Visual preferences for communicating modelling: a global analysis of COVID-19 policy and decision makers

Abstract

Effective communication of modelling results to policy and decision makers has been a longstanding challenge in times of crises. This communication takes many forms - visualisations, reports, presentations - and requires careful consideration to ensure accurate maintenance of the key scientific messages. Science-to-policy communication is further exacerbated when presenting fundamentally uncertain forms of science such as infectious disease modelling and other types of modelled evidence, something which has been understudied. Here we assess the communication and visualisation of infectious disease modelling results to national COVID-19 policy and decision makers in 13 different countries. We present a synthesis of recommendations on what aspects of visuals, graphs, and plots policymakers found to be most helpful in their COVID-19 response work. This work serves as a first evidence base for developing guidelines on the communication and translation of infectious disease modelling into policy.

Keywords: Communication; Modelling; Outbreak; Policy; Visualisation.

Fig. 1

There are many decisions to make when designing visuals for policy communication. Depicted are: (A) hand-drawn sketches, (B) an AI-generated graph displaying uncertainty, created with DALL·E 3, (C) graphic icons, used with permission from Microsoft.

Fig. 2

Example of the graph format used by South Korean modellers when presenting findings to decision makers. Slide depicts predictions on ICU bed numbers for the Omicron wave in April 2022. Reproduced with permission from the authors.

Fig. 3

Examples of modelling visuals from Canada (Public Health Agency of Canada, 2020). (A) Figure presented in one of the Epidemiology and Modelling COVID-19 Technical Briefings from the Public Health Agency of Canada in March 2021. Slide shows mathematical model-based long-range forecasting of the impact of the Alpha variant. The key message is displayed as a title. (B) The same visualisation one month on, in April 2021 with March data overlaid, again emphasising key findings in a clear manner. Reproduced with permission from the Public Health Agency of Canada.

Fig. 4

Example of a ‘watermelon slice’ diagram from members of the UK modelling consortium SPI-M-O. This figure is taken from an early modelling report of the University of Bristol and University of Exeter on the impact of opening schools in late April 2020 (Brooks-Pollock et al., 2020). Figure compares opening primary schools with opening secondary schools, when the infectiousness of children was unknown. Figures such as these, accompanied by clear summary statements, helped illustrate key modelling findings in official SPI-M-O Consensus Statements to policy. Reproduced with permission.

Fig. 5

Example of a ‘rainbow diagram’ generated by the UK modelling consortium SPI-M-O. Figure shows extrapolations for different plausible levels of the effective reproduction number Rt after an easing of restrictions in the UK in mid 2021. The figure was accompanied by clear summary statements outlining the major modelling findings (not shown). Taken from SPI-M-O Consensus Statement 9th June 2021 (SPI-M-O, 2021). Reproduced with permission from the SPI-M-O Secretariat.