[Editor's note: Both the letter to the editor by Stratford and Riddle and the response by Parrott and colleagues are commenting on the author manuscript version of the article that was published ahead of print on May 22, 2014.]
We thank Stratford and Riddle1 for their comments on our article2 and for providing us with this opportunity to clarify several points. They indicate that because the items of the Lower Limb Functional Index (LLFI) and Lower Extremity Functional Scale (LEFS) are different, the tools are likely measuring somewhat different theoretical constructs. Because the LLFI asks questions about how leg function affects several different aspects of the patient's life (sleep, appetite, daily activities), it is possibly a hybrid measure of lower limb dysfunction (ie, the LLFI may be a more comprehensive index of the negative effects of lower limb dysfunction). Because the 2 tools are measuring different things (LEFS: “moving about,” LLFI: more global sequelae of lower limb dysfunction), Stratford and Riddle suggest the comparison is problematic.
Differences in theoretical focus (or psychometric constructs) point precisely to the value of comparing the clinical usefulness of 2 tools designed to serve a similar purpose (in this case, the purpose is to describe lower limb function). Whether or not they are measuring exactly the same thing (and in this case, they likely are not), the question persists: Which of the tools would be more likely to identify self-reported limitations in function? The purpose of comparing LLFI and LEFS, therefore, is not theoretical, but practical. If, in fact, the LEFS focuses more tightly on lower extremity function (“moving about”), but in so doing may systematically indicate adequate function for a set of patients who could otherwise benefit from physical therapy, the LLFI may be preferred, as it captures a broader set of relevant constructs related to function in patients with HIV-related distal sensory polyneuropathy (DSP).
The key point of our comparison was to explore the practical utility in identifying patients with HIV-related DSP (who may benefit from physical therapy) because of the negative effects of lower limb problems. The comparison in our study is not psychometric—thus, the question of whether the 2 tools measure the same self-perception of the impact of HIV-related DSP on function is somewhat beside the point. As Stratford and Riddle point out, they likely do not, but this may be precisely the source of the practical utility of using one versus the other. Hence, a step toward a practical comparison of tool performance is warranted.
Stratford and Riddle are correct that Bland and Altman illustrated their graphical comparison approach by comparing 2 tests of peak expiratory flow rate, but the basic motivation for the development of their statistical technique was to devise a method to compare 2 tools when “neither provides an unequivocally correct measurement.”3 In short, the Bland-Altman technique provides a method for assessing whether 2 approaches agree when the true measures remain unknown. Indeed, in the context of patients with HIV disease, it is not clear that it is particularly meaningful to talk of a “true” measure of the lifestyle effects of lower limb dysfunction—especially given that a patient's self-representation of the disease (“what does this mean for me in my daily life?”) is multifactorial and dynamic. As we pointed out above, it is unclear precisely what should be measured for the purpose of best indicating problems with self-reported function. We are asking not what the “true” measure of the effects of lower limb dysfunction is, but which of these 2 approaches may better describe the level of function limitation in this population. Given that the Bland-Altman technique was designed specifically for instances of measurement uncertainty (in contrast to situations where there is a true criterion measure), its application in the context of treatment of patients with HIV disease seems particularly apt.
To address Stratford and Riddle's concerns about differences in tool metrics, we point out that the Binkley et al response4 to Gabel and colleagues' similar transformation5 was “there is no provision for converting the LEFS score to a percentage”4(p182) and that this lack of provision was considered by Binkley et al as adequate justification for a critique of Gabel and colleagues' procedure. However, Binkley and colleagues' flat contention does not provide a statistical justification. For the purposes of statistical comparison (as opposed to clinical use), conversion to a 100-point (100%) scale amounts to a simple linear transformation of the LEFS where X=a+bX (where a=0 and b=1/80). This type of transformation does not affect the LEFS distributional properties. We suspect that Binkley and colleagues' (and, similarly, Stratford and Riddle's) contention that LEFS should not be converted to a 100-point scale is for the purposes of clinical interpretation. There is no obvious reason that this clinical limitation should be similarly applied for statistical comparison and interpretation. Thus, we propose that the criticism that “metrics cannot be homogenized” (ie, statistically transforming the LEFS to a 100-point scale) is misplaced.
We asked the question: If LEFS and LLFI are set to the same scale via simple linear transformation, do they provide similar relative measures of the effects of HIV-related DSP on self-reported function? As we have already agreed that LLFI and LEFS are likely measuring somewhat different aspects of the negative effects of lower limb dysfunction, the measures need only be rendered statistically comparable in terms of their relative assessment of the effects of the HIV-related DSP to determine how the 2 constructs might perform differently. To this end, we found that—relative to the total possible range of each tool—the LEFS tends to rate patients with HIV-related DSP higher (proportionally) on the scale than the LLFI. Thus, the LLFI may be more likely to identify some patients with lower extremity functional limitations. Also, given that the disagreement between the 2 tools is more marked in patients with DSP (ie, greater differences between the 2 tools when the scaled LEFS is greater than LLFI), our findings suggest that the LEFS may be less likely to suggest a need for treatment, particularly among those patients with HIV-related DSP who may benefit from physical therapy.
We thank Stratford and Riddle for pointing out an unfortunate ambiguity in our figures that were posted online ahead of print. Corrections have been made and will appear in the printed version. To summarize, the LEFS values listed in the Table titled “LLFI and LEFS Scores by Foot DSP Diagnosis” are raw scores (ie, not scaled to a percentage score). The score ranges that appear in that table have been corrected so that they now list the raw score range. Corresponding text in the “Results” section of the text also was corrected. In the Bland-Altman analysis and related figures, we used the scaled LEFS score (percentage score). Thus, the values in the figures that Stratford and Riddle commented on were accurate. We since modified the labels of the graph axes to clarify that the calculated difference was based on “Scaled LEFS − LLFI.”
In sum, we thank Stratford and Riddle for their insightful comments. They have, we believe, highlighted precisely what our study does not do. Specifically, this is not a study of diagnostic accuracy or psychometric consistency. Indeed, given the likelihood that LLFI and LEFS are measuring somewhat different constructs, a diagnostic accuracy study would be inappropriate. Rather, ours is a study of one aspect of the relative tool performance in a particular population (patients with HIV disease, with and without DSP). As the purpose of this study was descriptive (to provide information on this particular population previously unavailable), we suggest that our findings regarding tool performance, while meaningful, are only suggestive. A different study design would be required to determine whether, in actual practice, use of the LLFI (versus LEFS) is more appropriate for indicating physical therapist intervention for patients with HIV disease. Our findings suggest that such a study is warranted. With respect for the methodological concerns discussed herein, we close by emphasizing the most important clinical bottom line of our study: patients with HIV-related DSP have lower quality-of-life scores (physical health summary) and lower self-reported function scores, as determined by either the LEFS or the LLFI, than patients with HIV disease but not DSP. The greatest impact of our study is that we have added to the limited body of literature that describes the negative impact of DSP in patients with HIV disease.
Footnotes
This letter was posted as a Rapid Response on July 31, 2014. at ptjournal.apta.org.
- © 2014 American Physical Therapy Association