Abstract
The determination of clinically significant cognitive change across time is an important issue in neuropsychology, and repeated assessments are common with older adults. Regression-based prediction formulas, which use initial test performance and demographic variables to predict follow-up test performance, have been utilized with patient and healthy control samples. Comparisons between predicted and observed follow-up performances can assist clinicians in determining the significance of change in the individual patient. In the current study, multiple regression-based prediction equations for the five Indexes and Total Score of the RBANS were developed for a sample of 146 community-dwelling older adults across a 2-year interval. These algorithms were then validated on a separate elderly sample (n = 145). Minimal differences were present between Observed and Predicted follow-up scores in the validation sample, suggesting that the prediction formulas are clinically useful for practitioners who assess older adults. A case example is presented that illustrates how the algorithms can be used clinically.
ACKNOWLEDGMENTS
The authors would like to extend their thanks to the Department of Family Medicine at the University of Oklahoma Health Sciences Center for their continued support of this project, and especially to Michelle Roberts for her assistance with data collection.
Notes
Unless otherwise noted, values represent means and standard deviations (in parentheses). Age is in years and Retest Interval is in days. Index scores were age-corrected scaled scores based on the RBANS manual.
All F tests are significant at p < .001. Index scores are age-corrected scale scores. a Standard error of the estimate, bUnstandardized beta weights for variables in the equation, including Time 1 Indexes and demographic variables. Refer to Method section for coding of education (Ed). Interval is retest interval in number of days.
Index scores are age-corrected standardized scores. Difference represents Observed Time 3 score minus Predicted Time 3 score. Superscripts represent degrees of freedom for t-tests (a = 142, b = 143, c = 141). * = p < .01. r = Pearson correlation between Observed Time 3 and Predicted Time 3 scores. d = effect size for difference between Observed Time 3 and Predicted Time 3 scores.
Difference Observed represents Observed Time 3 minus Observed Time 1. Predicted Time 3 scores are based on regression formulas presented in Table 2. Difference Predicted represents Observed Time 3 minus Predicted Time 3. SEest = standard error of the estimate from the regression equations of the Development sample. z = Difference Predicted ÷ SEest. *p < .05.
First published online: August 20, 2007.