Home energy efficiency and radon: An observational study

Abstract Exposure to radon gas is the second leading cause of lung cancer worldwide behind smoking. Changing the energy characteristics of a dwelling can influence both its thermal and ventilative properties, which can affect indoor air quality. This study uses radon measurements made in 470 689 UK homes between 1980 and 2015, linked to dwelling information contained within the Home Energy Efficiency Database (HEED). The linked dataset, the largest of its kind, was used to analyze the association of housing and energy performance characteristics with indoor radon concentrations in the UK. The findings show that energy efficiency measures that increase the airtightness of properties are observed to have an adverse association with indoor radon levels. Homes with double glazing installed had radon measurements with a significantly higher geometric mean, 67% (95% CI: 44, 89) greater than those without a recorded fabric retrofit. Those with loft insulation (47%, 95% CI: 26, 69) and wall insulation (32%, 95% CI: 11, 53) were also found to have higher radon readings. Improving the energy performance of the UK's housing stock is vital in meeting carbon emission reduction targets. However, compromising indoor air quality must be avoided through careful assessment and implementation practices.


| INTRODUC TI ON
Radon is a naturally occurring radioactive gas and has been identified as the second leading cause of lung cancer worldwide after tobacco smoking. It is estimated to cause between 3% and 14% of lung cancer deaths depending on average radon levels and smoking prevalence. 1 In the UK, 1100 annual deaths have been attributed to radon exposure in homes in a Public Health England (PHE) report, 2 while a recent international study put the UK figure at 2858 (95% CI: 219, 9419). 3 Radon is emitted from all soil and rock types at various concentrations and presents a continuous source of human radiation exposure, 4 though quick dilution in the atmosphere leads to low concentrations in open spaces. In enclosed spaces, however, concentrations can become relatively high as it enters through gaps and cracks in suspended floors, construction joints, or walls. 5 In accordance with the Paris Climate Change Agreement, governments are committed to limiting global average temperature rise to well below 2°C above pre-industrial levels. 6 To achieve this target, a variety of measures are required to reduce greenhouse gas (GHG) emissions. These measures can be made both at source through, for example, increased generation of renewable energy and at point of use (eg, through home energy efficiency (HEE) measures). The UK government has pledged to reduce GHG emissions by 80% (from the 1990 baseline) by 2050. 7 The domestic housing stock is one of the areas targeted, with HEE measures incentivized through schemes such as the Energy Company Obligation. A large body of evidence has been amassed in recent years on how such measures might impact on the health of building occupants. Hamilton et al 8  health impacts due to increased exposure to internally produced pollutants. 9 As with other internally produced pollutants, the air tightening of buildings may inhibit radon from leaving the indoor environment or increase the stack effect, causing it to accumulate. 10 A recent modeling study indicated that increasing the airtightness of English homes (without providing compensatory ventilation) would increase indoor radon concentrations by around 60%, resulting in an annual burden of 4700 life years lost and 278 deaths (at peak) per year. 11 While earlier empirical studies have investigated the impact of dwelling characteristics on indoor radon concentration measurements, they relied on smaller samples. 12 The interaction between indoor radon levels and the presence of energy efficiency attributes of dwellings has been studied previously, with Gunby et al using data from a national radon survey. 13,14 Associations were identified between indoor radon levels and the presence of double glazing and downdraft proofing using data from around 2000 dwellings with property information provided by the occupiers. This paper uses a substantially larger dataset of greater coverage and over a longer timescale than previous studies. This allows the relationship between various dwelling characteristics such as HEE interventions and indoor radon concentrations to be empirically derived for a UK setting.

| ME THODS
This study involves the analysis of radon measurements in approximately 470 000 UK homes held by PHE, matched to dwelling characteristics recorded in the Homes Energy Efficiency Database (HEED).
There were two main components to this analysis: • Dataset matching and processing.
• Statistical analysis and interpretation.
The aim of this study is to investigate any relationships that may exist between dwelling characteristics and indoor radon concentrations with a particular focus on the impact of energy efficiency interventions which modify the building envelope. The addition of glazing, loft and wall insulation, and downdraft proofing are considered.

| Indoor radon measurements
Public Health England holds over 525 000 radon measurements recorded in UK homes made over the period between 1980 and 2015. These radon measurements were collected over several measurement campaigns conducted by PHE (and previously the Health Protection Agency [HPA] and National Radiological Protection Board [NRPB]) for a variety of purposes. All valid radon measurements made by PHE are in the database. Since the database was established, the main sources have been, national and regional surveys, aimed at establishing the distribution of indoor radon levels in the UK; targeted programs, undertaken in areas of known higher radon risk and often including offers of free radon tests for householders; research programs investigating specific aspects of indoor radon; and radon measurements purchased by individual householders, landlords, and social housing providers.
In some of the above cases, there is a deliberate bias toward obtaining measurements from areas of higher radon risk since that is where most of the higher individual exposures and risks are incurred and where intervention to reduce radon is most likely to be required. The dataset therefore has a known, deliberate bias toward high radon areas but includes over 150 000 radon measurements made in areas of lowest radon risk (outside "radon-affected areas").
The measurement procedure is reported in greater detail elsewhere. 15 Briefly, measurements are usually made by two passive radon detectors (shown in Figure 1), placed by a member of the household (in accordance with instructions) in both the living room and an occupied bedroom. The detectors are left in place for three

