Defining Clinically Meaningful Thresholds for Patient-Reported Outcomes in Knee Arthroplasty

BACKGROUND: For primary knee arthroplasties, clinically meaningful thresholds of patient-reported outcomes that associate with patient satisfaction have not been defined appropriately.

METHODS: In this retrospective study of 26,720 primary total knee replacements registered in the Dutch Arthroplasty Register (2016-2018), receiver operating curve analysis was used to define minimal clinically important changes (MCICs) and patient acceptable symptom states (PASSs) with the anchor satisfaction. Patient-reported outcome measures were pain, European Quality of Life 5 Dimensions, Knee disability and Osteoarthritis Outcome Score, and Oxford Knee Score (OKS). Independent analyses were performed for groups, which showed statistically significant interactions with the (change in) score to achieve satisfaction in logistic regression.

RESULTS: In this cohort, 84.9% completed the anchor questions, of whom 71.3% with a satisfaction score ≥8. Good discriminative abilities (area under the receiving operator curve >0.8) for PASS were achieved by OKS ≥38.5, pain in activity ≤2.5, Knee disability and Osteoarthritis Outcome Score ≤33, and Quality of Life-Index ≥0.813. Discriminative abilities for MCIC were not good. If assessed per baseline tercile, discrimination improved (area under the receiving operator curve >0.8) and yielded different MCICs per preoperative tercile (preoperative OKS ≤19: MCIC ≥19.5; pre-OKS 20-27: MCIC ≥14.5; pre-OKS ≥28: MCIC ≥8.5). For MCIC, the tercile method produced an 11% improved accuracy compared to one threshold for every patient. For the PASS scores, tercile-specific did not improve the accuracy of predicting satisfaction. Demographics were not clinically relevant in determining thresholds.

CONCLUSION: Estimating the likelihood of satisfaction with surgery is critical in shared decision-making. Patients with more preoperative symptom severity require larger changes to report satisfaction. Both in the clinic and in science, such differences must be considered when predictions of satisfaction are attempted.