Article Text

Download PDFPDF

Original research
Patient appropriateness for total knee arthroplasty and predicted probability of a good outcome
  1. Gillian A Hawker1,2,
  2. Eric Bohm3,
  3. Michael J Dunbar4,
  4. Peter Faris5,
  5. C Allyson Jones6,
  6. Tom Noseworthy5,
  7. Bheeshma Ravi7,
  8. Linda J Woodhouse6,8 and
  9. Deborah A Marshall5
  10. BEST-Knee Study Team
    1. 1Department of Medicine, University of Toronto, Toronto, Ontario, Canada
    2. 2Women's College Research Institute, Women's College Hospital, Toronto, Ontario, Canada
    3. 3Department of Surgery, University of Manitoba, Winnipeg, Manitoba, Canada
    4. 4Department of Surgery, Dalhousie University, Halifax, Nova Scotia, Canada
    5. 5Department of Community Health Sciences, University of Calgary, Calgary, Alberta, Canada
    6. 6Department of Physical Therapy, University of Alberta, Edmonton, Alberta, Canada
    7. 7Department of Surgery, University of Toronto, Toronto, Ontario, Canada
    8. 8Department of Public Health and Community Medicine, Tufts University, Phoenix, Arizona, USA
    1. Correspondence to Dr Gillian A Hawker; g.hawker{at}


    Objectives One-fifth of total knee arthroplasty (TKA) recipients experience a suboptimal outcome. Incorporation of patients’ preferences in TKA assessment may improve outcomes. We determined the discriminant ability of preoperative measures of TKA need, readiness/willingness and expectations for a good TKA outcome.

    Methods In patients with knee osteoarthritis (OA) undergoing primary TKA, we preoperatively assessed TKA need (Western Ontario-McMaster Universities OA Index (WOMAC) Pain Score and Knee injury and Osteoarthritis Outcome Score (KOOS) function, arthritis coping), health status, readiness (Patient Acceptable Symptom State, depressive symptoms), willingness (definitely yes—yes/no) and expectations (outcomes deemed ‘very important’). A good outcome was defined as symptom improvement (met Outcome Measures in Rheumatology and Osteoarthritis Research Society International (OMERACT–OARSI) responder criteria) and satisfaction with results 1 year post TKA. Using logistic regression, we assessed independent outcome predictors, model discrimination (area under the receiver operating characteristic curve, AUC) and the predicted probability of a good outcome for different need, readiness/willingness and expectations scenarios.

    Results Of 1,053 TKA recipients (mean age 66.9 years (SD 8.8); 58.6% women), 78.1% achieved a good outcome. With TKA need alone (WOMAC pain subscale, KOOS physical function short-form), model discrimination was good (AUC 0.67, 95% CI 0.63 to 0.71). Inclusion of readiness/willingness, depressive symptoms and expectations regarding kneeling, stair climbing, well-being and performing recreational activities improved discrimination (p=0.01; optimism corrected AUC 0.70, 0.66–0.74). The predicted probability of a good outcome ranged from 44.4% (33.9–55.5) to 92.4% (88.4–95.1) depending on level of TKA need, readiness/willingness, depressive symptoms and surgical expectations.

    Conclusions Although external validation is required, our findings suggest that incorporation of patients’ TKA readiness, willingness and expectations in TKA decision-making may improve the proportion of recipients that experience a good outcome.

    • Knee Osteoarthritis
    • Outcomes research
    • Osteoarthritis, Knee
    • Orthopedic Procedures

    Data availability statement

    Anonymised data are available upon reasonable request.

    This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

    Statistics from

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


    • Total knee arthroplasty (TKA) is an effective treatment for knee osteoarthritis (OA), but a substantial proportion of recipients report little symptom improvement or dissatisfaction with results, questioning their surgical appropriateness.

    • TKA is a preference-sensitive procedure, performed to improve patients’ quality of life. Yet, to date, there has been little consideration of preferences or values of patients with knee OA in determining patient appropriateness for TKA. Incorporation of patients’ perspectives in TKA decision-making may improve outcomes.


