arXiv subjects

Lee W. Cohnstaedt

Publications and source records attributed to Lee W. Cohnstaedt.

9 recordsLinked to original sources

arXiv2026

A two-parameter, minimal-data model to predict dengue cases: the 2022-2023 outbreak in Florida, USA

Reliable and timely dengue predictions can provide an actionable lead time for targeted vector control and clinical preparedness, reducing preventable disease and health-system costs in at-risk communities. However, many forecasting approaches depend on site-specific covariates and entomological surveillance, which limits portability to data-sparse settings. In this work, we mathematically prove that a parabolic ICC structure, previously established for the basic SIR model, also holds in a substantially more complex model: a two-population (human-mosquito) four-serotype dengue transmission model with primary and secondary infections and mild/severe disease classes. To predict the number of new cases, we propose a data-parsimonious (DP) framework built on the incidence-cumulative cases (ICC) curve that requires only the human incidence time series of the target season and estimates only two parameters, thereby reducing estimation noise and computational burden. We further develop a Bayesian extension that accounts for case-reporting and fitting uncertainty, producing calibrated predictive intervals. The Bayesian model yields improved predictive performance compared to the parabolic ICC model. We evaluate the framework for dengue outbreaks in Florida in 2022-2023, where standardized clinical tests and reporting support accurate case determination.

stat.AP↗

arXiv2026

Optimal Scheduling of Dengue Vector Control

Dengue transmission is shaped by the population dynamics of the Aedes aegypti mosquito, making vector control a central strategy for disease mitigation. The impact of interventions such as larvicide, adulticide, and breeding-site reduction depends critically on their timing under fluctuating environmental conditions. We build on a high-fidelity, non-Markovian mechanistic model of the Aedes life cycle that captures stage-structured, temperature-dependent developmental delays, and mortality, and extend it to incorporate multiple vector control measures. Rather than using continuous abstract control amplitudes as in standard optimal control formulations, we introduce intervention-specific temporal profiles that better reflect operational practice. We then develop an adjoint-based gradient descent framework to compute the optimal timing of a sequence of interventions by minimizing the time-dependent dengue reproduction number, R0. Numerical simulations based on seasonal temperature data from Miami, Florida, show that appropriately timed combinations of interventions can substantially suppress transmission risk, with outcomes strongly influenced by seasonal temperature variation and intervention duration. We further propose embedding the resulting optimization framework within a Model Predictive Control architecture, yielding a closed-loop approach for real-time, surveillance-driven vector management under environmental and operational uncertainty.

math.DS↗

arXiv2026

EpidemIQs: Prompt-to-Paper LLM Agents for Epidemic Modeling and Analysis

Large Language Models (LLMs) offer new opportunities to accelerate complex interdisciplinary research domains. Epidemic modeling, characterized by its complexity and reliance on network science, dynamical systems, epidemiology, and stochastic simulations, represents a prime candidate for leveraging LLM-driven automation. We introduce EpidemIQs, a novel multi-agent LLM framework that integrates user inputs and autonomously conducts literature review, analytical derivation, network modeling, mechanistic modeling, stochastic simulations, data visualization and analysis, and finally documentation of findings in a structured manuscript, through five predefined research phases. We introduce two types of agents: a scientist agent for planning, coordination, reflection, and generation of final results, and a task-expert agent to focus exclusively on one specific duty serving as a tool to the scientist agent. The framework consistently generated complete reports in scientific article format. Specifically, using GPT 4.1 and GPT 4.1 Mini as backbone LLMs for scientist and task-expert agents, respectively, the autonomous process completes with average total token usage 870K at a cost of about $1.57 per study, successfully executing all phases and final report. We evaluate EpidemIQs across several different epidemic scenarios, measuring computational cost, workflow reliability, task success rate, and LLM-as-Judge and human expert reviews to estimate the overall quality and technical correctness of the generated results. Through our experiments, the framework consistently addresses evaluation scenarios with an average task success rate of 79%. We compare EpidemIQs to an iterative single-agent LLM, benefiting from the same system prompts and tools, iteratively planning, invoking tools, and revising outputs until task completion. The comparisons suggest a consistently higher performance of EpidemIQs.

