Screening for drinking problems by patient self-report. Even 'safe' levels may indicate a problem.

Background: Physicians often screen their ambulatory patients for serious drinking problems by asking questions related to the quantity of alcohol that they consume. Never previously reported is whether this "quantitative" approach to screening can be used to effectively screen ambulatory patients for the presence of a serious drinking problem.

Methods: The project interviewed 510 patients attending an inner city general medicine practice with the alcohol module of the Diagnostic Interview Schedule, revised for the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition. Collected data also included reported quantity, frequency, and recency of drinking. We then calculated the sensitivity, specificity, positive predictive values, and receiver operating characteristic curve for zero to two, three to five, six to 11, 12 to 23, and 24 or more standard drinks as reported by 155 patients who reported drinking within 30 days of their visit.

Results: Forty-eight of 155 active drinkers met the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition criteria for alcohol dependence or abuse. Only five patients with an active diagnosis failed to report drinking within 30 days of their visit. The calculated area under the receiver operating characteristic curve for reported quantity was 0.81. The sensitivities of reported consumption decline with increasing drinking, while the specificities and positive predictive values rise. The report of drinking between six and 12 drinks per week was associated with a positive predictive value of 0.54 for an active Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition, diagnosis.

Conclusions: Patient self-report of drinking can be used to screen actively drinking outpatients on the general medicine service for serious drinking problems. Further, in an urban general medicine outpatient population, even federally recommended levels of drinking may indicate a problem. Our data suggest that physicians' recommendations be adjusted for the setting in which they practice.