    • In a large prospective cohort study, we found that preoperatively patient-reported measures of TKA readiness, willingness and expectations significantly enhanced ability to discriminate those who did versus did not go on to experience a good TKA outcome compared with measures of TKA need alone. Given documented wide variability among participants with respect to preoperative measures of TKA need, readiness/willingness and surgical expectations, these findings have the potential to improve shared patient–physician TKA decision-making.


    • More explicit incorporation of patients’ perspectives in assessment of TKA appropriateness has the potential to improve TKA outcomes and the use of valuable healthcare resources.


    Total knee arthroplasty (TKA) is used widely to treat advanced symptomatic knee osteoarthritis (OA).1 2 Over 75,000 TKAs are performed annually in Canada for knee OA, with inpatient costs of approximately $1 billion.3 TKA rates are rising due to obesity and ageing, both risk factors for knee OA.1 4 5 While, on average, TKA is highly effective, 15%–30% of recipients report little or no symptom improvement and/or dissatisfaction with results.6 7 Among primary TKA recipients in an Ontario OA cohort, we found that changes in pain, function and health resource use after TKA varied substantially; while mean changes were positive, 23.6% experienced no meaningful change in symptoms and 17.1% were worse after surgery.8 More appropriate selection of TKA candidates offers the opportunity to improve both patient outcomes and the use of valuable healthcare resources.9

    TKA is a preference-sensitive procedure, performed to improve patients’ quality of life. Patients must weigh the benefits and risks of surgery in the context of their personal preferences and values.10 11 Thus, TKA appropriateness criteria should be used to determine if it is reasonable to consider operating given a patient’s particular circumstances.12 13 Yet, to date, there has been little or no involvement of patients with knee OA in the development or validation of TKA appropriateness criteria or decision-support tools.14–18 Most have been developed based on expert clinician consensus informed by the best available scientific evidence.9 15 16 18–20 Existing tools generally consider patients’ symptoms, comorbidity, radiographic severity and clinical examination (knee mobility and stability).17 21 While shared patient–physician decision-making is encouraged,16 22–24 no criteria explicitly incorporate patients’ expectations or preferences for care. While studies have shown that patients who meet existing criteria for appropriateness experience greater improvements in pain and function than those that do not, no significant differences have been found with respect to patient satisfaction with surgical results.15 25–28 Incorporation of patients’ perspectives in TKA appropriateness assessment may improve ability to identify at surgeon consultation those patients most likely to benefit from surgery.

    In a prior population-based cohort study, we showed that patient willingness to undergo TKA was the strongest predictor of subsequent receipt of surgery and that TKA willingness reflected patients’ perceptions of their candidacy for surgery, including perceived OA severity, coping efficacy and risks and benefits of surgery.29–31 In subsequent qualitative research to elucidate key stakeholders’ opinions about patient appropriateness for TKA, people with OA and orthopaedic surgeons agreed that it was important to consider four appropriateness domains: patient’s need for TKA (symptom severity, prior therapies tried and evidence of knee OA on clinical and radiographic examination), health status (surgical risk), psychological readiness and willingness for TKA and expectations of TKA.32–34 Participants with a prior knee replacement recalled the challenges of getting through the early postoperative period. They emphasised the importance of having a positive attitude and feeling mentally prepared to undergo surgery to experiencing a good surgical outcome.

    Informed by this work, the current study sought to establish the predictive validity, and relative weights, of valid and reliable measures of the four appropriateness domains (TKA need, health status, readiness and willingness and surgical expectations) for TKA ‘benefit’, defined as a composite dichotomous outcome incorporating symptom improvement and satisfaction with surgical results. In particular, we were interested in the contribution, if any, made by measures of readiness, willingness and expectations to our ability to discriminate, preoperatively, those who would versus would not experience a good TKA outcome.


    Setting and design

    This prospective cohort study recruited people with knee OA referred for elective TKA to two orthopaedic clinics responsible for approximately 60% of joint replacements in Alberta, Canada, between 2014 and 2016. Those aged 30 years or older, able to read and comprehend English, and confirmed to have knee OA on physical examination and imaging were eligible. Individuals with inflammatory arthritis were excluded. All 45 clinic surgeons agreed to participate.


