BACKGROUNDThe COVID-19 pandemic is both a global health crisis, and a civic emergency for national governments, including the UK. With currently no vaccine and no treatment, there is no medical solution. Consequently, the questions for evidence and policy are complex, draw on multiple streams, and are about management rather than medicine. This working paper sets out models and methods for a forthcoming study to analyse the interaction of evidence and policy in the first 100 days of the UK government’s response to COVID-19.MODELSDrawing on chaos theory and insights from the policy sciences, three models of evidence-based policy are discussed: a linear cipher model, a multiple streams model, and a melee model. The nonlinear melee model is adopted for the forthcoming study, in which multiple forms of evidence (science, economic, political, social) and actors interact in real time, and supposedly independent evidence streams concomitantly consider evidence from other streams, creating an apparently chaotic melee in which it is unclear where, how and by whom decisions originate or are made. This is normal, and much closer to how policy decisions emerge than the process represented by rational, ideal-type, linear models.METHODSDESIGN: An analysis of evidence in and of the policy response to COVID-19 by the UK government for the 100 days from 1st February 2020 to 11th May 2020 will be undertaken.DATA: Three data sources will be accessed: papers from the Scientific Advisory Group for Emergencies (SAGE) and its feeder groups and committees; formal government statements, information and guidance; BBC News coverage of government press briefings and further documents and/or coverage snowballing from them.ANALYSIS: A thematic content analysis, pre-structured by the four evidence streams in the melee model, will be undertaken to identify critical incidents for a strict contemporaneous analysis using only information available at the time of the incident, and referring only to the contemporaneous context for the incident. ASSUMPTIONS: Illustrative critical incidents will be sufficient to provide evidence for the explanatory utility of the melee model. Comprehensive coverage of all incidents would unnecessarily super-saturate the analysis.