Wood 1992.

Study characteristics
Methods	Aim of study: to determine the effects of an inpatient self‐administration of medication programme on compliance post discharge among elderly patients Study design: cluster‐RCT (2 wards intervention, 2 wards control, 1 intervention and 1 control per hospital) Number of arms/groups: 2
Participants	Description: patient/consumer Geographic location: UK Setting: hospital inpatient (pre‐discharge) from rehabilitation wards Inclusion criteria: rehabilitating with the ultimate aim of discharge to own home to live alone Exclusion criteria: nil mentioned Number of participants randomised: 33 (18 and 15) Number of participants included in analysis: 22 (11 and 11) Age: 85.4 ± 6.0 vs 84.8 ± 5.8 Gender: M:F = 1:8 (88.8% F) vs 1:6.5 (86.7%) Ethnicity: not specified Number of medications: number of dose‐taking events per day: 8.1 ± 5.0 vs 6.17 ± 5.0 Frailty/Functional impairment: not specified Cognitive impairment: abbreviated mental test score: 9.1 vs 9.4 Comorbidities: not specified
Interventions	Group 1 ‐ Inpatient self‐administration: 3 distinct phases ‐ phase 1: medicine containers labelled as for discharge, drugs handed to patient at appropriate times, and full supervision of medication selection and ingestion. After 7 days, or earlier if appropriate, patient moved to phase 2; phase 2: patient required to request medication at appropriate times. After 7 error‐free days, patient moves on; phase 3: patient becomes totally responsible for his/her own medication. Medicines stored in locked cupboard. Compliance checked by tablet count Group 2 ‐ Usualcare: discharge medicines issued by nursing staff immediately before discharge Co‐intervention: N/A Provider: care team ‐ including pharmacist and nurse Where: hospital inpatient (pre‐discharge) When and how often: up to 3 weeks, inpatient Intervention personalised: yes ‐ moved through phases only if able to self‐administer medications
Outcomes	Timing of outcome assessment: 2 weeks and 3 months post discharge Medication adherence (objective) : pill count: average percentage of non‐compliance calculated by pill count. No errors made, few errors (1% to 15% non‐compliance), and many errors (> 15%)
Notes	Trial registration: N/A Consumer involvement: not specified Funding source: not specified Dropout: 8 lost to follow‐up (4 and 4) plus 3 intervention could not be analysed ICC value unclear; unsuccessful contact with study authors likely due to age of study. Thus unit of analysis error exists
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Wards allocated ‐ randomisation method not stated
Allocation concealment (selection bias)	Unclear risk	Not specified
Blinding of participants and personnel (performance bias) All outcomes	High risk	Not blinded. Blinding was not possible for the type of intervention provided unless the researcher was not involved in providing the intervention; unclear if this had any impact on the outcome
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	Not specified whether home visit staff were blinded
Incomplete outcome data (attrition bias) All outcomes	High risk	4 lost to follow‐up from both groups, 3 intervention patients excluded because they had received new medications but transferred them to original containers = potentially unbalanced analysis
Selective reporting (reporting bias)	Unclear risk	Additional data in results that are not specified in methods (e.g. errors sufficient to be detrimental to health)
Other bias	Low risk	Recruitment bias (selective recruitment of cluster participants): low risk, recruitment occurred after allocation but all patients on the ward were included in the study (no exclusions), thus limited risk of recruitment bias