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

TL;DR

This study introduces a data-parsimonious incidence-vs-cumulative (ICC) framework for dengue forecasting by extending ICC from a basic SIR to a two-population SEIR model and pairing it with a Bayesian uncertainty mechanism. The method relies only on the current season's incidence data, achieving competitive predictive performance with substantially lower data and computational demands than traditional models. Applied to Florida's 2022–2023 outbreaks, it identifies four distinct outbreak seasons and demonstrates robust short-horizon forecasts, with Bayesian predictions providing calibrated uncertainty intervals and often improved accuracy over a censored-Poisson baseline. The approach offers practical value for dengue early detection in data-limited settings and newly affected regions, where extensive entomological or historical data are unavailable.

Abstract

Reliable and timely dengue predictions provide actionable lead time for targeted vector control and clinical preparedness, reducing preventable diseases and health-system costs in at-risk communities. Dengue forecasting often relies on site-specific covariates and entomological data, limiting generalizability in data-sparse settings. We propose a data-parsimonious (DP) framework based on the incidence versus cumulative cases (ICC) curve, extending it from a basic SIR to a two-population SEIR model for dengue. Our DP model uses only the target season's incidence time series and estimates only two parameters, reducing noise and computational burden. A Bayesian extension quantifies the case reporting and fitting uncertainty to produce calibrated predictive intervals. We evaluated the performance of the DP model in the 2022-2023 outbreaks in Florida, where standardized clinical tests and reporting support accurate case determination. The DP framework demonstrates competitive predictive performance at substantially lower computational cost than more elaborate models, making it suitable for dengue early detection where dense surveillance or long historical records are unavailable.

Figures (8)

• Figure 1: Time series of new cases in Florida for 2022-2023, highlighting the outbreak seasons.
• Figure 2: Fitted parabola and logistic function to the data at the beginning of the outbreak. As it is seen from parabola, L is prediction of the final size of the epidemic which has been used to fit the logistic curve of the epidemic trajectory.
• Figure 3: (a) Shows the logistic trajectory of the epidemic obtained by ICC-curve method and the values of prediction for four weeks ahead and the real reported cases; (b) Shows the result of the prediction based on the censored Poisson model. The prediction obtained in panel (a) is used as rate for the Poisson model. Also for each week winner epidemiological bins has been shown by orange region
• Figure 4: One– to four–week ahead predictions of weekly cumulative cases versus reported values in Florida (2022–2023). Each panel corresponds to a forecast horizon (1, 2, 3, or 4 weeks). Reported cases are on the horizontal axis; predicted cases are on the vertical axis. The red dashed line
• Figure 5: Forecasts of cumulative cases by week and their probability densities for four Florida outbreaks (2022–2023) at 1–4-week horizons. Black dots mark observed cases; lighter shading denotes higher probability density.