Turkish Version of the Ankle Fracture Outcome of Rehabilitation Measure (A-FORM-TR): A Multi-Centre Validation Study

Rasch measurement analysis serves as the primary framework for evaluating dimensionality and measurement properties of the A-FORM-TR. Examined: item fit (infit and outfit mean-square 0.5 to 1.5), category threshold ordering, person separation reliability (target greater than or equal to 2.0), local independence (residual correlations less than 0.30), unidimensionality (residual principal component analysis first contrast less than 2.0), and differential item functioning by sex, age group, treatment modality, centre, and time since injury. A priori decision rules: items showing substantive misfit will be reviewed for rescoring or removal; disordered thresholds will trigger response category collapsing and re-analysis; items showing significant DIF will be split or removed and the analysis re-run on the modified scale.

Confirmatory factor analysis testing the prespecified single-factor (unidimensional) structure established in the original A-FORM development by McPhail et al. (2014). CFA serves as a complementary structural check from a latent-variable perspective alongside the primary Rasch analysis. Model fit assessed by CFI and TLI (target greater than or equal to 0.95), RMSEA (target less than or equal to 0.08), and SRMR (target less than or equal to 0.08).

Pearson or Spearman correlations between the A-FORM-TR Summary Score and the AOFAS Ankle-Hindfoot Score and OMAS-TR. Pre-specified hypotheses: H1, strong positive correlation between A-FORM-TR and OMAS-TR (r greater than or equal to 0.60); H2, moderate positive correlation between A-FORM-TR and AOFAS (r between 0.40 and 0.60); H3, the psychosocial items of the A-FORM-TR correlate more weakly with the AOFAS than the physical-function items. Construct validity is considered supported if at least 75 percent of the a priori hypotheses are confirmed.

Item-level and scale-level Content Validity Index and the Kappa coefficient of agreement derived from the expert committee evaluation conducted in the pre-testing phase.

Percentage of participants scoring at the floor (lowest possible score) and ceiling (highest possible score) of the A-FORM-TR Summary Score. Effects are considered acceptable if less than 15 percent at either extreme.

Known-groups validity assessed by testing whether A-FORM-TR Summary Scores differ across pre-specified clinical subgroups. Pre-specified hypotheses: H4, patients in the 3-week to 3-month post-injury subgroup will score significantly lower (greater impairment) than patients in the 6- to 12-month subgroup; H5, A-FORM-TR Summary Scores will differ between surgically treated and conservatively treated patients (no a priori directional hypothesis). Analyses by independent-samples t-test or Mann-Whitney U as appropriate, with Cohen's d or effect size reporting. Added in response to peer review feedback from the original instrument developer.

Standard error of measurement (SEM) derived from the test-retest data, and the smallest detectable change (SDC) calculated as 1.96 x square root of 2 x SEM. Bland-Altman plots with limits of agreement are reported. Measurement error is evaluated as a property distinct from reliability, in line with the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) framework.

Ability of the A-FORM-TR to detect change over time, assessed against an anchor. Change scores from baseline to 3 months are correlated with a single-item Global Rating of Change (expected r greater than or equal to 0.50). The standardised response mean (SRM) and effect size are computed separately in patients reporting improvement and in those reporting no change (expected SRM greater than or equal to 0.80 versus less than 0.20). Discrimination between improved and stable patients is assessed by receiver operating characteristic (ROC) analysis, with the area under the curve (AUC) expected to be greater than or equal to 0.70. The minimal important change derived from the ROC optimal cut-point is compared against the smallest detectable change.

In a bilingual sub-study of approximately 30 Turkish Cypriot participants who are functionally literate in both Turkish and English, both the A-FORM-TR and the original English-language A-FORM are administered in counterbalanced order with at least 30 minutes between administrations. Agreement between the two language versions is assessed by Pearson or Spearman correlation (expected r greater than or equal to 0.80), intraclass correlation coefficient (expected greater than or equal to 0.80), and Bland-Altman analysis to detect systematic differences. Bilingual eligibility requires documented criteria beyond self-report. This subgroup is not representative of the main sample and this is reported as a limitation.

Comparison of the Turkish item calibration against the original English-language calibration reported by McPhail et al. 2014. The analysis is conditional on the level of source data available from the instrument developer: with raw item-level data, multi-group confirmatory factor analysis testing configural, metric and scalar invariance; with item difficulty estimates only, anchored Rasch calibration with comparison of item difficulty hierarchies. If neither is available, within-sample differential item functioning by sex, age group, treatment modality, recruiting centre and time since injury is reported together with a descriptive comparison against the published psychometric profile.