    Pre consultation, consenting patients completed a standardised questionnaire to assess appropriateness domains and contextual factors. In the absence of consensus on how best to measure the domains, we conducted a comprehensive search for reliable and valid self-complete questionnaires and selected for use those that were brief and validated for use in knee OA, ideally in the TKA setting (Table 1). To assess TKA need, participants reported their knee pain and disability (Western Ontario-McMaster Universities OA Index (WOMAC) pain subscale35 and Knee injury and Osteoarthritis Outcome Score physical function short-form, KOOS-PS36), perceived ability to cope with knee symptoms (four-item Arthritis Coping Efficacy Scale)37 and prior OA therapies (formal or informal exercise, physiotherapy, weight loss, acetaminophen, anti-inflammatories, codeine, joint injection and walking aids).38 To assess psychological readiness for TKA, participants completed the Patient Health Questionnaire (PHQ-8) to assess for symptoms of depression39 and the Patient Acceptable Symptom State (PASS) to assess the acceptability of their knee symptoms (acceptable or unacceptable)40 and reported their willingness to consider TKA if recommended (5-point Likert scale from ‘definitely willing’ to ‘definitely not willing’). To assess health status, we ascertained height/weight to calculate body mass index (BMI), physician-diagnosed comorbidities41 and pain or stiffness in the hips, contralateral knee or low back. To assess TKA expectations, participants completed the Hospital for Special Surgery Knee Replacement Expectations Questionnaire42; they rated the importance of 17 TKA outcomes as ‘very important’, ‘somewhat important’, ‘a little important’, ‘I do not expect this’ or ‘this does not apply to me’ and those deemed ‘very important’ were considered expectations. We also assessed level of social support (Lubben Social Support Scale)43 and prior joint replacement (hip or contralateral knee). Patient age and sex were obtained from clinic records. Evidence of knee OA on clinical and radiographic examination of the joint being considered for surgery were confirmed by the surgeon (yes/no) post consultation.

    Table 1

    Variables considered as measures of TKA appropriateness criteria

    One year post surgery, we reassessed knee symptoms in addition to patient global assessment (PGA) of change in knee pain and function (5-point Likert scale: ‘much better’ to ‘much worse’),44 overall satisfaction (Patient Satisfaction Scale for Primary Hip and Knee Arthroplasty, 4-point Likert scale: ‘very dissatisfied’ to ‘very satisfied’)45 and occurrence of TKA complications (verified in patients’ medical record).

    Primary outcome

    In our prior qualitative research,32–34 patients with OA and surgeons told us that both meaningful symptom improvement and satisfaction with surgical results should be considered if assessing post hoc whether a TKA had been appropriate. Thus, we defined a good TKA outcome (yes/no) as both meaningful improvement in OA symptoms and patient-reported satisfaction (very or somewhat) with TKA results overall. Symptom improvement was defined as fulfilment of the Outcome Measures in Rheumatology and Osteoarthritis Research Society International (OMERACT–OARSI) OA responder criteria46: PGA pain and function ‘much’ or ‘somewhat improved’ and a relative change in both WOMAC Pain Score and KOOS-PS function of ≥20% and an absolute change in these scores of ≥10 points/100.

    Exposures were our measures of the appropriateness domains. Individuals were considered to have ‘received recommended OA therapies’ (yes/no) if they had ‘ever tried’ formal exercise or physiotherapy and weight loss if overweight/obese and any analgesic. TKA willingness was categorised as definitely willing (yes/no). Covariates were age, gender, social support and prior joint replacement.

    Statistical analyses

    Variables were summarised as proportions, means and medians, as appropriate. KOOS-PS and WOMAC Pain Score were transformed to a 0–100 scale with higher scores indicating greater severity. Collinearity was assessed using a variance inflation factor of >447; as perceived importance of kneeling and going upstairs was collinear with perceived importance of squatting and going downstairs, respectively, the latter was excluded from analyses.

