Multi-pathogen situational assessment and forecasting of respiratory disease in Aotearoa New Zealand

TL;DR

Two models that were used in a 2025 New Zealand winter situational assessment programme for three respiratory pathogens, SARS-CoV-2, influenza and respiratory syncytial virus, are presented and it is concluded that real-time analyses performed reasonably well.

Abstract

Real-time analysis of epidemic trends and forecasts can help support public health planning and the response to seasonal respiratory disease. Here, we present two models that were used in a 2025 New Zealand winter situational assessment programme for three respiratory pathogens: SARS-CoV-2, influenza and respiratory syncytial virus (RSV). These models were run weekly from May to October 2025 on real-time disease surveillance data and provided a quantitative representation of the current epidemic trend, along with estimates of the epidemic growth rate and 28-day ahead forecasts of case incidence. Model results and interpretation were provided in weekly reports to public health partners as part of a trans-Tasman winter programme run by the Australia--Aotearoa Consortium for Epidemic Forecasting and Analytics (ACEFA). We compare in-season results that were included in these reports to a retrospective analysis of the complete data for the season. We conclude that real-time analyses performed reasonably well, and identify some areas for improvement in future winter situational assessment programmes.

Figures (9)

• Figure 1: Modelled trends in SARS-CoV-2 cases fitted to full season data. (A) Daily SARS-CoV-2 cases (points) and daily SARS-CoV-2 cases modelled using a Bayesian P-spline model, including a day-of-the-week effect (line and shaded region). (B) Modelled trend in SARS-CoV-2 cases (with day-of-the-week effect removed yet still included in the model). (C) Growth rate inferred from the modelled trend in SARS-CoV-2 cases (with day-of-the-week effect removed). The dashed line indicates a growth rate of zero reflecting the threshold between epidemic growth or decline. For all panels we include the model posterior distribution's median (line) and 50% and 95% credible intervals (dark shaded and light shaded regions).
• Figure 2: Modelled trends in hospitalisations fitted to full season data for: (A) SARS-CoV-2; (B) influenza; and (C) RSV. In the top panel for each pathogen we plot the daily number of hospitalisations (points) and modelled trend in hospitalisations (line and shaded regions) using a Bayesian P-spline model that does not include a day-of-the-week effect. In the bottom panel for each pathogen we plot the growth rate inferred from the modelled trend in hospitalisations. The dashed line indicates a growth rate of zero (the threshold between epidemic growth or decline). For all panels we include the model posterior distribution's median (line) and 50% and 95% credible intervals (dark shaded and light shaded regions).
• Figure 3: Modelled trends in SARS-CoV-2 cases reported in real-time over the course of the season. (A) Modelled trends in SARS-CoV-2 cases including day-of-the-week effect (lines and shaded region). (B) Modelled trends in SARS-CoV-2 cases with day-of-the-week effect removed (lines and shaded region). (C) Growth rate (lines and shaded region) inferred from the modelled trend in SARS-CoV-2 cases (with day-of-the-week effect removed). For all panels real-time estimates (coloured lines and shaded regions) are compared to 95% credible intervals of end-of-season estimates (black dashed lines). For all real-time estimates (coloured by round), we include the model posterior distribution's median (line) and 50% and 95% credible intervals (dark shaded and light shaded regions). We include the real-time model estimates up to the final day of data (for periods greater than the previous rounds final day of data). In (A) and (B) the daily SARS-CoV-2 cases for the final end-of-season dataset (black points) and the daily SARS-CoV-2 cases in the real-time datasets (points coloured by round) are connected by lines to highlight revisions to data used in real-time analyses. For example, during the fourth round (first pink) there are visible pink data points connected directly to the black end-of-season data; the data in pink is the data available when models were estimated, the data was revised upwards in the following round where it matched the end-of-season data. Note that when coloured points (and connecting lines) can not be seen then there has been no revisions to the data. When a coloured point can be seen and is the same colour as the real-time estimate for the same period (and connected directly to a black data point), it indicates that the number of cases for a specific day was revised (upwards in most instances) in the following round of data, but was not revised in any further rounds of data).
• Figure 4: Modelled trends in hospitalisations reported in real-time over the course of the season. Modelled trends in hospitalisations are shown for: (A) SARS-CoV-2; (B) influenza; and (C) RSV. For each pathogen we plot: (top panels) modelled trends in hospitalisations (lines and shaded region); and (bottom panels) growth rate (lines and shaded region) inferred from the modelled trend in hospitalisations. For all panels real-time estimates (coloured lines and shaded regions) are compared to 95% credible intervals of end-of-season estimates (black dashed lines). For all real-time estimates (coloured by round) we include the model posterior distribution's median (line) and 50% and 95% credible intervals (dark shaded and light shaded regions). We include the real-time model estimates up to the final day of data (for periods greater than the previous rounds final day of data). The daily number of hospitalisation for the final end-of-season dataset (black points) and the daily number of hospitalisations in the real-time datasets (points coloured by round) are connected by lines to highlight revisions to data used in real-time analyses. Note that when coloured points (and connecting lines) can not be seen then there has been no revisions to the data. Note that for SARS-CoV-2 hospitalisations some data were revised in multiple rounds and so many points can be seen for the same time point.
• Figure 5: In-season forecasts for daily SARS-CoV-2 cases at different origin dates (coloured lines and bands) alongside subsequently observed data (black points). Each coloured block in each panel shows a forecast generated using a specific origin date, corresponding to the left-hand side of the block (blue -- 4 May to 25 May; red -- 1 June to 22 June; yellow -- 29 June to 20 July; purple -- 27 July to 17 August; green -- 24 August to 14 September). Each panel shows the same data with forecasts from a different set of origin dates. For each forecast, we show the posterior median (line) and 50% and 90% credible intervals (dark shaded and light shaded regions).