Participants.
For this study we used data from three samples of subjects. The first sample consisted of 58 ACLR patients, who participated in a study conducted in 2015 to measure psychometric dimensions relevant to FoM after ACLR[20]. Patients were included from physical therapy practices when they were between 3 months and 3 years after their ACLR. The second sample of respondents were 169 healthy regular students at the HAN University of Applied Sciences. All first year physical therapy students attending a lecture were asked to complete the ACLR-PHOSA. At the start of the lecture, students were informed about the study, and asked to provide written informed consent. The third group of participants were 30 healthy junior soccer players that were enrolled in a training programme of the professional Soccer club Vitesse (Arnhem), additional to their regular educational programme.
Measurements.
All participants reported demographics (gender, age), and completed the PHOSA-ACLR using an online questionnaire (sample 1) or printed questionnaire (sample 2 and 3). The PHOSA-ACLR is comprised of photographic images of 12 sports situations invoking FoM in ACLR patients. In the introduction to the questionnaire participants are asked to imagine performing the movement depicted in the photograph and then they are asked to report the subjective FoM for each item on a scale from 0–10 (11 point rating scale). A score of 0 means ‘totally not damaging’, and 10 stands for ‘very damaging’. The average item score for the scale is computed. In a previous study the PHOSA-ACLR showed excellent reliability and validity in patients with ACLR[20].
Sample 1 (ACLR patients) was asked to report time since reconstruction and completed two additional questionnaires assessing self-reported functioning and fear of movement. Self-reported functioning was assessed using the validated Dutch version of the Knee Injury and Osteoarthritis Outcome Score (KOOS)[22]. The KOOS measures pain, other disease-specific symptoms, activities of daily life (ADL) function, sport and recreation function, and knee-related quality of life (QOL). For each scale the scores were recoded from 0–100, with 100 depicting no problems. Fear of movement was assessed using the Dutch version of the Tampa Scale of Kinesiophobia (TSK) [18, 23, 24]. The TSK measures generic fear of movement or (re)injury using 17 items. Item scores range from 1—4 where 1 = strongly disagree and 4 = strongly agree. Total score is the count of all 17 items after inversion of items 4, 8, 12 and 16. A score higher than 36 is considered kinesiophobia.
Healthy subjects in sample 2 and 3 were asked to complete two additional questions related to sports activities (which sports activities they were active in, and hours of sports per week), and three yes/no questions relate to physical condition (do you have a current medical condition, have you ever had previous lower extremity injuries, and do you know anyone with an ACL rupture).
Analysis.
Descriptive statistics are given for the participants in the three groups. For continuous variables mean score is reported, standard deviation, and 95% Confidence Interval for Means (95%CI) when appropriate. For dichotomous variables percentages are given for each group. For sample 1 ACLR related mean scores are given for functioning (KOOS), and Fear of Movement (TSK). In sample 2 and 3, reported sports activities were recoded into two categories: low risk ACL tear sports (code 1 = swimming, horseback riding, ballet, running, etc.); and high-risk ACL tear sports (code 2 = soccer, hockey, tennis, baseball, etc.).
Item Response Theory (IRT) models are used to determine the probability of an individual’s response to an item, given their latent ability, which in this case would be ACLR specific FoM [24]. In this study we applied a polytomous IRT-model called a Graded Response Model (GRM) to the PHOSA-ACLR in which the item parameters were estimated for each 11-point Guttman item used in this study. The GRM is an especially useful item response model when item response options can be conceptualized as ordered categories with a strongly restricted monotonicity, such as Guttman scales. Item locations are the polytomous equivalent to item difficulties in IRT models for dichotomous responses. A GRM is similar to IRT models with dichotomous outcomes, but within the GRM framework an item response scale reconstructs a polytomous scale in a series of m–1 dummy response dichotomies, where m represents the number of response options for a given item. Thus, an item rated on a 11-point scale has ten response dichotomies: category 1 versus category 2, category 2 versus category 3, etc. The graded response model estimates the probability of endorsing a response category for each item conditional on the latent trait, in which the latent trait is considered a weighted and standardized total scale score. Two models were tested: The first model was a 1-parameter (1-PL) model, in which only the item locations were estimated, and discrimination was set to one for all items (often referred to as a Rasch-model). The second model that was tested was a 2-parameter model in which in addition to the estimation of item locations the item discrimination-parameters were estimated. Item discrimination equates to the slope of the item response function, and is a measure of how sharply the response options discriminate between each other around their location parameters. Higher discrimination means that the items are more informative. To determine whether the added complexity of the 2PL model is necessary the fit of the two models were compared on their –2loglikelihood, which is χ2 distributed. By testing whether the chi-square is significant it was determined if the added complexity of the 2PL model significantly improved the model fit.
Differences between groups in PHOSA-ACLR. Differences between groups were tested with Univariate Analysis of Covariance (ANCOVA). Before running an ANCOVA possible covariates have to be identified. Covariates are variables that differ between groups under study and that are related to the dependent variables. Differences between samples in demographics were tested using χ2 for dichotomous variables, and ANOVA for continuous variables. Associations between demographic variables and PHOSA-ACLR scores were calculated using χ2 for dichotomous variables and Pearson correlation for continuous variables. Correlations were interpreted as small (r =.10), medium (r =.30), and large (r >.50)[25]. Normality of items distribution were tested using the Shapiro-Wilks test. Levene’s test of equality of error variance was used to test the hypothesis that the error variance of the dependent variable (PHOSA-ACLR) is equal across groups. F statistics and significance level are given for the corrected model comparing the three groups of participants controlling for covariates (ANCOVA). The same statistics are given for the independent variable (groups to be compared), and covariates. For significant co-variates partial eta square was computed to determine the percentage of variance attributed to that covariate. All analyses were done using SPSS version 26 (IBM Corporation). A p-value < 0.05 was used as an indication of statistical significance.