This project provides research relevant to the effect of (routine) HEE improvements on health in the population and complements evidence from more in-depth assessments of specific intervention programmes targeted at high-risk groups. Through its different components and analyses, it provides a number of new insights and implications for policy.
As with all research in this area, it provides evidence on a relatively selected subset of pathways/connections to health. Its primary focus has been on the consequences of HEE interventions operating through changes to the indoor environment, specifically changes in (primarily winter) indoor temperatures and indoor air quality.
There were two chief surprises from the study’s findings. The first was the apparently small impact that ‘routine’ HEE interventions have had on indoor temperatures (and hence on cold-related mortality/morbidity) across the housing stock in England. Our estimate is that, on average, such interventions (specifically those installed between 2002 and 2007) achieved an increase in SIT of only around 0.1 °C and have reduced the annual burden of cold-related mortality by around 300 deaths. This is, of course, a welcome benefit, but smaller than many might hope for and appreciably less than is theoretically achievable if all dwellings, particularly those for the most vulnerable people, received the full range of retrofit energy efficiency upgrades that could be installed in them.
As these changes were small, this presented a practical difficulty in the epidemiological investigation of the impact that such interventions had on mortality and morbidity. We concluded that direct empirical observation of pre–post change in mortality risk with data aggregated to the level of postcodes (which would further ‘dilute’ the effect of intervention-related changes made in an individual dwelling) would be beyond the limits of detection using national routine data, especially given other sources of imprecision in the classification of exposure changes. Although we are continuing to pursue studies to examine methods for linkage at the level of individual households, our assessment of intervention-related changes to the indoor environment and health for this project was based on an indirect approach.
Such indirect methods necessarily entail a range of uncertainties, and it is therefore worth considering the degree of confidence in our estimates and conclusions. For reasons already outlined above (see Chapter 3,The impact of home energy efficiency interventions: 2002–10) we believe that the estimate of change in indoor temperature, although not as robust or precise as might be obtained from direct empirical measurement, is nonetheless a reasonably accurate estimate of the average change associated with the different forms of energy efficiency interventions. As indoor temperature monitoring has mainly been undertaken only in the context of relatively small-scale surveys, there is no data source that provides measurements of temperature change following retrofits for even a sizeable minority of dwellings across England. Rarely have measurements been made both before and after the installation of energy efficiency measures. However, the general relationship between housing characteristics and temperature can be obtained from the data of those surveys and extrapolated to classify the changes likely to have occurred in other dwellings. It is noteworthy that the relationship we used for that classification is based on an empirical relationship that appears to be very similar in two independent data sources: the previously published evaluation of the Warm Front HEE scheme27 and our new analyses of data from the EFUS.30
What we learn from that relationship is that large changes in the engine efficiency characteristics of a dwelling are needed to achieve appreciable change in indoor temperatures, and that there is a plateau effect beyond which no further increase in temperature occurs with further energy efficiency improvement. This plateau begins at energy efficiency levels close to the average for the English stock. Thus, the shape of this relationship and the relatively modest nature of most HEE improvements are the principal reasons for the small temperature impact that energy efficiency improvements had. Furthermore, this evidence also suggests that appreciable temperature increases are likely to be seen only in the future following measures applied to dwellings with below-average energy efficiency levels. These are the dwellings in which there is potential for reduction of cold-related health impacts. For all other dwellings, the improvements in energy efficiency are taken as energy and cost savings (and consequent reductions of emissions to the environment).
The second surprise was the potential relative magnitude of health effects relating to changes in indoor air quality compared with those of temperature change. Almost all assessments to date have focused primarily on the short-term effects of energy efficiency improvements on temperature. Indeed, this was the intended primary focus of the current study. However, our analyses suggest that the impact of indoor air quality has the potential to be appreciably larger than that of changes in temperature, and that it might be negative as well as positive for health, depending on such factors as location (high radon area, area of high outdoor air pollution, etc.); the characteristics of household occupants; their behaviours and time activity patterns and the type of energy efficiency intervention. Risk may be further modified if interventions are not properly installed, operated and maintained. We have previously published on the potential risks for radon exposure,77 but similar considerations apply to a range of pollutants of both well-characterised (PM2.5 of indoor origin, STS) and less well-characterised (e.g. volatile organic compounds) risks to health. The risks associated with these exposures may take years or even decades to evolve fully and, thus, are not mainly visible as short-term changes in health (although some forms of impact may occur very quickly) but rather are expected to emerge over the medium to long term.
