Scoring Patient Outcomes No Matter How We Ask Them: Commentary on an article by Breanna A. Polascik, BS, et al.: "Crosswalks Between Knee and Hip Arthroplasty Short Forms. HOOS/KOOS JR and Oxford"

Linking Hip Disability and Osteoarthritis Outcome Score-Physical Function Short Form and PROMIS Physical Function

INTRODUCTION

Linking scores on patient-reported outcome measures can enable data aggregation for research, clinical care, and quality. We aimed to link scores on the Hip Disability and Osteoarthritis Outcome Score-Physical Function Short Form (HOOS-PS) and the Patient-reported Outcomes Measurement Information System Physical Function (PROMIS PF).

METHODS

A retrospective study was conducted from 2017 to 2020 evaluating patients with hip osteoarthritis who received routine clinical care from an orthopaedic surgeon. Our sample included 3,382 unique patients with 7,369 pairs of HOOS-PS and PROMIS PF measures completed at a single nonsurgical, preoperative, or postoperative time point. We included one randomly selected time point of scores for each patient in our linking analysis sample. We compared the accuracy of linking using four methods, including equipercentile and item response theory-based approaches.

RESULTS

PROMIS PF and HOOS-PS scores were strongly correlated ( r = -0.827 for raw HOOS-PS scores and r = 0.820 for summary HOOS-PS scores). The assumptions were met for equipercentile and item response theory approaches to linking. We selected the item response theory-based Stocking-Lord approach as the optimal crosswalk and estimated item parameters for the HOOS-PS items on the PROMIS metric. A sensitivity analysis demonstrated overall robustness of the crosswalk estimates in nonsurgical, preoperative, and postoperative patients.

CONCLUSION

These crosswalks can be used to convert scores between HOOS-PS and PROMIS PF metric at the group level, which can be valuable for data aggregation. Conversion of individual patient-level data is not recommended secondary to increased risk of error.

Crosswalks between the Oxford hip and knee scores and the HOOS-12 and KOOS-12 instruments

OBJECTIVE

To develop and validate bi-directional crosswalks between the Oxford Hip Score (OHS) and HOOS-12 summary impact score, and between the Oxford Knee Score (OKS) and KOOS-12 summary impact score.

METHODS

Data were sourced from the Australian Orthopaedic Association National Joint Replacement Registry Patient-Reported Outcome Measures Program. Patients undergoing primary joint replacement for osteoarthritis who completed the OHS and HOOS-12 or OKS and KOOS-12 instruments were included in the analysis. An equipercentile method was used to create four crosswalks, with the distribution of scores smoothed using log-linear models prior to equating. Crosswalk validity was assessed through comparison of actual vs derived scores, Pearson correlation coefficients, root mean square errors (RMSE) and Bland-Altman plots.

RESULTS

Paired OHS/HOOS-12 data and paired OKS/KOOS-12 data were available for 4,513 patients undergoing total hip replacement and 5,942 patients undergoing total knee replacement, respectively. Minimal differences were observed between actual and crosswalk-derived mean scores (actual OHS 27.55 vs derived OHS 27.56; actual HOOS-12 53.28 vs derived HOOS-12 53.31; actual OKS 27.34 vs derived OKS 27.34; actual KOOS-12 50.51 vs derived KOOS-12 50.58). High correlation was observed between actual and derived scores (Pearson's r for hip-specific instruments: 0.943-0.946; Pearson's r for knee-specific instruments: 0.925-0.931). Plotted actual vs mean derived scores also indicated robust concordance across the breadth of the instrument scales.

CONCLUSION

These crosswalks provide close approximations of actual OHS, OKS, HOOS-12 and KOOS-12 scores, as indicated by multiple validation metrics. They offer a resource for clinicians, researchers and arthroplasty registries to support PROMs score conversion and data harmonisation efforts.