Table 1 Summary of data sources for pediatric flu hospitalization forecasting.
Data source | Provider | Content | Temporal range | Granularity | Preprocessing | Usage in model |
|---|---|---|---|---|---|---|
CDC FluSurv-NET | CDC | Weekly pediatric influenza hospitalization rates by age group | 2019–2024 | State-level (10 states, including MA) | Zero imputation for weeks 18–39 (non-flu season); aligned by ISO week/year | Used to derive target variable for pediatric influenza hospitalization count |
HHS Protect Data Hub | HHS | Hospital capacity, flu and COVID-19 admissions | 2019–2024 | Facility-level, aggregated to state level | Median imputation for suppressed values (-99999); forward fill remaining missing values (~ 7%) | Used to derive target variable for pediatric influenza hospitalization count |
DOT Mobility Data | U.S. DOT | Daily mobility behavior (“not staying at home” metric) | 2019–2024 | State-Level | Mean by year and week. | Feature: proxy for human movement |
Neighboring State Hospitalizations | HHS Protect Data Hub | Derived Weekly pediatric flu hospitalization rates | 2019–2024 | State-level | Harmonized to MA format; aligned by week | Feature: regional burden indicator |
Public School Calendar | MA Dept. of Education | School holidays & breaks | 2024 | State-level | Coded as binary regressors | Feature: special event effects |
Prophet Implementation | Facebook Prophet (v1.1.4) | Time series forecasting | N/A | N/A | Fourier series (weekly, yearly); L-BFGS optimizer | Model for hospitalization trend prediction |
