Adaptive forecasting and alert evaluation for pediatric RSV and influenza hospitalization burden using public U.S. surveillance data: a time-series study

Pediatric influenza and respiratory syncytial virus (RSV) hospitalization surges strain hospitals and public health agencies. Post-pandemic changes in seasonality and surveillance practice have weakened fixed seasonal baselines. We evaluated whether public surveillance signals can support transparent warning evaluation for pediatric influenza and RSV hospitalization burden. We conducted a retrospective ecological time-series study using aggregated U.S. RESP-NET hospitalization surveillance and related public signal streams from October 2018 through March 2026. Outcomes were weekly pediatric hospitalization rates for influenza and RSV among children aged 0–4 and 5–17 years. We compared a prior-year same-week seasonal naïve baseline with expanding-window ridge models using seasonal/autoregressive terms and feasible public signals, including emergency department (ED) percent-visit indicators. Forecasts were evaluated by rolling-origin validation, with a finalized-data sensitivity excluding the most recent 4 weeks. We summarized mean absolute error (MAE), alert utility, onset lead time, and expected loss under training-quantile alert thresholds. For RSV among children aged 0–4 years, an ED-feasible ridge model improved matched-week forecasts versus the seasonal naïve baseline at horizon 1 under the finalized-data sensitivity (paired MAE improvement 3.49; 95% confidence interval [CI] 1.59 to 5.69; paired n = 104). Influenza gains were smaller and uncertain (ages 0–4 years, horizon 1: paired MAE improvement 0.60; 95% CI -0.24 to 1.24; paired n = 51). Direct matched comparisons of autoregressive/seasonal ridge models with versus without ED terms did not show a stable independent ED point-error gain. At a fixed 90th training-quantile alert threshold and miss: false-alarm cost ratio of 2.0, expected-loss ratios relative to the seasonal baseline were 0.280 for RSV ages 0–4 years and 0.545 for influenza ages 0–4 years. These findings support pathogen-specific, locally calibrated strategies for pediatric respiratory-virus hospitalization warning. Adaptive ridge models improved RSV warning metrics relative to a fixed prior-year baseline, whereas ED terms did not improve matched point-error accuracy beyond AR+season terms. ED signals are therefore a feasible syndromic layer for alert-policy evaluation, not evidence of an independent forecasting gain.

Authors

Institutions

Publication Details

Journal
BMC Public Health
Published
2026-11-14
DOI
https://doi.org/10.1186/s12889-026-30380-2
Primary Topic
Respiratory viral infections research
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Adaptive forecasting and alert evaluation for pediatric RSV and influenza hospitalization burden using public U.S. surveillance data: a time-series study

Yuying Zhang, Yuhong Lin, zhang jingrong, Siwen Chen et al.
BMC Public Health
Respiratory viral infections research
article

Adaptive forecasting and alert evaluation for pediatric RSV and influenza hospitalization burden using public U.S. surveillance data: a time-series study

Yuying Zhang, Yuhong Lin, zhang jingrong, Siwen Chen, Shaomei Lin, Hongwei Zheng
article en

Abstract

Pediatric influenza and respiratory syncytial virus (RSV) hospitalization surges strain hospitals and public health agencies. Post-pandemic changes in seasonality and surveillance practice have weakened fixed seasonal baselines. We evaluated whether public surveillance signals can support transparent warning evaluation for pediatric influenza and RSV hospitalization burden. We conducted a retrospective ecological time-series study using aggregated U.S. RESP-NET hospitalization surveillance and related public signal streams from October 2018 through March 2026. Outcomes were weekly pediatric hospitalization rates for influenza and RSV among children aged 0–4 and 5–17 years. We compared a prior-year same-week seasonal naïve baseline with expanding-window ridge models using seasonal/autoregressive terms and feasible public signals, including emergency department (ED) percent-visit indicators. Forecasts were evaluated by rolling-origin validation, with a finalized-data sensitivity excluding the most recent 4 weeks. We summarized mean absolute error (MAE), alert utility, onset lead time, and expected loss under training-quantile alert thresholds. For RSV among children aged 0–4 years, an ED-feasible ridge model improved matched-week forecasts versus the seasonal naïve baseline at horizon 1 under the finalized-data sensitivity (paired MAE improvement 3.49; 95% confidence interval [CI] 1.59 to 5.69; paired n = 104). Influenza gains were smaller and uncertain (ages 0–4 years, horizon 1: paired MAE improvement 0.60; 95% CI -0.24 to 1.24; paired n = 51). Direct matched comparisons of autoregressive/seasonal ridge models with versus without ED terms did not show a stable independent ED point-error gain. At a fixed 90th training-quantile alert threshold and miss: false-alarm cost ratio of 2.0, expected-loss ratios relative to the seasonal baseline were 0.280 for RSV ages 0–4 years and 0.545 for influenza ages 0–4 years. These findings support pathogen-specific, locally calibrated strategies for pediatric respiratory-virus hospitalization warning. Adaptive ridge models improved RSV warning metrics relative to a fixed prior-year baseline, whereas ED terms did not improve matched point-error accuracy beyond AR+season terms. ED signals are therefore a feasible syndromic layer for alert-policy evaluation, not evidence of an independent forecasting gain.

BMC Public Health
Fujian University of Traditional Chinese Medicine (CN), Beijing University of Chinese Medicine (CN), Quzhou City People's Hospital (CN), The 180th Hospital of PLA (CN), Dongzhimen Hospital Affiliated to Beijing University of Chinese Medicine (CN), Quanzhou Medical College (CN)
Respiratory viral infections research
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon