Psychometrics of the self-efficacy for physical activity scale among a Latina women sample

BACKGROUND: Even though Latinos have become a priority population for the promotion of physical activity in the United States, several widely used scales in physical activity promotion research have not been validated among this population, particularly in Spanish. This study aims to assess the validity and other psychometrics of the Self-Efficacy for Physical Activity scale among a sample of Spanish-speaking Latina women who participated in the Pasos Hacia La Salud intervention. We also explored alternatives for scale simplification. METHODS: Data from 205 women corresponding to baseline, 6-month, and 12-month time points were analyzed. Internal consistency was assessed. A series of Spearman correlations, t-tests, linear regressions, and logistic regressions were used to assess the concurrent and predictive validity of the Self Efficacy for Physical Activity scale against both self-report and accelerometer-measured physical activity, using both continuous and categorical outcome data. Item Response Theory and factor analysis methods were used to explore alternatives to simplify the scale. Psychometric tests were repeated with the simplified scale. RESULTS: Cronbach’s alpha for the original scale was .72, .76, and .78 for baseline, 6-month, and 12-month data respectively. All concurrent validity tests conducted with 6-month and 12-month data, but not with baseline data, were statistically significant. Self-efficacy at 6 months was also predictive of physical activity at 12 months for all tests except one. Based on plots of Option Characteristic Curves, a modified version of the scale was created. Psychometric results of the modified scale were similar to those of the original scale. CONCLUSIONS: This study confirmed the scale’s reliability and validity, and revealed that the scale’s accuracy improves when some response items are collapsed, which is an important finding for future research among populations with low literacy levels.