    Given the large number of potential predictors, we used logistic regression and best possible subsets variable selection to identify the fewest predictors with the best discriminative ability for our outcome, defined as the smallest Akaike’s information criterion.48 All-possible-subset regression then determined the best model of that variable number. All potential predictors, irrespective of appropriateness domain, were considered. Diagnostics, including a check of the residuals and tests for influential observations, were run.49 To control for potential clustering by surgeon, we used multivariable log Poisson regression50 to assess the unadjusted and adjusted risk ratios (RRs) for a good TKA outcome associated with the identified predictors. Receiver operating characteristic (ROC) curves were generated to examine and compare the discriminant ability of significant predictors of TKA outcome by appropriateness domain and overall based on the area under the ROC curve (AUC). Effron’s enhanced bootstrap was used to generate an optimism corrected AUC for the multivariable model.51 52

    To assess the potential usefulness of the findings in appropriateness assessment, we categorised each participant’s level of appropriateness (low to high) for each of the appropriateness domains (TKA need, health status, readiness/willingness, expectations) based on the variables in the final model. Multivariable logistic regression was then used to calculate the predicted probability of a good TKA outcome for various TKA appropriateness scenarios. Statistical significance was based on a two-sided p value<0.05.

    Patient and public involvement

    Public involvement was first initiated during the design stage of the study through interviews with patients and stakeholders. Patients and members of the public served as consultants for initial questionnaire design, methods of administration and time required for administration of the questionnaire. No patients were involved in setting the research question or the outcome measures, nor were they asked to advise on interpretation or writing up of results.


    Study sample

    Of 1374 consenting, eligible patients who completed preoperative assessments and underwent TKA for knee OA, 1275 patients completed the 1-year assessment. In total, 1053 (82.6%) patients with data to assess our TKA outcome were included (figure 1). Responses were missing for less than 10% of variables except for KOOS-PS, where scores could not be calculated for 12.6%.

    Figure 1

    Participant recruitment flow chart.

    Cohort characteristics pre TKA

    Participant characteristics pre TKA are shown in table 2. Mean age was 66.9 years (SD, 8.8) and 58.6% of participants were women. Mean WOMAC Pain Score was 57.2/100 (SD 17.3), mean KOOS-PS was 52.8/100 (SD 17.1) and 71.0% had received recommended OA treatments, as defined. The median PHQ-8 Score was 5/24 (IQR, 2–10); 25.8% had scores≥10 indicating moderate-to-severe depressive symptoms. Overall, 79.7% indicated their knee symptoms were unacceptable and 87.8% indicated definite willingness to undergo TKA at consultation. Mean BMI was 32.5 kg/m2 (SD 6.3), 38.6% had two or more non-musculoskeletal (MSK) comorbid conditions and 70.3% had MSK comorbidities. Most participants indicated that improved ability to walk and reduced knee pain were very important TKA outcomes (96.7% and 93.5%, respectively), while 75%–85% considered improved ability to perform daily activities, climb stairs, participate in recreational activities and change position very important. More variability was observed with respect to the importance of other outcomes.

    Table 2

    Preoperative characteristics of knee OA participants

    TKA outcome at 1 year

    In 23 participants (2.2%) experienced a TKA complication: 18 revisions (13 unspecified, 2 for infection and 3 for patellar resurfacing) and 10 manipulations under anaesthesia. Including these individuals, 79.5% (76.9–81.9) met OMERACT–OARSI criteria for response and 74.7% (72.1–77.4) and 16.7% (14.5–19.0) were very or somewhat satisfied with their surgical results, respectively. Overall, 78.1% (95% CI 75.4% to 80.5%) met our composite criteria for a good TKA outcome (table 3).

    Table 3

    Proportion (95% CI) that achieved net benefit from TKA at 1 year post TKA

    Results of stepwise variable selection

    Optimal outcome discrimination was achieved with 13 variables. The best 13-variable model included WOMAC Pain Score and KOOS-PS (measures of TKA need), PHQ-8 depressed mood, PASS knee symptom acceptability and definite TKA willingness (measures of readiness/willingness), troublesome contralateral knee or low back (measures of health status), five TKA expectations (improved psychological well-being and improved ability to go upstairs, kneel, perform recreational activities and perform activities of daily living) and patient age. All other factors, including prior OA treatment, BMI, non-MSK comorbidities, sex and level of social support, were not selected for inclusion.

