Table 1 Constituent models providing county-level hospitalization census predictions that are archived on CalCAT and included in the analysis
Model | Forecast update frequency | Forecast horizon | Methods/Approach | Documentation |
|---|---|---|---|---|
Columbia | Weekly | Up to 6 weeks | County level metapopulation model | [6] |
UCSF, COVID NearTerm | Daily | 2–4 weeks | Bootstrap-based method based on an autoregressive model | [7] |
UCB LEMMA | Daily | Up to 4 weeks | SEIR compartmental model with parameters fit using case series data of COVID-19 hospital and ICU census, hospital admissions, deaths, cases and seroprevalence | [8] |
CDPH Simple Growth | Daily | Up to 4 weeks | Assumes new cases grow exponentially according to the rate given by the latest ensemble R-effective. Assumes a fixed severity and average length of stay to generate hospitalizations | [4] |
CalCAT Ensemble | Daily | Up to 4 weeks | The ensemble forecast takes the median of all the forecasts available on a given date and fits a smoothed spline to the trend | [4] |
CA Baseline | Daily | Up to 4 weeks | Retroactive 7-day rolling average mean of past hospitalization values | Methods |