One of the difficulties of assessing these risks is the uncertainty about what changes to ventilation occur in practice. Building regulation gives guidance on the need to ensure adequate ventilation,139 but it is not clear how that guidance is followed. Energy efficiency interventions usually entail a reduction in air exchange because improving insulation normally cuts down air leakage through gaps in the building fabric and it is also often a deliberate objective to reduce uncontrolled ventilation and hence energy losses through the escape of heated air. An important note of caution is that such reduction will inevitably alter not only the dynamics of the entry of pollutants from the outdoor to the indoor environment but also the escape of pollutants from the indoor to the outdoor environment. Whether or not the net effect is beneficial or harmful depends on the relative balance of the changes across the range of pollutants. Although greater control over ventilation can be beneficial from an air quality perspective, if appropriate measures are implemented, operated and maintained, there is significant risk that the air quality may be made poorer by allowing the build-up of pollutants derived from indoor sources. In the worst circumstances, these risks appear more than sufficient to offset the benefits of improved winter indoor temperatures.
At present, there is insufficient research on the changes to indoor air quality associated with energy efficiency measures and how best to protect health. However, from our analyses, it would appear that health effects relating to air quality should be treated as being as important as, if not more important than, those of changes in temperature.
A further outcome of energy efficiency improvement is, of course, altered requirement and demand for energy. The metered energy use data suggest that, in addition to the modest effect of interventions on indoor temperatures, improved insulation of the dwelling fabric reduces heating energy use on average by around 4%, while improvement to the heating system may save around 10%. It is difficult to know how to translate these benefits into health terms. For very low income families living in FP, there may be some psychosocial benefit associated with a degree of alleviation of concern about managing the household budget, as well as the possibility that some of those cost savings may be translated into such things as a more varied (and hence healthier) diet – although they might equally be spent on unhealthy or health-neutral items.
Developing a model of the health impacts of energy efficiency interventions is useful not only because it can provide evidence to help inform policy decisions, but also because it provides a framework to explore an extended set of pathways of exposure and health impact, including ones that emerge over long time horizons (as with various air quality-related impacts). The model that we have implemented is still limited to effects relating to the indoor environment, and there are many areas in which further model development would be very desirable. There will always be value in more empirical studies to underpin and validate the functional relationships used in the model. The methodological work that we undertook has helped to reduce one key area of uncertainty, namely the degree of life shortening in cold- and heat-related mortality, although only as far as to provide evidence that the degree of life shortening in cold-related deaths is unlikely to be less than 6 months or so (and hence cannot be attributed to harvesting). It still remains unclear whether or not those who die of cold have a life expectancy similar to the average among the population of their age and hence lose a corresponding number of life-years. This is the implicit assumption if the RR of cold-related death derived from time-series studies is applied to the annual hazard of mortality in life table calculations. However, the probability is that those who are vulnerable to cold-related death are vulnerable because of underlying disease processes (such as atherosclerosis), which would leave them with a shorter life expectancy than average. Thus, some correction would be appropriate in calculations of YLL, but in the absence of clear evidence of what this correction should be, we can do little better than to test results using a range of assumptions. Until there are analyses of data from large cohort studies to compare survival in populations exposed to different temperature distributions (adjusting for all other major risk factors), it is unlikely that there will be substantial further clarification of this area of uncertainty.
Our evidence with regard to heat risk was not sufficient to suggest a clear basis for quantifying heat-related health burdens associated with housing characteristics but it does suggest a need to give greater weight to this issue, both because of the potential importance of housing adaptations as a response to climate change and because in many cases energy efficiency interventions appear to exacerbate the risk of overheating. For the moment, arguably the ‘best guess’ approach to estimating the mortality risk associated with overheating in dwellings is to assume that the well-established relationships between outdoor temperature and mortality also apply to the indoor environment.73 Thus, the temperature anomaly of one dwelling compared with the average for the region might suggest the degree to which an individual’s heat risk moves to the left or right on the (population-average) outdoor temperature–mortality function. As far as we can tell, the additional level of risk for heat-related death associated with improved energy efficiency is modest, but it may not remain so if ambient temperatures rise because of global warming. Therefore, it would be prudent to ensure that measures to limit overheating risk are built into housing improvement plans at the same time as those for improving energy efficiency.
