You are on page 1of 1

EDITORIAL

Reliability
Clinical measurements lie at the core of our prac- repeated. A perfect correlation would be an exact rep-
tice. We use them constantly to aid in diagnosis, etition of the same score both times the measure is
assess severity of impairment, monitor response to taken. In most formulas used to express this correla-
treatment, and ultimately guide our clinical deci- tion, such a perfect correlation would be 1.0. The clos-
sions. Crucial as our measurements are, they must be er to 1.0 a correlation is, the more reliable and hence
accurate. As the one and only Erle Peacock, MO, the more accurate the measure. For example, a 0.95
pointed out some years ago in this journal, "The correlation is far better than a 0.50 correlation. There
greatest advance in hand therapy in the last few are no absolute values, or cut-off points, assigned to
decades and certainly one of the greatest contribu- correlation scores to render them excellent, good, fair,
tions to the care of patients has been the introduction or poor. It is more a matter of a gradient of scores,
of the science of measurement." 1 where the higher the better, and the lower the poorer.
And at the very heart of the concept of accuracy is Another question that needs to be asked in any reli-
the concept of reliability. Webster defines reliability ability situation is whether we are talking about the
as, "suitable or fit to be relied on, or dependable." same person taking the measure more than once or
This is how we all understand it in the context of about different people taking the measure on different
everyday life. In our clinical settings, we need a con- occasions. Clearly, both these situations may arise clin-
cept of reliability with a slightly different spin to it, ically. If it is always the same person taking the meas-
like that offered by Portney and Watkins in their ure, then this is intra rater or intratester reliability, in con-
research textbook Foundations of Clinical Research: trast to interrater or intertester reliability, when we are
"Reliability can be conceptualized as reproducibility comparing measures taken by different people.
or dependability, and is the extent to which a meas- Many practitioners (myself included) are outspo-
ure is consistent and free from error." 2 When applied ken advocates of avoiding, whenever possible, the
to a clinical measurement, it becomes an important sharing of responsibility for repeated measures with
indicator of the accuracy of that measurement. our colleagues. We strongly recommend that one
Despite the obvious fact that outcomes deserve our person alone follow the patient and take all meas-
attention,* we cannot forget about the important ele- ures. One reason for this argument is that, typically,
ments that contribute to function, such as ROM. I can intrarater reliability is greater than interrater reliabil-
think of no area of our practice where accuracy is more ity. Therefore, if accuracy is the legitimate concern
critical than in measuring joint motion. Small changes that Peacock suggests, we should opt for the more
in motion trigger our clinical decisions to alter plans of reliable method. This is especially true if we are
treatment in the management of joint stiffness, for going to make clinical decisions based on these meas-
example. Such decisions require us to have confidence ures. Given that bias, I am usually more interested in
in the accuracy of our measures. the intrarater reliability value.
To comply with Peacock's description of our prac- I hope to see many more such manuscripts, so that
tice being based on science, we cannot casually in future issues of the JHT we can continue to docu-
assume that our methods-and in this case I am ment the accuracy of our methods.
referring to goniometry-are accurate. We need doc-
umented evidence. In this issue, we are offered four
reliability studies that contribute to that body of evi-
dence, three of which are goniometric investigations.
As a general statement, reliability is expressed sta-
tistically in the form of a correlation coefficient of References
some sort. In lay terms, this means a mathematical
1. Peacock EE Jr. Frontiers in wound healing research that may
representation that indicates how closely a measure affect the work of hand therapists. J Hand Ther. 1989;2:221-2.
comes to being duplicated when the measurement is 2. Portney LG, Watkins M. Foundations of Clinical Research. 2nd
ed. Upper Saddle River, NJ: Prentice Hall, 2000.

'Because of the increasing importance of outcomes as indicators of


the efficacy of hand therapy practice, this year's Special Issue of the Correspondence to Kenneth R. Flowers, PT, CHT, Editor, Journal
Journal (April-June 2001) will be dedicated to Outcomes and guest of Hand Therapy, 270 Danell Road, Radnor, PA 19087; e-mail: <ken.
edited by Joy c. MacDermid, BScPT, PhD. flowers@prodigy.net>.

January-March 200 I

You might also like