Cardiovascular disease (CVD) constitutes a major public health burden and is the greatest cause of mortality globally. Stroke alone is the second greatest cause of death [
With today’s trends of urbanization [
To date, four different explanations of an association between floor level and CVD have been proposed. Three explanations have been suggested by Panczak and coworkers [
The aim in this study was to investigate the association between floor level and CVD morbidity. We used stroke as our main outcome variable. To better understand the influence of a late-look bias [
We used the Health and Environment in Oslo study (HELMILO) which is a cross-sectional study of the inhabitants of Oslo, Norway, conducted in 2009/2010. The sample consists of five age groups between 39 and 85 years (born in 1924/25, 1940/41, 1955, 1960, and 1970). The sample had originally been drawn for inclusion in the Oslo Health Study (HUBRO) and encompassed all Oslo inhabitants in the chosen age groups, as registered in the Norwegian National Registry as of December 1999 (
Overview of the sample.
We excluded those who reported that they had lived less than one year at the current address (
The questionnaire contained the inquiry “Have you or have you had?” followed by a list of sixteen different disease outcomes. Response choices were “yes” and “no” as well as two additional alternatives of “yes” and “no” with an inscription above, “confirmed by a doctor.” The following three cardiovascular events were chosen for this study (our nomenclature in brackets): “stroke (cerebral infarction/haemorrhage, ministroke)” (stroke); “blood clot, phlebitis” (venous thromboembolism (VTE)), and “hardening of the arteries in the legs” (intermittent claudication (IC)).
Prevalent cases were defined as participants having or having had the disease, irrespective of whether the disease was reported to have been confirmed by a doctor. Those with missing values on both subquestions of disease (disease/doctor confirmed disease) remained missing in our new variable. An exception was the instances where values were missing at one disease inquiry, but complete on one of the other two disease inquiries. In these cases, missing values were interpreted as “no.” Participants who answered both “yes”
We included two available measures of socioeconomic status. Education level was reported as number of years of schooling. We split this variable into three education levels: 12 years or less; 13–16 years; and more than 16 years of education. Occupational status was initially given in nine categories at an ordinal scale. The participants chose one or more employment categories that they were or had been employed in. We used the highest status reported. In a new variable, we kept the first category: “administrative leader, politician.” Then, we collapsed the next two categories: “academic occupations (at least 4 years of high school or university education)” and “occupations with shorter high school or university education (1–3 years) and technicians.” A third category was produced from the following two occupational categories: “office and customer service occupations” and “sales, services, and care professions.” The fourth category, which we termed blue collar workers, constituted those who were or had been employed in “farming, forestry, or fishery occupations”; as “craftsman, builder, labourer, and so forth”; as “operator (machinist), driver, and so forth”; and in “elementary occupations without need for formal education.”
Chi-square tests were used to investigate the association between categorical variables such as prevalence of the three disease outcomes and floor level. We used logistic regression to model the odds for the outcomes as a function of floor levels, with “basement and 1st floor” as reference category. In multivariate models, we controlled for potentially confounding variables. The covariates were included stepwise in three blocks of belonging variables. Results are reported for all three models, as well as crude estimates (Model 0). The variable
Analyses were conducted in SPSS version 22.
HELMILO is approved by the Regional Committee for Medical and Health Research Ethics, Norway. We used anonymous data in the present study.
Population characteristics are reported in Table
Sample characteristics dependent on floor level.
