The Japanese version of Substance Abuse Subtle Screening Inventory and its psychometric assessment

Abstract

Objective: The purpose of our study was the development of the Japanese version of the Substance Abuse Subtle Screening Inventory (SASSI) and its psychometric assessment.

Method: We translated SASSI-3 into Japanese and refined it through a back-translation technique and focus group approach. Study 1 was a preliminary study to examine the internal consistency and discriminant function of subscales, sensitivity and specificity of the Japanese version SASSI-3 (SASSI-3J). The participants were 20 drug-dependent individuals and 23 non-dependent comparison individuals. Study 2 comprised three parts. The participants were university students recruited from three classes. The first part examined the internal consistency of subscales and the relationship between the subscale Face Valid Alcohol (FVA) and the Kurihama Alcoholism Screening Test (KAST) with 269 students. The second part examined the 2-week test-retest reliability with 51 students. The third part examined the robustness of SASSI-3J against response distortion with 65 students: participants completed the SASSI-3J under standard instructions, under instructions to minimize any alcohol problems they might have (fake good), and under instructions to exaggerate any alcohol problems they might have (fake bad). They also completed the Social Desirability Scale.

Results: 1) SASSI-3J showed relatively high sensitivity and specificity when only the Risk Prediction Scales were used or when Rule 9 was excluded from the decision tree; 2) The Risk Prediction Scales of SASSI-3J had high internal consistency, and among them, Face Valid Alcohol (FVA) showed moderately high concurrent validity with the KAST; 3) In both Study 1 and Study 2, the internal consistencies of OAT, SAT, DEF and SAM were moderate to low (0.05 to 0.57). This may not interfere with the subscales' discriminant power but does deny sufficient uni-dimensionality of these scales; 4) SASSI-3J clinical scale scores were shown to be robust against faking good/social desirability bias.