Titova 2017.

Study characteristics
Methods	RCT, single‐centred; follow‐up: 3 years (initially 2 years planned); control: usual care
Participants	Eligible: 199 Randomised:172, I: 91, C: 81 Completed: 100, I: 51, C: 49 Mean age: I: 74 years, C: 72 years Sex (% male): I: 43, C: 43 Inclusion criteria: admission due to COPD exacerbations; clinical diagnosis of COPD with GOLD stage III or IV; living in Trondheim municipality; ability to communicate in Norwegian; ability to sign informed consent Major exclusions: serious disease that might cause a very short life span (expected survival time < 6 months)
Interventions	Intervention components of COPD ‐ home intervention ‐ Call centre for support and communication with patients, home care nurses, co‐ordination between various levels of care ‐ Educational session for home care nurses and interactive e‐learning programme for patients ‐ Individualised self‐management plan for patients ‐ Joint visits at patients‘ homes by a specialist nurse who repeated the core element of the educational programme and reinforced specific health behaviours, as well as making necessary changes to patient's treatment programme Duration intervention: 24 months Disciplines involved: home care nurse, specialised nurse, GP
Outcomes	Number of hospital admissions caused by AECOPD (primary outcome), number of in‐hospital days due to AECOPD, all‐cause mortality, COPD‐related mortality, SGRQ, HADS, Patient Activation Measurement (PAM), use of medication, lung function, cost‐effectiveness
Notes	Dominant compone n t: stru c tured follow‐up Study temporarily stopped after 2 years' follow‐up for 8 months due to increased mortality rates in intervention group. REC concluded that mortality was not related to intervention. Follow‐up continued, intervention not continued
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: “it was decided by lottery that participants from District Pair 1 were assigned to the UC group, and participants from District Pair 2 were assigned to the IC group”; "the demography is quite similar according to age and disease panorama, i.e. the number of inhabitants 55–79 years old are the same in the two district pairs (Lerkendal/Heimdal; 15 800 and Østbyen/Midtbyen 15 200) Comment: randomisation was performed on district level, matched based on district size
Allocation concealment (selection bias)	High risk	Quote: "they were randomly allocated to either integrated care (IC) or usual care (UC) based on address of permanent residence Comment: insufficient detail on allocation concealment provided. Considering randomisation procedure on district level and following hospitalisation, it seems unlikely that participants and investigators could not foresee assignment to intervention or control conditions
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "the study was a prospective, open, single‐centre intervention study"
Blinding of outcome assessment (detection bias) All outcomes	High risk	Quote: "the study was a prospective, open, single‐centre intervention study"
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Quote: “an increased number of deaths were registered among the patients in the IC group compared to the UC group.”; “data on the causes of death were analysed, and the REC concluded that the increased number of deaths in the IC group was not related to the COPD‐home intervention, but could be explained by pre‐study poorer health status and higher age” Comment: imbalance in number of deaths could have resulted in overall healthier health status among intervention group members at follow‐up. Insufficient details provided to be conclusive regarding effect on true outcome
Selective reporting (reporting bias)	Low risk	Comment: published reports include all expected outcomes that were pre‐specified, with the exception of lung function and cost‐effectiveness as included in the clinical trial register
Other bias	High risk	Comment: not defined as cluster‐RCT; however potential clustering effect, considering that level of randomisation is region (4 clusters)