Enhancing long-term forecasting: Learning from COVID-19 models

Mathematical analysis of simple behavioral epidemic models

COVID-19 highlighted the importance of considering human behavior change when modeling disease dynamics. This led to developing various models that incorporate human behavior. Our objective is to contribute to an in-depth, mathematical examination of such models. Here, we consider a simple deterministic compartmental model with endogenous incorporation of human behavior (i.e., behavioral feedback) through transmission in a classic Susceptible-Exposed-Infectious-Recovered (SEIR) structure. Despite its simplicity, the SEIR structure with behavior (SEIRb) was shown to perform well in forecasting, especially compared to more complicated models. We contrast this model with an SEIR model that excludes endogenous incorporation of behavior. Both models assume permanent immunity to COVID-19, so we also consider a modification of the models which include waning immunity (SEIRS and SEIRSb). We perform equilibria, sensitivity, and identifiability analyses on all models and examine the fidelity of the models to replicate COVID-19 data across the United States. Endogenous incorporation of behavior significantly improves a model's ability to produce realistic outbreaks. While the two endogenous models are similar with respect to identifiability and sensitivity, the SEIRSb model, with the more accurate assumption of the waning immunity, strengthens the initial SEIRb model by allowing for the existence of an endemic equilibrium, a realistic feature of COVID-19 dynamics. When fitting the model to data, we further consider the addition of simple seasonality affecting disease transmission to highlight the explanatory power of the models.

Incorporating endogenous human behavior in models of COVID-19 transmission: A systematic scoping review

Background

During the COVID-19 pandemic there was a plethora of dynamical forecasting models created, but their ability to effectively describe future trajectories of disease was mixed. A major challenge in evaluating future case trends was forecasting the behavior of individuals. When behavior was incorporated into models, it was primarily incorporated exogenously (e.g., fitting to cellphone mobility data). Fewer models incorporated behavior endogenously (e.g., dynamically changing a model parameter throughout the simulation).

Methods

This review aimed to qualitatively characterize models that included an adaptive (endogenous) behavioral element in the context of COVID-19 transmission. We categorized studies into three approaches: 1) feedback loops, 2) game theory/utility theory, and 3) information/opinion spread.

Findings

Of the 92 included studies, 72% employed a feedback loop, 27% used game/utility theory, and 9% used a model if information/opinion spread. Among all studies, 89% used a compartmental model alone or in combination with other model types. Similarly, 15% used a network model, 11% used an agent-based model, 7% used a system dynamics model, and 1% used a Markov chain model. Descriptors of behavior change included mask-wearing, social distancing, vaccination, and others. Sixty-eight percent of studies calibrated their model to observed data and 25% compared simulated forecasts to observed data. Forty-one percent of studies compared versions of their model with and without endogenous behavior. Models with endogenous behavior tended to show a smaller and delayed initial peak with subsequent periodic waves.

Interpretation

While many COVID-19 models incorporated behavior exogenously, these approaches may fail to capture future adaptations in human behavior, resulting in under- or overestimates of disease burden. By incorporating behavior endogenously, the next generation of infectious disease models could more effectively predict outcomes so that decision makers can better prepare for and respond to epidemics.

Funding

This study was funded in-part by Centers for Disease Control and Prevention (CDC) MInD-Healthcare Program (1U01CK000536), the National Science Foundation (NSF) Modeling Dynamic Disease-Behavior Feedbacks for Improved Epidemic Prediction and Response grant (2229996), and the NSF PIPP Phase I: Evaluating the Effectiveness of Messaging and Modeling during Pandemics grant (2200256).

Parameter estimation in behavioral epidemic models with endogenous societal risk-response

Behavioral epidemic models incorporating endogenous societal risk-response, where changes in risk perceptions prompt adjustments in contact rates, are crucial for predicting pandemic trajectories. Accurate parameter estimation in these models is vital for validation and precise projections. However, few studies have examined the problem of identifiability in models where disease and behavior parameters must be jointly estimated. To address this gap, we conduct simulation experiments to assess the effect on parameter estimation accuracy of a) delayed risk response, b) neglecting behavioral response in model structure, and c) integrating disease and public behavior data. Our findings reveal systematic biases in estimating behavior parameters even with comprehensive and accurate disease data and a well-structured simulation model when data are limited to the first wave. This is due to the significant delay between evolving risks and societal reactions, corresponding to the duration of a pandemic wave. Moreover, we demonstrate that conventional SEIR models, which disregard behavioral changes, may fit well in the early stages of a pandemic but exhibit significant errors after the initial peak. Furthermore, early on, relatively small data samples of public behavior, such as mobility, can significantly improve estimation accuracy. However, the marginal benefits decline as the pandemic progresses. These results highlight the challenges associated with the joint estimation of disease and behavior parameters in a behavioral epidemic model.

