Combination of housing type (detached houses vs flats) and tenure (owned vs rented) in relation to cardiovascular mortality: findings from a 6-year cohort study in Japan

Abstract

Introduction

The WHO Housing and Health Guidelines have highlighted the impact of housing quality on cardiovascular diseases (CVDs), including pathways such as cold-induced hypertension. Major factors influencing housing quality include architectural type (detached houses vs flats) and tenure (owned vs rented), but few studies have examined their effects on CVDs.

Methods

46 850 occupants were included during the follow-up period from 1 January 2012 to 31 December 2017 in the Japan Gerontological Evaluation Study. By linking survey data with cause-of-death records, the Kaplan-Meier curves were constructed. Competing risk regression models were applied to calculate the subdistribution HRs (SHRs) for cardiovascular mortality risks across housing statuses, adjusted for demographics, socioeconomic factors and lifestyle behaviours. Sex-stratified analyses and Cox regression analyses were also conducted to calculate the HRs.

Results

A total of 38 731 participants (46.6% men) were analysed, with a mean age of 73.6 years and a median follow-up period of 2091 days. The cardiovascular mortality rate was 3.97 per 1000 person-years, with 2.3% experiencing CVD-related deaths. The Kaplan-Meier curve indicated higher cardiovascular mortality for those living in rental flats and owned detached houses compared with those in owned flats. Competing risk regression models indicated a significantly higher risk of cardiovascular deaths among occupants living in rental flats compared with those in owned flats (SHR=1.78; 95% CI 1.05–3.02). For men, the risk was notably higher (SHR=2.32; 95% CI 1.13–4.75), though not statistically significant in women. Sensitivity analyses using Cox regression supported these findings, showing higher risk estimates for men (HR=2.36; 95% CI 1.16–4.82).

Conclusions

Rental housing and detached houses are likely to have lower temperatures and greater temperature instabilities, raising blood pressure and increasing CVDs. Improving housing quality can contribute to cardiovascular health at the population level.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• The WHO Housing and Health Guidelines (2018) and a recent review (2023) summarised evidence linking housing and cardiovascular health, using blood pressure as an indicator.

• Although the short-term impacts of housing, such as cold-induced increases in blood pressure, have been examined, the long-term effects of housing on cardiovascular health remain insufficiently studied.

WHAT THIS STUDY ADDS

• In this 6-year cohort study targeting independent older adults aged ≥65 years, cardiovascular mortality was lowest among residents of owned flats, moderate among those living in owned detached houses and highest among individuals residing in rented flats.

• Living in rented flats significantly affected cardiovascular mortality even after adjusting for socioeconomic status.

• The HR for cardiovascular mortality associated with living in rented flats was higher in men, potentially due to two factors: men generally have higher blood pressure than women, and hypertensive individuals tend to be more sensitive to the effects of indoor environment (eg, low temperature).

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Housing is the foundation of life, where people spend more than half of their lives, making these findings relevant to many people.

• In particular, people living in rented flats should pay attention to indoor environmental quality, and policy interventions are necessary to overcome the split incentive problem and improve housing quality in rented flats.

• Improving housing quality has widespread benefits—not only in addressing cardiovascular health disparities but also in mitigating climate change.

Introduction

Housing plays an important role in the prevention of cardiovascular diseases (CVDs). The WHO Housing and Health Guidelines¹ issued in 2018 provide recommendations for improving cardiovascular health through better housing conditions. In 2020, the American Heart Association (AHA) published a scientific statement about the importance of housing to keep cardiovascular health and well-being.² This statement reviews studies on how housing status affects cardiovascular health from four key pathways—stability, quality/safety, affordability/accessibility and neighbourhood environment—and summarises research gaps and considerations for the future. Among these gaps, the necessity to examine whether and how housing structural features affect cardiovascular outcomes was pointed out in the quality and safety category. The most prominent difference in housing structural features is the housing type (detached houses or flats). Detached houses are exposed to outdoor conditions on all sides; in contrast, flats share walls with neighbouring units, leading to differences in indoor environmental quality. These environmental differences may, in turn, contribute to disparities in cardiovascular outcomes.

