Data analysis

Data were analyzed using SAS v9.4.^a Descriptive analysis statistics for independent variables included in the models were generated. These items included amputation level and primary prosthesis type because the authors’ expert opinions indicated that these factors would be most influential in evaluation of function. Age and sex were also included as independent variables. The occurrence of possible multicollinearity between all independent variables was assessed using the Pearson correlation coefficient. We also examined the variance inflation factor and tolerance for each of the CAPPFUL and logBBT models.18^,19 No high correlation coefficients between independent variables were observed; a moderate correlation between amputation level and prosthesis type was observed (r=0.59). The minimum tolerance and maximum variance inflation factor for the CAPPFUL model were 0.30 and 3.34, respectively. The minimum tolerance and maximum variance inflation factor for the logBBT model were 0.23 and 4.17, respectively. These values indicate that there is no multicollinearity between independent variables, thus satisfying the assumption of multiple linear regression analyses. DASH, CAPPFUL, BBT scores, and age were treated as continuous variables and mean values for each item were calculated (table 2). To estimate the unadjusted association between actual function and perceived function of the selected independent variables, we first conducted simple linear regression analyses. CAPPFUL score or BBT score, representing the actual function, was respectively treated as the dependent variable in the simple linear regression analyses. In these analyses, a number of categories were collapsed within an independent variable to make the results more interpretable and to account for low sample size in a given category. Categories within the amputation level were grouped as follows: amputation level 1 includes digit(s)/fingers and partial hand amputations, amputation level 2 includes wrist disarticulation and transradial (below elbow), and amputation level 3 includes elbow disarticulation, transhumeral (above elbow), and shoulder disarticulation. Categories within prosthesis type were grouped as follows: type 1 includes electrically powered; type 2 includes body-powered; type 3 includes hybrid; and type 4 includes passive and passive positionable. Only variables significantly associated with change in CAPPFUL or logBBT scores in the simple regression analyses were used in the multiple linear regression analysis.

Mean, SD, SE, and 95%CI of Continuous Variables

Two separate multiple linear regression analyses were then conducted to examine the adjusted association between CAPPFUL or BBT and DASH, amputation level, and prosthesis type. The same collapsed independent variables described previously were also used here. To avoid violation of linear regression assumptions, BBT scores were log transformed and residual analyses were applied.

A total of 61 participants were included in this study, but not all participants completed every outcome measure. In the linear regression analyses, participants without complete information for dependent variables and independent variables were excluded from the models. Details of the number of participants in the subgroup (n) can be found in the regression analysis (Table 3, Table 4, Table 5, Table 6).

Simple Linear Regression Models With CAPPFUL as Dependent Variable

Multiple Linear Regression Models With CAPPFUL as the Dependent Variable (n=32)

Simple Linear Regression Models With logBBT as the Dependent Variable

Multiple Linear Regression Models With logBBT as the Dependent Variable (n=31)