Longitudinal comparison of ED admissions for alcohol-related trauma and self-reported alcohol-related injury in a college population

Background/Purpose: Self-reported survey data is the most frequently used and cost-effective method for tracking health behaviors and outcomes in a university population. However, the validity of such data has repeatedly been called into question when the respondent is asked to self-report potentially damaging information. Using information collected at a large public university, we examine the correlation between Emergency Department (ED) data and data from annual random sample student surveys to examine the convergent validity of the self-reported information on alcohol-related injury over a seven year period.

Methods: The Pearson Correlation Coefficient was computed for the total sample, as well as demographic subsamples by sex, age category (<=8, 19, 20, >=21) and Caucasian students (numbers are too low for meaningful comparison for other race/ethnic groups).

Results/Outcome: Over the seven year period, 585 of 1,164 ED admissions for alcohol-related reasons were due to injury (50%); during the same time period, 1,932 of 16,373 survey respondents reported alcohol-related injury (12%). Both ED admissions for alcohol-related trauma and self-reported alcohol-related injury decreased significantly over the years (from 60% to 46% for ED reports and from 24% to 9% for self reports) and are highly correlated: ñ=.70, 48% shared variance. A strong correlation is evident for males (ñ=.68, shared variance=46%), for 19 year olds (ñ=.74, shared variance=54%), 20 year olds (ñ=.76, shared variance=58%) and for Caucasian students (ñ=.74, shared variance=54%).

Conclusions: The shared longitudinal pattern of ED data with self-report data provides an important validity check on the self-report data.