Jordhoy 2001.

Methods	Cluster‐RCT; Unit of allocation: Community healthcare districts of living; Stratified by: Pairs of districts, Inhabitants' age, type of area represented
Participants	Advanced cancer patients referred to the hospital Palliative Medical Unit (PMU) by healthcare districts. Setting / country: Norwegian Public Health Service (Hospital of Trondheim and community care close to the PMU (8 districts) / Norway Type of cancer: Any type Phase of care: Palliative care Sample size at randomisation: 434 patients (312 close family members)
Interventions	Comprehensive palliative care: In this home based palliative care program, the patients general practitioner (GP) and a community nurse were defined as the main professional caregivers. Treatment plans were set up in a meeting of the patient, the informal caregiver, the GP, the community nurse and the nurse or physician from the Palliative Medical Unit (PMU). Hospital service was offered on request, always at the PMU. The PMU has 12 inpatient beds, an outpatient clinic, and a multidisciplinary consultant team that works in and out of the hospital. The PMU consultant team organised the follow‐up. Predefined guidelines were used to keep the interaction at an optimum between services. The educational program for community professionals included bedside training and 6‐12 hours of lecture every 6 months. Follow‐up consultations by community staff were set up as routine. Multidisciplinary staff meetings were arranged weekly. For referrals and admission to nursing homes, conventional routines were followed. Control: Conventional care (without well‐defined follow‐up routines).
Outcomes	Patient: Psychologic distress, QoL Informal carer: Satisfaction Process: Place of death, use of hospital services, use of nursing home services, use of hospital services, proportion of readmission time in institutions (nursing homes and hospital)
Notes	Length of follow‐up: Until death (min: 2; max: 24) months
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote from ref #1: "Before opening of the trial, three clusters were allocated to intervention and three to conventional care (control). Eligible patients were assigned treatment according to the district (cluster) in which they lived."
Allocation concealment (selection bias)	High risk	Quote from reference #1: "A difference in distribution of diagnostic groups was probably related to lack of concealment of individual patient allocation, because the treatment assignment of individual patients could be identified from their address."
Blinding (performance bias and detection bias) All outcomes	High risk	Quote from ref #1: "All questionnaires, except the baseline forms, were distributed by mail." Comment: Since the patients could not be blinded to group assignment, and since they were responsible for completing the questionnaire, then the assessors could not be blinded.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote from ref #1: "If data from one assessment point were missing, then the mean of the two adjacent ones was used. HRQL scores were assumed to be zero for the time after death. For the patients who withdrew or dropped out before death during the first 4 months, the last value carried forward was used to impute the missing subsequent values. The latter approach might, however, introduce a bias if the main reason for drop‐out was deterioration. Hence, the analyses were repeated imputing worse possible scale/item score for the missing ones. The results were consistent with those that are presented." Quote from ref #1: Missing items were imputed for the EORTC QLQ‐C30 and the IES multi‐item scales, using the method advocated by the EORTC Quality‐of‐Life Study Group. If at least half of the items from a scale were completed, the values of missing ones were imputed as the mean value of the completed items. For the IES, which had a higher number of missing items, the analyses were made both with and without using imputation; imputation had a minor impact on the group means and did not alter the results concerning the comparisons between treatment groups. Comment: Sensitivity analysis was used both for missing items and missing participants to minimize the risk of bias. Such a procedure is valid.
Selective reporting (reporting bias)	Low risk	All outcomes described in Methods are reported in Results.
Other bias	Low risk	No evidence of any other bias.
Baseline outcomes similar?	Low risk	Quote from reference #1: "the treatment groups were comparable on a wide range of baseline data, including HRQL scores" See table 4. Quote from ref #1: "To adjust for possible baseline differences, the AUC calculation for each patient was based on changes from baseline (actual score ‐ baseline score), i.e., on the improvement or deterioration at 1 to 4 months compared with trial entry."
Baseline characteristics similar?	High risk	Quote from references #1 and #2: "Diagnoses were classified by traditional sharing of treatment responsibility among the departments at the University Hospital of Trondheim (groups A‐C, table 1). The distribution of patients to these groups differed significantly between the treatment groups. There was also a difference in the time from diagnosis to inclusion." Quote from ref #1: "At baseline, the groups differed for housing, access to informal help, home care nursing, and, slightly, for living situation (Table 2)." Comment: No mention of any statistical adjustment for baseline differences in patient characteristics was found.
Protected against contamination?	Low risk	Quote from ref #1: "Thus, to minimize the exposure of the control group to the experimental effect, a cluster randomised design was chosen." Comment: Cluster‐randomisation by community healthcare districts prevents contamination.