The result deserves more than one headline number
Age and Ageing published the pragmatic cluster-randomized trial on July 19. It enrolled 318 community-dwelling adults aged 75 or older through 58 community pharmacies in the Netherlands. Eligibility included at least 10 chronic medicines and multidose drug dispensing, although investigators later found that some participants did not meet the medicine threshold. In all, 304 supplied the six-month dispensing data used in the primary analysis. Pharmacies, rather than individual patients, were randomized to the intervention or usual-care group.
Intervention pharmacists received a day of training and a deprescribing toolbox, then used a five-step clinical medication review: learn the patient's current problems, goals, and preferences; analyze the pharmacotherapy; reach consensus with the general practitioner; agree a care plan with the patient; and monitor the action and its effects. That is a structured service, not a pop-up alert or an instruction to reduce every long medication list.
Stopped or reduced and total medicines are different measures
After six months, the intervention group averaged 2.5 stopped or reduced medicines per person, compared with 1.7 in usual care. The mean difference was 0.79 medicines, with a 95% confidence interval of 0.29 to 1.28. That was the trial's primary result. The endpoint was reconstructed from dispensing data rather than counted recommendations. A reduction could mean a lower dose or a switch to a lower-potency or less-complex alternative, and two researchers independently classified the medication histories.
The total medication count told a different story. It fell by an average of 0.30 in the intervention group and rose by 0.12 in the control group, but the between-group difference was not statistically significant. Reductions do not necessarily remove a medicine from a count, and new medicines were started at a similar rate in both groups. Those figures do not conflict: more deprescribing activity can coexist with little net change in current medicines.
Do not import the result into a nursing facility
This was Dutch primary-care research. The investigators excluded nursing-home residents, people with cognitive impairment, people with an estimated life expectancy of six months or less, and those whose repeat prescriptions came only from a hospital specialist. Patients were recruited after pharmacy randomization by clinicians who knew their group; participants, clinicians, and outcome assessors were not formally blinded. Participants may also have been more open to deprescribing than average, and follow-up lasted six months.
The groups did not differ significantly in reported health problems or health-related quality of life. Those were secondary outcomes, questionnaire data were missing for some participants, and the trial was not powered to detect small changes in them. The study therefore does not show that a medication-count target improves resident outcomes, that a particular medicine should be stopped, or that the same intervention would produce the same result in a U.S. nursing facility.
Keep five records behind the deprescribing metric
The trial's five-step review suggests a practical documentation test for an independent consultant pharmacist. This is an editorial translation of the research, not a clinical protocol: the software should preserve enough context to distinguish a recommendation from an authorized, implemented, and monitored change.
- Review context: the current regimen, source date, health problems, goals, preferences, and information that was unavailable.
- Recommendation: the medicine, reason for review, proposed stop, dose reduction, or lower-burden substitution, and the evidence used.
- Decision: who reviewed the proposal, what the prescriber and resident or representative decided, and when they decided it.
- Implementation: the authorized order, effective date, taper or staged change where applicable, dispensing status, and any medicine started in the same period.
- Follow-up: the observation or monitoring plan, what was actually reviewed, unresolved questions, later changes, and the owner and date of the next step.
Make the dashboard show flow, not just subtraction
A useful view keeps proposed, accepted, implemented, and sustained changes separate. It can show complete stops, dose reductions, substitutions, new starts, and the current total without treating any one measure as the outcome. It should also keep a declined recommendation visible without turning it into a failure or silently closing it.
Be equally careful with response and outcome language. A prescriber agreement is not implementation. A lower dose is not a discontinued medicine. An unchanged medication count is not proof that nothing improved. A follow-up note is not evidence of better health unless the underlying observation and method support that conclusion.
Test one realistic change sequence
Ask a vendor to demonstrate a synthetic case in which a pharmacist proposes a dose reduction, the prescriber modifies the recommendation, a different medicine is started later, and follow-up produces another adjustment. Then inspect the timeline, report, and export. Each view should preserve the original recommendation and the sequence of authorized changes rather than overwriting the story with the latest count.
Software may flag candidates, but it cannot replace resident-specific clinical judgment, authorize a medication change, supply missing preferences, or prove a clinical benefit. Start small: map one recent or synthetic review across stopped, reduced, started, and total measures, then note where the underlying trail becomes ambiguous.