Drawing the line on risky use of cannabis: Assessing problematic use with the ASSIST

Abstract

Health and social harms from cannabis use typically are assessed by comparing those who use to those who do not use. Recognizing that not all use of cannabis is necessarily problematic, we examine rates of self-reported harms as a function of frequency of use. Second, we assess the effectiveness of the cannabis portion of the Alcohol, Smoking and Substance Involvement Screening Test (ASSIST) as a screening tool for identifying problematic cannabis users. Data come from the Canadian Addiction Survey (CAS; N = 13,909) and the 2006 NWT Addictions Survey (2006 NWTAS; N = 1235). Results from both surveys indicate that harms are most likely among weekly and daily users. Although frequent users are at increased risk of harms, greater balance of sensitivity with specificity is obtained with the ASSIST screening tool using a somewhat higher threshold than what is suggested in clinical applications of the instrument. Implications for this higher threshold for public policy are discussed.

• Cannabis harms
• assessment
• addiction

Notes

Notes

1. The CAS was a collaborative initiative sponsored by Health Canada, the Canadian Centre on Substance Abuse, and the Canadian Executive Council on Addictions–which includes the Alberta Alcohol and Drug Abuse Commission, the Addictions Foundation of Manitoba, the Centre for Addictions and Mental Health, the Prince Edward Island Provincial Health Authority, and the Kaiser Foundation–the Centre for Addictions Research of British Columbia, and the provinces of Nova Scotia, New Brunswick and British Columbia. The 2006 NWT Addictions Survey was conducted by the NWT Bureau of Statistics on behalf of the GNWT Department of Health and Social Services and Health Canada.

2. Due to an error, one question (failed to control use) was not asked correctly and so has been excluded from analyses. This item is useful for distinguishing dependence from abuse. As such, this error is unlikely to significantly affect the results we present, as this symptom is unlikely to distinguish low-risk from moderate-risk users.

3. In some situations, specificity and sensitivity should not be weighted equally, in the sense that the cost of a false negative (i.e., a person indicates harm from drug use, but their score on the ASSIST is below the cut-point) may be greater (or less) than the cost of a false positive, such as in a disease context, where the cost of missing cues to a disease may be more costly than a false positive reading of a cue. In our context, we have assumed that false positives are as costly as false negatives. We also conducted the ROC separately for those aged 15–24 years and those aged 25+years to establish whether the same criteria applied to both young and older users. Although youth (age 15–24 years) were more likely to report harms than older participants (19.3% vs. 10.3%), areas under the ROC curves were very similar (79.6% vs. 81.6%) and both analyses indicated that the best balance between sensitivity and specificity is obtained with a cut-off of eight.

4. Note that the population estimates for weekly/daily users and for monthly users who report harms has moderate sampling variability, so population estimates should be interpreted cautiously.

5. We thank an anonymous reviewer for pointing out this possibility.