A recent review also shows the pathways from housing status to health,³ where housing quality and access, influenced by social determinants, lead to health hazards (eg, low temperature) and poor health outcomes (eg, CVDs). Low indoor temperatures may increase the risk of CVD primarily through raised blood pressure, as cold exposure causes vasoconstriction and increases cardiovascular strain.⁴ The review focused on tenure patterns as one of the factors affecting housing quality and revealed that occupants in rented housing have poor health status.5,8 The low quality of rented housing can be attributed to the ‘split incentive’ issue,⁹ where the owner, who does not reside in the property, may be less incentivised to invest in improvements that would primarily benefit the occupant. As a leading example of addressing this issue, the Housing Health and Safety Rating System (HHSRS)¹⁰ was developed in England and Wales, where improvement orders are issued to the owner if a property, including rented housing, is deemed hazardous. In New Zealand, the HHSRS framework has been adopted to improve the standards of rented housing and has now established the Healthy Homes Standards for rented housing.¹¹ However, in many countries, there are no established standards for rented housing.

Based on the review of prior studies, there is a possibility that housing type (detached houses or flats) and tenure pattern (owned or rented houses) may influence CVDs. However, few studies have examined the combination of these housing characteristics. From a policy implementation perspective, understanding the impact of this combination on CVDs is critical for identifying high-risk groups and designing effective, targeted interventions. Rather than addressing housing type or tenure in isolation, examining their combination can reveal vulnerable populations who may benefit most from housing-related health policies. Therefore, this study addresses the research question of how cardiovascular mortality is associated with the combination of housing type and tenure pattern. To answer this question, we analysed data from a long-term cohort database.

Methods

Data and study design

This 6-year prospective cohort study used data from the Japan Gerontological Evaluation Study (JAGES).¹² A baseline survey was conducted between July 2010 and January 2012. We recruited participants from 11 municipalities in Japan, including urban and rural areas in Hokkaido, Miyagi, Chiba, Yamanashi, Aichi and Nagasaki Prefectures. In seven smaller municipalities, complete enumeration was conducted, and in four larger municipalities, random sampling was conducted.

The inclusion criteria were community-dwelling adults aged 65 years and older who were both physically and cognitively independent. A self-reported questionnaire was distributed to 80 744 participants between 2010 and 2012, with a response rate of 64.3% (n=51 923). The exclusion criteria were as follows: invalid identifiers, sex or age information (n=5073); no linkage with mortality records (n=706); no agreement to participate in the survey (n=4038); lack of functional independence at baseline (n=1677); residence in row houses or other unclassified housing types (n=1212) and death prior to 1 January 2012 (n=486). Row houses and other housing types were excluded because they could not be clearly classified as either detached houses or flats, which were the two categories of housing type used in this study. The final analytical sample size was 38 731 participants. JAGES is designed for multiple purposes rather than focusing on a specific outcome and therefore was not established with a formal sample size calculation. Figure 1 shows the flow chart of the study population. As the different municipalities had different follow-up periods, we defined the valid follow-up period as 1 January 2012 to 31 December 2017, which was a common follow-up period.

Figure 1. Flowchart of the study participants.

To ensure data quality, multiple quality control procedures were implemented. The questionnaire was repeatedly revised and reviewed by domain experts. A set of frequently asked questions was developed beforehand to support standardised responses during the survey process. After data collection, responses were checked for completeness and logical consistency. A data use manual was prepared to standardise data handling procedures across researchers. Additionally, syntax files for statistical analyses were shared to ensure consistency and transparency in the analytical process.