Practical Implications
• There is a need to ensure that appropriate measures are put in place to assess and address possible increases in radon exposure post-intervention.
• While energy efficiency measures are likely to provide a net benefit in terms of energy savings and warmer homes, care should to be taken to mitigate against reductions in air quality when installing interventions that increase the airtightness of homes.
• Increases in integrated population exposure to radon will lead to a rise in radon-related lung cancer rates.
• Energy efficiency interventions in radon-affected areas should be coupled with radon risk assessment strategies and monitoring to check that radon levels are not negatively impacted.
• Efforts should be made, where necessary, to reduce high indoor radon concentrations to below the Public Health England target level of 100 Bq/m 3 . months and then returned to PHE who calculate an annual average household radon exposure (Bq/m 3 ), which reflects typical occupancy patterns and seasonal corrections. 16 At the radon Action Level (200 Bq/m 3 ), an average 3-month measurement is expected to have an uncertainty no greater than 15%, while for measurements in the ranges 46-140 and 460-1400 Bq/m 3 , the acceptable uncertainty is 25%, 17 which includes uncertainties relating to the occupancy patterns and seasonal corrections.

| The homes energy efficiency database
The version of HEED used in this study comprises information on approximately 16.4 million UK dwellings and includes house-level characteristics such as age and type (eg, detached and semi-detached).
Uptake of HEE measures is also included such as the installation of loft and wall insulation, boiler replacement, downdraft proofing, and the addition of double glazing. The data are broadly representative of the English housing stock, although flats are underrepresented. 18 There are also substantial amounts of data missing in HEED. 19 Information within the database used was compiled by the Energy Savings Trust and contains data collected between 1993 and 2016 from multiple sources; installers, industry accreditation bodies, energy suppliers, government-funded programs, local authorities, and home surveys. 20 Homes which have had multiple HEE interventions have multiple entries within HEED, meaning that energy efficiency retrofits can be tracked over time.

| Data matching and processing
Radon measurements were corrected for average outdoor radon concentrations by subtracting 4 Bq/m 3 from all indoor measurements. Negative measurements were removed from any subsequent analysis. The radon and HEED datasets were matched using the postal address of the property. After the matching process, the address of the property was removed and anonymized to leave postcode district (the first four characters). The match resulted in a sample size of 470 689 homes. For homes with multiple radon measurements (some 20 000 homes), the match was made to the earliest measurement chronologically. This allowed the analysis to focus on HEE, since a second radon measurement typically follows radon mitigation measures applied to the home. The dwelling postcode district was also used to match to urban/rural classification (rucomb) using Office for National Statistics data. 21 Processing of the matched radon-HEED data was performed, such that only HEE interventions made prior to a radon measurement were classified as retrofit data. Given that the radon measurement program began 13 years before HEED was initiated, only 15.6% (73 550) of the radon measurements follow any HEE intervention.
Of these, the radon measurement follows the most recent HEED entry by an average of 3.8 years (1392.5 days). Figure 2 presents a histogram of the time in years after which the radon measurement followed a HEE retrofit. For cases where a radon measurement precedes the most recent HEE intervention, time-invariant information (such as dwelling age and type) is used. However, information concerning HEE (such as wall insulation type) is treated as "missing data."

| Statistical analysis
Radon concentrations in homes are generally observed to be lognormally distributed. 22 Geometric means and standard deviations of indoor radon concentration have been calculated within groups F I G U R E 1 Passive monitors used by Public Health England to measure indoor radon concentrations. © Crown Copyright, 2013. Public Health England F I G U R E 2 Time in years that a radon measurement follows the most recent energy efficiency intervention made in a home of data to estimate radon variability between various dwelling characteristics. The following dwelling characteristics have been studied: 1. Those considered invariant with time: • Dwelling type.
• Number of bedrooms.
• Heating system type.
This paper focuses on HEE measures that modify the building envelope (fabric interventions) with results regarding heating system shown in Appendix S1. The data allowed for analysis of the specific HEE measures (eg, the thickness of loft insulation) and also the binary condition of whether a HEE measure had been installed or not, and their combination, and association with average radon levels.

| RE SULTS
Radon measured in the full sample (470 689 homes) is observed to be log-normally distributed with a geometric mean of 46.6 Bq/m 3 and an arithmetic mean of 96.0 Bq/m 3 after subtracting for outdoor radon. Figure 3 shows the distribution of radon measurements for the full dataset.

