Drug and alcohol use in Canada: General population (ages 15+)

Data source

This dashboard features data from the Canadian Substance Use Survey (CSUS), covering the 2023 cycle to the most recent 2025 cycle. The table below shows the overall sample size for each cycle, including both unweighted and weighted numbers.

CSUS has evolved over time. Details on methodological changes can be found in the technical notes from the 2023 CSUS. The 2025 iteration of CSUS followed the same methodology for respondent recruitment as in 2023.

The previous CSUS 2023 dashboard and an older blog on alcohol use by the general population have been archived.

Table 1: Overall sample sizes of each cycle of the Canadian Substance Use Survey (CSUS)

Data notes

• Due to rounding, percentages may not add up to exactly 100.
• Data were suppressed if the denominators were < 30 based on Statistics Canada guidelines using survey design effects. Due to small sample sizes, certain combinations of indicator categories and data breakdowns are not presented in the dashboard.
• The confidence intervals shown in the text descriptions were calculated using the modified Wilson method, with 500 bootstrap weights.
• CSUS uses random digit dialling (RDD) to recruit respondents. Because no comprehensive list links telephone numbers to individuals or households, RDD provides no information about non-respondents. This means standard non-response adjustments cannot be applied. This produces artificially narrow confidence intervals that may overstate precision and increase susceptibility to flagging differences as statistically significant when none exist in the population. To address this, we recommend considering effect size and corroborating evidence before drawing conclusions about differences.
• For continual improvement, some changes were made to the questionnaire between 2023 and 2025 to ensure continued relevance and reduce stigma. These changes included reordering of questions, separating substances that were grouped together, and adding clarifying plain-language text. Where question changed between years, differences should be interpreted with caution as changes in results may reflect better data capture, rather than true changes in behaviours over time.