Abstract
BACKGROUND: Anterior cruciate ligament (ACL) registries do not all use the same patient-reported outcome measures, limiting comparisons and preventing pooling of data for meta-analysis. Our objective was to create a statistical crosswalk to convert cohort and registry mean Knee Injury and Osteoarthritis Outcome Scores (KOOS) to International Knee Documentation Committee-Subjective Knee Form (IKDC-SKF) scores and vice versa to allow these comparisons. METHODS: Data from 3 ACL registries were pooled (n = 14,412) and were separated into a training data set (70% of the sample) or a validation data set (30% of the sample). The KOOS and the IKDC-SKF scores were available prior to the operation and at 1, 2, and 5 or 6 years postoperatively. We used equipercentile equating methods to create crosswalks in the training data set and examined accuracy in the validation data set as well as bootstrapping analyses to assess the impact of sample size on accuracy. RESULTS: Preliminary analyses suggested that crosswalks could be attempted: large correlations between scores on the 2 measures (r = 0.84 to 0.94), unidimensionality of scores, and subpopulation invariance were deemed sufficient. When comparing actual scores with crosswalked scores in the validation data set, negligible bias was observed at the group level; however, individual score deviations were variable. The crosswalks are successful for the group level only. CONCLUSIONS: Our crosswalks successfully convert between the KOOS and the IKDC-SKF scores to allow for a group-level comparison of registry and other cohort data. CLINICAL RELEVANCE: These crosswalks allow comparisons among different national ligament registries as well as other research cohorts and studies; they also allow data from different patient-reported outcome measures to be pooled for meta-analysis. These crosswalks have great potential to improve our understanding of recovery after ACL reconstruction and aid in our ongoing efforts to improve outcomes and patient satisfaction, as well as to allow the continued analysis of historical data.