Read the 86% as an exposure finding—not an adverse-event rate
The Journal of the American Medical Directors Association published the U.S. cohort study online ahead of print on July 7. Researchers derived the cohort from Long-Term Care Data Cooperative electronic medication administration records dated January 1, 2022, through January 31, 2025, and linked those records with Medicare beneficiary data. Among 48,260 residents who received nirmatrelvir/ritonavir, 85.9% had at least one day when a medicine on the study's potential-interaction list was also administered.
For this study, a “potential drug–drug interaction” meant same-day administration. The researchers did not determine that the combination was inappropriate, that staff missed a planned adjustment, or that the resident experienced bleeding, a fall, myopathy, an arrhythmia, or another adverse outcome. The finding measures overlapping exposure. It is not an 86% medication-error or harm rate.
The common overlaps show why one generic alert is not enough
Atorvastatin was administered during treatment for 28.1% of residents, amlodipine for 21.6%, and apixaban for 15.9%. Median overlap for those medicines was five to six days. Multiple overlaps were common: 60.9% of residents had at least two potentially interacting medicines on one or more treatment days, and 35.3% had at least three.
The study says the most frequent potential interactions generally involved temporary withholding, dose adjustment, or monitoring rather than an absolute contraindication. That summary is not a universal hold list. The current prescribing information directs clinicians to review all medicines, determine whether interruption, adjustment, or additional monitoring is needed, and weigh the treatment benefit against whether an individual interaction can be managed. The exact plan still depends on the current sources and the resident-specific decision.
Give every potential interaction a start–change–restart clock
A five-day antiviral course can create a longer coordination task. The interacting medicine may need an action before the first dose, monitoring during treatment, and a planned reassessment after the last dose. A consultant pharmacist should be able to see one chain rather than reconstruct it from an alert, an order message, an eMAR exception, and a later progress note.
For each potential interaction, keep these links together:
- Trigger and evidence: the antiviral order, symptom or diagnosis date, interaction source and version, and the date the review was performed.
- Exposure timeline: the complete prescription, nonprescription, vitamin, and herbal list available for review; the relevant medicine's last and next scheduled administrations; and what remains unverified.
- Authorized plan: the medicine-specific continue, interrupt, adjust, substitute, monitor, or alternative-treatment decision; who authorized it; and when it takes effect.
- Execution and monitoring: what was actually administered, held, or changed; the ordered observation or laboratory plan; exceptions; and the responsible person or team.
- Close-out: the intended reassessment or resumption point, the actual action and time, any reason for a different outcome, and the next follow-up owner.
Test software on the handoff after the alert fires
Use a synthetic resident already receiving atorvastatin, amlodipine, and apixaban when an after-hours nirmatrelvir/ritonavir order arrives. Keep the test bounded: it evaluates interaction handoffs, not the complete prescribing decision; the current label separately requires review of indication and treatment timing and selection of the appropriate renal dose. The exercise is not to invent one correct clinical response. It is to see whether the system can present the current medication evidence, identify the source and age of each interaction rule, route the case to an authorized reviewer, and record a different plan for each medicine without implying that software made the decision.
Then move the clock forward. Can nursing staff see the effective action before the next administration? Can the pharmacist distinguish a proposed change from an authorized order and from what the eMAR shows actually happened? Does the restart or reassessment remain on a worklist after the antiviral course disappears from the active list? Finally, inspect the report and export: neither should convert a potential interaction into a confirmed adverse event or erase an override and its rationale.
Keep the study limits beside the headline
The analysis covered residents who received nirmatrelvir/ritonavir, not residents who were untreated or given another antiviral, so it cannot measure access or underuse. Only 16.2% of treated residents were classified as long stay, so the 85.9% finding should not be presented as a long-stay nursing-home estimate. Its data ended in January 2025. The interaction set came from a February 2024 NIH list, and current management should be checked against current prescribing information and other appropriate resources rather than copied from a historical research table.
Electronic administration records are a strength because they show documented administration rather than an order alone. They still could not show whether clinicians deliberately accepted an overlap, followed a recommended plan, or observed harm. The same-day definition may also miss interaction risk that continues after the last overlapping administration. Those limits are why a screening result should remain visibly labeled as such and linked to the authorized human decision.
Audit one interaction clock before adding more alerts
Run the synthetic case through the real after-hours route used by one facility. Count the missing links between detection, authorization, administration, monitoring, and close-out. Repair the field, handoff, or worklist where the chain breaks. The study supports better coordination around a common, time-sensitive interaction problem. It does not support a bigger alert count, a blanket medication stop, or an automated conclusion about an individual resident.