Fig 1.
Illustration of the reporting delay structure, with elements of the 2-dimensional array.
Horizontal axis represents the report delay and vertical axis the specimen date. Complete data per specimen date correspond to the sum of each row across the reporting delays. Each cell represents the case count for a given specimen date and reporting delay. Case counts are unknown in real-time when d > T-t, represented here by blue cells. The lower triangular part of the matrix, represented by the yellow cells, are the observed data, which we refer to as the reporting triangle.
Table 1.
Summary of key model structures, assumptions, and characteristics to compare for each model. The model runtime refers to the time taken for the nowcast model to fit, perform inference and post-processing to occur. The posterior samples are the number of samples taken from the model fit, with the burn-in referring to the number of warm up samples taken before estimating the nowcast.
Fig 2.
The backfilling of norovirus cases over the Winter 2023/2024 season.
Without looking at the final revised cases more recent trends appear to tail off due to reporting delays. (a) daily counts of cases. The solid colour lines show the “initial” count of cases uploaded by the end of the week, the dotted black line shows the final “revised” counts uploaded after the week’s end. (b) The count of cases reported by the end of each week denoted by “initial”, with the additional cases reported after the weeks end denoted by “revised”. The end date for each week was taken as a Sunday, to produce a nowcast of data from the previous week.
Fig 3.
Time delay distribution of days between specimen date and report date.
Includes complete data from 02-10-2023 to 10-03-2024.
Fig 4.
Daily predictions from all models with 50% and 90% prediction intervals against initial and final reported count of cases.
Fig 5.
Weekly predictions from all models with 50% (box) and 90% (whiskers) prediction intervals against initial and final reported count of cases.
The weekly predictions are created as the sum of sample predictions per week.
Table 2.
Breakdown of overall model scores by temporal granularity. The daily granularity shows the average daily score over the time series. The weekly granularity shows the average weekly score over the time series. The most optimal score by temporal granularity and scoring metric is in bold.
Fig 6.
Daily count of final and initial reported cases (a) with daily mean model scores for each prediction week.
The Weighted Interval Score (b), Weighted Interval Skill Score (c), Bias (d) and Coverage deviation (e) are given across models and time.
Fig 7.
Model scores, Weighted Interval Score (WIS), weighted interval skill score, bias and coverage deviation, averaged over each day of prediction.
A Monday has near complete data, whereas a Sunday has many cases not yet reported. The scores are the average over the evaluation period.