Outcome variable

The outcome variable was cardiovascular mortality from 1 January 2012 to 31 December 2017. We obtained the date of death from the municipalities’ long-term care insurance system database. The cause of death information was provided by the Ministry of Health, Labour and Welfare, a Japanese governmental agency. We then combined the questionnaire, date of death and cause of death using key variables including sex, date of birth, date of death and municipalities. Researchers were not provided with, nor did they handle, any personally identifiable information such as names. Cardiovascular mortality was determined using the corresponding causes of death codes from the vital statistics data. Cardiovascular mortality includes acute myocardial infarction, ischaemic heart diseases, arrhythmias and conduction disorders, heart failure, cerebral infarction and related diseases. Information about cardiovascular mortality (I00–I99) was evaluated according to the 10th revision of the International Classification of Diseases.¹³

Exposure variable

The exposure variable was the housing status. We asked the participants, ‘What types of residence do you live in?’ with five types of answers (owned house, privately rented house, municipality-managed house, company-owned house and others). We also asked, ‘What are the types of your house?’ with four types of answers (detached house, row house, flat and other). After excluding those who lived in row houses and other housing types, we created the housing status variables as follows (owned detached houses/owned flats/rented flats/rented detached houses and others) by combining the two questions on ownership and housing types.

Covariates

We selected the covariates from the previous study,⁸ the guideline¹⁴ and the Life’s Essential 8 suggested by AHA.¹⁵ We used the following covariates which were obtained at baseline: sex (men/women), age (65–69/70–74/75–79/80–84/≥85 years), household equivalent income (<1.00/1.00–1.99/2.00–2.99/3.00–3.99/≥4.00 million Japanese Yen (JPY)), self-rated economic status (high/middle–high/middle–low/low), education level (low (≤9 years)/middle (10–12 years)/high (≥13 years)), marital status (no partner/having partner), depression (< 5/5–9/≥10), body mass index (BMI) (<18.5/18.5–24.9/≥25.0), drinking status (current/past/never), smoking status (current/past/never), vegetable and fruit intake (<once per day/≥once per day), daily physical activity (<30/30–59/60–89/≥90 mins), heart diseases (yes/no), stroke (yes/no), dyslipidaemia (yes/no), diabetes (yes/no), sleep disorder (yes/no), self-rated health (very good/good/bad/very bad) and regions (11 municipalities). We determined household equivalent income by dividing household income by the number of household members. We used the self-rated economic status from the question ‘To what extent are you concerned if you need to incur unexpected expenses?’ with the four answers (not at all worried (high)/a little worried (middle–high)/quite worried (middle–low)/very worried (low)). Depression status was measured using the Geriatric Depression Scale.¹⁶ Hypertension was excluded as a covariate because it is assumed to function as a mediator when assessing the impact of housing status on cardiovascular mortality.

Statistical analysis

First, we obtained descriptive characteristics and described the Kaplan-Meier curve. Second, a competing risk regression model was used to calculate the subdistribution HRs (SHRs) and 95% CIs of the onset of cardiovascular mortality by housing status. We used competing risk regression models by the Fine and Gray model,¹⁷ which focused on the subdistribution hazard using the Stata ‘stcrreg’ command. We used non-cardiovascular mortality as competing events. Failure to account for competing risks can lead to overestimation of the incidence of the event.¹⁸ For statistical modelling, we included sex and age in model 1. In model 2, we added income, self-rated economic status, education level, marital status, depression, BMI, drinking status, smoking status, vegetable and fruit intake, daily physical activity, heart diseases, stroke, dyslipidaemia, diabetes, sleep disorder, self-rated health and region, in addition to model 1. We conducted analyses including both sexes, as we did not confirm any significance of the interaction term. However, some previous studies reported differences in cardiovascular mortality by sex;^{19 20} therefore, we also conducted sex-stratified analysis. To address potential reverse causation and confounding by pre-existing conditions, we conducted two sensitivity analyses: (1) excluding participants who died within the first year of follow-up and (2) excluding those with a prior diagnosis of heart diseases or stroke. To further assess the robustness of our findings, we also conducted two additional analyses: (3) applying standard Cox regression models to calculate HRs for cardiovascular mortality and (4) using inverse probability weighting (IPW), treating cardiovascular death as a binary outcome. To reduce selection bias due to missing values, we applied multiple imputations using chained equations. We generated 20 datasets and combined the results according to Rubin’s rule.²¹ All analyses were performed using Stata 18.0 (Stata Corporation LP, Windows version). The significance level was set at p<0.05. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for reporting.