Floor level | Total | |||||
---|---|---|---|---|---|---|
Basement and 1st floor | 2nd-3rd floor | 4th-5th floor | 6th–10th floor | ≥11th floor | ||
| ||||||
Type of housing | ||||||
Block apartment/terraced flat | 35.0 | 41.8 | 94.6 | 97.4 | 80.8 | 48.1 |
Detached house/villa | 36.2 | 27.3 | 0.3 | 0.5 | 11.1 | 25.9 |
Un/semidetached house | 26.0 | 27.8 | 0.5 | 0.9 | 4.0 | 22.8 |
Other residences | 2.8 | 3.1 | 4.5 | 1.2 | 4.0 | 3.1 |
Period of residence | ||||||
1–10 years | 37.7 | 39.6 | 49.4 | 48.6 | 31.3 | 40.4 |
>10 years | 62.3 | 60.4 | 50.6 | 51.4 | 68.7 | 59.6 |
| ||||||
Age (year of birth) | ||||||
1924-25 | 9.4 | 8.5 | 6.6 | 13.2 | 18.2 | 8.9 |
1940-41 | 28.8 | 23.2 | 28.1 | 32.5 | 31.3 | 26.0 |
1955 | 20.1 | 22.6 | 19.1 | 16.7 | 21.2 | 21.2 |
1960 | 22.0 | 24.2 | 20.1 | 19.8 | 12.1 | 22.7 |
1970 | 19.7 | 21.4 | 26.0 | 17.7 | 17.2 | 21.2 |
Gender (men) | 46.0 | 46.4 | 48.9 | 47.4 | 54.5 | 46.7 |
Living with someone (yes) | 79.3 | 78.5 | 61.5 | 54.2 | 58.6 | 75.7 |
Country of origin | ||||||
Norway | 80.8 | 81.7 | 77.7 | 74.3 | 76.8 | 80.6 |
Other western countries | 6.4 | 6.1 | 6.2 | 7.5 | 2.0 | 6.2 |
Other nonwestern countries | 12.8 | 12.2 | 16.2 | 18.2 | 21.2 | 13.2 |
| ||||||
Education (in years) | ||||||
≤12 | 32.0 | 28.6 | 32.7 | 40.8 | 52.5 | 30.9 |
13–16 | 36.2 | 33.4 | 35.1 | 27.6 | 31.3 | 34.3 |
>16 | 31.8 | 38.0 | 32.2 | 31.6 | 16.2 | 34.9 |
Occupational status | ||||||
Leader, politician | 17.4 | 17.6 | 13.5 | 13.9 | 14.1 | 16.9 |
Occ. req. higher education | 42.6 | 45.3 | 41.6 | 36.8 | 28.3 | 43.5 |
Office, sales, care, etc. | 29.2 | 27.6 | 32.4 | 35.8 | 39.4 | 29.1 |
Blue collar and farming | 10.8 | 9.5 | 12.5 | 13.4 | 18.2 | 10.5 |
| ||||||
Body mass index (BMI) | ||||||
Normal and underweight | 52.5 | 54.5 | 55.0 | 50.7 | 46.5 | 53.7 |
Overweight | 36.7 | 35.4 | 33.7 | 33.7 | 41.4 | 35.6 |
Obese | 10.8 | 10.1 | 11.3 | 15.6 | 12.1 | 10.7 |
Physical activity (PA) level | ||||||
1 (sedentary) | 9.9 | 9.0 | 11.2 | 14.9 | 15.2 | 9.8 |
2 | 32.0 | 30.4 | 29.7 | 34.2 | 36.4 | 31.0 |
3 | 33.2 | 35.8 | 38.3 | 37.0 | 26.3 | 35.2 |
4 | 20.5 | 19.7 | 14.7 | 9.9 | 16.2 | 19.0 |
5 (very active) | 4.5 | 5.0 | 6.2 | 4.0 | 6.1 | 5.0 |
Smoking status | ||||||
Current smoker | 15.5 | 14.5 | 18.8 | 18.2 | 21.2 | 15.5 |
Former smoker | 38.5 | 37.5 | 36.5 | 41.7 | 28.3 | 37.8 |
Never smoked (RG) | 46.0 | 48.0 | 44.7 | 40.1 | 50.5 | 46.7 |
Alcohol consumption | ||||||
Frequent | 41.5 | 43.1 | 43.2 | 41.7 | 31.3 | 42.4 |
Moderate | 42.1 | 40.2 | 37.5 | 35.8 | 41.4 | 40.4 |
Infrequent | 16.4 | 16.7 | 19.2 | 22.4 | 27.3 | 17.2 |
Fatty fish consumption | ||||||
High | 4.7 | 4.5 | 4.0 | 3.5 | 6.1 | 4.5 |
Moderate | 64.2 | 63.7 | 62.8 | 67.7 | 62.6 | 63.9 |
Low | 31.1 | 31.8 | 33.3 | 28.8 | 31.3 | 31.6 |
Vegetable consumption | ||||||
High | 42.7 | 44.9 | 39.8 | 37.3 | 37.4 | 43.3 |
Moderate | 53.8 | 52.5 | 56.8 | 56.1 | 59.6 | 53.6 |
Low | 3.6 | 2.6 | 3.4 | 6.6 | 3.0 | 3.2 |
Fruit consumption | ||||||
High | 44.1 | 46.5 | 44.3 | 39.4 | 32.3 | 45.1 |
Moderate | 47.9 | 47.1 | 46.8 | 50.2 | 60.6 | 47.6 |
Low | 8.0 | 6.4 | 8.9 | 10.4 | 7.1 | 7.4 |
All values are given in percentages.
