The Achilles Tendon Total Rupture Score
The Achilles Tendon Total Rupture Score
The ATRS was published in 2007, and advocated by the authors as the only validated PROM available to evaluate patients following a rupture of their Achilles tendon. There have been no subsequent validation studies. Therefore, this study represents the first paper to investigate aspects of validity, outside of the developing centre.
We investigated aspects of internal consistency, convergent validity and responsiveness of the ATRS, using a sample of 64 patients. The ATRS was found to have high internal consistency at each time point (Cronbach's alpha between 0.89 and 0.95). This finding is also consistent with the original development article, which reported a Cronbach's alpha of 0.96. However, a result above 0.90 has been debated within the literature as being an indication that the outcome is too homogeneous. The implication being that further item reduction may be appropriate.
Convergent validity of the ATRS was evaluated against the DRI score. Correlation coefficients between the DRI and ATRS demonstrated statistically significant correlations between the two scores at each time point. However the confidence intervals around this were wide, with only the six and nine month time points demonstrating a correlation coefficient of at least 0.7. These wide confidence intervals may be the result of the limited sample size of this study. Alternatively, they may reflect an element of heterogeneity amongst the sample. For example the inclusion of patients managed both operatively and non-operatively and patients with co-morbidities such as asthma or diabetes, may affect the distribution of PROMs scores.
These results do however provide some evidence that the ATRS is measuring similar aspects of outcome when compared to the DRI. The main limitation with this methodology was highlighted by the developing authors of the score, who acknowledged that this element of validity should be interpreted with caution as there was no existing gold-standard PROM with which to compare.
Further exploring correlations of the ATRS with the DRI we next investigated if the ATRS correlated more strongly with aspects of the three DRI sub-divisions. To further analyse aspects of convergent validity the ATRS was also correlated against the EQ-5D and two of its subdivisions evaluating 'mobility' and 'usual activities'. Again the confidence intervals were large across the time points and scales evaluated. There size of the correlations did not fulfil the pre-defined criteria of 0.7 within the EQ-5D or its subdivisions. Within the DRI this criteria was met by the second and third sub-divisions of the DRI at the six and nine month time points. These results were anticipated by the authors to an extent because the EQ-5D measures more generic quality of life, as opposed to the alternative construct of physical activity, measured by the ATRS. Again the key limitation of these correlations is that the three scores are measuring only similar constructs as opposed to exact constructs and with the large confidence intervals reported, a larger sample may be required. A 'foot and ankle' specific PROM may provide a more exact construct for comparison, but as described in the methods section there is also a distinct lack of robustly-developed outcome measures in this area.
The more specific ATRS outcome measure demonstrated greater responsiveness than the more generic DRI and EQ-5D scores at each time point. These results were in keeping with the original development article. The level of responsiveness was only marginal in comparison to the DRI and EQ-5D up until the three month time point, with greater levels of responsiveness evident at the six month and nine month time points. This may by representative of the greater ceiling effects seen within the EQ-5D and DRI scores. There are many methods available to determine responsiveness. This method was used as opposed to more routinely reported effect sizes because it does not require parametric assumptions.
Discussion
The ATRS was published in 2007, and advocated by the authors as the only validated PROM available to evaluate patients following a rupture of their Achilles tendon. There have been no subsequent validation studies. Therefore, this study represents the first paper to investigate aspects of validity, outside of the developing centre.
We investigated aspects of internal consistency, convergent validity and responsiveness of the ATRS, using a sample of 64 patients. The ATRS was found to have high internal consistency at each time point (Cronbach's alpha between 0.89 and 0.95). This finding is also consistent with the original development article, which reported a Cronbach's alpha of 0.96. However, a result above 0.90 has been debated within the literature as being an indication that the outcome is too homogeneous. The implication being that further item reduction may be appropriate.
Convergent validity of the ATRS was evaluated against the DRI score. Correlation coefficients between the DRI and ATRS demonstrated statistically significant correlations between the two scores at each time point. However the confidence intervals around this were wide, with only the six and nine month time points demonstrating a correlation coefficient of at least 0.7. These wide confidence intervals may be the result of the limited sample size of this study. Alternatively, they may reflect an element of heterogeneity amongst the sample. For example the inclusion of patients managed both operatively and non-operatively and patients with co-morbidities such as asthma or diabetes, may affect the distribution of PROMs scores.
These results do however provide some evidence that the ATRS is measuring similar aspects of outcome when compared to the DRI. The main limitation with this methodology was highlighted by the developing authors of the score, who acknowledged that this element of validity should be interpreted with caution as there was no existing gold-standard PROM with which to compare.
Further exploring correlations of the ATRS with the DRI we next investigated if the ATRS correlated more strongly with aspects of the three DRI sub-divisions. To further analyse aspects of convergent validity the ATRS was also correlated against the EQ-5D and two of its subdivisions evaluating 'mobility' and 'usual activities'. Again the confidence intervals were large across the time points and scales evaluated. There size of the correlations did not fulfil the pre-defined criteria of 0.7 within the EQ-5D or its subdivisions. Within the DRI this criteria was met by the second and third sub-divisions of the DRI at the six and nine month time points. These results were anticipated by the authors to an extent because the EQ-5D measures more generic quality of life, as opposed to the alternative construct of physical activity, measured by the ATRS. Again the key limitation of these correlations is that the three scores are measuring only similar constructs as opposed to exact constructs and with the large confidence intervals reported, a larger sample may be required. A 'foot and ankle' specific PROM may provide a more exact construct for comparison, but as described in the methods section there is also a distinct lack of robustly-developed outcome measures in this area.
The more specific ATRS outcome measure demonstrated greater responsiveness than the more generic DRI and EQ-5D scores at each time point. These results were in keeping with the original development article. The level of responsiveness was only marginal in comparison to the DRI and EQ-5D up until the three month time point, with greater levels of responsiveness evident at the six month and nine month time points. This may by representative of the greater ceiling effects seen within the EQ-5D and DRI scores. There are many methods available to determine responsiveness. This method was used as opposed to more routinely reported effect sizes because it does not require parametric assumptions.
