Sample 1 included 58 ACLR patients (mean age 25.9 years; range 17–56; SD = 8.2; 43% male). Average time since reconstruction was 15.5 months (range 3–36, SD = 8.0). In Table 1 mean levels of self-reported functioning (KOOS), and fear of movement (TSK) are given.
Sample 2 included 169 regular students (mean age = 19.2, SD = 2.0; 48% male). Students participated frequently in sports activities. Only 6 students (3.4%) reported that they did not actively participate in sports activities. On average PT students reported to be active in sports during 5.6 hours a week. Sample 3 comprised of 30 soccer players from Vitesse (mean age = 18.3, range – 17–20; SD = 3.2; 94% males). In this group of soccer players only two persons were female. This group reported an average of 14.9 hours of active sporting per week. The three groups of participants differed in average age (F(2,254) = 51.2, p < 0.001), with ACLR patients being of higher age compared to the student groups. The three groups of participants differed in gender (χ2 = 21.4, p < .001). The group of junior soccer players were predominantly male, with only two females. Group 2 differed in percentage high/low risk sports compared to both other groups, where all participants in group 1 and 3 were considered high rick athletes.
Assessment of Graded Response Model
Prior to analysis, the PHOSA items were examined for accuracy of data entry, and missing values. No uni- or multivariate outliers were found. The dimensionality of the PHOSA was tested with Principal Axis Factoring, which showed a clear unidimensional fit for the PHOSA items (TLI = 0.90, RMSR = 0.05) with factor loadings between 0.69 and 0.89. Descriptive statistics for the PHOSA items are summarized in Table 3.
Table 3
Phosa items, observed range, means, standard deviations, and median score.
Sports related activity | Observed Range | Mean score (SD) | Median |
1. Running | 0–10 | 3.67 (2.73) | 3.0 |
2. Landing after jumping | 0–10 | 4.39 (2.90) | 4.0 |
3. Squats | 0–10 | 3.11 (2.63) | 3.0 |
4. Lateral lunging | 0–10 | 4.76 (2.83) | 5.0 |
5. Single leg jump | 0–10 | 3.88 (2.71) | 4.0 |
6. Sliding | 0–10 | 3.98 (2.79) | 4.0 |
7. Bring to a halt | 0–10 | 5.11 (2.85) | 6.0 |
8. Hop | 0–10 | 4.11 (2.84) | 4.0 |
9. Lunge | 0–10 | 3.77 (2.78) | 4.0 |
10. Start to sprint | 0–10 | 4.09 (2.84) | 4.0 |
11. Jumping on a trampoline | 0–10 | 4.06 (2.77) | 4.0 |
12. Pivoting movement | 0–10 | 5.68 (2.86) | 6.0 |
The graded response model was used to calibrate item parameter estimates for the PHOSA given its Guttman-scaling; i.e. ordered polytomous response format. The best fitting GRM was first assessed by comparing two models, (1) a one-parameter model (Rasch) with location as parameter (and discrimination assumed fixed to 1); and (2) a two-parameter model, with both location and discrimination as parameters. The two-parameter model (-2LL = -5485.05) showed a significantly better fit than the one-parameter model (-2LL = -5515.36), x2(1) = 30.31, p = 0.00). Subsequent analyses pertain to the fit of the two-parameter model in which both location and discrimination are estimated. In Table 4 the item difficulty locations are sorted, from top to bottom, in ascending order of increasing difficulty.
Table 4
location parameters of the PHOSA items, sorted from lowest difficulty to highest.
| β.1 | β.2 | β.3 | β.4 | β.5 | β.6 | β.7 | β.8 | β.9 | β.10 | β |
6. Sliding | -2.73 | -2.03 | -1.10 | -0.20 | 0.43 | 1.15 | 1.80 | 2.64 | 3.50 | 4.73 | 1.69 |
3. Squats | -2.42 | -1.15 | -0.19 | 0.80 | 1.59 | 2.30 | 2.70 | 3.75 | 5.82 | 7.25 | 2.18 |
1. Running | -3.30 | -2.04 | -0.63 | 0.30 | 0.86 | 1.79 | 2.73 | 3.85 | 5.39 | 5.96 | 2.40 |
11. Jumping | -4.31 | -2.56 | -1.32 | -0.35 | 0.49 | 1.24 | 1.92 | 3.52 | 4.42 | 6.19 | 2.42 |
9. Lunge | -3.40 | -2.03 | -0.97 | -0.07 | 0.70 | 1.63 | 2.53 | 3.47 | 5.34 | 7.82 | 2.47 |
4. Lateral Lunging | -4.48 | -3.45 | -2.49 | -1.60 | -0.40 | 0.65 | 1.36 | 2.76 | 4.12 | 5.67 | 2.49 |
5. Leg jump | -3.95 | -2.33 | -1.33 | -0.28 | 0.61 | 1.72 | 2.58 | 3.89 | 4.73 | 5.76 | 2.56 |
12. Pivoting | -5.92 | -5.32 | -3.56 | -2.59 | -1.91 | -1.03 | 0.37 | 1.44 | 2.80 | 4.03 | 2.58 |
7. Bring to Halt | -5.08 | -3.84 | -3.01 | -1.79 | -0.98 | -0.08 | 1.00 | 2.14 | 3.85 | 5.84 | 2.59 |
2. Landing | -4.73 | -3.12 | -1.74 | -0.69 | 0.31 | 1.02 | 1.67 | 3.14 | 5.23 | 7.28 | 2.92 |
10. Start sprint | -4.46 | -3.28 | -1.81 | -0.40 | 0.63 | 1.49 | 2.77 | 3.92 | 5.88 | 8.81 | 3.39 |
8. Hop | -4.69 | -3.09 | -2.11 | -0.53 | 0.68 | 1.43 | 2.78 | 4.12 | 6.07 | 7.67 | 3.48 |
β.1 = Beta.1 represents the threshold between response category 0 and 1, Beta.2 represents the threshold between response category 1 and 2, etc. The final Beta indicates the item’s overall difficulty; the ability (theta) associated with the weighted threshold between response category 0 and 10. |
How these locations relate to the thresholds within each item are summarized in Item Information Curves in Fig. 1.
