Abstract
INTRODUCTION AND AIMS: Recall bias is a concern in self-reported alcohol consumption, potentially accounting for varying risk estimates for injury in emergency department (ED) studies. The likelihood of reporting drinking for the same 6-h period each day of the week for a full week preceding the injury event is analysed among injured ED patients. DESIGN AND METHODS: Probability samples of patients 18 years old and older were interviewed in two ED sites in Vancouver and one in Victoria, BC (n = 1191). Generalized estimating equation modelling was used to predict the likelihood of reporting drinking for the same 6-h period prior to the injury event for each day of the week, compared to day 7 as the reference recall day, for a full week preceding the event. Recall by frequency of drinking and frequency of heavy drinking was analysed. RESULTS: Drinking was significantly more likely to be reported for each of the first 3 days of recall compared to 7-day recall and highest for 1-day recall (odds ration 1.55; = 0.002). Patients who reported ≥ weekly drinking and 5+ drinking < monthly were significantly more likely to report drinking for each of the first 3 days of recall (compared to 7-day recall). DISCUSSION: Findings suggest the first 3 days prior to injury may be a less biased multiple-matched control period than longer periods of recall in case-crossover studies. CONCLUSION: Length of accurate recall may be important to consider in case-crossover analysis and other study designs that rely on patient self-report such as the Timeline Followback. [Cherpitel CJ, Ye Y, Stockwell T, Vallance K, Chow C. Recall bias across 7 days in self-reported alcohol consumption prior to injury among emergency department patients.