Objectives: To describe the prevalence of unplanned hospitalisations caused by adverse drug reactions (ADRs) in older veterans and to examine the association between this outcome and polypharmacy after controlling for comorbidities and other patient characteristics.

Design: Retrospective cohort.

Setting: US Veterans Affairs Medical Centers.

Participants: 678 randomly selected unplanned hospitalisations of older (aged 65 years or older) veterans between 1 Oct 2003 and 30 Sep 2006.

Measurements: Naranjo ADR algorithm, ADR preventability and polypharmacy (0-4, 5-8 and 9 or more scheduled medications).

Results: 70 ADRs involving 113 drugs were found in 68 (10%) hospitalisations of older veterans, of which 25 (36.8%) were preventable.  Extrapolating to the population of more than 2.4 million older veterans receiving care during the study period, 8000 hospitalisations might have been unnecessary.  The most common ADRs that occurred were bradycardia (n = 6; beta-blockers, digoxin), hypoglycaemia (n = 6; sulfonylureas, insulin), falls (n = 6; antidepressants, ACE inhibitors) and mental status changes (n = 6; anticonvulsants, benzodiazepines).  Overall, 44.8% of veterans took 9 or more outpatient medications and 35.4% took 5 to 8.  Using multivariable logistic regression and controlling for demographic, health-status, and access-to-care variables, polypharmacy (9 or more and 5-8) was associated with greater risk of ADR-related hospitalisation (adjusted odds ratio (AOR) = 3.90; 95% CI, 1.43 to 10.61 and AOR = 2.85; 95% CI, 1.03 to 7.85, respectively).

Conclusions: ADRs, determined using a validated causality algorithm, are a common cause of unplanned hospitalisation in older veterans, are frequently preventable, and are associated with polypharmacy.