| Radon and dwelling characteristics
The geometric means and standard deviations are presented as a function of dwelling characteristics in Table 1. As mentioned in the methods section, substantial amounts of dwelling information in HEED are classed as "missing data". For example, 42.7% of the homes are missing information on the type of dwelling. Despite this, the matched dataset is large enough to yield sufficiently sized samples.
The data suggest that certain dwelling characteristics appear to be associated with higher average radon measurements. Bungalows are observed to have significantly higher radon levels than other dwell- In terms of dwelling tenure, homes in the "other" category have the highest radon levels, although this is only a relatively small sample (~700 homes), which includes multi-ownership properties such as care homes, second homes, and holiday rentals. Council and social housing properties are observed to have lower radon levels 23 -this may be because this tenure type is composed of a higher proportion of flats with no contact with the ground. Number of bedrooms appears to have little influence on radon levels. Radon concentrations by geographic location (region) and urban/rural class are shown in Appendix S2.

| Radon and energy efficiency measures in homes
The full dataset was grouped by various HEE measures installed within homes (inclusive of all other interventions). The results presented in Table 2 indicate that homes that have undergone HEE  showing radon measurements by heating system type are presented in Appendix S1.

| Independent and combinations of energy efficiency measures
The results presented in Section 3.  Table 3 presents the geometric means and standard deviations for independent HEE interventions. The results here support findings presented in the previous section.
Double glazing (Glz) is the intervention that has the single greatest influence on indoor radon, followed by the addition of loft (LI) and wall insulation (WI). Downdraft proofing (DP) appears to have less of an association with indoor radon and may in fact be associated with reduced levels in some cases, although this sample suffers from low statistics.
Various combinations of HEE intervention are shown in Table 4, and histograms showing normalized distributions of ln(radon measurement) for a variety of HEE measures are shown in Figure 4.
The results seem to suggest that interventions have a cumulative effect on radon levels, since a combination of all HEE interventions (DP + Glz + LI + WI), yield the highest geometric mean.
Quantile-Quantile (Q-Q) plots are shown in Figure 5   Note: "Missing data" refer to there being no entry in Home Energy Efficiency Database (HEED) or the radon measurement being pre-HEED intervention, whereas "unknown" was flagged up this way by the building surveyor/data entry professional.

TA B L E 1 (Continued)
foundations which means that radon is able to enter the home more easily from the underlying soil.
Our results are in general agreement with findings in previous work which used a smaller dataset (N ~ 40 000) and showed that homes with double glazing have radon concentrations 66% higher than those without, 12 similar studies exist in France and Switzerland. 28,29 Our findings add weight to previous modeling work which showed that the air tightening of the English housing stock could raise radon levels by an average of 57%. 11  while other types of downdraft proofing may be detrimental.
HEED does not provide data on the types of downdraft proofing that were installed in homes which makes drawing any firm conclusions difficult.
In the UK, building regulations introduced in 2002 prescribe the levels of ventilation required to maintain a healthy indoor environment. 33 Purpose provided ventilation (PPV) such as trickle vents and extract fans may be used to reach these minimum ventilation TA B L E 3 Geometric means and standard deviations for radon measurements grouped by independent home energy efficiency interventions

| Strengths, limitations, and future work
The main strength of this study is that it uses empirical data from a large sample (470 689) of UK homes measured over a long period of time . It is the largest dataset of its kind and allows radon measurements from subsets of dwellings to be analyzed without the In (Radon measurement) retrofit" sample used for base comparison. It is assumed that the majority of the homes in the "no recorded retrofit" sample have not had a HEE retrofit applied, which may be justified, since the average year in which radon measurements were made for these data sample is 1996 and HEE uptake did not become commonplace in the UK until the early 2000s. 18  to identify homes for follow-up radon measurements where a measurement was made prior to, but not after retrofit.
As this was an observational study, it was not possible to control for various confounding variables. Unlike other studies, 12

| CON CLUS IONS
The matched Radon-HEED dataset has provided a rich resource to observe, at a national level, how indoor radon concentrations vary with an increasingly energy-efficient housing stock. The findings suggest that homes that have undergone certain fabric energy-efficient retrofits are likely to have higher indoor radon concentrations than those without, which is likely to have consequences for other indoor pollutants. Double glazed windows were observed to have the largest association with indoor radon levels, 67% (95% CI: 44, 89) higher than dwellings with no recorded retrofit, while loft (47%, 95% CI: 26, 69) and wall insulation (32%, 95% CI: 11, 53) also have relatively strong associations. With an ever more energy-efficient stock, this could result in a substantial rise in integrated population exposure and, hence, radon-related lung cancer rates. This implies the importance of radon risk assessment and monitoring in conjunction with HEE improvements, especially in radon-affected areas. The data matching process has helped identify homes that may be subject to further study. Obtaining additional radon measurements following a retrofit (where a prior radon measurement already exists) coupled with modeling work will further enhance our understanding of the relationship between HEE and indoor radon levels. This paper does not seek to discourage the installation of HEE measures. On the contrary, the UK must meet its carbon emission commitments to help mitigate anthropogenic warming of the climate and doing so while reducing exposure to indoor air pollutants will offer both climate change and health benefits.

CO N FLI C T O F I NTE R E S T
The authors have no conflict of interests to declare.