First of all, almost all self-report measures assess drug consumption retrospectively, it is just a question of the timescale (days, weeks; you want years). Based on this logic, you could use any establieshed instrument.

Article The Kreek-McHugh-Schluger-Kellogg scale: A new, rapid method...

Of course, the longer the time period and the more regular and vague an activity the less respondents use a recall-and-count approach and more of a heuristic guessing approach. This is problematic, because it means that the accuracy of the reported frequency varies as a function of the frequency of use.

Second, keep in mind that most instruments only assess a few drugs, mainstream drugs (e.g., caffeine, sugar) and very new drugs (e.g., Azstarys) are usually not considered.

Are you concerned with whether the respondent will answer honestly? If so, you could research the "random response method." It is a way of collecting the data where it will not be known for sure which people answered "yes" because they really did use drugs illegally. Perhaps you know this. I just wanted to be certain.

(See "random response method," "random response technique," "random response sampling," "random response," "randomized response.")