Hall 2002.

Methods	RCT. Unit of randomisation: Patient. Duration: Two batches of telephone surveys ‐ September 1999 to September 2000.
Participants	Clinical setting: USA, Primary Care Patients of two different HMO health plans. Numbers of patients followed up in intervention group: 515 in capitated plan and 670 in mixed incentive group. Numbers of patients followed up in control group: capitated plan controls = 382, mixed incentive controls = 351. Characteristics of intervention providers: The disclosure interventions were devised by the authors with the assistance of 'an expert panel'. The nature of the person carrying out the telephone follow up to clarify the disclosure and ensure comprehension was not recorded. Characteristics of participants: The sample was stratified to ensure a roughly even number of subscribers who had been with the plans for either between two and four, or more than four years. Of 4024 patients initially approached, 3844 were successfully contacted; 15.6% of those contacted were ineligible (leaving 3246 eligible) and a further 22.2% refused to take part, resulting in contacts with 62.3% of 3844 potentially eligible contacts (n = 2394). Further attrition reduced the final sample to 1918 patients, which was 59% of eligible patients.
Interventions	Intervention targeted patients Content of intervention: Revelation of physician incentives. Patients provided with two different disclosures of incentive payments to doctors tailored for health plan members of two different types of plan: a capitated plan (payment by capitation and mixed incentive bonus); and an open plan point of service (physicians paid by discounted fee for service and mixed incentive bonus). The disclosure contained information on the structure, direction and possible effects of the incentives in both plans (more versus less care). In the capitation plan intervention more emphasis was given to cost saving aspects of the plan. The description of the physician incentives emphasised more positive features (promoting health, eliminating unnecessary care). These were written in simple English and further reinforced by a subsequent phone call in which the details of the incentives of each plan were read out to subjects, their comprehension checked by a set of simple questions and any errors corrected and checked again. Intervention (disclosure of incentives) took place four to six weeks after baseline measurements and final measurement occurred one month after the intervention. Control group received no training.
Outcomes	Trust as a primary outcome (Table 1). Baseline measurements: Trust, education, demographic measures and health status, knowledge of incentives, number of years with doctor, physician visit volume, prior disputes with and choice of physician or insurer. Outcomes at final measurement (1 month post intervention) ‐ Trust of physician, trust in insurer (using two previously validated scales, Hall 2002b and Zheng 2002)., knowledge of physician incentives, history of disputes with insurer.
Funding	Study funded by Robert Wood Johnstone Foundation.
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Equal number from each plan selected. Sequence generation by computer (information from author).
Allocation concealment (selection bias)	Unclear risk	No mention in text.
Blinding of participants and personnel (performance bias) All outcomes	Unclear risk	No mention in text.
Blinding of outcome assessment (detection bias) All outcomes	Unclear risk	No mention in text.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	Poorly described in text.
Selective reporting (reporting bias)	Low risk	Most outcomes reported.
Other bias	Unclear risk	Protection against contamination: restricted the study to one person per household to avoid contamination. Intention‐to‐treat analysis: Not appropriate. Potential for unit of analysis error for some outcomes?: Yes. Acknowledged and adjusted for. Used a mixed modelling technique. The regressions adjusted for baseline differences between the groups that were associated with trust scores as well as differences in the time of survey.