    Predictors of a good TKA outcome: result of log Poisson regression modelling

    The unadjusted and adjusted RRs for a good TKA outcome associated with identified predictors are shown in table 4. In the multivariable model, the probability of a good TKA outcome was significantly higher for those with greater TKA need (adjusted RR per 10-unit increase WOMAC Pain Score 1.03, 95% CI 1.01 to 1.05; adjusted RR per 10-unit increase KOOS-PS 1.06, 95% CI 1.03 to 1.08), greater TKA readiness and willingness (unacceptable knee symptoms adjusted RR 1.14, 95% CI 1.03 to 1.27; definitely willing adjusted RR 1.20, 95% CI 1.05 to 1.37), fewer symptoms of depressed mood (adjusted RR per 10-unit increase in PHQ-8 0.94, 95% CI 0.89 to 0.99) and who considered it ‘very important’ that TKA improve their ability to go upstairs or perform recreational activities (adjusted RR and 95% CI 1.15, 1.02 to 1.30 and 1.10, 1.01 to 1.20, respectively). Those who considered it ‘very important’ that TKA improve ability to kneel or psychological well-being had a significantly lower probability of achieving a good outcome (adjusted RR and 95% CI 0.93, 0.87 to 0.99 and 0.92, 0.86 to 0.99, respectively).

    Table 4

    Preoperative predictors of a good TKA outcome*: results of multivariable robust Poisson regression

    Discriminant ability of appropriateness domains for achievement of a good TKA outcome

    Model discrimination for TKA need (WOMAC Pain Score and KOOS-PS) was good (AUC 0.67, 95% CI 0.63 to 0.71). Inclusion of readiness/willingness (PASS and TKA willingness), symptoms of depression and TKA expectations (importance of improved kneeling, stair climbing, psychological well-being and ability to perform recreational activities) significantly improved model discrimination (p=0.01), giving an optimism corrected AUC of 0.70 (0.66–0.74) (figure 2).

    Figure 2

    Area under the receiver operating characteristic (AUROC) curves by TKA appropriateness domain. The corresponding AUROC curves and 95% CIs are as follows: TKA need (WOMAC Pain Score and KOOS-PS) 0.67 (0.63–0.71); TKA readiness and willingness (PASS knee symptoms and definite willingness and PHQ-8 depressive symptoms) 0.61 (0.57–0.64); TKA expectations (importance of four TKA outcomes: improvements in going upstairs, performing recreational activities, kneeling and psychological well-being) 0.57 (0.53–0.62); and for all appropriateness domains combined 0.70 (0.66–0.74). KOOS-PS, Knee injury and Osteoarthritis Outcome Score physical function short-form; PASS, Patient Acceptable Symptom State; PHQ-8, Patient Health Questionnaire; TKA, total knee arthroplasty; WOMAC, Western Ontario-McMaster Universities Osteoarthritis Index.

