Concurrent and retrospective reports of alcohol consumption across 30, 90 and 366 days: interactive voice response compared with the timeline follow back.

Objective: To compare near-concurrent and retrospective reports of alcohol consumption and to determine whether a simple adjustment function could be applied to the retrospective reports to correct any distortions in reporting.

Method: Reports of daily alcohol consumption collected via an Interactive Voice Response (IVR) system were aggregated over 30, 90 and 366 days and compared with the corresponding retrospective reports gathered with the Timeline Follow Back (TLFB). Comparisons of the IVR and TLFB reports for five consumption variables were made for each time period for 33 men.

Results: Significant differences between reporting methods were found for four of the five variables tested. On average, the TLFB yielded an underestimate of alcohol consumption compared to the IVR. This underestimate associated with TLFB reports was not dependent on the length of the recall period. Regression analyses indicated that the relationship between the two modalities of reporting was dependent on the quantity of drinks reported, thus precluding the application of a simple, linear metric to the TLFB data to model near real-time daily data obtained through the IVR more accurately.

Conclusions: The retrospective TLFB consistently and significantly underestimated alcohol consumption compared to the aggregated daily reports. This underestimate, however, was stable across the three reporting periods. With one exception, no simple linear function could be applied to the retrospective TLFB consumption reports to more accurately reflect the IVR data gathered daily. Evidence presented here also suggests that underreporting on the TLFB is not simply a matter of consciously attempting to conceal consumption.