Design: Population-level retrospective cohort with prescription sequence symmetry analysis
Population: 2,297,942 community-dwelling adults aged ≥66, Ontario
Primary outcome: Prevalence, incidence and temporal association of 65 candidate cascades
Effect: 24 prioritised; iron→laxative 11.9%; corticosteroid→antipsychotic sequence ratio 2.55 (2.47–2.64)

A population study in the BMJ (10 September) used Ontario data on 2.3 million community-dwelling adults aged 66 and over to test 65 prescribing cascades agreed by an international expert panel — where a drug's side effect is treated as a new problem with a second drug.

Twenty-four cascades met all three priority criteria: the first drug was common, the drug pair was common, and the second drug followed the first more often than chance. The most frequent were iron supplement then laxative (11.9% of new iron users), statin then a pain reliever (10.9%) and cholinesterase inhibitor then a sleep agent (10.3%). The strongest time links were corticosteroid then antipsychotic, laxative then antidiarrhoeal, and cholinesterase inhibitor then antiemetic.

The design detects a pattern, not a cause in any individual, and some second prescriptions will be appropriate. But the list is a practical checklist for medication review: when a new drug appears, ask what was started in the weeks before it.

Before prescribing a laxative, ask whether an iron supplement was recently started.
Before adding an analgesic for aches, consider whether a statin may be the cause.
Insomnia, nausea or new agitation after a cholinesterase inhibitor or corticosteroid may be drug effects.
At medication review, ask of each drug: was this started to treat another drug's side effect?

Why it matters

It turns a familiar principle into a named, ranked list that fits a ten-minute medication review.

The statistics, in plain English

A sequence ratio compares how often drug B follows drug A with how often A follows B. A ratio of 2.55 means B came after A about two and a half times as often as the other way round, which suggests A may be prompting B. It is a population signal, not proof for any one patient.