Reply to comment on “Shoulder Pain and Disability Index: Italian cross-cultural validation in patients with non-specific shoulder pain”

Dear Editor,

We are very happy that our paper has given a significative contribution to a fruitful debate in the field of clinimetrics and validation of outcome measures, within the Italian community of clinicians and researchers. Here follows some evidence that supports our published results point by point.

Structural validity

As reported in the discussion (page 8), we are aware that sample size is a current weakness of the study according to COnsensus-based standards for the Selection of health Measurement INstruments (COSMIN) guidelines, which is the reason why a post-hoc power analysis was performed. Although the power analysis was good, root mean square error of approximation, comparative fit index and normed fit index performance were affected, as expected, raising arguments over oversampling suggested by guidelines, maybe compensating poorness of standard methodologies accepted as a rigid gold standard for testing validity. The unique index we found reaching acceptable level of robustness is the post-hoc Chi-squared p-value, and testing the model had no significant residuals. Nevertheless, poor-fit-to-the-model remains a limitation to be overcome by improving the experimental design, rather than by statistical approaches. Moreover, also the trans-cultural adaptation of Shoulder Pain and Disability Index (SPADI) in Thai (N = 44) faces the same problem of not complying by COSMIN guidelines (DOI 10.1186/s12955-015-0333-2-), thus raising concerns over acknowledged clinimetrics of SPADI as a whole.

The primary focus should be on stimulating more researchers to get involved in validation of outcome measures in the Italian community, with the aim to improve the culture of using valid measurement also in the rehabilitation field to take the rehabilitation care quality delivered to patients a step further. We will be glad whether our results will be confirmed or not by new studies improving current one.

With regard to Figure 2, the numbers reported are not the factors of principal component analysis, but the scores of confirmatory factor analysis (CFA). Differences in raw values between right and left part of the diagram are due to the different techniques used (i.e. Rotated Component [RC], Maximum Likehood [ML]) to test hypotheses. As a result, similarities obtained by different techniques (e.g. domanda_1.1: RC1= 0.779, ML1= 0.7) confirm the existence of two factors. Specifically, scores represent single item saturation, the critical value of which is 0.4 and, as illustrated in Figure 1, all of the items included in a factor are significantly higher than 0.4. From a statistical point, an item could saturate in more than one factor (e.g. domanda_1.4), thus the decision to allocate an item in one factor, rather than a different one is not quantitative, but qualitative. In other words, items composing the factor were chosen and not guessed. Finally, allocation of an item in a factor did not change overall saturation.

Construct validity

Limitation in sample size affects also correlation coefficients weight and thus the testing of construct validity. Furthermore, in this case, the issue has to be solved by redesigning a study which complies better to COSMIN guidelines, as no statistical approaches are available to overcome claimed limitation.

Internal consistency

We thank authors’ letter for highlighting misleading reports; nevertheless, analyses were run on both scale factorisations (i.e. original, extracted) highlighting that the internal consistency was good (>.80) to excellent (>.90) in both cases. If the editor will allow the paper to be amended, we will be ready to improve it. The following table shows results with regard to Figure 2.

Items grouped according to factors in the left part of Figure 2.

	Cronbach’s alpha
Pain subscale (7 items)	0.86
Disability subscale (6 items)	0.89
Total scale (13 item)	0.89
	Cronbach’s alpha
Pain subscale (7 items)	0.86
Disability subscale (6 items)	0.89
Total scale (13 item)	0.89
	Cronbach’s alpha
Pain subscale (7 items)	0.86
Disability subscale (6 items)	0.89
Total scale (13 item)	0.89
	Cronbach’s alpha
Pain subscale (6 items)	0.91
Disability subscale (7 items)	0.84
Total scale (13 item)	0.88
	Cronbach’s alpha
Pain subscale (6 items)	0.91
Disability subscale (7 items)	0.84
Total scale (13 item)	0.88
	Cronbach’s alpha
Pain subscale (6 items)	0.91
Disability subscale (7 items)	0.84
Total scale (13 item)	0.88

Measurement error

We thank the authors for notifying errors in Table 5 and inconsistencies in the report between abstract and conclusions. The SEM reported is not the standard error of measurement, but the Structural Equation Modeling (SEM) used for the CFA, while Minimally Detectable Changes (MDCs) were not calculated.

Therefore, conclusions on SPADI-I are the same already reported both in the abstract (“can be considered as a useful tool in daily clinical practice for assessing musculoskeletal non-specific shoulder pain because of its good internal consistency and validity. Further studies should focus on other psychometric proprieties such as test re-test reliability, responsiveness and clinical interpretability to improve the available clinimetrics of the tool.”) and page 9 (“The main limitations of our study are worth noting: the lack of the re-test phase limiting information about reliability of the scale.”).

Claiming this, we were not in the position of measuring any error, and in fact we did not.

If the editor approves, we will be glad to amend the paper by removing the last sentence of the Results section in the abstract (“Standard error of measurement and minimally detectable change were calculated”) and reedit Table 5 without the last two columns (i.e. SEM, MDC).

We would like to thank again authors’ comment that gave us the opportunity to go further into details of our contribution. However, we do not agree on strong conclusion suggested. As known, scientific knowledge is not a deterministic process, neither an inquisition court, but an open space of debate where transparency of empiric arguments lead to the best temporary shared evidence. As pointed out, we never argued that the present paper is the final evidence offering the full picture on SPADI; conversely, we believe it represents the first jigsaw piece of a puzzle, which needs to be completed. Overall, we expect new findings on the topic to confirm or disconfirm our first step.