    Cohort distribution by level of appropriateness

    TKA need was considered low if WOMAC Pain Score and KOOS-PS were in the lowest tertiles of preoperative scores (WOMAC Pain Score≤45 and KOOS-PS≤44), high if both scores were in the uppermost tertiles (WOMAC Pain Score≥60 and KOOS-PS≥62) and moderate otherwise. Readiness/willingness was considered high if knee symptoms were unacceptable and the individual indicated definite willingness for TKA and moderate otherwise. Symptoms of depression were categorised as per PHQ-8 recommendations39: no depressed mood (high appropriateness) if ≤4, mild depressed mood (moderate appropriateness) if 5–9 and moderate/severe depressed mood (low appropriateness) if ≥10. TKA expectations that were positively associated with our TKA outcome in our multivariable regression model (improved ability to go upstairs and perform recreational activities) were considered ‘realistic’, while those that were negatively associated with our outcome (improved ability to kneel or improved psychological well-being) were considered ‘unrealistic’. For each participant, we assigned an expectation score of 1 for each realistic expectation deemed ‘very important’ and a score of −1 for each unrealistic expectation indicated as ‘very important’. Participant expectation scores were then summed (−2 to +2) and appropriateness was considered to be low if the summary score was <0 (preponderance of unrealistic expectations), moderate if 0 and high if >0, indicating a preponderance of realistic expectations (table 5). Using this approach, 211 participants (20.0%) had low need, 630 (59.8%) had moderate need and 212 (20.1%) had high need. Readiness/willingness was moderate in 299 (29.1%) and high in 728 (70.9%). Almost half (n=496, 47.2%) had no depressive symptoms (high appropriateness), 285 (27.1%) had mild symptoms (moderate appropriateness) and 271 (25.8%) had moderate or severe symptoms (low appropriateness). Over half of the participants (n=547, 52.0%) had both realistic and unrealistic expectations (moderate appropriateness), while 130 (12.4%) and 376 (35.7%) had a preponderance of unrealistic and realistic expectations (low and high expectations), respectively.

    Table 5

    Estimated probability of a good TKA outcome by level of TKA appropriateness

    Predicted probability of a good TKA outcome by level of TKA appropriateness

    The predicted probability of a good TKA outcome ranged from 44.4% (33.9–55.5) for those with low TKA need, moderate readiness/willingness, moderate/severe symptoms of depression and a preponderance of unrealistic expectations to 92.4% (88.4–95.1) for those with high TKA need, moderate readiness and willingness, mild depressive symptoms and predominantly realistic expectations (table 5).


    In prior qualitative research, patients and surgeons agreed that evidence of demonstrable TKA need and fitness for surgery were important to consider in assessing patient appropriateness for TKA.32–34 In addition, people with OA stressed that patients should be ready and willing to undergo TKA and have a positive attitude and realistic expectations for surgery.32 In a large cohort of knee OA TKA recipients, we asked if patients who met these criteria would be more likely than those who did not to experience a good outcome. One year post surgery, 78% achieved a good TKA outcome, defined as patient-reported symptom improvement and satisfaction with results. The model that included preoperative measures of TKA need, readiness/willingness and surgical expectations had good discrimination for our outcome, with a model AUC of 0.70 and discrimination was significantly better than for measures of TKA need alone. Although external validation is required, our findings support the need for more explicit consideration of patients’ preferences and values in TKA decision-making.

    There is strong consensus that TKA should be offered to individuals with knee symptoms that are negatively affecting quality of life despite a trial of non-surgical therapies.15 16 25–27 However, only 71% of our TKA recipients had received recommended OA therapies. Although we did not find that receipt of prior recommended OA therapies was associated with TKA outcome, it is reasonable to expect that non-surgical therapies have been exhausted before a costly and potentially risky procedure is performed. Furthermore, preoperative WOMAC Pain Score and KOOS-PS spanned the full range and one-fifth of participants reporting their symptoms as ‘acceptable’. Low symptom scores at surgery may reflect patient avoidance of activities that exacerbate knee symptoms, use of analgesic therapies or simply less TKA need. Participants with low symptom scores were less likely than those with higher scores to report unacceptable knee symptoms, suggesting the latter may be the case.

    Our findings raise concerns regarding the prior use of the RAND appropriateness methodology to develop TKA appropriateness criteria.9 20 The RAND approach is based on expert consensus and best-available evidence from literature review. There is no incorporation of patients’ perspectives and validation of preliminary criteria is not based a priori on demonstration that patients deemed appropriate versus uncertain or inappropriate prior to surgery are more likely to achieve ‘net benefit’ from TKA. The revised Spanish criteria and those from the American Academy of Orthopaedic Surgery (AAOS) were developed in this manner.15 16 18 Using the Spanish criteria, Escobar et al found that appropriate candidates were more likely than those deemed inappropriate to experience minimally important improvements in pain and function at follow-up, but outcomes were similar for those deemed appropriate versus uncertain and level of satisfaction was similar across groups.15 Furthermore, both the Spanish and AAOS criteria ask surgeons to make assumptions regarding the patients whose TKA appropriateness they are assessing. In development of the Spanish criteria, experts were instructed to assume that the patient had unilateral knee symptoms, a BMI<40 kg/m2, adequate social support and received appropriate knee OA management and their expectations had been addressed appropriately.18 The AAOS TKA Criteria similarly asked clinical experts to assume that the patient had failed other relevant appropriate treatments.16 Our findings question the validity of these assumptions.

