so we are effectively comparing the effects of the covariate across the range of satisfaction scores. Patient satisfaction was significantly and positively related to ‘Operative success’. As success declined, the OR of a poor satisfaction score increased. A poor operative
< 0.001 < 0.001
outcome had an OR of 845 (95% CI 652 to 1097) of a low satisfaction score when compared with a case with excellent operative success. Similarly, poor health had a higher OR of low satisfaction (3.3 (95% CI 2.6 to 4.2)) relative to a patient in good health. Satisfaction was also related to pre- and post-operative function as represented by the EQ5D scores (OR 1.52 (95% CI 1.38 to 1.68) and OR 0.09 (95% CI 0.08 to 0.10), respectively). Males had a lower risk of dissatisfaction relative to females (OR 0.86 (95% CI 0.81 to 0.91)), as did not having a previous diagnosis of depression (OR 0.83 (95% CI 0.74 to 0.92)) (Table III). Age was not a significant predictor of satisfaction within the ordinal regression model. The regression diagnostics indicated that post-operative variables had a larger influence upon satisfaction than pre-operative variables. For the regression models proportionality assumptions were broadly met except where the number of cases in a class was low, specifically the ‘much worse’ success group. THE BONE & JOINT JOURNAL
PATIENT SATISFACTION WITH TOTAL KNEE REPLACEMENT CANNOT BE PREDICTED FROM PRE-OPERATIVE VARIABLES ALONE
Age (1.0) 0.04
0.02
Pre-operative EQ-5D index (1.0)
0.03
0.33
0.09
Post-operative EQ-5D index (0.89)
-0.21
-0.55
0.57 Satisfaction (0.51)
Success (0.72)
Fig. 1 Final structural equation modelling (SEM) model for male total knee replacement (TKR) patients after removal of all non-significant variables and pathways. The numbers alongside arrows represent the proportional change in standard deviation in the response (end of arrow) relative to unit change in the standard deviation of the predictor. Values in parentheses represent the proportion of the observed variability which is explained by the suggested model.
Structural equation modelling (SEM). Full models including all hypothesised explanatory variables and pathways did not adequately describe the variation in satisfaction observed in the data. Z values for all variables included in the final SEM path model (Fig. 1) were significant (p < 0.001) with the exception of the pathway from age to satisfaction in females where significance was marginal (Z = 1.81, p = 0.07). The variable was kept in the final model however, as it was more significant for males (Z = 2.60, p = 0.009). Whilst these data show levels of significance for key covariates the coefficients are not standardised in relation to each other so are difficult to compare across pathways. Standardised coefficients for the best fit models for males are shown in Figure 1. The coefficients on the path lines represent the proportional change in standard deviation of the response in relation to a unit standard deviation change in the hypothesised driving variable. The final models and model path coefficients for females were similar to those for the males and are therefore not shown. The Root Means Square Error of Association (RMSEA) for the TKR model was 0.064; the Comparative Fit Index (CFI) was 0.97 and the Bollen–Stine bootstrap probability (PBS) for this model was 0.0, indicating that the model was a very good fit for the data. Patient satisfaction was strongly dependent on ‘operative success’. Higher levels of satisfaction and ‘operative success’ were related to higher levels of post-operative function as represented by the EQ-5D score. Pre-operative function had a positive impact on post-operative function and both pre-operative function and age had slight but significant impacts on satisfaction, with older patients being more satisfied. In the context of these relationships the level of VOL. 95-B, No. 10, OCTOBER 2013
1363
pre-operative general health and a diagnosis of depression were not found to be important. The direct and indirect contributions of pre-operative and post-operative covariates on patient satisfaction were estimated by the product coefficient method (Table IV). The indirect effect of the post-operative EQ-5D on satisfaction was the sum of the products of coefficients for its relationship with ‘success’ and that for ‘success’ with satisfaction (i.e. 0.55 × 0.57) (Fig. 1). Using these methods pre-operative variables (age and pre-operative EQ-5D) contributed little direct or indirect effect up on patient satisfaction. In contrast, the post-operatively measured ‘success’ and EQ-5D score contributed much more to the perceived satisfaction, with a unit change in ‘success’ resulting in a > 0.55 unit change in patient satisfaction for both males and females. These results demonstrate the greater influence of post-operative variables upon satisfaction when compared with pre-operative variables and highlight the significant role indirect effects have upon this outcome.
