Improving Alcohol Consumption Self-Report Measurement

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): A methodological study is proposed to investigate sources and kinds of errors in self-report measures of alcohol consumption: specifically, retrospective summary drinking pattern measures for use in epidemiological and services research. We hypothesize that features of the "true" drinking pattern (defined as mean and variability of quantities per occasion, frequency of drinking, and beverage choices), will affect both random and systematic error (bias) in standard measures of consumption, reducing their reliability and validity. Beyond average volume, drinking above a 5-drink/day-quantity threshold (a common pattem indicator) has been implicated in risks of health outcomes of alcohol abuse: social consequences, morbidity, and mortality. Three distinct types of self-report summary drinking measures are selected for study: the beverage-specific Knupfer Series, the Graduated Frequencies measures--both assessing pattem and volume of drinking--and the simple quantity-frequency (QF) measure. These standard measures appear together in several Alcohol Research Group (ARG) national alcohol telephone and in-person surveys, allowing within-subjects comparisons. In the first phase of the research, a diary pilot phase, we will refine a new diary format including innovative serving-size and ethanol-content assessment techniques developed in the ARG Center. Sensitization effects in the retrospective summaries from the preceding 4-week diary series will also be investigated and a noninvasive physiological measure (a wrist-worn transdermal alcohol sensor, or TAS) will be used to validate the diary self reports. In the main data collection phase (n = 500), a screening strategy assures adequate inclusion of frequent drinkers. Various statistics of the drinking distribution based on each summary measure will be compared with similar statistics dedved from by-occasion reports in daily diaries. TheTAS again will be used for 300 cases to model errors of self-report and study estimated blood alcohol concentrations (EBACs). Random effects and Bayesian approaches will be used to analyze events, comparing physiological-and self-report-based estimates within individuals. Modeling of errors will include covariates like age, sex, ethnic group, and education, plus context of drinking and attitude toward drunkenness, hypothesized to bias and degrade summary reports more than diaries, with TAS-data assumed free of self-report biases. The study will advance theory about measurement error in surveys of complex human health behaviors; it is designed to generate applied findings that will aid calibration of existing survey measures and help develop better measures of alcohol-intake pattern essential for improvement of epidemiological, clinical and services research studies.