Predicting Function.

To the Editor:

In the article by Beissner et al entitled "Muscle Force and Range of Motion as Predictors of Function in Older Adults" (June 2000), the use of the words "predict" and "predictor" is somewhat puzzling. To predict is to say something about an occurrence in advance of the actual occurrence. The authors of this study have not demonstrated that muscle force and range of motion are predictors, because all their measurements were done at the same time (actually a week apart, according to the report). They have demonstrated a relationship, but in order to demonstrate prediction, the predicting phenomena have to be measured in advance of the object of prediction.

For what I believe is a good study of this type, see the article by Rantanen et al entitled "Midlife Hand Grip Strength as a Predictor of Old Age Disability"[1] in the Journal of the American Medical Association. The grip strength measurements were done 21 years or more in advance of the disability measurements. Using this type of study, we can really say something meaningful to our patients about exercise and the future quality of their lives.

Stephen W Werner, PT 2316 Lawndale Ave Evanston, IL 60201

References

[1] Rantanen T Guralnik JM, Foley D, et al. Midlife hand grip strength as a predictor of old age disability. JAMA. 1999;281:558-560.

Author Response:

Mr Werner raises an excellent point regarding the term "predictor," which may lead to some confusion on the part of readers. We used the term "predictor" in the statistical sense of the word. In regression analyses, a set of independent variables is used to "predict" or estimate an outcome. From the results, it is possible to identify which of the independent variables (eg, upper- or lower-extremity muscle force, upper- or lower-extremity range of motion) are most important in predicting the outcome (function score). Furthermore, the regression results allow us to estimate how much of an improvement in function we would expect from a given improvement in the independent variables.

As we noted in the "Discussion" section of our article, one of the limitations of our research is that the data were cross-sectional. However, a cross-sectional design using regression analyses is, in our view, an efficient and practical approach that can be used to generate estimates that are useful in the context of what we are studying. True "predictions" of future states are, as Mr Werner notes, best based on longitudinal data. Likewise, judgments about the value of interventions such as exercise are best made by directly studying these variables. The growing body of research regarding the functional implications of particular exercise regimens will serve the profession well in promoting exercise as a means to a higher quality of life.

Katherine Beissner, PT, PhD Professor and Chair Department of Physical Therapy Ithaca College Ithaca, NY 14850 ([e-mail])

Heidi Holmes, MA Project Manager Pathways to Life Quality Study Gerontology Institute Ithaca College