The ICCs in Fig. 1 indicate that items 8 and 10 are most informative in the range of -2 and 2, whereas item 12 is most informative below − 2, and item 9 above + 3. Item 6 is least informative in general. The thresholds in Table 2 correspond to the intersection between curves. At these intersection it is equally likely a person will be classified into adjacent categories, and therefore to obtain one of two successive scores on the item. The mean of the threshold locations within an item is represented by the location parameter. The locations range from 1.69 to 3.48 logits, where “Hopscotch” (PHOSA_8) was the most difficult item and “Sliding” (PHOSA_6) was the easiest item (Table 4).
In Fig. 2 the distributions of ability estimates in the sample are summarized in a kernel density plot showing an approximately normal distribution slightly negatively skewed.
The kernel density plot shows that relatively high PHOSA-scores are rare, whereas most of the respondents are either slightly above average or well below average. The Test Item Information Function, visualized in Fig. 3 shows that the test is most informative (has its peak-information) for people within a standardized PHOSA-range of -2 to + 2 standard deviations, with an evenly distributed peak in that range, indicating that the test is consistently reliable in this entire range.
Difference between groups in FoM assessed with the PHOSA-ACLR
Levene’s test of variance was significant (F (2,253) = 5.1); p = 0.007) indicating that the error variance is not equal across groups. The one sample Shapiro Wilks was significant indicating that the distribution of PHOSA-ACLR scores was not normal. Table 5 depicts average PHOSA-ACLR scores for patients of each of the three samples, without any correction of covariates.
Table 5
Average PHOSA score, range, standard deviation (S.D.) and 95% Confidence Interval for patients after ACLR, regular students, and soccer players.
| ACLR N = 58 | Regular students N = 169 | Soccer N = 30 | Total N = 257 |
Average PHOSA | 3.1 | 4.9 | 2.6 | 4.2 |
Range | 0–10 | 0–10 | 0–10 | 0–10 |
Standard Deviation (SD) | 2.3 | 2.0 | 2.7 | 2.3 |
95% CI | 2.5–3.7 | 4.6–5.2 | 1.6–3.6 | 3.9–4.5 |
Median | 2.5 | 5.5 | 1.6 | 4.8 |
On average, regular students reported the highest level of FoM assessed using the PHOSA-ACLR comparable with FoM reported in soccer players. Regular students on average reported higher levels of FoM compared to ACLR patients and soccer players. Within the healthy samples (samples 2 and 3) students participating in high risk sports did not differ from students with low risk sports activities in average PHOSA-ACLR scores; mean PHOSA-ACLR scores in the high risk group is 2.06 (SD 0.22), compared to 2.4 (SD = 0.23) in the low risk group (F (2,190) = 4.47, p = ns). A chronic health condition was reported by 17 students, and these students did not report different PHOSA-ACLR scores when compared to students without chronic condition. Over one third of the students (N = 73; 38%) reported to know someone with an ACL rupture. These students did not differ in mean PHOSA-ACLR from students who did not know anyone with an ACL rupture. Number of hours spent on sports on average per week showed weak association with PHOSA-ACLR (N = 198; r = − .30, p < .001). Students spending more time on sports showed on average lower scores on PHOSA-ACLR.
For all three samples gender, age and previous injury are known and analyzed as possible covariates. Average PHOSA-ACLR score for males is 3.8 (SD = 2.3) compared to 4.7 (SD = 2.2) in females (t(254) = 2.98, p < .0001). Age of participants showed a small but significantly reversed correlation with PHOSA-ACLR (r = -0.21, p < .001). Previous injury at the knee was reported by 194 participants (including all participants from sample 1). Mean PHOSA-ACLR for these participants was 4.32 (SD = 2.3) and was not significantly different from the average score of 4.0 (SD = 2.6) in those participants not reporting previous injury (t(254) = 0.97, p = 0.35). Therefore, only gender and age are considered as covariates in the ANCOVA for differences between the three groups, as these are related to PHOSA-ACLR, and differ between groups. In the ANCOVA analysis the covariate was unrelated to PHOSA-ACLR, with gender being the sole covariate with a significant relation to PHOSA-ACLR (F(1,255) = 4.01; p = .04). Gender explained 1.6% of the variance in PHOSA-ACLR independently from the other variables in the equation (partial eta squared = 0.016). There was a significant difference between the three groups in PHOSA-ACLR after controlling for age and gender (F (2,255) = 17.1, p < .001). Differences in groups explained 12% of variation in PHOSA-ACLR (Partial Eta Squared = 0.12). Post-hoc pairwise comparison adjusted for multiple comparisons (Sidak) showed that the students differed from both the ACLR group (lower FoM, mean difference = 1.6, p < .001), and from soccer players (higher FoM, mean difference = 2.0, p < .001). As the distribution of scores on the PHOSA-ACLR was shown to be not normal, the ANCOVA was repeated using normal transformation of PHOSA-ACLR, with similar results (data not shown).