Patient and public involvement

JAGES collaborates with many municipalities across Japan, involving a large number of members of the public in its research. In particular, the implementation of the questionnaire survey has been made possible through the cooperation of numerous municipal staff members. As the study is intended to inform policy, the development of the questionnaire incorporated input from administrative officials in participating municipalities. The results will be disseminated through this publication, press releases and municipal public relations channels.

Results

A total of 38 731 participants were included in the analysis. Table 1 presents the descriptive characteristics of the study participants (men: 46.6%). The mean age of the study participants at baseline was 73.6 (SD=6.0) years. The descriptive characteristics with missing values and those stratified by sex are shown in online supplemental tables S1 and S2. The descriptive characteristics stratified by housing status are also shown in online supplemental table S3. Residents with higher household equivalent income are more likely to live in owned detached houses or owned flats.

Table 1. Descriptive characteristics of the study sample (n = 38 731).

	No event		CV death		Other death		Total
	n	row%	n	row%	n	row%	n	col%
Housing status
Owned detached houses	30 179	89.9	763	2.3	2620	7.8	33 562	86.7
Owned flats	1510	91.8	22	1.3	114	6.9	1645	4.2
Rented flats	2146	88.1	68	2.8	222	9.1	2436	6.3
Rented detached houses and others	957	88.0	29	2.6	103	9.4	1088	2.8
Sex
Men	15 477	85.7	528	2.9	2047	11.3	18 052	46.6
Women	19 315	93.4	353	1.7	1011	4.9	20 679	53.4
Age
65–69 years	11 278	95.6	101	0.9	412	3.5	11 791	30.4
70–74 years	10 763	93.4	127	1.1	637	5.5	11 527	29.8
75–79 years	7548	88.5	224	2.6	757	8.9	8529	22.0
80–84 years	3810	79.7	235	4.9	733	15.3	4778	12.3
≥85 years	1393	66.1	194	9.2	519	24.6	2106	5.4
Income
<1.00 million JPY	5048	88.4	144	2.5	521	9.1	5713	14.8
1.00–1.99 million JPY	12 255	89.6	321	2.3	1104	8.1	13 680	35.3
2.00–2.99 million JPY	8468	90.1	200	2.1	726	7.7	9395	24.3
3.00–3.99 million JPY	5274	90.8	120	2.1	414	7.1	5808	15.0
≥4.00 million JPY	3746	90.6	96	2.3	293	7.1	4134	10.7
Self-rated economic status
Low	5204	89.4	140	2.4	477	8.2	5821	15.0
Middle–low	9721	89.9	224	2.1	868	8.0	10 813	27.9
Middle–high	16 135	90.3	403	2.3	1334	7.5	17 873	46.1
High	3732	88.3	114	2.7	379	9.0	4224	10.9
Education level
Low (≤ 9 years)	15 694	88.5	463	2.6	1583	8.9	17 740	45.8
Middle (10–12 years)	12 598	91.0	262	1.9	979	7.1	13 838	35.7
High (≥13 years)	6500	90.9	157	2.2	496	6.9	7153	18.5
Marital status
No partner	9142	88.0	313	3.0	933	9.0	10 387	26.8
Having partner	25 650	90.5	568	2.0	2126	7.5	28 344	73.2
Depression (GDS)
<5 points	25 696	91.1	555	2.0	1946	6.9	28 197	72.8
5–9 points	6982	87.2	234	2.9	790	9.9	8006	20.7
