Kruis 2014.

Study characteristics
Methods	Cluster‐RCT (40 clusters); follow‐up: 24 months; control: usual care; 40 clusters of primary care teams
Participants	Eligible: 22698 Randomised: 1086, I: 554, C: 532 Completed: 810, I: 419, C: 391 Mean age: I: 68 years, C: 68 years Sex (% male): 54, I: 51, C: 57 Inclusion criteria: clinical diagnosis of COPD according to GOLD criteria, if possible and necessary (no spirometry data available) verified by available spirometry data or spirometry assessment Major exclusion criteria: terminally ill patients, dementia or cognitive impairment, inability to fill in Dutch questionnaires, hard drug or alcohol abuser
Interventions	Two‐day training of multi‐disciplinary team on all IDM components of intervention before implementation intervention. During training, the team redesigns the care process and defines responsibilities of different caregivers, and is trained in how to use feedback on process and outcome data to implement guideline‐driven integrated health care. The team sets up a time‐contingent individual practice plan, agreeing on steps to be taken to integrate a COPD IDM programme into daily practice. Practice‐tailored feedback reports are provided at baseline and at 6 and 12 months to each team. After 6 and 12 months, a refresher course is provided for all teams simultaneously to enable them to learn from each other’s experiences. Intensity of the IDM programme for individual patients depended on health status, personal needs, and preferences Intervention components ‐ Access to patient healthcare provider portal for process and outcome measures ‐ Optimal medication adherence ‐ Proper diagnosis ‐ Motivational interviewing ‐ Smoking cessation ‐ Self‐management ‐ Dietary intervention Duration intervention: 12 months Disciplines involved: GP, practice nurse, physiotherapist and dietician, consulting pulmonary physician
Outcomes	CCQ, SGRQ‐C, EQ‐5D, SF‐36, smoking behaviour (guided smoking attempts), IPAQ, SMAS‐30, MRC Dyspnoea, number of moderate exacerbations, number of severe exacerbations, level of care integration (PACIC and ACIC), satisfaction with healthcare providers, costs, healthcare utilisation, costs of productivity loss
Notes	Dominant component: self‐management Additional comment: practices affiliated to Primary Care Research Network ‐ signed agreement to collaborate in scientific research
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "the same blinded researcher randomised matched clusters in pairs by using a computer generated list in four blocks of 10"
Allocation concealment (selection bias)	Low risk	Quote: “the clusters were matched and randomised by a researcher who was blinded to the identity of the practices”
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "because of the nature of the intervention, participating healthcare providers and patients could not be blinded"
Blinding of outcome assessment (detection bias) All outcomes	Low risk	Quote: "blinded research nurses assessed outcomes to minimise detection bias. Patients were instructed not to report on their type of management to these research nurses"
Incomplete outcome data (attrition bias) All outcomes	Low risk	Comment: missing outcome data was balanced in numbers across intervention (n = 135) and control groups (n = 141), with similar reasons for missing data across groups
Selective reporting (reporting bias)	Low risk	Comment: all outcomes included in the protocol were reported, with the exception of ACIC (assessment of chronic illness care) and level of healthcare providers' satisfaction (intervention group only). Missing of outcomes most probably does not impact the quality of the evidence
Recruitment bias	High risk	Quote: “the GPs checked the selected patients against the formal inclusion and exclusion criteria before the recruitment procedure started” Comment: study flowchart suggests that patients were recruited after the cluster had been randomised
Baseline imbalance between groups	Low risk	Comment: most baseline characteristics did not differ significantly between intervention group and usual care group, although participants in the intervention group were significantly less likely to be male and had significantly higher functional CCQ scores
Loss to follow‐up of clusters	Unclear risk	Comment: insufficient information provided on practice level
Adequate analysis methods for CRT	Low risk	Quote: “we used linear mixed model analyses to assess differences within and between groups for all continuous outcomes, correcting for baseline scores, age, sex, proportion of patients with MRC score above 2, and clustering of patients per general practice. We used baseline scores as a dependent variable, the cluster was represented by a random effect, and the within patient covariance structure was unstructured. For dichotomous outcomes, we used logistic link generalised linear mixed models for repeated measurements to analyse differences within and between groups at all time points, correcting for the same covariates"