Lapane 2011.

Methods	Cluster‐randomised controlled trial with nursing home as the unit of randomisation
Participants	3538 residents of 25 nursing homes in Virginia, USA, recruited between 2003 and 2004 Medicare‐ and Medicaid‐certified nursing homes with contracts with Omnicare pharmacies, ≥ 50 geriatric beds and few short‐stay residents were considered for inclusion. 73.9% women 39.0% aged ≥ 85 years
Interventions	Intervention group: GRAM software used to identify resident‐specific medications that may contribute to delirium and falls risk. Pharmacy automatically generated GRAM report within 24 hours of nursing‐home admission. For those who triggered GRAM resident assessment protocols for delirium or falls risk, an automatic report was sent to the pharmacist to coincide with a monthly visit to the nursing home. A medication review was then undertaken at the visit and a proactive monitoring plan was initiated by the care home staff to assess for medication adverse effects. Control group: nursing homes did not receive the triggered pharmacist visit or proactive monitoring plan.
Outcomes	Incidence of delirium, measured using the NH‐CAM (Dosa 2007) Fall events, measured using MDS records Hospital admissions, measured using MDS records. Mortality, measured using MDS records. The trial used resident months (defined as the number of days from date of first assessment to the first outcome occurrence, the last date in the nursing home, the death date or 31 December 2004), rather than individuals as its unit of outcome measurement. Results applied only to new admissions during 2004. All outcomes were recorded electronically by participating care‐home staff over a 12‐month period.
Notes	Funding source: Agency for Healthcare Research and Quality and the National Institutes of Health Center for Research Resources.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Method of allocation sequence generation not provided.
Allocation concealment (selection bias)	Unclear risk	Unclear if all care homes recruited prior to randomisation.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants or personnel (or both) aware of allocation to intervention or routine care.
Blinding of outcome assessment (detection bias) All outcomes	High risk	Outcomes assessed using data from the minimum dataset and assessments were made by staff aware of allocation.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	1 cluster lost. No information on intention‐to‐treat analysis.
Selective reporting (reporting bias)	Low risk	No evidence of selective outcome reporting.
Other bias	Low risk	Only 1 cluster was lost. Poisson regression accounting for the cluster design was used.