The figures were similar among block apartment residents alone (data not shown). However, participants in block apartments living on the 0–3rd floor were, for instance, less educated and more often of nonwestern origin compared to the same floor levels in the full sample.
The floor level values had a range from 0 to 33, median and mode of 2, and mean of 2.33. Only 32 residents reported to reside above the 12th floor (data not shown). A total of 19.2% of residents who reported to live on the 11th floor or higher also reported that they did not reside in multistory buildings (data not shown). Parallel figures for those living on the 6th–10th floor were 2.6%.
In Table
Prevalence of disease by floor level and chi-square tests of difference (all types of housing).
Total | Basement and 1st floor | 2nd-3rd floor | 4th-5th floor | 6th–10th floor | ≥11th floor | | |
---|---|---|---|---|---|---|---|
Stroke | 4.6% ( | 4.8% ( | 4.0% ( | 5.4% ( | 7.3% ( | 11.1% ( | <0.001 |
Venous thromboembolism (VTE) | 4.9% ( | 4.6% ( | 4.8% ( | 4.5% ( | 9.0% ( | 10.1% ( | <0.001 |
Intermittent claudication (IC) | 4.8% ( | 4.9% ( | 4.6% ( | 4.4% ( | 6.4% ( | 13.1% ( | 0.001 |
Results from the regression models are shown in Table
Results from logistic regression models including all types of housing (
Basement and 1st floor | 2nd-3rd floor | 4th-5th floor | 6th–10th floor | ≥11th floor | | |||||
---|---|---|---|---|---|---|---|---|---|---|
OR | 95% CI | OR | 95% CI | OR | 95% CI | OR | 95% CI | |||
| ||||||||||
Model 0 | 1 (ref.) | 0.836 | (0.683–1.023) | 1.139 | (0.855–1.517) | 1.578 | (1.061–2.345) | 2.500 | (1.311–4.766) | <0.001 |
Model 1 | 1 (ref.) | 0.912 | (0.742–1.122) | 1.206 | (0.895–1.623) | 1.281 | (0.849–1.932) | 1.936 | (0.990–3.785) | 0.001 |
Model 2 | 1 (ref.) | 0.915 | (0.744–1.126) | 1.193 | (0.886–1.608) | 1.240 | (0.821–1.873) | 1.846 | (0.945–3.605) | 0.002 |
Model 3 | 1 (ref.) | 0.907 | (0.736–1.118) | 1.172 | (0.868–1.583) | 1.183 | (0.779–1.795) | 1.876 | (0.953–3.691) | 0.002 |
| ||||||||||
Model 0 | 1 (ref.) | 1.046 | (0.859–1.274) | 0.984 | (0.725–1.336) | 2.043 | (1.415–2.951) | 2.332 | (1.191–4.566) | <0.001 |
Model 1 | 1 (ref.) | 1.129 | (0.924–1.379) | 1.052 | (0.770–1.439) | 1.814 | (1.242–2.647) | 1.961 | (0.985–3.906) | 0.007 |
Model 2 | 1 (ref.) | 1.138 | (0.932–1.391) | 1.046 | (0.765–1.431) | 1.768 | (1.210–2.585) | 1.873 | (0.940–3.733) | 0.012 |
Model 3 | 1 (ref.) | 1.143 | (0.935–1.398) | 1.043 | (0.761–1.429) | 1.720 | (1.174–2.518) | 1.873 | (0.939–3.738) | 0.015 |
| ||||||||||
Model 0 | 1 (ref.) | 0.935 | (0.770–1.136) | 0.882 | (0.649–1.199) | 1.312 | (0.864–1.993) | 2.917 | (1.597–5.327) | <0.001 |
Model 1 | 1 (ref.) | 1.007 | (0.825–1.228) | 0.950 | (0.639–1.303) | 1.111 | (0.722–1.708) | 2.410 | (1.288–4.509) | 0.003 |
Model 2 | 1 (ref.) | 1.015 | (0.832–1.238) | 0.940 | (0.685–1.290) | 1.083 | (0.703–1.668) | 2.292 | (1.226–4.283) | 0.005 |
Model 3 | 1 (ref.) | 1.020 | (0.835–1.246) | 0.951 | (0.692–1.308) | 1.067 | (0.691–1.647) | 2.318 | (1.237–4.345) | 0.005 |
Model 0: crude associations.
