Purpose: To identify key model uncertainties to establish a research agenda for model improvement.
Methods: Thirteen experts with varying expertise were interviewed to identify uncertainty sources. The sources were grouped into assumption-clusters, which were scored on their scientific rigor using four so-called pedigree criteria and on the anticipated influence on model outcomes, all using a discrete 0-4 ordinal scale. Scoring was done in a structured expert elicitation workshop following the Numeral, Unit, Spread, Assessment and Pedigree (NUSAP)-approach. The scoring for the influence on end-result was done for the total DALY estimate, total COI estimate, and pathogen-ranking based on DALYs and COI. Those pathogens with low overall pedigree scores (low scientific rigor) and large influence on the outcome are key assumptions that require further study.
Results: The interviews resulted in 117 uniquely mentioned assumptions and uncertainties, grouped into 15 key assumption-clusters. The pedigree and influence scores elicited with NUSAP showed that no assumption scored highly unrigorous and/or influential. The assumption for the correction of seemingly asymptomatic shedding of pathogens in the population has been identified as the most critical among the 15 for all four outcomes. Specifically for the COI outcome, the use of GE incidence data from 1999, although corrected for pathogen specific trends, is considered a priority issue for model improvement.
Significance: Insight in priority uncertainties should be obtained in modeling studies, both for quantitative and qualitative sources. NUSAP is useful to this end to generate research agendas.