We recognise that our modelling methods are also currently limited because they do not incorporate any quantification of the possible impacts that changes in fuel use would have on outdoor air quality (although this could be done using, for example, emission dispersion models). Neither do they explicitly consider health effects relating to climate change. However, the model can yield estimates of change in energy use and CO2 emissions that are potentially important outputs in their own right as markers of progress towards climate change mitigation targets and, in the case of the costs of energy, as a measure of action towards the alleviation of FP for disadvantaged groups.100
Multicriterion decision analysis is a framework that can usefully allow a range of parameters (criteria) to be considered in any decision process of policy choices. This seems particularly apposite with regard to HEE, the rationale for which is driven by more than one policy objective. Recently The Lancet commissions have emphasised the increasingly urgent need to address responsibilities for global health in all facets of policy development140 and, in the case of housing policy, those responsibilities can be met at the same time as addressing important local public health objectives relating to winter mortality and energy security. Indeed, there is a commonly applicable principle that actions aimed at the transition towards a low carbon economy (which includes the improvement of the energy efficiency of the housing stock) have the potential for substantial cobenefits for health.13,141,142 Thus, there is potential value in attempting to align policy objectives and to consider the value of policy actions not in terms of single outcomes but using multiple criteria of costs and benefits, as the MCDA process does. Policies that may not be justified in terms of the impact on a single criterion may be more persuasive when several criteria are considered, such as the potential impact that housing could have on health, health inequalities, FP, GHG emissions and energy security. The value attached to the attainment of each of those criteria, and how they might influence policy choices, can be usefully explored using the MCDA, although our experience in engaging decision-makers and other stakeholders in project workshops suggests that it is best applied as a tool for discussion and iterative exploration of options rather than as a method for arriving at a ‘single best choice’.
This alignment of policy objectives was behind our illustrative comparison of the consequences of perpetuation of WFPs compared with similar value investments in HEE improvements. This was intended not as a challenge to government policy, but simply to illustrate the possibilities of alternative strategies. In the example we presented there are reasons to favour energy efficiency investments if judged against multiple outcomes including health and contribution towards environmental sustainability. From both a health perspective and to meet climate change mitigation goals, there is a need for transformative change in the UK housing stock.
The evidence from our qualitative study provides insights on how this might be approached in terms of the factors most likely to motivate householders. Framing the issue in terms of climate change or environmental sustainability would most likely result in slow adoption of HEE upgrades. Only a small minority of householders principally frame HEE technologies in this way. However, similar to subsidised HEE, this framing has the potential to generate common goods that may be more beneficial in the long run to ensure both sustainability and equality goals. Although consumerist framings might improve short-term uptake rates, they might have significant costs in the longer term of eroding the ‘common good’ of commitment to environmental sustainability.
Study limitations
This research attempted to address questions of considerable methodological complexity relating to the multiple pathways by which HEE interventions may have an impact on health. Numerous factors make this endeavour challenging including, but not limited to, the relatively modest effect of most HEE interventions on the indoor environment, the consequent small changes in health risks at an individual level, the rarity of attributable health outcomes (e.g. mortality), the different (sometimes very long) time courses for the development of different health outcomes, the complexity of gathering data of sufficient detail to characterise the timing, type and consequences of interventions for specific dwellings or groups of dwellings, the quantification of the associated changes to human exposures, the modifying effect of variations in human behaviour and the fact that data have to be combined from different data sources with different spatial units.
The small risks and rarity of attributable outcomes was the main reasoning for a study based on data from a very large number of dwellings. A brief consideration of cold-related mortality as one of the outcomes of interest illustrates the challenge. Mortality is a rare outcome and cold-related mortality is rarer still, but our particular interest was in the potential impact that small/modest intervention-related changes have on the indoor environment as a modifier of the risk of cold-related mortality. To study this, very large data sets are needed with linkage across dwellings for which we have not only health data but also evidence on the intervention and its effect on the indoor environment.
The changes attributable to HEE interventions appear to be so small that analysis without robust exposure classification and one-to-one data linkage (rather than by small spatial units) is very unlikely to be fruitful. Unsurprisingly, data on ideal detail, specificity and precision about the dwelling and energy efficiency interventions are not available from routine sources and so exposure classification, in particular, has to be based on building physics and other models using imperfect data to predict changes. Moreover, such models predict only ‘average’ patterns and not individual-level responses, which may be modified by, for example, personal behaviours (e.g. variations in choice about home heating). Such behavioural influences may mean that intervention changes may vary appreciably from one dwelling to another, even if the dwelling characteristics and interventions are otherwise identical and this also adds complexity to model validation, as any empirical data gathered to test model predictions has to recognise such variations.
