Overall a very small proportion of patients had cannabis use documented, especially compared to prevalence estimates in the general population [12]. This is consistent with other types of behavioural risk factor data, such as alcohol use or smoking, which tend to be missing often or not recorded in a structured or easily extractable way in primary care EMRs [13–16]. Relative to all patients within the Alberta CPCSSN database, patients with a record of cannabis use were more likely to have a chronic condition. These findings may suggest an informed presence bias [17], where patients seeking care more frequently tend to have more opportunities for discussions around current substance use or screening. Further, these findings may indicate the use of cannabis as an alternative, patient-initiated treatment option for chronic conditions such as depression; Rotermann et al. found that cannabis use for some self-defined medical purposes in addition to non-medical purposes was higher among Canadians who reported fair to poor physical or mental health compared to those in better general or mental health [18].

The lack of observed cannabis use records in the EMR data may be due to multiple factors—providers may not ask all patients about cannabis use, patients may be unwilling to disclose it due to persisting stigma around social acceptability, or this information may be recorded in areas of the EMR that CPCSSN is unable to extract from (e.g. narrative notes). Alternatively, perhaps a patient-provider discussion of cannabis use occurred but was not recorded, particularly prior to its legalization [6, 7, 12]. Various text terms were used but without specific information related to amount, frequency, or method, which is necessary to help clinicians evaluate potential drug-drug interactions or to monitor high-risk use. Lastly, there is insufficient information to indicate if the cannabis is medically indicated or used for non-medical purposes. While cannabinoid medications with a Drug Information Number (DIN) will be recorded similarly to other types of medication in the EMR, herbal cannabis does not have a DIN and, until 2018, was subject to additional documentation required by Health Canada and regulatory bodies (which would not be part of the CPCSSN database). After legalization in Canada, patients can now obtain cannabis from retail sources without a prescription from a healthcare provider, although many still report using it for medical purposes; this information would also not be included as medication data in the CPCSSN database unless disclosed by the patient. However, a recent Canadian survey suggests that most cannabis is purchased from either legal storefronts or a legal online source (approximately 54%) and very few respondents reported obtaining a medical document from a healthcare provider (an estimated 3% of Canadians) [19], meaning that the EMR database is likely missing most uses of cannabis apart from instances obtained through medical professionals (which appear to make up a very small proportion of cannabis use) or when elicited and recorded by physicians during clinical visits. The overall lack of cannabis use information accessible in our primary care EMR database presents challenges not only for secondary uses but may also potentially contribute to gaps in clinical care.