cs.SI↗

arXiv2025

A Data-Parsimonious Model for Long-Term Risk Assessments of West Nile Virus Spillover

Many West Nile virus (WNV) forecasting frameworks incorporate entomological or avian surveillance data, which may be unavailable in some regions. We introduce a novel data-parsimonious probabilistic model to predict both the timing of outbreak onset and the seasonal severity of WNV spillover. Our approach combines a temperature-driven compartmental model of WNV with nonparametric kernel density estimation methods to construct a joint probability density function and a Poisson rate surface as function of mosquito abundance and normalized cumulative temperature. Calibrated on human incidence records, the model produces reliable forecasts several months before the transmission season begins, supporting proactive mitigation efforts. We evaluated the framework across three counties in California (Orange, Los Angeles, and Riverside), two in Texas (Dallas and Harris), and one in Florida (Duval), representing completely different ecology and distinct climatic regimes, and observed strong agreement across multiple performance metrics.

stat.AP↗

arXiv2025

Probabilistic Assessment of West Nile Virus Spillover Risk Using a Compartmental Mechanistic Model

This paper presents a novel probabilistic approach for assessing the risk of West Nile Disease (WND) spillover to the human population. The assessment has been conducted under two different scenarios: (1) assessment of the onset of spillover, and (2) assessment of the severity of the epidemic after the onset of the disease. A compartmental model of differential equations is developed to describe the disease transmission mechanism, and a probability density function for pathogen spillover to humans is derived based on the model for the assessment of the risk of the spillover onset and the severity of the epidemic. The prediction strategy involves making a long-term forecast and then updating it with a short-term (lead time of two weeks or daily). The methodology is demonstrated using detailed outbreak data from high-case counties in California, including Orange County, Los Angeles County, and Kern County. The predicted results are compared with actual infection dates reported by the California Department of Public Health for 2022-2024 to assess prediction accuracy. The performance accuracy is evaluated using a logarithmic scoring system and compared with one of the most renowned predictive models to assess its effectiveness. In all prediction scenarios, the model demonstrated strong performance. Lastly, the method is applied to explore the impact of global warming on spillover risk, revealing an increasing trend in the number of high-risk days and a shift toward a greater proportion of these days over time for the onset of the disease.

stat.AP↗

arXiv2023

Assessing Dengue Risk Globally Using Non-Markovian Models

Dengue is a vector-borne disease transmitted by Aedes mosquitoes. The worldwide spread of these mosquitoes and the increasing disease burden have emphasized the need for a spatio-temporal risk map capable of assessing dengue outbreak conditions and quantifying the outbreak risk. Given that the life cycle of Aedes mosquitoes is strongly influenced by habitat temperature, numerous studies have utilized temperature-dependent development rates of these mosquitoes to construct virus transmission and outbreak risk models. In this study, we advance existing research by developing a mechanistic model for the mosquito life cycle that accurately accounts for the non-Markovian nature of the process. By fitting the model to data on human dengue cases, we estimate several model parameters, allowing the development of a global spatiotemporal dengue risk map. This risk model employs temperature and precipitation data to assess the environmental suitability for dengue outbreaks in a given area. Furthermore, we demonstrate how to reduce the model to the corresponding differential equations, enabling us to utilize existing methods for analyzing the system and fitting the model to observations. This approach can be further applied to similar non-Markovian processes that are currently described with less accurate Markovian models.

