Test characteristics of a drug CAGE questionnaire for the detection of non-alcohol substance use disorders in trauma inpatients.

Abstract

Non-alcohol substance use disorders (drug use disorders [DUDs]) are common in trauma patients. To determine the test characteristics of a 4-item drug CAGE questionnaire to detect DUDs in a cohort of adult trauma inpatients. Observational cross-sectional cohort of 1,115 adult patients admitted directly to a level-one trauma center between September, 1994 and November, 1996. All participants underwent both a 4-item drug CAGE questionnaire and the substance use disorder section of a structured psychiatric diagnostic clinical interview (SCID) (DSM-IIIR criteria), administered by staff unaware of their clinical status. Sensitivity, specificity, positive (PPV) and negative predictive values (NPV), positive (LR+) and negative likelihood ratios (LR-), and the area under the receiver operating curve (AUC) were calculated for each individual question and the overall questionnaire, using SCID-generated DUD diagnoses as the standard. Performance characteristics of the screen were also compared across selected sociodemographic, injury mechanism, and diagnostic sub-groups. Subjects with DUDs were common (n = 349, 31.3%), including cannabis (n = 203, 18.2%), cocaine (n = 199, 17.8%), and opioids (n = 156, 14.0%). The screen performed well overall (AUC = 0.90, 95% CI: 0.88-0.91) and across subgroups based on age, sex, race, marriage status, income, education, employment status, mechanism of injury, and current/past DUD status (AUCs 0.75-1.00). Answering any one question in the affirmative had a sensitivity = 83.4% (95% CI: 79.1-87.1), specificity = 92.3% (95% CI: 90.2-94.1), PPV = 83.1%, LR+ = 10.8. The 4-item drug CAGE and its individual questions had good-to-excellent ability to detect DUDs in this adult trauma inpatient population, suggesting its usefulness as a screening tool.