This means that there are limits to what can be tested using existing data and it places greater dependence on the evidence of models to address important research questions: models to characterise the effect of interventions on dwelling performance, models of occupant exposures, models of the associated health risks and so on. All entail assumptions and uncertainties, even when there are good empirical data, but sometimes the evidence is limited and/or indirect. To make the evidence more interpretable, we have tried to use models that are based on fairly simple clear empirical relationships, such as the relationship between HEE U-value and (population-average) SIT. As already discussed, the evidence of two completely independent data sets appears fairly consistent with regard to this relationship, which helps to strengthen inferences based on it. But it is just a model and hence is in some degree inaccurate. The evidence underpinning other important relationships is subject to similar uncertainties. We have referred already to the evidence relating to ventilation changes, which is arguably less robust than that for temperature but is clearly important for health.
Therefore, to draw inferences about the impacts that HEE measures have requires considerable care and expertise in interpreting empirical and model data, but also the putting together of an evidence jigsaw, in which the pieces come from research in different disciplinary domains (building physics, mathematical modelling, exposure science, epidemiology, etc). This makes for a complex synthesis that can be difficult to judge by someone from any one disciplinary background and difficult to present transparently to decision-makers.
Despite these many challenges, we believe that the analyses that we and others have produced do make it possible to draw defensible conclusions about the impacts that HEE interventions have and at least to define some of the important questions that need to be considered in any programme development. The evidence is not beyond doubt, but it is reasonable to conclude, for example, that, on average, individual energy efficiency interventions achieve only modest changes to winter indoor temperatures, that ventilation changes are important (and need to be further studied) and that HEE improvement represents an important (and integrated) way to help achieve multiple benefits relating to areas such as health, FP, climate change and inequalities – even if there are uncertainties about the magnitude of effects and about potential risks if programmes are not carefully designed and implemented.
For many questions about HEE impacts, model evidence is, therefore, likely to remain a core element of research, even in observational studies. This has implications for future research agendas but also for the presentation of evidence to decision-makers, because of the indirect nature of some part of such evidence and the additional uncertainties it entails. This is an aspect of the evidence that we had only limited capacity to explore in our engagements with stakeholder groups, including in the MCDA discussions.
Further research
The comments of the preceding section suggest a very large research agenda spanning multiple disciplinary areas. Without attempting a comprehensive analysis, the following as principal categories of research need are suggested:
- Further empirical evidence on the basic relationships between dwelling characteristics, interventions, indoor environment and human exposures. This is needed to strengthen the understanding of the ‘chain of causation’ linking HEE interventions to health and especially to improve the development of models of impact.
- Continued efforts to develop, validate and iteratively improve models that reflect those causal links through to health. Validation is always a challenge because of the data requirements, which may be costly to meet, but it is important for model credibility; for example, validation of models of ventilation and indoor air would require detailed surveys of dwelling characteristics as well as the monitoring of air exchange characteristics and pollutant levels for large samples of dwellings sufficient to test model results across a range of archetypes and occupant behaviours.
- Monitoring of before-and-after changes in a large sample of dwellings that undergo energy efficiency interventions to strengthen the observational evidence on the consequences of those interventions. Given the scale of proposed changes to the housing stock to meet health and environmental objectives, such monitoring would seem crucial to ensure that programmes of energy efficiency upgrades are well designed and implemented and to avoid the potential risks of unintended adverse consequences.
- Further exploration of data linkage studies between health and housing data, but recognising the importance of accurate exposure classification and high resolution linkage as a complement to primary intervention studies and trials.
- Further participatory research with stakeholders on their evidence needs for policy development, especially in the context of uncertainties, so as to improve specific evidence provision and the form and methods of presentation of greatest value to user groups.
- Further qualitative and quantitative research into the motivations and behaviours of householders in relation to the uptake of HEE interventions given the potential value of accelerated expansion of such measures to help achieve health, environmental (climate mitigation) and energy security goals.
In summary, this study has provided new insights into the relationship between HEE interventions, the indoor environment and health. Despite various difficulties and limitations, it has yielded important messages for policy, including the need for much larger-scale action if the full potential benefits are to be realised for reducing cold-related burdens, the need for much greater focus on the potential implications for indoor air quality, the value for decision-making of considering the impact that interventions have with regard to multiple criteria, and the potential value of framing of national policies to capitalise on the goodwill of people to contribute to actions in support of environmental sustainability.