Table 1.
Demographics of confirmed mpox cases from the mpox linelist (as of 31 August 2022 and 31 December 2022).
For statistical disclosure, counts under 10 are removed.
Fig 1.
Coverage of the key variables used in the nowcasting during the study period.
Black points show the data, the blue line is a binomial GAM smooth through the data. (A) shows the proportion of reported cases with specimen date recorded, plotted by reporting date. (B) shows the proportion of reported cases with symptom onset date recorded, plotted by reporting date.
Fig 2.
Latest nowcast by specimen date.
Left panel–instantaneous growth rate, the shaded region indicates 95% confidence intervals, with the solid line indicating the median estimate. Right panel–nowcast modelled incidence, the shaded region indicates 95% prediction intervals of the model, with the solid line indicating the median nowcast.
Fig 3.
Latest nowcast by symptom onset date.
Left panel–instantaneous growth rate, the shaded region indicates 95% confidence intervals, with the solid line indicating the median estimate. Right panel–nowcast modelled incidence, the shaded region indicates 95% prediction intervals of the model, with the solid line indicating the median nowcast.
Fig 4.
Growth rate by date of report.
Left panel–instantaneous growth rate, the shaded region indicates 95% confidence intervals, with the solid line indicating the median estimate. Right panel–nowcast modelled incidence, the shaded region indicates 95% prediction intervals of the model, with the solid line indicating the median nowcast.
Fig 5.
Latest regional nowcast by specimen date.
Upper panels–instantaneous growth rate. Lower panels–nowcast modelled incidence. In the growth rate plots, the shaded regions indicate 95% confidence intervals, with the solid lines indicating median estimates. The shaded regions in the nowcasting plots are 95% prediction intervals of the model, with the solid line indicating the median nowcast.
Table 2.
* indicates the best performing model in each pair.
Table 3.
Knots scoring.
Table 4.
Parametric versus non-parametric scoring.
Table 5.
Data modification scoring.
Fig 6.
Specimen date historic nowcasting.
Left panel–nowcast generated on 2 June 2022. Right panel–nowcast generated on 25 July 2022. Old data (red) are those available at the time the nowcast was produced. Backfilled data (blue) is the true value of the data. The shaded region (grey) is the 95% prediction interval of the nowcasting model, with median value marked by the dashed line.
Fig 7.
Symptom onset date historic nowcasting.
Left panel–nowcast generated on 2 June 2022. Right panel–nowcast generated on 25 July 2022. Old data (red) are those available at the time the nowcast was produced. Backfilled data (blue) is the true value of the data. The shaded region (grey) is the 95% prediction interval of the nowcasting model, with median value marked by the dashed line.
Fig 8.
Performance of the non-parametric and parametric nowcasting models by symptom onset date.
The grey bars indicate the data available at the time of the nowcast, the white bars are the complete data, and the lines are the nowcasting projection, with 95% prediction intervals in the ribbon. Each panel shows a different lead time, where the nowcasting model is trying to predict that many days prior to the current date.
Fig 9.
Performance of the non-parametric and parametric nowcasting models by specimen date.
The grey bars indicate the data available at the time of the nowcast, the white bars are the complete data, and the lines are the nowcasting projection, with 95% prediction intervals in the ribbon. Each panel shows a different lead time, where the nowcasting model is trying to predict that many days prior to the current date.
Fig 10.
Bias in the nowcasting model over time, averaged over lead times from 0 days to 8 days.
The upper panel shows the results for specimen date nowcasting, the lower panel the results for symptom onset date nowcasting.