    Others have made recommendations on appropriate use of TKA based on statistical modelling of the predictors of TKA outcomes.17 28 These studies have variably found that younger patient age, greater BMI and the presence of comorbidities were associated with worse TKA outcomes. Despite wide variability in age, BMI and comorbidities, none of these variables independently predicted our TKA outcome. A possible explanation for this finding is that these factors influence TKA outcome through their effects on TKA need, readiness and willingness and expectations for surgery.31 53

    Appropriateness criteria should inform shared decision-making with the patient, not replace it.54 The extent to which our identified criteria are considered and weighed in the TKA decision is unclear and, to our knowledge, no existing TKA decision-support tools incorporate the criteria we have identified in this study as important. In our prior work,33 orthopaedic surgeons were generally supportive of the use of a TKA decision support tool, but not if it produced a score or threshold to determine appropriateness as clinical judgement should ultimately guide decision-making. Thus, in the current study, we did not categorise participants as appropriate or inappropriate for surgery, but rather considered TKA appropriateness on a spectrum, from low to high, for each appropriateness domain. Although there were small numbers of participants in many cells resulting in wide CIs around our probability estimates, categorising participants post hoc using this approach demonstrated more than twofold variability in the predicted probability of a good TKA outcome. Incorporation of these findings into a TKA decision-support tool has the potential to improve identification of those patients most likely to benefit from surgery and is the focus of ongoing work.

    Patients’ expectations for TKA, as assessed in this study, were associated with both increased and decreased probability of a good TKA outcome. This finding is consistent with the premise that dissatisfaction with results may result, in part, from failure to achieve preoperative expectations. Many factors may influence both individuals’ expectations for TKA, as well as the likelihood that the expectations are achievable, including age, social support, OA severity and comorbidities. In other words, what might be unrealistic in one patient may be totally realistic in another. Given this, assessment of whether patients’ expectations are realistic must take these contextual factors into consideration. This is likely best done by the surgeon or another skilled clinician with deep understanding of TKA at the time of consultation regarding TKA.

    Study strengths include our focus on people with knee OA undergoing primary TKA, which represents the majority of TKA recipients, and inclusion of TKA recipients across 45 arthroplasty surgeons, who perform approximately 60% of all TKAs in the province of Alberta, Canada. Appropriateness domains were assessed preoperatively using simple, brief, validated measures. Our sample size, with broad representation of patients by age and gender, enabled consideration of multiple predictor variables. As patients with severe preoperatively symptoms may report comparable, or even greater, improvements as compared with patients with lesser severity and be very satisfied with their TKA results, despite higher postoperative pain and disability scores, we chose a dichotomous outcome that incorporated symptom improvement (the journey), final status (the destination) and satisfaction with results. Follow-up at 1 year and data completeness were high, and analyses controlled for clustering among surgeons. There are also study limitations. Our final model had only good discriminant ability for our outcome. Inclusion of additional factors, such as findings on knee examination or imaging, may further enhance discrimination and should be considered in future work. Participating surgeons were unaware of patients’ study responses; whether they elicited and discussed patients’ readiness, willingness or TKA expectations is unknown. Finally, our study was performed in Alberta, thus results may not be generalisable to other provincial healthcare systems.

    In conclusion, this prospective cohort study confirmed that incorporation of patients’ readiness, willingness and expectations for TKA with TKA need improved ability to discriminate, preoperatively, those TKA recipients who did versus did not experience a good TKA outcome at 1-year follow-up. These findings highlight the need for explicit incorporation of patients’ preferences and values in assessment of TKA appropriateness and suggest that doing so has the potential to improve the proportion of TKA recipients who experience a good surgical outcome.

