Ultra-rapid screening for substance-use disorders: the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST-Lite).

Abstract

The Alcohol, Smoking and Substance Involvement Screening Test (ASSIST 3.0; index test) is a structured interview for alcohol, tobacco, cannabis, stimulants, sedatives and opioid use disorders in general medical settings. Perceived administration time deters routine use. This study releases a short-form: the ASSIST-Lite. Diagnostic accuracy study among 2082 adults recruited from general medical (70%) and specialist mental health/addiction treatment services (22%). Current DSM-IV substance dependence (MINI International Neuropsychiatric Interview) and moderate-severe tobacco dependence (Fagerstrom Nicotine Dependence Test) were reference standards. Exploratory factor and item-response theory models re-calibrated ordinal test items. Items for the ASSIST-Lite were selected by diagnostic accuracy evaluation (area under the receiver-operating characteristic [AUC] curve [≤0.7]), sensitivity, specificity, positive and negative predictive values [PVP, NVP], kappa, likelihood ratios [LR+, LR-], and clinical utility index [CU+, CU-]). For each substance an item pair was selected (AUC [0.8-1.0], sensitivity [0.8-1.0], specificity [0.7-0.8], PVP [0.8-1.0], NVP [0.7-1.0], kappa [0.5-0.9], LR+ [2.5-5.9], LR- [0.0-0.2], CU+ [0.7-0.9], and CU- [0.5-0.8]). Gender, age and recruitment setting (specialist mental health versus general medical) did not moderate accuracy, with the exception of opioids (AUC <0.7, participants ≥59 years). Male opioid users had more severe substance involvement scores that females (differential item functioning analysis, P=0.00). There was no evidence of differential accuracy between countries (AUC range, 0.8-1.0). The ASSIST-Lite is an ultra-rapid screener which has been optimised for general medical settings. Optionally, a criterion question can be added to capture hazardous drinking, and to capture use of another type of mood-altering substance.