A medicine that is difficult to supply is not necessarily the medicine whose absence would cause the greatest clinical harm. For manufacturers and distributors discussing shortages with health-system partners, those are different questions: how vulnerable is supply, and how important is the medicine if supply fails?
A single disruption score can obscure the distinction. A product with fragile supply but workable alternatives calls for a different conversation from a clinically essential medicine with more stable supply. Neither deserves to disappear behind an overall ranking.
The Canadian approach described in an FIP webinar on pharmaceutical supply-chain data connects supply-chain vulnerability with a separate assessment of clinical importance. The presentation describes expert scoring, linked data and AI-assisted checks for guideline changes, rather than AI replacing clinical judgment. Its most useful lesson for technology owners is how to preserve those distinct signals when turning analysis into priorities.
Separate supply vulnerability from clinical importance
The webinar describes linking 13 data sources to understand real-world utilization, incoming shortage reports, medicine use and drug characteristics. Wholesalers, manufacturers and clinicians contributed to expert review of the approach. The integration matters because supply information and clinical importance answer different questions, even when they concern the same medicine.
Health Canada identifies the underlying tool as part of the University of Toronto-led At-Risk Medicines project. Its vulnerability assessment uses past shortage reports, supply-chain factors and market dynamics. Clinical importance is assessed by clinicians against nine criteria. Vulnerability concerns the possibility of supply failure; clinical importance concerns what is at stake if the medicine becomes unavailable.
Keeping the assessments visible allows a decision-maker to distinguish a vulnerable medicine needing active supply mitigation from a clinically important medicine needing protection even when its supply appears less exposed. When both assessments are high, the reasons for attention reinforce each other. When they differ, that difference is useful information, not a defect to be averaged away.
The same separation should apply to consequences. The webinar distinguishes clinical impact, health-system impact and cost, including the work involved in switching patients to alternatives. Operational disruption is not itself a measure of observed patient harm. A large switching workload and a deterioration in clinical outcomes are different findings and should remain separate in reporting.
Read the Canadian list as a planning signal
Health Canada’s Critical and Vulnerable Drug List methodology assesses medicines by medicinal ingredient and route of administration. Health Canada and other experts validated the assessment. For a manufacturer or distributor, that means interpreting the designation at the ingredient-and-route level before relating it to individual products or stock positions.
The published selection process considers clinical-criticality scores above the 75th percentile, comparable international lists and appropriate consultation recommendations. It then applies a shortage-vulnerability threshold of 10% or greater. These are list-selection rules, not a formula for collapsing clinical importance and vulnerability into a single score.
There is also an exception pathway. Health Canada considered medicines below the vulnerability threshold when they had at least two specified risk factors: infectious-disease use, a sole manufacturer or an intravenous formulation. The process therefore retains explicit judgment about particular supply vulnerabilities rather than relying exclusively on a numerical cutoff.
List membership is not a shortage alert. Health Canada states that inclusion means neither that a shortage currently exists nor that one is expected. The list is also not linked to regulatory requirements. A shared dashboard should preserve that distinction: a prioritization designation, a current shortage report and a local inventory concern should not become interchangeable labels.
Health Canada says the list will be updated from time to time to keep the assessments aligned with the Canadian context, including new medicines entering the market. Technology owners should therefore retain the assessment version and review date alongside the designation. Partners need to know which assessment they are discussing, not just its current colour on a dashboard.
Alternatives and local circumstances change the response
Clinical importance is broader than the seriousness of the disease being treated. The published Canadian clinical-importance framework includes comparator limitations and unmet needs, alongside disease severity, population size, clinical guidelines, effectiveness, safety, patient-reported outcomes, type of medical service and public-health interest. The availability and limitations of alternatives are therefore part of the reasoning, not an afterthought once supply fails.
Consider a planning discussion about a medicine with constrained supply. The existence of another product should prompt a clinical review of whether it is a suitable alternative, while supply teams examine whether that alternative is itself vulnerable. It should not automatically reduce the original medicine’s priority.
The webinar highlights the possibility of a domino effect when demand shifts to an alternative and that medicine also enters shortage. It describes using shortage reports, predictive information and clinical importance together to support consideration of alternatives. The practical implication is to examine the supply position of potential substitutes alongside the original disruption, rather than treating substitution as the end of the supply-chain problem.
Local scope matters too. The presentation describes viewing the tool through a hospital’s needs, such as focusing on intravenous medicines or chemotherapy, and interpreting risk through the relevant clinical lens. A national planning signal and an institution’s immediate priorities need not be identical.
For shared reporting, keep the national assessment intact and add local context separately: the services affected, the alternative under review, its supply position and the health-system owner responsible for interpreting clinical implications. That makes differences between partners understandable without allowing each local adjustment to silently rewrite the underlying assessment.
AI can check evidence without owning the decision
In the webinar’s account, clinicians independently scored medicines against clinical criteria without knowing the weighting. Multiple reviewers assessed medicines, and an AI tool double-checked scores for changes in guidelines. That is a bounded assistance role: checking whether evolving guidance may warrant another look at a clinical assessment.
A guideline-change flag should lead to review, not an automatic declaration that a medicine has become more or less important. The responsible reviewer needs to establish whether the guidance applies to the relevant indication and route, whether it changes the assessment, and whether the score should be revised. The model’s output is an input to that decision, not its clinical authority.
For technology owners, preserve the chain from the guideline passage to the flagged issue, the reviewer’s interpretation and the approved assessment. Keep a change in evidence distinguishable from a change in score and from a decision to act. Those events may be related, but they are not the same event.
Likewise, evaluate supply prediction, operational response and clinical outcomes separately. Earlier recognition of vulnerability or a changed purchasing decision does not, by itself, establish that patient harm was reduced. Clinical accountability remains with the people and institutions making clinical decisions; manufacturers and distributors contribute supply knowledge without turning a prioritization model into patient-level allocation advice.
The value of AI-assisted pharmaceutical decision support is not that it produces an unquestionable ranking. It is that it can help people keep supply vulnerability, clinical importance and changing evidence in view together. If you are connecting these signals across partner systems, Digiventi’s Data & AI services are a relevant starting point for discussing the data and application work.