Why Similar Policies Resulted In Different COVID-19 Outcomes: How Responsiveness And Culture Influenced Mortality Rates

In the first two years of the COVID-19 pandemic, per capita mortality varied by more than a hundredfold across countries, despite most implementing similar nonpharmaceutical interventions. Factors such as policy stringency, gross domestic product, and age distribution explain only a small fraction of mortality variation. To address this puzzle, we built on a previously validated pandemic model in which perceived risk altered societal responses affecting SARS-CoV-2 transmission. Using data from more than 100 countries, we found that a key factor explaining heterogeneous death rates was not the policy responses themselves but rather variation in responsiveness. Responsiveness measures how sensitive communities are to evolving mortality risks and how readily they adopt nonpharmaceutical interventions in response, to curb transmission. We further found that responsiveness correlated with two cultural constructs across countries: uncertainty avoidance and power distance. Our findings show that more responsive adoption of similar policies saves many lives, with important implications for the design and implementation of responses to future outbreaks.

Validation framework for epidemiological models with application to COVID-19 models

Mathematical models have been an important tool during the COVID-19 pandemic, for example to predict demand of critical resources such as medical devices, personal protective equipment and diagnostic tests. Many COVID-19 models have been developed. However, there is relatively little information available regarding reliability of model predictions. Here we present a general model validation framework for epidemiological models focused around predictive capability for questions relevant to decision-making end-users. COVID-19 models are typically comprised of multiple releases, and provide predictions for multiple localities, and these characteristics are systematically accounted for in the framework, which is based around a set of validation scores or metrics that quantify model accuracy of specific quantities of interest including: date of peak, magnitude of peak, rate of recovery, and monthly cumulative counts. We applied the framework to retrospectively assess accuracy of death predictions for four COVID-19 models, and accuracy of hospitalization predictions for one COVID-19 model (models for which sufficient data was publicly available). When predicting date of peak deaths, the most accurate model had errors of approximately 15 days or less, for releases 3-6 weeks in advance of the peak. Death peak magnitude relative errors were generally in the 50% range 3-6 weeks before peak. Hospitalization predictions were less accurate than death predictions. All models were highly variable in predictive accuracy across regions. Overall, our framework provides a wealth of information on the predictive accuracy of epidemiological models and could be used in future epidemics to evaluate new models or support existing modeling methodologies, and thereby aid in informed model-based public health decision making. The code for the validation framework is available at https://doi.org/10.5281/zenodo.7102854.

A system dynamics model of the COVID-19 pandemic considering risk perception: A case study of Iran

The new coronavirus disease 2019 (COVID-19) has become a complex issue around the world. As the disease advancing and death rates are continuously increasing, governments are trying to control the situation by implementing different response policies. In order to implement appropriate policies, we need to consider the behavior of the people. Risk perception (RP) is a critical component in many health behavior change theories studies. People's RP can shape their behavior. This research presents a system dynamics (SD) model of the COVID-19 outbreak considering RP. The proposed model considers effective factors on RP, including different media types, awareness, and public acceptable death rate. In addition, the simplifying assumption of permanent immunity due to infection has been eliminated, and reinfection is considered; thus, different waves of the pandemic have been simulated. Using the presented model, the trend of advancing and death rates due to the COVID-19 pandemic in Iran can be predicted. Some policies are proposed for pandemic management. Policies are categorized as the capacity of hospitals, preventive behaviors, and accepted death rate. The results show that the proposed policies are effective. In this case, reducing the accepted death rate was the most effective policy to manage the pandemics. About 20% reduction in the accepted death rate causes about 23% reduction in cumulative death and delays at epidemic peak. The mean daily error in predicting the death rate is less than 10%.

Prediction models for the impact of the COVID-19 pandemic on research activities of Japanese nursing researchers using deep learning

AIM

This study aimed to construct and evaluate prediction models using deep learning to explore the impact of attributes and lifestyle factors on research activities of nursing researchers during the COVID-19 pandemic.

METHODS

A secondary data analysis was conducted from a cross-sectional online survey by the Japanese Society of Nursing Science at the inception of the COVID-19 pandemic. A total of 1089 respondents from nursing faculties were divided into a training dataset and a test dataset. We constructed two prediction models with the training dataset using artificial intelligence (AI) predictive analysis tools; motivation and time were used as predictor items for negative impact on research activities. Predictive factors were attributes, lifestyle, and predictor items for each other. The models' accuracy and internal validity were evaluated using an ordinal logistic regression analysis to assess goodness-of-fit; the test dataset was used to assess external validity. Predicted contributions by each factor were also calculated.

RESULTS

The models' accuracy and goodness-of-fit were good. The prediction contribution analysis showed that no increase in research motivation and lack of increase in research time strongly influenced each other. Other factors that negatively influenced research motivation and research time were residing outside the special alert area and lecturer position and living with partner/spouse and associate professor position, respectively.