≥10 points	2115	83.6	92	3.7	322	12.7	2528	6.5
BMI
Low (<18.5)	2192	80.9	104	3.8	414	15.3	2710	7.0
Middle (18.5–24.9)	24 738	90.1	612	2.2	2094	7.6	27 445	70.9
High (≥25.0)	7862	91.7	165	1.9	550	6.4	8577	22.1
Drinking status
Current drinker	12 646	89.7	301	2.1	1151	8.2	14 098	36.4
Past drinker	1079	81.1	49	3.7	202	15.2	1330	3.4
Non-drinker	21 067	90.4	531	2.3	1705	7.3	23 303	60.2
Smoking status
Current smoker	3506	84.8	124	3.0	503	12.2	4133	10.7
Past smoker	9553	85.8	333	3.0	1252	11.2	11 138	28.8
Non-smoker	21 732	92.6	423	1.8	1304	5.6	23 460	60.6
Vegetable and fruits intake
<Once per day	6727	88.2	232	3.0	669	8.8	7628	19.7
≥Once per day	28 065	90.2	649	2.1	2389	7.7	31 104	80.3
Daily physical activity
<30 mins	11 028	86.3	405	3.2	1353	10.6	12 785	33.0
30–59 mins	12 443	90.7	289	2.1	981	7.2	13 713	35.4
60–89 mins	5674	92.1	102	1.7	383	6.2	6159	15.9
≥90 mins	5647	93.0	85	1.4	341	5.6	6074	15.7
Hypertension
No	16 486	89.0	417	2.3	1613	8.7	18 515	47.8
Yes	18 306	90.6	464	2.3	1446	7.2	20 216	52.2
Heart diseases
No	29 843	90.7	598	1.8	2457	7.5	32 897	84.9
Yes	4949	84.8	283	4.9	601	10.3	5834	15.1
Stroke
No	34 318	90.0	849	2.2	2985	7.8	38 153	98.5
Yes	474	81.9	32	5.5	73	12.6	579	1.5
Dyslipidaemia
No	29 838	89.0	822	2.5	2865	8.5	33 525	86.6
Yes	4955	95.2	59	1.1	193	3.7	5206	13.4
Diabetes
No	29 289	90.2	713	2.2	2486	7.7	32 487	83.9
Yes	5503	88.1	168	2.7	572	9.2	6244	16.1
Sleep disorder
No	32 195	89.8	823	2.3	2833	7.9	35 852	92.6
Yes	2597	90.2	58	2.0	225	7.8	2880	7.4
Self-rated health
Very good	4388	93.7	55	1.2	241	5.1	4685	12.1
Good	24 679	91.2	564	2.1	1818	6.7	27 061	69.9
Bad	5089	83.3	214	3.5	806	13.2	6110	15.8
Very bad	636	72.6	47	5.4	193	22.0	876	2.3
Total	34 792	89.8	881	2.3	3058	7.9	38 731	100

Each response was the average of 20 imputed datasets.

BMI, body mass index; CV, cardiovascular; GDS, Geriatric Depression Scale; JPY, Japanese Yen.

The median follow-up period was 2091 (range: 29–2191) days. The mortality rate per 1000 person-years was 3.97 for cardiovascular mortality and 13.8 for other causes of mortality. During the follow-up period, 2.3% had an onset of cardiovascular death, and 7.9% had other causes of death. The details of the causes of cardiovascular mortality are shown in online supplemental table S4. Figure 2 shows the Kaplan-Meier curve for cardiovascular mortality according to housing status. Compared with occupants living in owned flats, those living in owned detached houses and rented flats had higher cardiovascular mortality. Kaplan-Meier curves stratified by sex are also presented in online supplemental figure S1. Significant differences were observed among men based on the log-rank test, whereas no significant differences were found among women.

Figure 2. Kaplan-Meier curve for housing status and cardiovascular mortality (n = 36,912). Note: Log-rank test: p = 0.007.

