Handgrip prediction models for children, adults and the elderly

This paper describes the development of a set of multiple linear regression equations for predicting maximal voluntary static handgrip strength of males and females, 5-89 years of age, in each hand. Equations are presented for predicting handgrip strength of the left and right hands from: (i) overall body descriptive variables alone--age, sex, body weight and stature, (ii) overall descriptive variables plus pinch strength variables, and (iii) pinch strength variables alone. A data set on 182 persons living in the United States was used for this study. Two-thirds of the observations were used to develop the models and the rest for validating them. Models were found with 3-5 variables, R2 ranging from 0.80-0.87, and root mean square error from 4.7-5.8 kg. Among the most common predictor variables that appeared in the models were: stature, age, sex, and the stronger pinches (chuck, lateral and pulp-2). Model validation indicated a greater tendency to over-predict left hand strength but no directional tendency for right hand strength. Also prediction accuracy (mean absolute percentage error) ranged from 14.9-21.0%, with about equal magnitudes in the two hands. The regression model coefficients were refined by recombining the data set of 182 observations and performing the appropriate regressions analysis.