Population-Level Administration of AlcoholEdu for College: An ARIMA Time-Series Analysis

Abstract

Autoregressive integrated moving averages (ARIMA) is a powerful analytic tool for conducting interrupted time-series analysis, yet it is rarely used in studies of public health campaigns or programs. This study demonstrated the use of ARIMA to assess AlcoholEdu for College, an online alcohol education course for first-year students, and other health and safety programs introduced at a moderate-size public university in the South. From 1992 to 2009, the university administered annual Core Alcohol and Drug Surveys to samples of undergraduates (Ns = 498 to 1032). AlcoholEdu and other health and safety programs that began during the study period were assessed through a series of quasi-experimental ARIMA analyses. Implementation of AlcoholEdu in 2004 was significantly associated with substantial decreases in alcohol consumption and alcohol- or drug-related negative consequences. These improvements were sustained over time as succeeding first-year classes took the course. Previous studies have shown that AlcoholEdu has an initial positive effect on students' alcohol use and associated negative consequences. This investigation suggests that these positive changes may be sustainable over time through yearly implementation of the course with first-year students. ARIMA time-series analysis holds great promise for investigating the effect of program and policy interventions to address alcohol- and drug-related problems on campus.

Notes

Note. Figures in parentheses indicate data obtained from the university's office of institutional research and not from the Core Surveys because of missing data. N/A = data not available.

^a The difference between the Core Survey sample and the entire university student body is statistically significant (O², p < .05).

Note. The type of ARIMA model used is indicated by the number of autoregressive terms (p), the number of nonseasonal differences (d), and the number of lagged forecast errors in the prediction equation (q). Estimate = estimated value of the intervention parameter; SE = standard error of the intervention estimate; NT = no transformation—i.e., implementation of AlcoholEdu did not cause a significant transformation in the ARIMA model. ARIMA = autoregressive integrated moving average.