Discussion While pre-operative variables such as gender, a diagnosis of depression, pre-operative general health status and preoperative EQ-5D scores were related to patient satisfaction their predictive capacity was low. In contrast postoperative variables such as the post-operative EQ-5D score and the perception of operative success were more important when both their direct and indirect effects upon satisfaction were modelled. Limitations of this study include the possibility that we did not consider all variables that could potentially influence satisfaction identified through our structured literature review. However, our model fit was good despite inevitable omissions indicating that our simplified models adequately explained a significant proportion of the variability for satisfaction as an outcome. It may be that the EQ5D is not as sensitive as other health utility measures in this population and the use of other health metrics may have yielded different results. Patients were assessed at six months following surgery, a time point that has been identified as a valid time at which to assess functional improvements and patient satisfaction after knee replacement.32 The ordinal scales used to assess satisfaction and success have not been validated but mirror similar scales used for assessing patient reported satisfaction in national cohorts.10,12,23 While the content of these scales differs, their construct is similar with patients asked to rate their experience ranging from bad to good with responses presented as an ordered Likert scale. The benefit of these scales is that they give a simple representation of the patient’s perceptions of the results of surgery. In this respect they may be more important than a number of frequently used validated outcome scores that fail to appreciate an individual patient’s experiences and instead focus on hard symptomatic endpoints. By using national PROMS data with a cohort of patients in excess of 22 000 we have moved
1364
P. N. BAKER, S. RUSHTON, S. S. JAMESON, M. REED, P. GREGG, D. J. DEEHAN
Table IV. Direct and indirect pathways to effect from the Structural Equation Modelling analysis. The values represent the impact of a units standard deviation change in the driving variables on the patient’s satisfaction with total knee replacement Male Variables Pre-operative Pre-operative EuroQol 5D Age Post-operative Post-operative EuroQol 5D Success
Female
Direct
Indirect
Direct
Indirect
0.029 0.018
-0.115 -0.024
0.033 0.014
-0.126 -0.061
-0.214 0.573
-0.315
-0.259 0.552
-0.313
interpretation away from sampling and towards population reporting; a magnitude of sampling where statistical tests of goodness of fit are not relevant.33 Nonetheless a cohort of this magnitude has allowed us to use analytical approaches that are not suitable for the analysis of smaller datasets and as a result we have been able to quantify the relationships between satisfaction, operative success, functional capacity and patient characteristics. This study found that only 22% of patients gave the top ‘excellent’ rating when asked to describe the results of their knee replacement. This occurred despite 71% of patients perceiving their knee to be ‘much better’ following surgery. This demonstrates the inherent differences between satisfaction and operative success. Success measures the patient’s perception of whether they have symptomatically improved following surgery whereas satisfaction measures the extent to which they are happy with this improvement. Patient expectation is a key determinant of satisfaction.11,21 There were a significant number of patients (1592) who were dissatisfied despite reporting symptomatic improvements, possibly, we might speculate, because their expectations had not been fulfilled. This group were generally younger, had poorer mental and physical health and had poorer pre-operative function that the total population undergoing TKR, which is consistent with previous studies.9,11,12,22-25 It is likely that within this analysis ‘operative success’ was a surrogate marker for the fulfilment of expectation. This may explain why this variable was such a strong predictor of satisfaction and confirms the important role that patient expectation has upon outcome. The modelling techniques we employed allowed us to examine the combined direct and indirect effects of factors upon satisfaction. This confirmed that the primary determinant of a patient’s level of satisfaction was their post-operative perception of whether their operation was a success. While a number of other factors were also significant in the final models their effects upon satisfaction occurred through an indirect influence upon the ‘operative success’ variable. The findings of the ordinal regression and SEM differed in respect to the presence of age, depression and general health, which were significant in one model but not the other. This probably reflects the interdependence between variables. While gender was significant in the
