Background: An accessible inventory of fair tests of health policies would be of value when: developing evidence-informed health policies; preparing systematic reviews of effects of health policies; and examining the methods used in randomized trials of health policies. Objectives: 1) To prepare a searchable database of randomized trials of health policies defined as delivery, financial and governance arrangements of health systems; implementation strategies; and public health interventions targeted at populations. 2) To collect examples of other research designs which have yielded convincingly unbiased estimates of the effectiveness of such interventions, in order to assess the value of extending the inventory. 3) To identify early cluster randomised trials (CRTs) to add to the James Lind Library. Methods: Selection criteria were refined following test searches of Medline and the Cochrane Central Register of Controlled Trials (CENTRAL) and pilot screening. We then searched electronically records submitted by the Cochrane Effective Practice and Organisation of Care (EPOC), Consumers and Communication (COMMUN), and Public Health (PUBHEALTH) Groups to CENTRAL. References in key papers and unpublished databases found through personal contacts were hand searched. Search results are screened by one person with a second opinion being sought where there is uncertainty. Results: Electronic searches of CENTRAL resulted in 1085 titles for screening. 300 titles were identified from reviews and personal contacts, 106 of which meet the inclusion criteria. We describe the main categories of interventions in the included trials. Conclusions: The results of hand searching, and searching CENTRAL, suggest that between 500 and 700 health policy CRTs have been published. A wide range of health policies have been evaluated using CRTs, but the number of such trials is uncertain and these trials are difficult to identify. Further searching is necessary to make the inventory as complete as possible.