1. Introduction
Shoulder instability arises from traumatic or atraumatic causes. Its classification includes anterior, posterior or multidirectional instability. Pathologic conditions that may contribute to instability in the glenohumeral joint include joint laxity, labral tears, ligament injuries, impaired muscular control as well as bone defects within the humeral head or in the glenoid [
1,
2]. Shoulder instability may cause pain, popping and crepitus, a sensation of the joint giving way, limb weakness, repeated episodes of joint dislocation, reduced participation in activities of daily living and sports, and a deterioration in quality of life. Deterioration of joint function may develop over time [
2].
Apart from imaging and physical examinations of shoulder instability, medical history with patient feedback should be considered to assess the patient’s condition and recovery. In the orthopedic literature the most frequently described patient-reported outcome measures (PROMs) regarding shoulder instability are the Western Ontario Shoulder Instability Index (WOSI), the Rowe Score, the Constant Score and the Oxford Shoulder Instability Score (OSIS). All of these disease-specific questionnaires demonstrate appropriate validity, reliability and responsiveness for patients with shoulder instability [
3,
4,
5,
6].
The Western Ontario Shoulder Instability Index (WOSI), developed by Kirkley et al. in 1998, contains 21 items grouped in four domains concerning the symptoms of instability, everyday functioning and quality of life. The WOSI contains instruction for users clarifying every item [
3,
7]. High validity and reliability of the WOSI was demonstrated, with better responsiveness than other shoulder measurement tools—the Disabilities of the Arm, Shoulder and Hand tool (DASH), the American Shoulder and Elbow Surgeons Standardized Shoulder Assessment Form (ASES), UCLA Shoulder Rating Scale, Constant Score, Rowe Rating Scale and a global health instrument—the SF12 [
3]. Rouleau et al., in their systematic review, demonstrated that out of 25 questionnaires used to assess shoulder instability, the WOSI has the best psychometric features [
8]. Whittle et al., in their systematic review, also strongly recommend using WOSI for patients with shoulder instability [
6]. To our knowledge, the WOSI has been translated to Swedish [
9], German [
10,
11], Japanese [
12], Italian [
13], Brazilian, Portuguese [
14], Dutch [
15,
16], French—validated on a Canadian and Swiss population [
17], Spanish [
18], Danish [
19], French—validated on a French population [
20], Turkish [
21], Hebrew [
22], Arabic [
23,
24] and European Portuguese [
25]. In Poland, translation and cultural adaptation of the WOSI was conducted by Bejer et al., in 2019 [
26]. It met the guidelines of MAPI Research Institute and had the involvement of the author of the source version [
27]. This multi-staged study resulted in the creation of a well-translated and culturally adapted version of the Polish questionnaire [
26].
The aim of this study was to evaluate psychometric properties—reliability, validity and responsiveness of the Polish version of the WOSI. This research was undertaken because no disease-specific instrument currently available in Poland enables evaluation of various aspects of functioning and quality of life in individuals with shoulder instability.
4. Discussion
Adaptation of measurement tools allows the global exchange of results in a standardized manner, thereby enabling reliable international comparisons to be made, i.e., assessing the treatment strategies used and the impact of selected factors on them. The WOSI is one of most frequently used questionnaires for shoulder instability with very good psychometric properties [
6,
8]. It has been validated in twelve languages and in fourteen cultures so far [
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25]. Bejer et al., (2019) published the results of the translation and cultural adaptation of the WOSI into Polish [
26].
To the best of our knowledge, our research project was the first attempt to assess the psychometric properties of the Polish version of the WOSI. The majority of the hypotheses specified in the methodology were proven. The findings show that the psychometric properties of the WOSI-PL correspond to those reported for the original version of the tool and WOSI adaptations developed in other countries [
3,
9,
10,
11,
12,
13,
14,
15,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25]. A summary of the psychometric properties of all language versions of the WOSI questionnaire is shown in
Table S1. The evidence presented in the current study supports the validity of the Polish version of the questionnaire as a highly reliable and responsive tool enabling assessment of the quality of life in individuals after arthroscopic repair for shoulder instability.
In terms of reliability of the WOSI-PL, both internal consistency and reproducibility were assessed. The Cronbach’s alpha coefficient for the WOSI-PL was found to be 0.94 (
n = 74), which demonstrates good coherence between the different questions and indicates an excellent internal consistency. It also exceeds 0.90, which is the recommended threshold when a questionnaire is used in a clinical setting [
40]; not exceeding 0.95 which may indicate item redundancy [
34]. The Cronbach’s alpha coefficient in other language versions of the WOSI ranged from 0.84 (
n = 85) in the Japanese version [
12] to 0.97 (
n = 81) in the European Portuguese versions [
25]. The original research of the WOSI does not report this coefficient [
3].
Positive ratings for test–retest reliability can be reported if ICC is ≥0.70 [
34], so the present findings, showing ICC values of 0.99, also reflect good repeatability (interval 48–72 h) of the WOSI-PL (
n = 71). These values are similar to those reported in the case of the original WOSI (ICC at 2 weeks was 0.95) [
3] and are consistent with other studies reporting repeatability of this tool using a similar time interval. In the Turkish version, the ICC was 0.97 [
21], and in the European Portuguese version the ICC was 0.97 [
25]. In both of these cases, the time interval was 72 h [
21,
25]. In studies with a longer time interval, the ICC values were slightly lower: the Swedish version [
9] (ICC = 0.94; 2 months;
n = 32), the German version by Drerup et al. [
11] (ICC = 0.87; 10 days;
n = 29), the version for European and North American French-speaking populations by Gaudelli et al. [
17] (ICC = 0.87; 6–14 days;
n = 144) or the French version by Perrin et al. [
20] (ICC = 0.88; 7 days;
n = 27).
