Evaluation of FluSight influenza forecasting in the 2021–22 and 2022–23 seasons with a new target laboratory-confirmed influenza hospitalizations

Nature Communications, Journal Year: 2024, Volume and Issue: 15(1)

Published: July 26, 2024

Accurate forecasts can enable more effective public health responses during seasonal influenza epidemics. For the 2021-22 and 2022-23 seasons, 26 forecasting teams provided national jurisdiction-specific probabilistic predictions of weekly confirmed hospital admissions for one-to-four weeks ahead. Forecast skill is evaluated using Weighted Interval Score (WIS), relative WIS, coverage. Six out 23 models outperform baseline model across forecast locations in 12 18 2022-23. Averaging all targets, FluSight ensemble 2

Language: Английский

---

A multi-city COVID-19 categorical forecasting model utilizing wastewater-based epidemiology

The Science of The Total Environment, Journal Year: 2025, Volume and Issue: 960, P. 178172 - 178172

Published: Jan. 1, 2025

The COVID-19 pandemic highlighted shortcomings in forecasting models, such as unreliable inputs/outputs and poor performance at critical points. As remains a threat, it is imperative to improve current approaches by incorporating reliable data alternative targets better inform decision-makers. Wastewater-based epidemiology (WBE) has emerged viable method track transmission, offering more metric than reported cases for outcomes like hospitalizations. Recognizing the natural alignment of wastewater systems with city structures, ideal leveraging WBE data, this study introduces multi-city, wastewater-based model categorically predict Using hospitalization six US cities, accompanied other epidemiological variables, we develop Generalized Additive Model (GAM) generate two categorization types. Hospitaization Capacity Risk Categorization (HCR) predicts burden on healthcare system based number available hospital beds city. Hospitalization Rate Trend (HRT) trajectory growth rate these categorical thresholds, create probabilistic forecasts retrospectively risk trend category cities over 20-month period 1, 2, 3 week windows. We also propose new methodology measure change points, or time periods where sudden changes outbreak dynamics occurred. explore influence predictor hospitalizations, showing its inclusion positively impacts model's performance. With study, are able capacity disease trends novel useful way, giving decision-makers tool

Language: Английский

---

The variations of SIkJalpha model for COVID-19 forecasting and scenario projections

Epidemics, Journal Year: 2023, Volume and Issue: 45, P. 100729 - 100729

Published: Nov. 16, 2023

We proposed the SIkJalpha model at beginning of COVID-19 pandemic (early 2020). Since then, as evolved, more complexities were added to capture crucial factors and variables that can assist with projecting desired future scenarios. Throughout pandemic, multi-model collaborative efforts have been organized predict short-term outcomes (cases, deaths, hospitalizations) long-term scenario projections. participating in five such efforts. This paper presents evolution its many versions used submit these since pandemic. Specifically, we show is an approximation a class epidemiological models. demonstrate how be incorporate various complexities, including under-reporting, multiple variants, waning immunity, contact rates, generate probabilistic outputs.

Language: Английский

---

Infectious disease surveillance needs for the United States: lessons from Covid-19

Frontiers in Public Health, Journal Year: 2024, Volume and Issue: 12

Published: July 15, 2024

The COVID-19 pandemic has highlighted the need to upgrade systems for infectious disease surveillance and forecasting modeling of spread infection, both which inform evidence-based public health guidance policies. Here, we discuss requirements an effective system support decision making during a pandemic, drawing on lessons in U.S., while looking jurisdictions U.S. beyond learn about value specific data types. In this report, define range decisions are required, elements needed these calibrate inputs outputs transmission-dynamic models, types by state, territorial, local, tribal authorities. We actions ensure that such will be available consider contribution efforts improving equity.

Language: Английский

---

Combining models to generate consensus medium-term projections of hospital admissions, occupancy and deaths relating to COVID-19 in England

Royal Society Open Science, Journal Year: 2024, Volume and Issue: 11(5)

Published: May 1, 2024

Mathematical modelling has played an important role in offering informed advice during the COVID-19 pandemic. In England, a cross government and academia collaboration generated medium-term projections (MTPs) of possible epidemic trajectories over future 4-6 weeks from collection epidemiological models. this article, we outline collaborative approach evaluate accuracy combined individual model against data period November 2021-December 2022 when various Omicron subvariants were spreading across England. Using number statistical methods, quantify predictive performance for both MTPs, by evaluating point probabilistic accuracy. Our results illustrate that produced ensemble heterogeneous models, closer fit to than models periods growth or decline, with 90% confidence intervals widest around peaks. We also show MTPs increase robustness reduce biases associated single projection. Learning our experience epidemic, findings highlight importance developing cross-institutional multi-model infectious disease hubs outbreak control.

Language: Английский

---

Flusion: Integrating multiple data sources for accurate influenza predictions

Epidemics, Journal Year: 2024, Volume and Issue: 50, P. 100810 - 100810

Published: Dec. 25, 2024

Over the last ten years, US Centers for Disease Control and Prevention (CDC) has organized an annual influenza forecasting challenge with motivation that accurate probabilistic forecasts could improve situational awareness yield more effective public health actions. Starting 2021/22 season, targets this have been based on hospital admissions reported in CDC's National Healthcare Safety Network (NHSN) surveillance system. Reporting of through NHSN began within few as such only a limited amount historical data are available target signal. To produce presence system, we augmented these two signals longer record: 1) ILI+, which estimates proportion outpatient doctor visits where patient influenza; 2) rates laboratory-confirmed hospitalizations at selected set healthcare facilities. Our model, Flusion, is ensemble model combines machine learning models using gradient boosting quantile regression different feature sets Bayesian autoregressive model. The were trained all three signals, while was signal, admissions; jointly multiple locations. In each week produced quantiles predictive distribution state current following weeks; prediction computed by averaging predictions. Flusion emerged top-performing 2023/24 season. article investigate factors contributing to Flusion's success, find its strong performance primarily driven use from These results indicate value sharing information across locations especially when doing so adds pool training data.

