Methods: Data and sample: NESARC-III is a cross-sectional dataset containing national data of individuals, aged 18 years and older living in the United States. The final NESARC-III dataset contained 36,309 participants. This study sampled participants, aged 18- to -25-years-old, that reported past-year college enrollment. The final sample of this study contained 1,230 individuals.
Measurement development and analysis: The following steps were taken in the scale development process: systematic literature review, preliminary development of item pool, expert review of initial scale, and exploratory factor analysis (EFA). Based on thematic findings from the systematic review and pre-existing measures of these risk factors (i.e., PES-M, PFC-B, and CFCS), an initial item pool of questions was created from NESARC-III survey questions that aligned with the alternative rewards, commitment and consistency/congruence, and delay discounting constructs. EFA was used to examine whether the selected NESARC-III questions adequately load measures of alternative rewards, commitment and consistency, and delay discounting.
Results: The alternative reward variables reflected an index and not a scale. Scree plots indicated one-factor models were a good fit for commitment and consistency and delay discounting. The rotated factor loadings were as follows: delay discounting ranged from .461 to .768 (.461, .646, .735, .768) and congruence ranged from .562 to .678 (.562, .568, .642, .654, .678). Communality scores were as follows: delay discounting ranged from .339 to .657 (.339, .511, .609, .657) and commitment and consistency ranged from .346 to .580 (.346, .446, .513, .514, .580. Correlations matrix revealed variables with 11% (n=1) of minimal importance (±.30), 22% (n=2) of importance (±.40), and 67% (n=6) with practical significance (±.50).
Implications: Findings suggest creating reliable and valid measures using pre-existing data is feasible. : This paper describes the benefits and challenges associated with developing measures using national secondary datasets. These findings may inform future studies hoping to study constructs not initially measured in preexisting secondary datasets.