Table 2A presents the results of the competing risk regression models. After adjusting for confounders, those living in rented flats were significantly associated with a higher risk of the onset of cardiovascular death (SHR=1.78; 95% CI 1.05–3.02) compared with those living in owned flats. Table 2B presents the results of the sex-stratified model. For men, those living in rented flats were significantly associated with a higher risk of cardiovascular death (SHR=2.32; 95% CI 1.13–4.75), although not statistically significant in women (SHR=1.29; 95% CI 0.59–2.82).

Table 2. Associations between housing status and cardiovascular death using competing risk regression models (n=38 731).

(A) All
	Model 1				Model 2
	SHR	95% CI		P value	SHR	95% CI		P value
Housing status (Ref: owned flats)
Owned detached houses	1.40	0.89	2.20	0.145	1.35	0.84	2.15	0.212
Rented flats	2.11	1.26	3.52	0.004	1.78	1.05	3.02	0.034
Rented detached houses and others	1.71	0.95	3.08	0.075	1.39	0.76	2.56	0.288

(B) Sex-stratified
	Model 1				Model 2
	SHR	95% CI		P value	SHR	95% CI		P value
Men (n=18 052)
Housing status (Ref: owned flats)
Owned detached houses	1.77	0.95	3.32	0.073	1.79	0.94	3.40	0.077
Rented flats	2.53	1.26	5.09	0.009	2.32	1.13	4.75	0.021
Rented detached houses and others	2.06	0.93	4.58	0.076	1.80	0.79	4.08	0.159
Women (n=20 679)
Housing status (Ref: owned flats)
Owned detached houses	1.02	0.52	1.97	0.960	0.93	0.47	1.85	0.841
Rented flats	1.65	0.78	3.52	0.193	1.29	0.59	2.82	0.515
Rented detached houses and others	1.30	0.55	3.10	0.547	1.01	0.41	2.51	0.978

Model 1: (A) sex and age adjusted. (B) Age adjusted.

Model 2: adjusted for income, self-rated economic status, education level, marital status, depression, BMI, drinking status, smoking status, vegetable and fruit intake, daily physical activity, heart diseases, stroke, dyslipidaemia, diabetes, sleep disorder, self-rated health and regions in addition to model 1.

BMI, body mass index; Ref, reference; SHR, subdistribution HR.

Online supplemental tables S5 and S6 present the results of sensitivity analyses excluding participants who either died within the first year of follow-up or had a prior diagnosis of heart diseases or stroke. In both models, living in rented flats was significantly associated with a higher risk of cardiovascular death (SHR=1.81; 95% CI 1.04–3.17; SHR=2.54; 95% CI 1.10–5.88). Similarly, in both models, men living in rented flats showed significantly elevated SHR (SHR=2.73; 95% CI 1.25–5.97; SHR=3.91; 95% CI 1.11–13.73). Online supplemental tables S7A and S7B show the results of the sensitivity analysis using Cox proportional hazard regression. Those living in rented flats were significantly associated with a higher risk of cardiovascular death (HR=1.83; 95% CI 1.08–3.09). When comparing the results of the Cox proportional hazards model with those of competing risk models, the latter provided more conservative estimates. The sex-stratified models also show a similar tendency to the main analysis. Those living in rented flats had a higher risk of cardiovascular death for men (HR=2.36; 95% CI 1.16–4.82) but not for women (HR=1.35; 95% CI 0.62–2.92). Online supplemental table S8 shows the results of the sensitivity analysis using IPW. The results indicate that living in rented flats was significantly associated with a higher probability of cardiovascular mortality (average treatment effect (ATE)=0.02; 95% CI 0.00–0.04), with a more pronounced association observed among men (ATE=0.04; 95% CI 0.01–0.08).

Discussion

This study analysed the association between housing status and cardiovascular mortality based on a JAGES cohort study of 38 731 occupants. The results showed that 1) during the median follow-up period of 2091 days, 881 occupants (2.3%) experienced cardiovascular death; 2) the lowest mortality was seen in those living in owned flats, followed by those in owned detached houses, with rented flats having the highest cardiovascular mortality; 3) adjusted competing risk models revealed a significant association between rented flat living and increased cardiovascular death risk (SHR=1.78); 4) men in rented flats showed a particularly higher cardiovascular death risk (SHR=2.32), but non-significant trend for women and 5) sensitivity analyses using Cox proportional hazards model supported these findings. Previous studies primarily investigated the relationship between the risk of CVDs and housing as a proxy indicator of socioeconomic status (SES).^{22 23} This study revealed that housing status affected deaths due to CVDs even after adjusting SES.

