gms | German Medical Science

Introduction: During epidemic or pandemic situations caused by the outbreak of an emerging infection, studies aiming to evaluate diagnostic tests are faced with the challenges of changing disease characteristics and changing prevalence in the population, as well as the time pressure due to tests being needed urgently both for individual patient care and as part of public health interventions. At the same time, infectious disease models, which are often used to make decisions about interventions, are parametrised on the basis of studies (e.g. for measuring seroprevalence), which require good knowledge of diagnostic test accuracy. We present a joint approach to diagnostic test evaluation and infectious disease modelling, in which predictions from infectious disease models are used in interim analyses for test evaluation, and interim results from the evaluations are used for better model parametrisation.

Methods: Our simulation study is based on the use case of the SARS-CoV-2 pandemic and has three layers; an epidemic layer that simulates infection spread through a population, and is the basis for a modelling layer that simulates infectious disease modelling efforts in the simulated world as well as for an evaluation layer that represents a diagnostic test accuracy study. This study has an adaptive design that allows for interim analyses and early stopping due to efficacy or futility. Models from the modelling layer are used to make decisions about public health interventions. Their parametrisation is initially based on assumed diagnostic test accuracy parameters, but is updated with the interim diagnostic accuracy study results and then the final results as soon as they are available. Modelling also provides a predicted prevalence that is used for sample size adjustment in the diagnostic accuracy study.

Results: Differences in assumed diagnostic test accuracy have a large impact on the results of infectious disease models, especially in the early stages of the pandemic or epidemic. This leaves a large range of output thresholds where incorrectly parametrised models would not trigger interventions but correctly parametrised models would. This range becomes smaller as the outbreak progresses, and inaccuracies in diagnostic test accuracy are mitigated by more data for model fitting. Adaptive designs provide earlier and more valid estimates, and reduce the risk of misinformed decisions substantially. Sample size re-estimation in diagnostic accuracy studies ensures that the studies are powered appropriately even under rapidly changing prevalences.

Discussion: Inaccurate assumptions of diagnostic test accuracy parameters in infectious disease models can cause poorly informed and delayed decisions about public health interventions, especially in the early stages of an epidemic or pandemic. This decision-making process can be improved by faster diagnostic test studies that use adaptive designs that provide interim results about diagnostic test accuracy and re-estimate sample sizes based on predicted prevalences.

Conclusion: Integrating modelling results into diagnostic accuracy studies and vice versa improves the quality of both, and allows for better and faster decisions about population-level interventions.

The authors declare that they have no competing interests.

The authors declare that an ethics committee vote is not required.