Model 1: age, gender, living with someone, and country of birth (sociodemographics).
Model 2: Model 1 + education and occupational status (SES).
Model 3: Model 2 + BMI, PA, smoking status, alcohol consumption, and diet (health behaviors).
In the fully adjusted models, the effect measures were generally attenuated. The associations between floor level and stroke were no longer statistically significant. Residents on the 6th–10th floor had increased odds of having or having had VTE (OR: 1.720; 95% CI: 1.174–2.518) and residents on the 11th floor or higher had increased odds of having or having had IC (OR: 2.318; 95% CI: 1.237–4.345). The test of trend showed a statistically significant increase in prevalence of
When we investigated block apartment residents alone, the prevalence of VTE differed significantly across floor levels (
Prevalence of disease by floor level and chi-square tests of difference (block apartment residents only).
Total | Basement and 1st floor | 2nd-3rd floor | 4th-5th floor | 6th–10th floor | ≥11th floor | | |
---|---|---|---|---|---|---|---|
Stroke | 5.6% ( | 5.9% ( | 5.4% ( | 5.2% ( | 7.3% ( | 7.5% ( | 0.481 |
Venous thromboembolism (VTE) | 5.9% ( | 5.7% ( | 6.0% ( | 4.5% ( | 9.0% ( | 8.8% ( | 0.012 |
Intermittent claudication (IC) | 5.8% ( | 6.6% ( | 6.0% ( | 4.3% ( | 6.3% ( | 10.0% ( | 0.051 |
The associations found between residing on higher floors and CVD among all residents (Table
Results from logistic regression models (block apartment residents only (
Basement and 1st floor | 2nd-3rd floor | 4th-5th floor | 6th–10th floor | ≥11th floor | | |||||
---|---|---|---|---|---|---|---|---|---|---|
OR | 95% CI | OR | 95% CI | OR | 95% CI | OR | 95% CI | |||
| ||||||||||
Model 0 | 1 (ref.) | 0.907 | (0.676–1.215) | 0.874 | (0.619–1.234) | 1.249 | (0.805–1.937) | 1.293 | (0.545–3.068) | 0.363 |
Model 1 | 1 (ref.) | 0.921 | (0.682–1.243) | 1.030 | (0.723–1.467) | 1.151 | (0.734–1.807) | 1.170 | (0.482–2.842) | 0.377 |
Model 2 | 1 (ref.) | 0.930 | (0.689–1.257) | 1.054 | (0.739–1.503) | 1.166 | (0.741–1.833) | 1.138 | (0.469–2.764) | 0.354 |
Model 3 | 1 (ref.) | 0.935 | (0.689–1.269) | 1.076 | (0.750–1.544) | 1.134 | (0.715–1.798) | 1.215 | (0.494–2.991) | 0.316 |
| ||||||||||
Model 0 | 1 (ref.) | 1.056 | (0.790–1.411) | 0.765 | (0.533–1.098) | 1.615 | (1.069–2.439) | 1.574 | (0.700–3.540) | 0.715 |
Model 1 | 1 (ref.) | 1.082 | (0.806–1.452) | 0.876 | (0.607–1.263) | 1.568 | (1.030–2.386) | 1.475 | (0.645–3.371) | 0.516 |
Model 2 | 1 (ref.) | 1.086 | (0.809–1.459) | 0.895 | (0.620–1.293) | 1.563 | (1.025–2.383) | 1.446 | (0.632–3.311) | 0.484 |
Model 3 | 1 (ref.) | 1.091 | (0.810–1.467) | 0.900 | (0.621–1.303) | 1.517 | (0.992–2.321) | 1.440 | (0.626–3.310) | 0.491 |
| ||||||||||
Model 0 | 1 (ref.) | 0.903 | (0.684–1.192) | 0.633 | (0.444–0.904) | 0.949 | (0.603–1.496) | 1.570 | (0.732–3.368) | 0.573 |
Model 1 | 1 (ref.) | 0.905 | (0.680–1.204) | 0.741 | (0.515–1.066) | 0.881 | (0.553–1.405) | 1.424 | (0.646–3.139) | 0.551 |
Model 2 | 1 (ref.) | 0.918 | (0.690–1.222) | 0.761 | (0.529–1.096) | 0.894 | (0.559–1.428) | 1.380 | (0.626–3.039) | 0.519 |
Model 3 | 1 (ref.) | 0.945 | (0.708–1.262) | 0.796 | (0.551–1.151) | 0.886 | (0.552–1.423) | 1.418 | (0.637–3.153) | 0.549 |
Model 0: crude associations.