Housing, the space where people spend more than half of their lives, is regarded as a key social determinant of health (SDoH). A recent review²⁴ on SDoH explores how SDoH, such as economic, environmental and psychosocial factors, contribute to CVD risk, morbidity and mortality. It highlights the role of housing in shaping cardiovascular outcomes. Also, a recent report²⁵ issued by the AHA and the American College of Cardiology outlines the impact of social determinants on cardiovascular health. It identifies social, economic and environmental factors that influence CVD outcomes at individual, interpersonal and community levels. One of the key domains reviewed was housing quality and instability, with a particular focus on homelessness.^{26 27} Homeless individuals face approximately three times the risk of CVDs and higher cardiovascular mortality compared with non-homeless individuals.²⁶ However, among non-homeless individuals, there are few studies that examine cardiovascular mortality by housing status.

When considering housing status, housing type and tenure are fundamental and indispensable housing characteristics. Regarding housing type, although no significant differences were observed in this analysis, there was a general trend showing that cardiovascular mortality was higher in owned detached houses compared with owned flats. This is likely due to differences in indoor environmental quality resulting from the structural differences between detached houses and flats. Compared with blocks of flats, where each unit is typically surrounded by neighbouring units on all sides, detached houses, with all sides exposed to the outdoors, are more susceptible to the outdoor environment, making the indoor environment more prone to deterioration and instability. For example, previous research in Japan on detached houses has shown that indoor temperatures tend to be lower²⁸ and less stable.²⁹ Additionally, it has been demonstrated that such low indoor temperatures can elevate blood pressure (BP),³⁰ and temperature instability can increase BP variability,³¹ both of which are considered contributing factors to cardiovascular mortality. Regarding housing tenure, compared with owned flats, rented flats tend to have lower housing quality due to the issue of the split incentive, as explained in the background. In fact, according to the Housing and Land Survey in Japan, the percentage of homes with double window sashes or double-glazed windows is 38% for owned houses, while it is less than half that, at 15%, for rented houses. A recent study³² in China showed that occupants living in a rented house had an average indoor temperature 1.76°C lower than owner-occupiers. Another study³³ revealed that the occupants from rented flats were significantly more unlikely to be satisfied with their dwelling and to report their dwellings to be suitably warm in winter than the respondents from owner-occupied flats. Furthermore, owner-occupied flats showed less variation, especially for indoor temperature, indicating more stable indoor conditions.

In summary, one possible explanation for the findings of this study is illustrated in figure 3. In general, owned detached houses are similar in performance to owned flats but tend to experience a decline in indoor environmental quality, such as lower indoor temperatures and increased temperature instability, due to their structural characteristics. Compared with owned flats, rented flats tend to have lower indoor environmental quality due to differences in insulation performance, such as walls and windows. This decline in indoor environmental quality may lead to an increase in BP and greater BP variability, potentially contributing to CVDs.

Figure 3. Possible linkage between housing status and cardiovascular death through indoor environmental quality and blood pressure. CVDs, cardiovascular diseases.

