Ontario cohort maps the prescribing cascades most likely to harm older adults
Health administrative data from Ontario have now produced one of the clearest population-level maps of potentially inappropriate prescribing cascades in community dwelling older adults. Investigators examined 2,297,942 people alive on January 1, 2022, and used a 2025 international expert Delphi consensus to assemble 65 cascades for study. The work was designed to move beyond theoretical concern and identify which drug-to-drug sequences are common enough, temporally linked enough, and persistent enough to deserve priority in day-to-day prescribing and deprescribing efforts.
The study population was large, older, and clinically relevant. A total of 54.3% were female, and the analysis focused on incident users of the initial and subsequent drug combinations during the accrual period. The central question was not simply whether a second medication followed the first, but whether the sequence was frequent in the population, occurred often enough to matter at the bedside, and showed a temporal pattern that made a causal cascade plausible rather than accidental.
Twenty four of the 65 cascades met all three prioritization criteria.
Cardiovascular drugs dominated the starting point for the highest priority sequences
Among cascades that cleared the prespecified thresholds, cardiovascular drugs were the most prevalent starting drugs, with class ranges spanning 19.0% to 65.9%. In older adults, that pattern is clinically intuitive. Cardiovascular therapies are widely used, often layered over years, and many can produce symptoms or laboratory changes that prompt additional prescribing rather than dose reduction or discontinuation. A cascade begins when the adverse effect of one medication is treated as a new disease.
The highest incidence cascade was iron supplement to laxative, at 11.9% (95% CI 11.7% to 12.1%). Next came hydroxymethylglutaryl coenzyme A reductase inhibitor to pain reliever, at 10.9% (10.7% to 11%), and cholinesterase inhibitor to sleep agent, at 10.3% (9.9% to 10.7%). Those three sequences are a sharp reminder that the downstream prescription may be a response to a medication effect, not a new independent diagnosis. Constipation after iron therapy, muscle symptoms after statin therapy, and sleep disruption after cholinesterase inhibitor treatment each sit squarely in the territory where a prescriber may add a second drug instead of revisiting the original one.
Temporal sequence was strongest for corticosteroid, laxative, and cholinesterase inhibitor cascades
The strength of temporal association, measured with adjusted sequence ratios, was highest for corticosteroid to antipsychotic at 2.55 (95% CI 2.47 to 2.64), laxative to antidiarrheal at 2.53 (2.43 to 2.65), and cholinesterase inhibitor to antiemetic drug treatment at 2.24 (1.90 to 2.65). These are not subtle signals. A sequence ratio above 1 means the second drug followed the first more often than would be expected by chance alone, and ratios in this range suggest a reproducible prescribing pattern rather than isolated clinical noise.
One sentence captures the bedside consequence: a symptomatic treatment can lock in the original prescribing problem.
Corticosteroids can precipitate neuropsychiatric effects that may lead to antipsychotic prescribing; laxatives can overshoot into diarrhea that then leads to antidiarrheal treatment; cholinesterase inhibitors can trigger gastrointestinal adverse effects that then result in antiemetic use. Mechanistically, each cascade reflects a familiar clinical trap. The initial drug alters physiology, the adverse effect is treated as a separate disorder, and the therapeutic burden expands while the original trigger may remain untouched.
Sex-stratified results were mostly similar, with one notable divergence
The adjusted sequence ratios remained consistent when stratified by sex in nearly all cascades. One exception stood out: the selective serotonin reuptake inhibitor or serotonin-norepinephrine reuptake inhibitor to overactive bladder drug cascade was stronger in male older adults, with an adjusted sequence ratio of 1.21 (1.10 to 1.33), than in female older adults, where it was 1.02 (0.96 to 1.09). The sex-specific difference suggests that prescribing patterns, symptom recognition, or downstream treatment choices may differ by patient group, even when the initiating antidepressant exposure is similar.
This type of divergence matters because a population average can conceal a narrower subgroup pattern. In older men, urinary symptoms may be more likely to prompt pharmacologic treatment after antidepressant exposure, while in women the sequence may be less pronounced or less consistently coded in administrative data. The study does not resolve the reason for the difference, but it identifies a sequence where male older adults appear more vulnerable to the cascade pattern.
Why the prioritized list matters for everyday prescribing
The value of this work lies in prioritization. Sixty five cascades were catalogued, but 24 crossed all three thresholds: population prevalence of the initial drug at least 5%, cascade incidence at least 1%, and adjusted sequence ratio above 1 with the lower 95% confidence bound above the corresponding threshold for each measure. Those criteria filter out rare or weak patterns and elevate the cascades most likely to generate substantial patient-level and system-level harm.
In practical terms, the list identifies where clinicians should pause before adding a second drug. If a patient taking an iron supplement develops constipation, the next prescription may be a laxative, but the better question is whether the iron regimen can be changed. If a patient on a statin reports pain, the answer is not automatically a pain reliever; the statin itself may deserve reassessment. The same logic applies to cholinesterase inhibitors, corticosteroids, antidepressants, and other commonly used agents in older adults.
The broader pathophysiology is straightforward. Older adults have reduced physiologic reserve, more comorbidity, and greater exposure to polypharmacy. A medication adverse effect can masquerade as a new symptom complex, and every added drug increases the probability of interactions, adverse effects, and nonadherence. Prescribing cascades therefore function as both a pharmacologic and a geriatric syndrome problem.
Population data point toward deprescribing and safer prescribing policy
The investigators conclude that the prioritized list could inform prescribing and deprescribing initiatives and policy development to improve drug safety in older adults at both individual and population levels. The evidence base here is administrative rather than clinical adjudication, so the findings identify strong patterns of association rather than direct proof of harm in every individual case. Even so, the scale of the cohort and the consistency of the temporal signals make these cascades difficult to dismiss as random noise.
A population study of this size rarely produces a cleaner message: many of the drugs most often used to treat older adults can also start the chain that leads to another prescription. The remaining question is not whether cascades occur, but which ones clinicians should actively interrupt before they become routine. That is where the study leaves practice, with a short list of high-priority sequences and a concrete target for safer medication review in older adults.