Abstract
Objectives:
To assess for the presence of recall bias between prospectively and retrospectively collected patient-reported outcome (PRO) scores in hip arthroscopy (HA).
Methods:
Patients that underwent HA between 2015-2021 and provided pre-operative baseline responses for the International Hip Outcome Tool-12 (iHOT-12), the Hip Disability and Osteoarthritis Outcome Score Physical Shortform (HOOS-PS), and the modified Harris Hip Score (mHHS) were eligible for recruitment. Eligible participants were asked to complete a study-specific survey and the same preoperative PROs retrospectively, requiring them to recall the status of the hip prior to surgery. Agreement between the prospective and retrospective scores was assessed using a two-way mixed effects intraclass coefficient (ICC) model. Paired t-tests were used to compare the mean scores of both groups and linear regression models were used to help identify associations between score discrepancies and patient characteristics.
Results:
A total of 222 patients were approached for participation, of whom 104 (60.6% female) completed study requirements and were included for analysis. The mean duration of symptoms before surgery was 24.5 ± 31.8 months and the mean duration to recall (from the day of surgery) for the retrospective completion of hip surveys was 29.1 ± 21.8 months. The iHOT-12 had a low amount of agreement between prospective and retrospectively collected scores (ICC=.452; P =.000). HOOS-PS scores also had a low amount of agreement (ICC=.458; P =.000). The mHHS had a moderate amount of agreement (ICC=.621; P =.000). Mean scores for iHOT-12 (41.5 ± 22.6 vs. 35.3 ± 17.7; P <.01), HOOS-PS (29.5 ± 18.9 vs 40.7 ± 17.9; P <.001), and mHHS (62.9 ± 16.5 vs 55.1 ± 15.3; P <.001) were all significantly different prospectively vs. retrospectively. The average changes in score observed for the iHOT-12, HOOS-PS, and mHHS were -6.2, 11.2, and -7.8, respectively. Multiple linear regression identified duration to recall and sex as significant predictors of the absolute difference between prospective and retrospectively collected iHOT-12 data while sex alone was predictive of a difference in HOOS-PS data.
Conclusions:
The retrospective collection of patient-reported outcomes for hip arthroscopy procedures, which requires patient recall of their preoperative condition, is subject to bias and inaccuracy. Recalled PROs consistently reflected worse pain/function than their prospectively recorded counterpoints; therefore, retrospective patient recall is an unreliable source of clinical data and the prospective collection of iHOT-12, mHHS, and HOOS-PS data should be prioritized.
The mean scores of the three PROs collected prospectively and retrospectively. The mean changes in score for the iHOT-12, HOOS-PS, and mHHS were -6.2, 11.2, and -7.8, respectively. Note that a lower score on the iHOT-12 and mHHS axis indicates worse function/most pain while a lower score on the HOOS-PS axis indicates better hip condition.
Scatterplot demonstrating the relationship between the prospectively collected PROs and recalled (retrospective) PROs for the iHOT-12, HOOS-PS and mHHS surveys. Points falling outside the areas bound by the parallel lines represent patients with recall scores that differ from preoperative scores to a greater extent than the minimal clinically important difference (MCID). Note that the currently accepted HOOS MCID (9.1 points) was used in the middle frame in lieu of a HOOS-PS MCID as no such value has been established.
A summary of patient demographics stratified by patient-reported outcome.