Regarding demographic factors, men were strongly affected by housing status. One possible explanation for this is based on two factors: men generally have higher BP than women, and BP in hypertensive individuals tends to be more susceptible to the influence of the indoor environment. According to the Japanese Society of Hypertension Guidelines,³⁴ men in their 60s and 70s have higher systolic BP than women by 4.4 mm Hg and 2.5 mm Hg, respectively. Furthermore, previous studies³⁵ have suggested that while non-hypertensive residents experience a 2.2 mm Hg reduction in systolic BP following insulation retrofitting of houses, hypertensive patients experience a much larger reduction of 7.7 mm Hg, indicating their greater susceptibility to environmental influences. Therefore, the impact of housing status was more pronounced among men. In addition, as shown in online supplemental figure S1, cardiovascular mortality was higher among men, which may have contributed to the increased statistical power in this group. The Kaplan-Meier curves for women (online supplemental figure S1B) suggest that differences by housing status began to emerge in the later stages of the follow-up period. Therefore, with a longer follow-up, similar differences may become apparent among women as well.

To prevent cardiovascular death, it is important to maintain a minimum indoor temperature of 18°C, as recommended by the WHO,¹ to stay warm. However, due to the physical changes associated with ageing such as thermal blunting, older adults tend to become less sensitive to cold.³⁶ This makes it insufficient to rely solely on occupants for heating. Therefore, improving the insulation performance of homes should be prioritised. In fact, the WHO also recommends installing insulation in new housing and retrofitting insulation in existing housing to prevent cardiovascular morbidity and mortality.¹ This recommendation is supported by previous studies: one showing reduced hospital admissions for ischaemic heart disease among people aged 65 and over³⁷ and another showing lower mortality among older adults with a history of circulatory hospitalisations,³⁸ following home insulation retrofits. However, due to the issue of split incentives of rented housing, it may be necessary to incorporate policies inspired by successful examples from the UK and New Zealand. In Japan, gradual improvements are being made, such as the introduction of a duty to endeavour for energy efficiency labelling for buildings, including rented housing, starting in April 2024, indicating a step in the right direction. Furthermore, such initiatives for promoting the adoption of high-quality housing could contribute not only to improving cardiovascular health but also to enhancing planetary health by mitigating climate change through reduced energy consumption.

Strengths and limitations

The strength of this study lies in the use of high-precision cohort data, including cause of death, linked to the national long-term care insurance system, alongside a housing-related survey targeting the same cohort. There are limitations to this study. First, it has not accounted for differences in rented housing types. Rented housing includes various categories, such as public rented flats and private rented flats, and it is known that indoor environmental conditions can differ based on these categories.^{39 40} Therefore, a more detailed analysis considering these classifications is necessary. Second, unmeasured confounding factors may exist due to the nature of the observational study. For example, regarding SES, previous studies⁴¹ have shown that SES in older age is determined by a complex interplay of multiple factors, making complete adjustment challenging. However, in this study, three indicators of SES—income, self-rated economic status and education level—were included in the analysis to adjust for their influence on the greatest extent possible. Third, this study focused on older adults in Japan, and since housing conditions vary across countries, caution is required when generalising the findings to other contexts. Finally, there is a lack of objective data regarding housing conditions and residents’ health. In the discussion section, we explained potential pathways, using indoor temperature and BP as an example. Access to such objective data would allow for more precise analysis, but distributing temperature sensors and BP monitoring devices to a large sample is not practically feasible. Recently, the number of homes equipped with Home Energy Management Systems and attached temperature sensors has been increasing. Moreover, health data such as BP are increasingly being collected through smartphone applications. Therefore, overcoming this limitation will require efforts to integrate large-scale data on residential environments with health-related big data. Furthermore, while recent studies^{42 43} show that particulate matter 2.5 µm or less in diameter (PM_2.5) contributes to CVDs, most studies use outdoor PM_2.5 data. A comprehensive evaluation, including indoor PM_2.5 levels, is necessary for future studies.

Conclusions

The combination of housing type (detached houses vs flats) and tenure (owned vs rented) affected cardiovascular mortality after adjusting for covariates, including SES. Men living in rented flats showed a particularly higher risk of cardiovascular mortality. One possible cause of these findings might be the difference in indoor environmental quality, influenced by the structural characteristics and thermal insulation level of outer walls and windows. These findings can contribute to the development of public health policies and the improvement of cardiovascular health.

Data availability statement

Data were obtained from a third party and are not publicly available.