A warm home has been sold to the British public as a health technology: insulate the walls, lift the Energy Performance Certificate rating, and the cardiovascular dividends will arrive on schedule. A study published on 20 October 2026 in the Journal of Epidemiology & Community Health went looking for one of those dividends, a lower rate of new hypertension diagnoses, and could not find it.
The authors, Anthony Lim, Isobel Braithwaite, Rhodri Johnson and colleagues, linked Welsh housing and health records for 323,557 adults aged 40 and over. People whose homes had cavity wall insulation had a slightly higher hazard of a first hypertension diagnosis than people whose homes had none. Better EPC ratings pointed the same way, weakly and without holding up in sensitivity tests. Neither result supports the assumption that efficiency metrics translate into less hypertension.
Why anyone expected the opposite
The background is not in dispute. Cold homes are associated with higher hypertension risk, and public health guidance treats adequate indoor warmth as a housing standard for that reason. A decade of policy has treated energy efficiency measurement and improvement as the lever that makes homes warmer. If the lever worked as advertised, the health data should show it.
The gap the authors set out to test is the middle of that chain. Efficiency, as scored by the metrics governments actually use, is meant to stand in for warmth. Whether it does is an empirical question, and one that has been answered mostly with modelling rather than with linked records.
This study is an attempt to check the proxy against a hard outcome in a real population. That framing is worth holding on to, because the result is less about insulation than about whether a certificate predicts a diagnosis.
What the researchers measured
Two exposures: whether a dwelling had wall insulation and which type, and its EPC rating, treated as a proxy for household energy efficiency. One outcome: the first hypertension diagnosis coded by a general practitioner.
The models adjusted for age, sex, smoking status, education, employment, tenure, disability, rural or urban classification, deprivation, building size and property type. That list matters. Deprivation and tenure are the variables most likely to confound any housing and health relationship, and both were in the model.
What was not in the model is what the inside of those homes was like in January. The authors say as much in their conclusion, calling for further data on the quality of efficiency retrofits and their effects on the indoor environment.
Who was included, and who was not
The setting was Wales, using the Secure Anonymised Information Linkage Databank. Energy Performance Certificate records, the Welsh Demographic Service, GP data and the 2021 Census were linked into one retrospective longitudinal study.
The primary population was adults aged 40 and over living in dwellings with a valid EPC and registered with a GP contributing data as of 21 March 2021. The EPC requirement is the structural limit: those dwellings covered 37% of the at-risk population. Homes without a certificate skew towards stock nobody has assessed, often older buildings and the private rented sector.
The headline population is still large, with a median follow-up of 51 months (IQR 30 to 87). The analysis used Cox proportional hazard models with stabilised inverse probability weighting to account for who ends up with an EPC rating. That weighting is the correct tool for the selection problem. It does not convert an observational dataset into a trial.
The cavity wall result
Cavity wall insulation, compared with no insulation, was associated with a higher hazard of hypertension: HR 1.07, 95% CI 1.04 to 1.11. The authors describe the evidence as moderate. The confidence interval sits above 1, so sampling noise alone is an unlikely explanation.
In plain terms, the diagnosis rate in the insulated group ran about 7% higher than in the uninsulated group over the follow-up, holding the modelled covariates constant. That is small for one household and substantial for a national retrofit programme. It is also the opposite sign from the policy expectation.
Three explanations are available and the study cannot separate them. Insulation may proxy for something unmeasured, such as a particular housing stock or the kind of household that takes up a grant. Retrofit work may create damp or ventilation problems when poorly executed, a pathway the authors explicitly leave open. Or the association may be a residual artefact of who gets an EPC and who visits a GP. The paper reports an association and does not assign a cause.
The EPC result was weaker still
For EPC ratings the direction matched, with weak evidence of an increased hazard of hypertension at better efficiency levels. It did not hold in sensitivity analyses. A result that dissolves under alternative model specifications is not a basis for policy in either direction.
Put the two findings together and the honest summary is narrow. The study found no evidence that better home energy efficiency, measured the way policy measures it, reduces hypertension risk. It did not find robust evidence that efficiency raises risk either.
Where the design limits bite
Hypertension is largely silent. The outcome here is a GP-coded diagnosis, so anyone who did not see a GP during follow-up does not appear as a case, whatever their blood pressure was doing. Contact with primary care is patterned by age, work and deprivation, and those are patterned by housing.
Follow-up is short for a condition that develops over decades. A median of 51 months captures incidence, not lifetime risk. EPC rating describes a dwelling's modelled performance, not the temperature its occupants reach, and the gap between the two is precisely what a heat pump's rated COP and your actual winter bill tend to expose. The study also cannot see retrofit quality, ventilation or indoor air, the same blind spot that shows up in work on indoor air in the homes of preterm infants.
And it remains observational. Households that install insulation differ from households that do not in income, tenure, building age and baseline health. Inverse probability weighting handles the EPC selection, not the entire confounder set.
What this means if your house is cold
Nothing here says a warm home is unimportant. The case for avoiding cold indoor temperatures is older and broader than this dataset. What the study undercuts is the shortcut: certificate rating up, health outcome down. The mechanism is presumably the indoor environment, and the rating is a proxy for the mechanism rather than the mechanism itself.
For a reader, the practical takeaway is narrower than a retrofit decision. The things worth watching are the ones the study could not measure: whether insulation went in with ventilation, whether any room turned damp afterwards, and whether the rooms you actually sit in stay warm. A house can score well on paper and still be miserable in February.
The pattern echoes heat research, where timing matters as much as the average. Hot nights and their effect on sleep and cardiac strain do not show up in an annual figure either, and neither does a cold bedroom in a dwelling rated C.
Fact, interpretation and prediction
Fact: in this Welsh cohort of 323,557 adults, cavity wall insulation was associated with a hazard ratio for incident hypertension of 1.07 (95% CI 1.04 to 1.11), and EPC rating showed weak, inconsistent evidence in the same direction.
Interpretation: certificates are probably a poor stand-in for the lived indoor environment, and retrofit quality, ventilation and damp are the variables that would explain any real effect. That is my reading of the result, not a result of the study.
Prediction: the next useful paper on this question links retrofit records to indoor temperature and humidity sensors rather than to EPC bands. If no linked-record study of that kind appears by the end of 2028, this prediction fails on its own terms. Until then, expect this one to be cited by people arguing that grant programmes should be judged on delivered warmth.
The finding is inconvenient for a policy that has spent years counting certificates. It is not an argument against insulation. It is an argument for measuring the thing you claim to be fixing, and for treating a proxy as a proxy until someone shows it earns its place.
Comments
No comments yet. Be the first to share your thoughts.
Leave a comment