Summary

An effective instrument for situational awareness and decision-support for policymakers is infectious disease modelling. The COVID-19 attemodelling mpts, however, ran into a number of difficulties, from bad data to shifting laws and people’s behaviour.

We offer a narrative review with a systematic methodology that statistically evaluated prospective, data-driven modelling studies of COVID-19 in the USA in order to draw useful conclusions from the substantial body of COVID-19 modelling research that is now accessible. We examined 136 publications and concentrated on model characteristics that are crucial for decision-makers. We’ve made a record of the forecasted window.

For each research, the technique, prediction goal, datasets utilised, and geographic resolution. We also discovered that a sizable portion of publications did not disclose restrictions (36%), uncertainty (50%), or performance evaluation (25%). We suggest adopting the EPIFORGE 2020 model reporting criteria and developing an information-sharing system that is appropriate for the quick-moving field of infectious disease outbreak science in order to close some of these gaps.

Introduction

With more than 6 million confirmed fatalities worldwide as of September 7, 2022, the COVID-19 pandemic has escalated into an unparalleled public health disaster due to its protracted effects on health and disruption of economic and social life.
1 Mathematical modelling of present and future patterns of infectious disease outbreaks has traditionally been a useful tool to support planning and response efforts during a pandemic. Models for nowcasting and forecasting can help with situational knowledge of the present and upcoming trends in disease propagation.

While scenario modelling and long-term forecasts can provide insight into potential outcomes of a set of assumptions. Modeling insights can help people reduce their personal risks, and they can also improve policymakers’ decision-making when they want to minimise harm to a population as a whole.

These observations have traditionally been made in peer-reviewed literature, which is a priceless resource for disseminating the most recent scientific modelling. A staggering number of research publications have been published during the COVID-19 pandemic: roughly 125 000 within 10 months of the first confirmed case, of which about 30 000 are preprints. 2 Journals prioritised the speedy exchange of COVID-19 information in this crowded publishing environment, but there is a trade-off between expediting peer review and guaranteeing high-quality research.

Preprints played a significant part in the dissemination of COVID-19 research as well. Preprints frequently received media attention, attracted big viewers on social media sites like Twitter, and occasionally led to serious misunderstandings. 2 With regard to COVID-19 modelling specifically, the use of models for guiding response operations has received criticism mostly due to a few notably inaccurate estimates at the beginning of the epidemic and a lack of clarity around the kind of knowledge that models can and cannot offer.

Due in large part to the quick rate of publishing on preprint sites and in peer-reviewed publications, literature reviews that seek to synthesise COVID-19 modelling works, released up to the time of this Series issue, provide an incomplete, fragmented picture of modelling work. To the best of our knowledge, the majority of currently published reviews are either narrative in nature and do not provide a strategy to assess a representative group of articles,

They are systematic but only cover a small time period (e.g., up to July 2020). 10, 11, 12 The only outliers we discovered were one narrative review that included 50 of the most frequently referenced publications and one systematic review that covered 242 papers until November 2020.

Preprints were only included in one review13, and all reviews are only valid for publications that were published before August 2020,7, 8, 9, 10, or in 2020. 12, 13 Numerous of these assessments are centred on the goals and methods of the models8, 9, and 12, while ignoring other aspects of modelling that are essential for communicating research to the general public and decision-makers.

Building on earlier work, we propose a narrative review with a systematic approach in this Series paper. This technique addresses the difficulties in synthesising a large body of work with objective criteria in order to produce the most representative and educational sample of publications. Our assessment includes articles up to August 20, 2021, which includes 8 months of 2021 that other reviews have not yet covered. We concentrate on modelling aspects that have been overlooked in previous work, such as input data, uncertainty,

Performance evaluation and constraints are essential for science translation and allow models to give the general public and decision-makers information. In order to draw sound and convincing conclusions concerning trends and areas in need of development with regard to modelling COVID-19 and upcoming pandemics, we give a quantitative evaluation of each of these components.