Language: Английский

---

Discovering Time-Varying Public Interest for COVID-19 Case Prediction in South Korea Using Search Engine Queries: Infodemiology Study (Preprint)

Published: June 21, 2024

The number of confirmed coronavirus disease (COVID-19) cases is a crucial indicator policies and lifestyles. Previous studies have attempted to forecast using machine learning techniques that utilize previous case counts search engine queries predetermined by experts. However, they limitations in reflecting temporal variations associated with pandemic dynamics. We propose novel framework extract keywords highly COVID-19, considering their occurrence. aim relevant based on query expansion. Additionally, we examine time-delayed online behavior related public interest COVID-19 adjust for better prediction performance. To capture semantics regarding word embedding models were trained news corpus, the top 100 words "Corona" extracted over 4-month windows. Time-lagged cross-correlation was applied select optimal time lags correlated from expanded queries. Subsequently, EleastcNet regression after reducing feature dimensions principal component analysis time-lagged features predict future daily counts. Our approach successfully depending phase, encompassing directly such as its symptoms, societal impact. Specifically, during first outbreak, linked past infectious outbreaks similar those exhibited high positive correlation. In second phase pandemic, community infections emerged, government's control frequently observed third delta variant “economic crisis” “anxiety” appeared, fatigue. Consequently, windows outperformed methods most 1-14 day ahead predictions. Notably, our showed significantly higher Pearson correlation coefficients than solely predictions 9-11 days (P=.021, P =.004, P=.004), contrast heuristic- symptom-based sets. This study proposes case-prediction model automatically extracts embedding. relied static or heuristic queries, even without prior expert knowledge. results demonstrate capability track shifts changes pandemic.

Language: Английский

---

Democratizing Infectious Disease Modeling: An AI Assistant for Generating, Simulating, and Analyzing Dynamic Models

medRxiv (Cold Spring Harbor Laboratory), Journal Year: 2024, Volume and Issue: unknown

Published: July 17, 2024

Abstract Understanding and forecasting infectious disease spread is pivotal for effective public health management. Traditional dynamic modeling an essential tool characterization prediction, but often requires extensive expertise specialized software, which may not be readily available in low-resource environments. To address these challenges, we introduce AI-powered assistant that utilizes advanced capabilities from OpenAI’s latest models functionality. This enhances the accessibility usability of simulation frameworks by allowing users to generate or modify model configurations through intuitive natural language inputs importing explicit descriptions. Our prototype integrates with established open-source framework called Compartmental Modeling Software (CMS) provide a seamless experience setup analysis. The AI efficiently interprets parameters, constructs accurate files, executes simulations controlled environment, assists result interpretation using analytics tools. It encapsulates expert knowledge adheres best practices support ranging novices modelers. Furthermore, discuss limitations this assistant, particularly its performance complex scenarios where it might inaccurate specifications. By enhancing ease supporting ongoing capacity-building initiatives, believe assistants like one could significantly contribute global efforts empowering researchers, especially regions limited resources, develop refine their independently. innovative approach has potential democratize health, offering scalable solution adapts diverse needs across wide-range geographies, languages, populations.

Language: Английский

---

COVID-19 trends across borders: Identifying correlations among countries

Journal of International Medical Research, Journal Year: 2024, Volume and Issue: 52(7)

Published: July 1, 2024

Objectives To enhance the accuracy of forecasting future coronavirus disease 2019 (COVID-19) cases and trends by identifying analyzing correlations between daily case counts different countries reported January 2020 2023, to uncover significant links in COVID-19 patterns nations, allowing for real-time, precise predictions spread based on observed correlated countries. Methods Daily each country were tracked 2023 identify nations. Current data obtained from reliable sources, such as Johns Hopkins University World Health Organization. Data analyzed Microsoft Excel using Pearson’s correlation coefficient assess strength connections. Results Strong (r > 0.80) revealed numerous across various continents. Specifically, 62 nations showed with at least one (connected) per nation. These indicate a similarity over past 3 or more years. Conclusion This study addresses gap country-specific within methodologies. The proposed method offers essential real-time insights aid effective government organizational planning response pandemic.

Language: Английский

---

Increasing situational awareness through nowcasting of the reproduction number

Frontiers in Public Health, Journal Year: 2024, Volume and Issue: 12

Published: Aug. 21, 2024

Background The time-varying reproduction number R is a critical variable for situational awareness during infectious disease outbreaks; however, delays between infection and reporting of cases hinder its accurate estimation in real-time. A nowcasting methods, leveraging available information on data consolidation delays, have been proposed to mitigate this problem. Methods In work, we retrospectively validate the use algorithm 18 months COVID-19 pandemic Italy by quantitatively assessing performance against standard methods R. Results Nowcasting significantly reduced median lag from 13 8 days, while concurrently enhancing accuracy. Furthermore, it allowed detection periods epidemic growth with lead 6 23 days. Conclusions augments awareness, empowering better informed public health responses.

Language: Английский

---