We analyzed an SEM and MDC
95 to assess the error associated with applications of the WOSI-PL. Our findings recorded an SEM of 1.41% and MDC
95 of 3.90%. This indicates that a patient must change at least 3.9 points (on a scale from 0 to 100) to detect a significant change in shoulder function that can be considered independent of measurement error. For the WOSI-PL domains, the SEM ranged from 1.82 to 3.10%, resulting in MDC
95 which varied from 5.05 to 8.60%. These results show a slightly lower measurement error than other validations; however, they are still consistent with results of other researchers: Cacchio et al. (SEM of 3.4% and MDC
95 of 9.3%) [
13], Van der Linde et al. (SEM of 8.3% and SDC
95 of 23%) [
16], Wiertsema et al. (SEM of 6.2% and SDC
95 of 17.2%) [
15], Perrin et al. (SEM of 5.7% and MDC
95 of 15.9%) [
20] and Torres et al. (SEM of 3.10% and MDC
95 of 8.60%) [
25]. Yuguero et al. found, in the Spanish version, a greater measurement error than in the other versions (SEM = 23% and MDC = 76%) [
18]. Other WOSI validation studies did not calculate SEM or MDC [
3,
9,
10,
11,
12,
14,
17,
21,
22,
23].
Construct validity of the WOSI-PL was tested by having eighteen a priori hypotheses. They determined the connections between the WOSI-PL (Total and domain scores), the DASH (Total scores) and the SF-36 (Total and subscale scores). They revealed a stronger association between the WOSI-PL (which is a disease specific PROM) and the DASH (which measures a similar construct but is region specific), but a weaker connection between the WOSI-PL and the SF-36 (which represent a less convergent construct: global health). Sixteen out of 18 a priori assumed hypotheses (89%) were confirmed, which, according to Terwee et al. [
34], indicates a high construct validity of the questionnaire. The authors of the original version showed results similar to ours. The WOSI Total score was strongly correlated with the DASH questionnaire (r = 0.77), weaker with the SF-12 Physical Component (r = 0.66) and weakest with the SF-12 Mental Component (r = 0.12). The authors of all published language versions of the WOSI pointed to the correct construct validity of their language versions [
9,
10,
11,
12,
13,
14,
16,
17,
18,
19,
20,
21,
22,
23,
24,
25].
To assess responsiveness of WOSI-PL, that is, the questionnaire’s ability to detect clinically important changes over time, distribution-based methods were used (ES and SRM) in the group of patients subjected to physiotherapy after Test 1. According to Husted et al. [
38], all domains and the Total score of the WOSI-PL showed a moderate or large degree of responsiveness (Total: ES = 0.44, SRM = 1.26). Only eight previous studies have determined the responsiveness of the WOSI [
3,
9,
13,
17,
18,
22,
23,
24]. The WOSI SRM value of 0.93 represents a large degree of responsiveness as reported by Kirkley et al. [
3]. The researchers demonstrated moderate (SRM = 0.65; ES = 0.62) [
41] to large degree of responsiveness (up to SRM = 2.94, ES = 3.17 [
24]) in other studies of the WOSI [
3,
9,
13,
17,
18,
22,
23,
24]. Information on the value of MCID for the WOSI-PL can be applied to determine if the observed changes are meaningful for patients and to estimate the number of patients who achieve a change greater than the MCID following a specific intervention. Our data indicate that the MCID amounts to 126.43 pts. (6% of the 0-2100 WOSI-PL scale) for the anchor-based method, and 174.05 (8% of the 0-2100 WOSI-PL scale) for the distribution-based method. These findings are comparable to the results of studies shown by Kirkley et al. (MCID = 10%) [
42] and significantly lower than those reported by Cacchio et al. (MCID = 19%) [
13]. The potential reasons for this difference in the findings of the Italian researchers could be related to the fact that a different population of patients was tested and different methods were applied to calculate the MCID.
One limitation of the study is the lack of generalizability of the patients in the study. Only patients who had been treated with arthroscopic repair for shoulder instability were included. The psychometric properties of WOSI-PL therefore may not be generalizable to patients treated conservatively by physiotherapy or by open stabilization. Furthermore, the limited sample size of the female group makes it difficult to perform separate analyses assessing the reliability and validity of the Polish WOSI relative to gender. In line with the literature guidelines followed by our study [
34], further research should include a confirmatory factor analysis (CFA), which examines whether the data fit an a priori hypothesized factor structure. We wanted to provide the most accurate sample size for this analysis, since it is recommended that CFA should take into account a group of at least 200–300 subjects [
43]. Up until now, factor structure of the language versions of the WOSI was only assessed by researchers from the Netherlands [
16] and Spain [
18]. The results of the factor analyses (confirmatory factor analyses CFA, exploratory factor analyses EFA) did not allow the researchers to explicitly confirm the validity of the four domains in the relevant language versions of WOSI. The researchers also emphasise it is necessary to continue the related research and analyses.