Confirmatory factor analysis of the Autonomy over Tobacco Scale (AUTOS) in adults

Abstract

INTRODUCTION: The Autonomy over Tobacco Scale (AUTOS), a 12-item self-administered questionnaire, was designed to measure autonomy in three correlated lower-order symptom domains: withdrawal, psychological dependence, and cue-induced craving. The factor structure of the AUTOS remains an open question; confirmatory analyses in adolescents supported the hierarchical structure, while exploratory analyses in adolescents and adults yield single-factor solutions. Here we seek to determine whether the hypothesized hierarchical structure is valid in adult smokers. METHODS: The AUTOS was administered to two independent convenience samples of adult current smokers: a calibration sample recruited in the US for online studies, and a confirmation sample drawn from the prospective Nicotine Dependence in Teens study in Montreal. We tested competing hierarchical and single-factor models using the robust weighted least-squares (WLSMV) estimation method. RESULTS: A single-factor model that allowed correlated error variances between theoretically related items fit well in the calibration sample (n = 434), chi(2)SB(52) = 165.71; chi(2)/df = 3.19; SRMR = 0.03; CFI = 0.96; NNFI = 0.95; RMSEA = 0.07 (95% CI: 0.06, 0.08). Reliability of the single factor was high (omegaB = 0.92) and construct validity was adequate. In the confirmation sample (n = 335), a similar model fit well: chi(2)SB(53) = 126.94; chi(2)/df = 2.44; SRMR = 0.04; CFI = 0.95; NNFI = 0.93; RMSEA = 0.07 (95% CI: 0.05, 0.08). Reliability of the single factor was again high (omegaB = 0.92) and construct validity was adequate. CONCLUSION: The AUTOS is unidimensional in adult smokers.