Masterclass: Setting Operational Thresholds for Heat Early Warning Systems

Many warning systems have different levels. When issuing warnings for health there needs to be reasonable confidence in the forecast conditions while also allowing a reasonable lead time in which responding organizations have time to implement predefined actions in response to any high temperatures. This may mean a pragmatic approach to coming to consensus about what lead time is deemed reasonable.

There is no definite answer to this question, but again it will need to be a question that all stakeholders play a role in answering yet it may be determined by the forecasting capability of the met service producing the forecast.

If you don't have daily health data readily available, you can develop a heat alert systems based on meteorological data only. Within the masterclass we highlight the full spectrum of warning system types in Lecture 1, which goes through each type of system and how it is dependent on hazard, impact and vulnerability. Ultimately whatever type of system you go for will depend on what data are available and what is going to have the most benefit to elicit action. A simple threshold based on temperature (or temperature-humidity index) is sufficient as a heat alert system which can be further developed when you can and if it's useful to do so. Moving towards a threshold based on health impacts is advantageous, therefore it's important to work with health partners where possible to build structures to allow for data capture and the development of health surveillance systems. If no health data are accessible, working with partners can help to decide a most appropriate meteorological threshold, based on their experience, this is what we saw in lecture 1 as the “subjective approach”.

A range of health data can be used to inform an alert system. All-cause mortality and all cause emergency hospital admissions are often used as a good indicator of acute heat impacts in at the population level. Some health systems have real time surveillance data that can indicate heat impacts, either in general, by age group (in the elderly) or by diagnoses (heat-related).  

We know from the extensive evidence base that high temperatures exasperate a number of different diseases (as recorded by ICD codes), for example, admissions for respiratory, injuries and musculoskeletal causes can increase during heatwaves. However, looking at specific outcomes in isolation may weaken the sensitivity of the signal you see when assessing the overall temperature-impact relationship.  

However, looking at total mortality (daily data) will be informative and will show whether there is a peak in deaths associated with a heatwave. The analysis of the temperature-mortality relationship requires epidemiological methods and so should be done in collaboration with the local public health department or academic partner.

There is no single suggestion for special resolution. This will depend on the spatial resolution of the data used. If your different data sets are at different resolutions, you may need to aggregate your data which will again influence your assessment. Overall this decision will be influenced by the data available, but should be driven by user need. You can go from detailed resolution in a location level to regional one, it would depend on what is more suitable for end user taking action.

The threshold needs to be suitable for a given population. It is not possible to have thresholds tailored to individual characteristics (see answer above) but advice can be targeted to certain groups (e.g. by housing type, by age group, disability etc.). Further, the decision of issuing an alert must be as quick as possible, and evaluating to many criteria may delay the decision. It is something to analyze with partners.

It is good practice to validate thresholds used by updating temperature and health impact relationships at reasonable intervals – e.g. every 5 years. This would allow you to capture any changes in in the epidemiology and allow you to adjust your thresholds accordingly should there be any change observed. Heat alert systems should be regularly reviewed in order to ensure that the latest scientific evidence is being applied. Regular updates ensure the system is still relevant to the target population. There may be no additional action needed to consider climate change- except to consider that the hot season may start earlier and last for longer. Where external (non-weather) factors are concerned these should be assessed on an ongoing basis. An example of this is the COVID-19 pandemic and how the disease may overlap and exacerbate the impacts seen during heat episodes. Any additional information or factors that can help make decisions about issuing a warning should be considered as and when they arise.

This may be a reasonable starting point if this is the only data at hand. However, all efforts should be made to establishing ways of working to facilitate, with partners, a means of capturing relevant health data to allow further development of a heat early warning system. This may involve working with the wider health and social care system including capacity and consensus building. The operational choice of a threshold consider how often the heat alerts will be triggered.

If your warning is focused on health then we should use what we know about the health impacts to drive this answer. We know that the impacts from heat occur between 0-1 days from the onset of high temperatures. We also know that respite from high temperatures overnight is an important risk modifier for heat.

There is no agreed definition of a heatwave, and this is complicated further by different definitions used by different sectors even within one country. For example, In England there is a meteorological definition of a heatwave which is when a region exceeds a climatologically derived threshold for at least three consecutive days. This differs to when the Heat Health Watch Alert is issued which is when temperatures reach health and climatological derived thresholds for two consecutive days and the intervening night.

Therefore, the answer to this will be dependent on what the purpose of the system is and what you know about the relationships used to derive your thresholds.

Example from England: The Heatwave plan for England is a plan intended to protect the population from heat-related harm to health. It aims to prepare for, alert people to, and prevent, the major avoidable effects on health during periods of severe heat in England. It recommends a series of steps to reduce the risks to health from prolonged exposure to severe heat for the National Health Service (hospitals), local authorities, social care homes, schools, and other public agencies, professionals working with people at risk, individuals, local communities and voluntary groups.

The plan is underpinned by the Heat Health Watch Alert system operates from 1 June to 15 September. When an alert is issued, this will trigger levels of response from the NHS, government and public health system, and communicate risks as outlined within the plan. In addition, advice and information for the public and for health and social care professionals, particularly those working with at-risk groups. This includes both general preparation for hot weather and more specific advice when a severe heatwave is forecast.

Understanding the relationships between the hazard and the observed impacts is vital. Researchers should work right across the whole process with practitioners to ensure that research questions are reflective of operational considerations that will determine the utility of these systems. It's important to remember that operations questions may differ to purely academic questions. Thus, researcher working closely with practitioners is key in both developing and improving these systems. Also in the validation stage there is a gap that the academic sector can work on.

Formal engagement with end users from the beginning of the process will provide insight into user needs which is vital for an effective warning system. Such engagement may involve workshops, seminars or hackathons. Continue dialogue and commitment to stay in touch.