stat.AP↗

arXiv2021

A Non-Markovian Model to Assess Contact Tracing for the Containment of COVID-19

COVID-19 remains a challenging global threat with ongoing waves of infections and clinical disease which have resulted millions of deaths and an enormous strain on health systems worldwide. Effective vaccines have been developed for the SARS-CoV-2 virus and administered to billions of people; however, the virus continues to circulate and evolve into new variants for which vaccines may ultimately be less effective. Non-pharmaceutical interventions, such as social distancing, wearing face coverings, and contact tracing, remain important tools, especially at the onset of an outbreak. In this paper, we assess the effectiveness of contact tracing using a non-Markovian, network-based mathematical model. To improve the reliability of the novel model, empirically determined distributions were incorporated for the transition time of model state pairs, such as from exposed to infected states. The first-order closure approximation was used to derive an expression for the epidemic threshold, which is dependent on the number of close contacts. Using survey contact data collected during the 2020 fall academic semester from a university population, we determined that even four to five contacts are sufficient to maintain viral transmission. Additionally, our model reveals that contact tracing can be an effective outbreak mitigation measure by reducing the epidemic size by more than three-fold. Increasing the reliability of epidemic models is critical for accurate public health planning and use as decision support tools. Moving toward more accurate non-Markovian models built upon empirical data is important.

stat.AP↗

arXiv2021

A windowed correlation based feature selection method to improve time series prediction of dengue fever cases

The performance of data-driven prediction models depends on the availability of data samples for model training. A model that learns about dengue fever incidence in a population uses historical data from that corresponding location. Poor performance in prediction can result in places with inadequate data. This work aims to enhance temporally limited dengue case data by methodological addition of epidemically relevant data from nearby locations as predictors (features). A novel framework is presented for windowing incidence data and computing time-shifted correlation-based metrics to quantify feature relevance. The framework ranks incidence data of adjacent locations around a target location by combining the correlation metric with two other metrics: spatial distance and local prevalence. Recurrent neural network-based prediction models achieve up to 33.6% accuracy improvement on average using the proposed method compared to using training data from the target location only. These models achieved mean absolute error (MAE) values as low as 0.128 on [0,1] normalized incidence data for a municipality with the highest dengue prevalence in Brazil's Espirito Santo. When predicting cases aggregated over geographical ecoregions, the models achieved accuracy improvements up to 16.5%, using only 6.5% of incidence data from ranked feature sets. The paper also includes two techniques for windowing time series data: fixed-sized windows and outbreak detection windows. Both of these techniques perform comparably, while the window detection method uses less data for computations. The framework presented in this paper is application-independent, and it could improve the performances of prediction models where data from spatially adjacent locations are available.

cs.LG↗

arXiv2013

A hierarchical network approach for modeling Rift Valley fever epidemics with applications in North America

Rift Valley fever is a vector-borne zoonotic disease which causes high morbidity and mortality in livestock. In the event Rift Valley fever virus is introduced to the United States or other non-endemic areas, understanding the potential patterns of spread and the areas at risk based on disease vectors and hosts will be vital for developing mitigation strategies. Presented here is a general network-based mathematical model of Rift Valley fever. Given a lack of empirical data on disease vector species and their vector competence, this discrete time epidemic model uses stochastic parameters following several PERT distributions to model the dynamic interactions between hosts and likely North American mosquito vectors in dispersed geographic areas. Spatial effects and climate factors are also addressed in the model. The model is applied to a large directed asymmetric network of 3,621 nodes based on actual farms to examine a hypothetical introduction to some counties of Texas, an important ranching area in the United States of America (U.S.A.). The nodes of the networks represent livestock farms, livestock markets, and feedlots, and the links represent cattle movements and mosquito diffusion between different nodes. Cattle and mosquito (Aedes and Culex) populations are treated with different contact networks to assess virus propagation. Rift Valley fever virus spread is assessed under various initial infection conditions (infected mosquito eggs, adults or cattle). A surprising trend is fewer initial infectious organisms result in a longer delay before a larger and more prolonged outbreak. The delay is likely caused by a lack of herd immunity while the infections expands geographically before becoming an epidemic involving many dispersed farms and animals almost simultaneously.

q-bio.PE↗