CONCLUSIONS

Deep learning is a research method enabling early prediction of unexpected events, suggesting new applicability in nursing science. To continue research activities during the COVID-19 pandemic and future contingencies, the research environment needs to be improved, workload corrected by position, and considered in terms of work-life balance.

Coupling an individual adaptive-decision model with a SIRV model of influenza vaccination reveals new insights for epidemic control

Past seasonal influenza epidemics and vaccination experience may affect individuals' decisions on whether to be vaccinated or not, decisions that may be constantly reassessed in relation to recent influenza related experience. To understand the potentially complex interaction between experience and decisions and whether the vaccination rate is likely to reach a critical coverage level or not, we construct an adaptive-decision model. This model is then coupled with an influenza vaccination dynamics (SIRV) model to explore the interaction between individuals' decision-making and an influenza epidemic. Nonlinear least squares estimation is used to obtain the best-fit parameter values in the SIRV model based on data on new influenza-like illness (ILI) cases in Texas. Uncertainty and sensitivity analyses are then carried out to determine the impact of key parameters of the adaptive decision-making model on the ILI epidemic. The results showed that the necessary critical coverage rate of ILI vaccination could not be reached by voluntary vaccination. However, it could be reached in the fourth year if mass media reports improved individuals' memory of past vaccination experience. Individuals' memory of past vaccination experience, the proportion with histories of past vaccinations and the perceived cost of vaccination are important factors determining whether an ILI epidemic can be effectively controlled or not. Therefore, health authorities should guide people to improve their memory of past vaccination experience through media reports, publish timely data on annual vaccination proportions and adjust relevant measures to appropriately reduce vaccination perceived cost, in order to effectively control an ILI epidemic.

Effect of mandating vaccination on COVID-19 cases in colleges and universities

BACKGROUND

With the introduction of COVID-19 vaccines, many colleges and universities decided to mandate vaccination for all students and employees. The objective of this paper is to empirically investigate the effect of the mandate policy on Fall 2021 COVID-19 cases in institutions of higher education.

METHOD

We construct a unique dataset of a sample of 94 colleges and universities in the east and southeast regions of the United States, 41 of which required vaccination prior to Fall 2021. A difference-in-differences analysis is conducted, considering vaccine requirement as a policy implemented only in a sub-group of these institutions. We control for several factors, including state-level case per capita and student population.

RESULTS

Our analysis shows that mandatory vaccination substantially decreased cases in institutions of higher education by 1,473 cases per 100,000 student population (95 CI: 132, 2813).

CONCLUSIONS

The results suggest that a COVID-19 vaccine requirement is an effective policy in decreasing cases in such institutions, leading to a safer educational experience.

Quantifying the COVID-19 endgame: Is a new normal within reach?

Eradication of COVID-19 is out of reach. Are we close to a "new normal" in which people can leave behind restrictive non-pharmaceutical interventions (NPIs) yet face a tolerable burden of disease? The answer depends on the ongoing risks versus communities' tolerance for those risks. Using a detailed model of the COVID-19 pandemic spanning 93 countries, we estimate the biological and behavioral factors determining the risks and responses, and project the likely course of COVID-19. Infection fatality rates have fallen significantly due to vaccination, prior infections, better treatments, and the less severe Omicron variant. Yet based on their estimated tolerance for deaths, most nations are not ready to live with COVID-19 without any NPIs. Across the world the increased transmissibility of Omicron, combined with the decay of immunity, leads to repeated episodes of reinfections, hospitalizations, and deaths, complicating the emergence of a new normal in many nations. © 2022 The Authors. System Dynamics Review published by John Wiley & Sons Ltd on behalf of System Dynamics Society.

Behavioral responses to risk promote vaccinating high-contact individuals first

How should communities prioritize COVID-19 vaccinations? Prior studies found that prioritizing the elderly and most vulnerable minimizes deaths. However, prior research has ignored how behavioral responses to risk of disease endogenously change transmission rates. We show that incorporating risk-driven behavioral responses enhances fit to data and may change prioritization to vaccinating high-contact individuals. Behavioral responses matter because deaths grow exponentially until communities are compelled to reduce contacts, with deaths stabilizing at levels that oblige higher-contact groups to sufficiently cut their interactions and slow transmissions. More lives may be saved by vaccinating and taking those high-contact groups out of transmission chains earlier because the remaining groups will take more precautions while waiting for their turn for vaccination. These findings are especially important considering the need for further vaccination in many countries, the emergence of new variants, and the expected challenge of distributing new vaccines in the coming months and years. © 2022 The Author. System Dynamics Review published by John Wiley & Sons Ltd on behalf of System Dynamics Society.