Expert elicitation on the relative importance of possible SARS-CoV-2 transmission routes and the effectiveness of mitigations

Abstract

Objectives: To help people make decisions about the most effective mitigation measures against SARS-CoV-2 transmission in different scenarios, the likelihoods of transmission by different routes need to be quantified to some degree (however uncertain). These likelihoods need to be communicated in an appropriate way to illustrate the relative importance of different routes in different scenarios, the likely effectiveness of different mitigation measures along those routes, and the level of uncertainty in those estimates. In this study, a pragmatic expert elicitation was undertaken to supply the underlying quantitative values to produce such a communication tool.

Participants: Twenty-seven individual experts from five countries and many scientific disciplines provided estimates.

Outcome measures: Estimates of transmission parameters, assessments of the quality of the evidence, references to relevant literature, rationales for their estimates and sources of uncertainty.

Results and conclusion: The participants' responses showed that there is still considerable disagreement among experts about the relative importance of different transmission pathways and the effectiveness of different mitigation measures due to a lack of empirical evidence. Despite these disagreements, when pooled, the majority views on each parameter formed an internally consistent set of estimates (for example, that transmission was more likely indoors than outdoors, and at closer range), which formed the basis of a visualisation to help individuals and organisations understand the factors that influence transmission and the potential benefits of different mitigation measures.

Keywords: COVID-19; infection control; public health.

Figure 1

Main known transmission pathways for SARS-CoV-2 from a person with infection to a susceptible person, and potential mitigation measures (in green) that might affect each pathway.

Figure 2

Division of the transmission diagram into different pathways to allow participants to select those that reflected their expertise.

Figure 3

Example of elicited responses to a variable where participants fell into three 'camps': estimates of the proportion of infectious virus emitted via (A) aerosols (<10 μm), (B) small particles (10–100 μm), and (C) large particles (>100 μm) in the exhaled breath from a person infected with SARS-CoV-2 who was not talking. Checking the free text rationales indicated that some thought that transmission occurred almost entirely through larger particles, some almost exclusively by aerosols, and some that it is more evenly spread across all three sizes of particle. Many mentioned the uncertainties caused by problems in measuring the virus. Numbers on right hand axis show individual responses.

Figure 4

Example of an 'average distribution' from three participants in round 1 (blue) and the 70% confidence interval extracted from that to be displayed to participants in round 2 (green). In this example, the variable elicited is the percentage loss of infectious virus during transmission from a susceptible person’s hand to their mucous membranes (given no mitigations in place).

Figure 5

Participants’ estimates of the percentage increase that each of a range of different activities would cause in exhalation of infectious virus by an infected person: (A) speaking, (B) speaking loudly, (C) coughing, (D) singing, (E) exercising, (F) eating and (G) infected with a variant of concern. In each case, the top graph shows individual estimates, the middle graph the average of those estimates, and the bottom graph a cumulative probability distribution, with (in green) the median and interquartile range. Values in parentheses to the right show each participant’s estimate of the percentage likelihood of the true value lying below, within and above the range identified in round 1, shown by the orange lines. Values in parentheses in the bottom graph show 25th, 50th and 75th percentile estimates from the combined responses. The vertical dashed line shows the 100% reference case (that is, no different from silent breathing).

Figure 6

Participants’ estimates of the percentage of infectious virus carried in large (>100 μm) airborne particles that would NOT reach a susceptible person’s mucous membranes if the infected and susceptible person were (A) within 2 m of each other and (B) further than 2 m apart from each other in an unventilated room. In each case, the top graph shows individual estimates, the middle graph the average of those estimates, and the bottom graph a cumulative probability distribution, with (in green) the median and interquartile range. Values in parentheses to the right show percentage below, within, and above the range. Values in parentheses on the left of the bottom graph show 25th, 50th and 75th percentile estimates from the combined responses.

Figure 7

Participants’ individual estimates (top graphs) and pooled estimates (middle and bottom graphs) of the percentage reduction in infections virus transferred from a person’s contaminated hands to their mucous membranes A) under normal conditions, (B) if the person was wearing a face shield, (C) if the person was wearing gloves, (D) & (E) if the person was wearing a face covering (participants divided into two distinct camps on this question), (F) intervals elicited for the first time in round 2 from individuals (top graph, blue) and pooled (bottom graph, green) if the person were practising good hand hygiene. For graphs (A) to (E) the top graph shows individual estimates, the middle graph the average of those estimates, and the bottom graph a cumulative probability distribution, with (in green) the median and inter-quartile range. Values in parentheses to the right show percentage below, within and above the range. Values in parentheses in the bottom graph show 25th, 50th and 75th percentile estimates from the combined responses. The vertical dashed line shows the 100% reference case. For graph (F) the figures on right hand axis show individual responses in the upper graph and the 25th, 50th and 75th percentile values in the lower graph.