Lisby 2015.

Study characteristics
Methods	Randomised controlled trial
Participants	108 participants randomised (53 to medication review and 55 to control) Participants were admitted to an orthopaedic ward at a regional hospital in Denmark Mean age: 80.5 years 29% male Mean number of drugs: 6.7
Interventions	Systematic medication review by a clinical pharmacist and a clinical pharmacologist. A clinical pharmacist obtained medication history through medical records, electronic prescribing system, registry of drug purchase and interview after a ward physician had prescribed in‐hospital medication. Subsequently, the case was discussed with a clinical pharmacologist, and a note with comments and recommendations for medication changes was prepared and handed directly to the physician responsible for the ward round. Ward physicians were not obliged to follow these recommendations. No co‐interventions were reported. The participants assigned controls were treated routinely, that is, a ward physician obtaining medication history, performing a review and prescribing the in‐hospital medication.
Outcomes	Primary outcome: time to first unscheduled physician contact (general practitioner, emergency department, ambulatory care or hospital) after discharge from the orthopaedic department Secondary outcomes: admission time, time to first readmission, number of readmissions, emergency department visits, visits to outpatient care clinic, general practitioner contacts if first contact with general practitioner included medication issues, contacts with physicians outside working hours, medical specialist contacts, quality of life assessment, mortality All outcomes had 3 months of follow‐up
Notes	Funding: The Health Insurance Foundation in Denmark
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Eligible participants were randomly assigned to intervention or control by a computer‐generated code.
Allocation concealment (selection bias)	Unclear risk	Allocation concealment not described.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Trial described as not blinded.
Blinding of outcome assessment (detection bias) Mortality (all‐cause)	Low risk	Mortality data were taken from a non‐biased national register (participants identifiable through unique patient‐specific social security number).
Blinding of outcome assessment (detection bias) Hospital readmissions (all‐cause)	Low risk	Readmission data were likely gathered from a non‐biased national register.
Blinding of outcome assessment (detection bias) Hospital emergency department contacts (all‐cause)	Low risk	Emergency department contacts were likely gathered from a non‐biased national register.
Blinding of outcome assessment (detection bias) Health‐related quality of life	Low risk	Questionnaires including pre‐paid envelopes were sent by postal mail. Because of the design of the intervention, it was not possible to blind the participants toward group assignment. We assessed participant knowledge about the assignment as unlikely to influence the outcome.
Incomplete outcome data (attrition bias) Mortality (all‐cause)	Low risk	No loss to follow‐up was described; all data should be readily available from a non‐biased national register.
Incomplete outcome data (attrition bias) Hospital readmissions (all‐cause)	Low risk	No loss to follow‐up was described; all data should be readily available from a non‐biased national register.
Incomplete outcome data (attrition bias) Hospital emergency department contacts (all‐cause)	Low risk	No loss to follow‐up was described; all data should be readily available from a non‐biased national register.
Incomplete outcome data (attrition bias) Health‐related quality of life	High risk	Fifty‐five (54%) of 102 participants returned the questionnaire. Loss to follow‐up: intervention: 50% (25/50), control: 57% (30/52).
Selective reporting (reporting bias)	Low risk	Relevant outcomes were included in trial and were similar to information provided on www.clinicaltrials.gov.
Contamination bias	High risk	No cluster‐randomisation.
Other bias	Low risk	No evidence of other types of bias.