ordinal regression, with females having lower levels of satisfaction than males in keeping with previous analyses,9,20 when the SEM analysis was performed separately for males and females it demonstrated that the pathways and variables influencing satisfaction are similar irrespective of gender. The pre-operative OKS have previously been shown not to predict post-operative satisfaction.14 Similarly our analyses based on pre-operative variables alone did not adequately explain the variability in the satisfaction outcome and as such were not suited to making predictions. It was only when both pre- and post-operative variables were considered together that a robust prediction of satisfaction could be made. There is clearly merit in attempting to predict which patients are likely to exhibit high levels of satisfaction following surgery in order to support patient selection and clinical decision making. However, the findings of this study suggest that this process is complex and any predictions of satisfaction based on pre-operative variables alone are likely to be heavily modified by the outcome of surgery. This must be appreciated before these predictive models are used in clinical practice otherwise certain patient groups risk being unfairly excluded from the benefits of surgery. These data demonstrate that even in the ‘best’ patient with the most favourable patient characteristics, experiencing a large functional improvement from a highly successful operation does not guarantee the highest levels of satisfaction. However, in contrast in the ‘worst’ patient with the most unfavourable patient characteristics and poor functional improvement, a failure to produce a successful operation will almost certainly result in a dissatisfied patient. Finally we have demonstrated that there is a spectrum of satisfaction after knee replacement that varies significantly dependent upon both pre- and post-operative factors. Patient satisfaction is increasingly being used as a comparative measure when assessing the quality of care following joint replacement in European healthcare systems.34,35 The most important determinants of satisfaction are the patient’s perception of the success of their operation and post-operative function. Pre-operative variables have a minimal influence upon post-operative satisfaction. This brings into question the appropriateness of restricting access to care based on arbitrary pre-operative thresholds. THE BONE & JOINT JOURNAL
PATIENT SATISFACTION WITH TOTAL KNEE REPLACEMENT CANNOT BE PREDICTED FROM PRE-OPERATIVE VARIABLES ALONE The authors would like to thank the patients and staff of all the hospitals in England and Wales who have contributed data to the National Joint Registry. We are grateful to the Healthcare Quality Improvement Partnership (HQIP), the NJR steering committee and the staff at the NJR centre for facilitating this work. This work was funded by a fellowship from the National Joint Registry. The authors have conformed to the NJR’s standard protocol for data access and publication. The views expressed represent those of the authors and do not necessarily reflect those of the National Joint Registry Steering committee or the Health Quality Improvement Partnership (HQIP) who do not vouch for how the information is presented. No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
1365
15. No authors listed. NHS North Lincolnshire. Prior approval scheme 2009/10. http:// www.northlincolnshire.nhs.uk/documents/download964.aspx (date last accessed 6 June 2013). 16. Baker P, Petheram T, Jameson S, et al. The association between body mass index and the outcomes of total knee arthroplasty. J Bone Joint Surg [Am] 2012;94-A:1501– 1508. 17. Dawson J, Fitzpatrick R, Murray D, Carr A. Questionnaire on the perceptions of patients about total knee replacement. J Bone Joint Surg [Br] 1998;80-B:63–69. 18. No authors listed. Euroqol (EQ-5D) score. http://www.euroqol.org/home.html (date last accessed 6 June 2013).
This article was primary edited by D. Rowley and first-proof edited by G. Scott.
19. Streiner DL, Norman GR. Health Measurement Scales: a practical guide to their development and use. New York: Oxford University Press, 2008.
References
20. Ethgen O, Bruyère O, Richy F, Dardennes C, Reginster JY. Health-related quality of life in total hip and total knee arthroplasty: a qualitative and systematic review of the literature. J Bone Joint Surg [Am] 2004;86-A:963–974.
1. Kurtz S, Ong K, Lau E, Mowat F, Halpern M. Projections of primary and revision hip and knee arthroplasty in the United States from 2005 to 2030. J Bone Joint Surg [Am] 2007;89-A:780–785.
21. Bullens PH, van Loon CJ, de Waal Malefijt MC, Laan RF, Veth RP. Patient satisfaction after total knee arthroplasty: a comparison between subjective and objective outcome assessments. J Arthroplasty 2001;16:740–747.
2. No authors listed. National Joint Registry for England and Wales: 8th Annual report, 2011. http://www.njrcentre.org.uk (date last accessed 6 June 2013).
22. Anderson JG, Wilson RL, Tsai D, Stulberg SD, Chang RW. Functional outcome and patient satisfaction in total knee patients over the age of 75. J Arthroplasty 1996;11:831–840.
3. No authors listed. Inside government. 2011-12 tariff information spreadsheet. http:/ /www.dh.gov.uk/en/Publicationsandstatistics/Publications/PublicationsPolicyAndGuidance/DH_124356 (date last accessed 6 June 2013).
23. Scott CE, Howie CR, MacDonald D, Biant LC. Predicting dissatisfaction following total knee replacement: a prospective study of 1217 patients. J Bone Joint Surg [Br] 2010;92-B:1253–1258.
4. Jones CA, Voaklander DC, Johnston DW, Suarez-Almazor ME. Health related quality of life outcomes after total hip and knee arthroplasties in a community based population. J Rheumatol 2000;27:1745–1752.
