|
Abstract:
|
The National Survey on Drug Use and Health (NSDUH) provides annual data on alcohol use, tobacco use, illicit drug use, substance use disorders, mental health, receipt of services for behavioral health conditions, and other related measures of interest (e.g., risk perceptions related to substance use). When analyzing cross-sectional survey data such as NSDUH's, the "trend" depicts the general underlying pattern of change of an outcome variable over time in a finite population. This presentation will summarize the current methods used for analyzing trends in NSDUH and the trend testing methods used by other federal agencies. A variety of methods for trend analysis has been used, such as pairwise testing and statistical regression. However, no formal guidelines have been available for NSDUH and other studies on how to select an appropriate method in terms of fitness for use under certain constraints. Although the literature contains alternative methods, open questions remain about inclusion of survey design features into complex regression and time series models, and the sensitivity of analyses to outliers, moving time windows, and the availability of new data each year.
|
