Hultberg 2005.

Methods	Controlled before and after study with 12 months follow‐up of 3 intervention sites and 4 control sites. Collaborating partners Lead agency: Primary care. Strategic involvement (policy making and service planning): Primary health care, social services. Commissioning (implementing strategy taking account of resources available): Primary health care, social services. Operational (providing services directly): Primary health care, secondary health care, social services. Set in Sweden.
Participants	138 participants with musculoskeletal disorder in Goteburg (107 attending DELTA intervention health centres and 31 attending control health centres outside the DELTA trial area) who completed 3 interviews at baseline, 6 months and 12 month follow‐up. Intervention group 36% male and 21% aged 16 ‐ 30, 50% aged 31 ‐ 50, 29% aged 51 ‐ 65 years. Control group 19% male and 25% aged 16 ‐ 30, 65% aged 31 ‐ 50, 10% aged 51 ‐ 65 years.
Interventions	Collaboration consisted of a co‐financed collaborative care model to intensify rehabilitation through multidisciplinary teams (health centre physicians and nurses with occupational therapists, physiotherapists, social workers and social insurance officers) based in health centres. They had access to a joint budget from a common administrative body. They met weekly to discuss the rehabilitation of individual patients.
Outcomes	Pain level measured by the Visual Analogue Scale (VAS) Long term or repeated sick leave Health‐related quality of life measured through EuroQol 5 dimensions instrument (EQ‐5D)
Notes	Power calculation was not reported but they aimed for a sample size of 450 patients. Smaller sample size achieved than aimed for despite study recruiting 8 months longer than planned. Potential for selection bias. The total healthcare cost for an average patient in the intervention was 1979 Euro and 1286 Euro for the control group. Overall risk of bias was high.
Risk of bias
Bias	Authors' judgement	Support for judgement
Allocation concealment (selection bias)	High risk	Unlikely due to selection methods, as patients were recruited to intervention and control arms by their physicians, who would have been aware of their status.
Blinding (performance bias and detection bias) All outcomes	High risk	Unlikely that assessments were conducted blind due to study design. Some assessments were conducted in the intervention and control health centres.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Follow‐up rates adequate.
Selective reporting (reporting bias)	High risk	Linked paper Hultberg 2002 (p 5), together these papers report on three projects from a wider set of 20 DELTA projects in Goteborg.
Other bias	High risk	Low recruitment rate (aimed to recruit 450 but achieved 167). Control group small at 39 patients. P117, even though the planned recruitment period had been extended by 8 months, they only managed to recruit about half the targeted sample size. A large proportion of those invited declined to participate.
Randomisation adequately described/protected?	High risk	Not randomised
Protection against contamination?	Low risk	Adequate as intervention delivered at health centres
Follow‐up rate adequate?	Low risk	Follow‐up rate 83% (84% in intervention and 79% in control group)
Reliable primary outcome measure?	Unclear risk	The authors concluded that this co‐financing model was not associated with better patient outcome for patients with musculoskeletal disorders but the authors questioned their own findings. P122
Groups measured at baseline?	Low risk	P118 table2 Control group small compared to intervention group though groups appeared balanced at baseline. Not clear about recruitment process.
Appropriate choice of controls (CBA studies only)?	Low risk	Adequate choice of control areas but low recruitment rates for controls
Contemporaneous data collection (CBA studies only)?	Low risk	Yes
IS THE STUDY AT LOW RISK OF BIAS?	High risk	OVERALL RISK OF BIAS WAS HIGH