This study develops improved data for, and then conducts, a sequence of time-series, fixed and random effects analyses of alcohol-related mortality by cause for the US, individual states, and states grouped into five regions. An innovative feature is the use of more accurately estimated ethanol conversion factors specific to year and state to estimate per capita ethanol consumption from reported wine, beer and spirits sales data. Time series predictors include both these improved estimates of per capita ethanol consumption over a 50-year time span and, for several aims, detailed drinking pattern measures from three population survey series extending over approximately 20 years. Survey data include 5 waves (1979-2000) of the National Alcohol Survey, and the annual state-representative Behavioral Risk Factor Surveillance System surveys from 1984 to present and the National Household Survey on Drug Abuse surveys, 1985 to 2000. Outcome measures include age-standardized and specific mortality rates for liver cirrhosis, pancreatitis, other alcohol related cancers, heart disease (IHD), all explicitly alcohol-related causes and sub-categories such as psychosis, dependence and polyneuropathy, all external causes and the subcategories of suicide, homicide, all accidents and motor vehicle accidents, and all cause mortality. The study will develop and analyze 50-year time series for most outcomes (a 33-year series for several more specific causes) in Auto-Regressive Integrated Moving-Average (ARIMA) models and in pooled time-series models of the US states. A unique feature of this work is that it combines representative population survey data with aggregate data to investigate the effect of drinking patterns in conjunction with per capita consumption. Pattern measures include rates of current drinking, heavy drinking (ever and frequently consuming 5+ drinks per occasion), and concentration of consumption (e.g. the Gini index). A number of confounding variables such as state smoking rates will be controlled. An important additional aim is to use the survey-based drinking measures to characterize consumption patterns of age-, gender-, and ethnicity-based subgroups, specific risk behaviors that cannot be inferred from aggregate consumption data but are identified in mortality statistics. Accounting for variation both in drinking and mortality rates in these subgroup cohort analyses should provide much greater precision in estimating how alcohol consumption influences mortality and is expected to yield new information on health disparities in, and the time course of, alcohol-related mortality for African American and Hispanic populations.
Showing the most recent 10 out of 16 publications