24. Blackburn J, Qureshi A, Amirfeyz R, Bannister G. Does preoperative anxiety and depression predict satisfaction after total knee replacement? Knee 2011:Epub.
5. Rissanen P, Aro S, Slätis P, Sintonen H, Paavolainen P. Health and quality of life before and after hip or knee arthroplasty. J Arthroplasty 1995;10:169–175.
25. Hawker G, Wright J, Coyte P, et al. Health-related quality of life after knee replacement. J Bone Joint Surg [Am] 1998;80-A:163–173.
6. Drewett RF, Minns RJ, Sibly TF. Measuring outcome of total knee replacement using quality of life indices. Ann R Coll Surg Engl 1992;74:286–290.
26. Christensen RHB. Ordinal: Regression Models for Ordinal Data: R package version, 2011. http://www.cran.r-project.org/package=ordinal/ (date last accessed 6 June 2013).
7. Fitzgerald JD, Orav EJ, Lee TH, et al. Patient quality of life during the 12 months following joint replacement surgery. Arthritis Rheum 2004;51:100–109. 8. Bachmeier CJ, March LM, Cross MJ, et al. A comparison of outcomes in osteoarthritis patients undergoing total hip and knee replacement surgery. Osteoarthritis Cartilage 2001;9:137–146. 9. Baker PN, van der Meulen JH, Lewsey J, Gregg PJ. The role of pain and function in determining patient satisfaction after total knee replacement: data from the National Joint Registry for England and Wales. J Bone Joint Surg [Br] 2007;89-B:893–900. 10. Robertsson O, Dunbar M, Pehrsson T, Knutson K, Lidgren L. Patients satisfaction after knee arthroplasty: a report on 27,372 knees operated on between 1981 and 1995 in Sweden. Acta Orthop Scand 2000;71:262–267. 11. Bourne RB, Chesworth BM, Davis AM, Mahomed NN, Charron KD. Patient satisfaction after total knee arthroplasty: who is satisfied and who is not? Clin Orthop Relat Res 2010;468:57–63. 12. Noble PC, Conditt MA, Cook KF, Mathis KB. The John Insall Award: Patient expectations affect satisfaction with total knee arthroplasty. Clin Orthop Relat Res 2006;453:35–43. 13. No authors listed. Health & Social Care Information Centre. HES: Patient Reported Outcome Measures: Report 2009-2010. http://www.hscic.gov.uk/hes (date last accessed 6 June 2013). 14. Judge A, Arden NK, Price A, et al. Assessing patients for joint replacement: can pre-operative Oxford hip and knee scores be used to predict patient satisfaction following joint replacement surgery and to guide patient selection? J Bone Joint Surg [Br] 2011;93-B:1660–1664.
VOL. 95-B, No. 10, OCTOBER 2013
27. Bender R, Grouven U. Ordinal logistic regression in medical research. J R Coll Physicians Lond 1997;31:546–548. 28. Gould W. Interpreting logistic regression in all its forms. Stata Technical Bulletin 2000;124:19–29. 29. Olsson U, Foss T, Troye SV, Howell RD. The performance of ML, GLS and WLS estimation in structural equation modelling under conditions of misspecification and non-normality. Structural Equation Modelling 2000;7:557–595. 30. Bollen KA, Stine RA. Bootstrapping goodness of fit measures in structural equation models. Sociological Methods and Research 1992;21:205–229. 31. No authors listed. R Development Core Team. A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, 2010 http:// www.R-project.org (date last accessed 6 June 2013). 32. Judge A, Arden NK, Kiran A, et al. Interpretation of patient-reported outcomes for hip and knee replacement surgery: identification of thresholds associated with satisfaction with surgery. J Bone Joint Surg [Br] 2012;94-B:412–418. 33. Burnham KP, Anderson DR. Model selection and multimodel inference: a practical information-theoretic approach. Second ed. Springer, 2002:488. 34. No authors listed. Danish Knee Arthroplasty Register: 2010 Annual Report, 2010. http://www.knee.dk/groups/dkr/pdf/DKRreportEnglish2010.pdf (date last accessed 7 June 2013). 35. Rolfson O, Karrholm J, Dahlberg LE, Garellick G. Patient-reported outcomes in the Swedish Hip Arthroplasty Register: Results of a nationwide prospective observational study. J Bone Joint Surg [Br] 2011;93-B:867–875.