Model 1: age, gender, living with someone, and country of birth (sociodemographics).
Model 2: Model 1 + education and occupational status (SES).
Model 3: Model 2 + BMI, PA, smoking status, alcohol consumption, and diet (health behaviors).
A protective crude association appeared when we investigated block apartment residents separately (Table
When we included the interaction terms between floor level and period of residence (1–10 years versus >10 years), we did not find evidence that time lived at the different floor levels modified any of the associations (
In this study, we found significant crude associations between floor levels and the three outcomes. The association disappeared for stroke when we included the confounding variables but remained statistically significant for participants with IC and a history of VTE living in upper floors compared to the lowest level (0-1st floor). We found statistically significant linear trends (gradients) for all outcomes in adjusted model, but no such trends were observed among individuals living in block apartments. For block apartment residents, the effect measures were attenuated and not statistically significant, except for higher odds of history of VTE in residents of the 6th floor or higher when we erased the highest category border at the 10th floor.
Our results are in opposition to the only other previous study on the association between floor level and CVD that we have found. Panczak and coworkers (2013) reported a negative association between floor level and stroke mortality, as well as total CVD mortality, in a longitudinal study of the entire adult Swiss population of residents in buildings with four floors or more. The present study differs from the Swiss study by investigating a different population. In addition, our study is cross-sectional; we control for relevant health behaviors; we investigate additional outcomes such as VTE and IC; and we investigate the association between floor level and CVD for all types of housing and for block apartment residents only.
Vertical variations of air pollution [
A large proportion of individuals with low socioeconomic status resided on higher floors. SES is a particularly important aspect with regard to the development of CVD [
The latter view is supported by a prospective study of elder Americans finding an unexplained increased stroke risk in individuals living in multistory buildings [
The higher occurrence of CVD at higher floors may also fit a recent hypothesis about the impact of atmospheric electricity on the health of residents at higher floor levels [
The statistically nonsignificant interaction between period of residence (1–10 years versus more than 10 years) and floor level indicates that exposure time to floor level is not associated with more CVDs (causal criteria). However, we cannot rule out a causal link. The time measure is rough, and we cannot ascertain that the floor level of the participants’ previous residence was different. In fact, settlement patterns do show consistencies; it is, for instance, common to settle close to people one considers similar in terms of SES [
The estimated associations between floor level and stroke were consistently lower than for the other CVD outcomes. It is possible that a late-look bias [
Several factors may affect the generalizability of the results across time and place. Selection into floor level and type of housing may depend strongly on social, economic, and cultural environment (e.g., the housing market and housing policies) which change over time and differ greatly across countries and even cities. Oslo, which is the capital of Norway, has its own distinctive characteristics with regard to distribution of type, standard, and age of buildings, which may affect the public perception of the social status of different housing types. For instance, the tallest buildings in Oslo were built in the 1960s and 1970s and have a monotonous and “brutal” design [
There are two major strengths in this study. The study is based on a large representative sample of an urban population, and we were able to include confounding factors known to be associated with risk of CVD. Nevertheless, the study would have improved if we had been able to include variables on income and the height of the block apartment buildings. In addition, it is still possible that an unknown variable confounds the association between residence floor level and the outcome variables.
An important limitation is the cross-sectional design, which cannot ascertain causality. The design can also be affected by a late-look bias [
In this study, we found that floor level was positively associated with CVD history in adult inhabitants of Oslo. The findings may point to residual confounding by building height and socioeconomic status and underline a need to understand possible associations between residing in tall buildings and CVD. A causal effect of residing on higher floor levels per se seems less likely but could stem from a poorer psychosocial environment on higher floor levels. The alternative explanation of an impact of atmospheric electrical parameters may also be considered. We found no statistically significant negative associations between floor level and CVD, which questions the impact of higher levels of air pollution, environmental noise, or less use of stairs at lower floor levels on CVD, but the study design limits the ability to identify such effects. As the present study contradicts a previous study from Switzerland, more studies are needed to fully understand the association between floor level and CVD and to disentangle possible causal mechanisms.
The authors declare that there are no competing interests regarding the publication of this paper.