Official Statistics

Cold mortality development update: October 2026

Published 22 October 2026

Applies to England

In February 2026 we published the first cold mortality monitoring report. The report summarises cold-associated mortality during winter cold episodes to inform public health actions. It also supports the Adverse Weather and Health Plan to protect the population from cold-related harm to health.

We published the report as official statistics in development to improve quality, enhance policy relevance and interoperability, and respond to user feedback. We also published a development plan to explain to explain how we would make these improvements.

This development report explains the development work that was done to strengthen methods, outputs and user value. As a result of this work, we have made 2 changes to the statistics:

• Starting with the 2025 to 2026 report, the cold mortality monitoring report will be published as official statistics, rather than official statistics in development

• Starting with the 2025 to 2026 report, cold episodes will be defined using regional temperatures rather than an England-wide average temperature

Methodological refinement and testing

Since the release of the first report, we have carried out further methodological refinement and testing, focusing on 3 main areas:

1. Assessing whether a using a regional definition of cold episodes would improve the statistics
2. Testing different lag periods
3. Assessing the feasibility of including additional circulating respiratory viruses in the model

Regional definition of cold episodes

Cold episodes will now be defined using regional temperatures rather than an England-wide average temperature. This change better reflects geographical variation in temperature exposure.

Table 1 shows the number of cold days identified in each region, comparing the new regional definition with the England-wide approach. During winter 2024, the England-wide approach identified three cold episodes comprising 11 cold days. In contrast, the regional definition identified a greater number of cold days across most regions. The North East showed the largest difference, with 13 additional days meeting the cold episode threshold.

Table 1. Number of cold days identified using England-wide and regional definitions of cold episodes, winter 2024 to 2025

Region	Cold days identified using England-wide approach	Cold days identified using regional definition	Difference
England	11	11	0
North East	21	31	10
North West	13	15	2
Yorkshire and The Humber	20	29	9
West Midlands	17	23	6
East Midlands	17	23	6
East of England	15	19	4
London	10	9	-1
South East	10	9	-1
South West	6	1	-5

These findings indicate that an England-wide average can mask important regional variation in cold exposure, potentially underestimating cold weather impacts in colder regions.

Assessment of different lag periods

After testing different lag periods, we will continue to use a 14-day lag for cold-associated mortality.

There is known to be a delay between when cold weather occurs and when effects on mortality appear, known as lagged effects. Exposure to cold weather can increase the risk of death many days later, for example by triggering heart problems, worsening existing illnesses, or increasing the chance of infections that take time to develop. The lag period is the length of time that deaths are counted after each cold episode.

The model was run using a range of lag periods (7, 14, 21, 28 and 35 days), with lags assessed at 7-day intervals to align with the influenza surveillance data, which is only available at a weekly level. Model fit improved progressively with longer lag periods. However, this was accompanied by a reduction in the number of observations available for analysis due to the increased lag requirement.

While longer lag periods have been used in studies, the primary purpose of this model is to support operational surveillance. Therefore, consideration was given not only to statistical fit but also to the interpretability and practical usefulness of the outputs. At longer lag periods, cold weather episodes increasingly overlap, making it more difficult to attribute observed impacts to specific temperature events. This introduces greater uncertainty when interpreting results and reduces the operational value of the findings.

A 14-day lag was therefore selected as the most appropriate compromise between model performance and operational utility. This lag period is sufficiently long to capture delayed impacts while maintaining a clear temporal relationship between cold weather episodes and observed outcomes, enabling more meaningful interpretation for public health decision-making.

Adjusting for RSV and other respiratory viruses

We assessed the feasibility of including additional circulating respiratory viruses in the model, and concluded that no further adjustment for RSV or other respiratory viruses will be undertaken

Consultation with subject matter experts indicated that RSV is highly seasonal and closely aligned with broader seasonal patterns already represented within the model. Consequently, inclusion of RSV was not expected to substantially improve model performance or interpretation.

Although additional infectious disease modelling could potentially provide greater separation of the effects of influenza and other respiratory pathogens, this would require a dedicated modelling framework and specialist expertise beyond the scope of the present work. Therefore, no further adjustment for RSV or other respiratory viruses will be undertaken.

Communication, user engagement and gathering feedback

Feedback from stakeholders on the first publication of the Cold Mortality Monitoring Report has primarily been positive. Feedback has indicated that the publication addresses an important evidence gap and provides useful information for understanding the impacts of cold weather on mortality.

Some stakeholders have requested additional levels of detail and further demographic breakdowns. However, several of the requested variables are not currently collected through death registration data and therefore cannot be incorporated into the statistics at present.

Consistent with the commitments set out in the development plan, we will continue to actively seek user feedback on the usefulness, interpretation and application of these statistics. Feedback will be considered as part of the ongoing development of the report and supporting materials.

We will also seek to improve communication around what the statistics represent, how they should be interpreted, and the limitations of the methodology. This will help minimise misunderstanding and ensure users are able to make appropriate use of the findings. This aligns with the development plan commitment to improve user value, support intelligent use of the data, and respond transparently to user feedback.