Statistical methods in rehabilitation literature: A survey of recent publications

Abstract

Objective: To document the use of statistical methods in the recent rehabilitation research literature. Design: Descriptive survey study. Methods: All 1,039 articles published between January 1990 and December 1993 in the American Journal of Physical Medicine and Rehabilitation and the Archives of Physical Medicine and Rehabilitation were reviewed for the use of statistical methods. Results: There were 682 (66%) research articles in the sample that included systematic data collection and analysis. The most frequently encountered analytic tests and techniques included: (1) analysis of variance, (2) t tests, (3) correlation analysis, (4) contingency table analysis, (5) regression, and (6) non-parametric tests. Results were compared with results of a review of the 1982 rehabilitation literature by Wainapel and Kayne. Our study showed an increased sophistication in statistical methodology as well as a more intense use of such methods. In addition, there was a large number of relatively obscure and poorly documented tests encountered in the more recent literature. There was also a lack of adherence to a standardized format for describing statistical methods. Conclusions: The following recommendations are made: (1) Training curricula for rehabilitation professionals should provide instruction in the most commonly-encountered statistical methods and should be revised as needed to reflect changes in statistical method usage. (2) When less common statistical tests are applied, the responsibility of the authors to fully describe and justify their methods should be recognized. (3) Critical assessment of the literature is facilitated when statistical methods are reported in a standardized format.