    Data availability statement

    Anonymised data are available upon reasonable request.

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by the research ethics boards of the universities of Alberta (PRO-00051108) and Calgary (REB 14-1294) and Women’s College Hospital (REB 2014-0092). Patients and surgeons provided written informed consent.


    *Members of the BEST-Knee Study Team: Dr. Gillian A. Hawker (University of Toronto, Toronto, ON), Dr. Deborah A. Marshall (University of Calgary, Calgary, AB), Dr. Eric Bohm (University of Manitoba, Winnipeg, MB), Dr. Michael J. Dunbar (Dalhousie University, Halifax, NS), Dr. Peter Faris (University of Calgary, Calgary, AB), Dr. C. Allyson Jones (University of Alberta, Edmonton, AB), Dr. Tom Noseworthy (University of Calgary, Calgary, AB), Dr. Bheeshma Ravi (University of Toronto, Toronto, ON), Dr. Linda Woodhouse (University of Alberta, Edmonton, AB & Tufts University, Phoenix, AZ, United States). Edmonton Bone and Joint Centre, Edmonton, Alberta, Canada: participating study surgeons (Dr. Gordon Arnett, Dr. Robert Balyk, Dr. Jeffery Bury, Dr. John Cinats, Dr. Donald Dick, Dr. D'Arcy Durand, Dr. Lee Ekert, Dr. Don Glasgow, Dr. Robert Glasgow Sr., Dr. Gordon Goplen, Dr. Ben Herman, Dr. Catherine Hui, Dr. Larry Hunka, Dr. Hongxing Jiang, Dr. William C. Johnson (deceased), Dr. Frank Kortbeek, Dr. Guy Lavoie, Dr. Mitch Lavoie, Dr. Paul K. Leung, Dr. James Mahood, Dr. Edward Masson, Dr. Richard McLeod, Dr. James McMillan (deceased), Dr. Greg O’Connor, Dr. David Otto, Dr. Carlo Panaro, Dr. Paulose Paul, Dr. Gordon Russell, Dr. Don Weber, Dr. Colleen Weeks, Dr. Andrea Woo (FP, screening)), clinic staff (Jane Squire Howden and Candace Kenyon), and research staff (Anne-Marie Adachi, Jessica Beatty, Shakib Rahman, Braden Woodhouse). Alberta Hip & Knee Clinic, Calgary, Alberta, Canada: participating study surgeons (Dr. Greg Abelseth, Dr. Kelley De Souza, Dr. John Donaghy, Dr. Paul Duffy, Dr. Kelly Johnston, Dr. Robert Korley, Dr. Raul Kuchinad, Dr. Michael Monument, Dr. Maureen O'Brien, Dr. James Powell, Dr. Shannon Puloski, Dr. Ed Rendall, Dr. Alex Rezansoff, Dr. Raj Sharma, Dr. James Stewart, Dr. Scott Timmerman, Dr. Jason Werle), clinic staff (Tanya Reczek), and research staff (Jeffrey Depew), Bukky Dada (Department of Community Health Sciences, University of Calgary) and Ian Stanaitis (Women’s College Hospital/University of Toronto).



    • Contributors GAH and DM conceived of the study and are guarantors of this study. All authors contributed to the study design and data collection, critically revised the article, approved the final version for submission and had full access to all the data in the study. GAH was responsible for the data analysis and drafted the article.

    • Funding This study was funded by an Operating Grant from the Canadian Institutes of Health Research, grant #MOP-312807. The funder had no role in the design and conduct of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript; or the decision to submit the manuscript for publication.

    • Competing interests GAH has received research support as the Sir John and Lady Eaton Professor and Chair of Medicine, Department of Medicine, University of Toronto. DM has received salary and research support as a Canada Research Chair and as the Arthur J.E. Child Chair Professor. All other authors declare no other relationships or activities that could appear to have influenced the submitted work.

    • Provenance and peer review Not commissioned; externally peer reviewed.