Housing conditions and risk of COPD: a Danish cohort study, 2000–2018 | BMC Public Health

Study design

The study was a cohort study combining individual-level data from the Representative National Health and Morbidity Survey 2000 (DHMS 2000), the Danish Population Register, the Building and Dwelling Register, the Danish National Patient Register and the Danish Prescription Register.

Environment and study population

A total of 16,688 people (≥ 16 years) took part in the DHMS 2000, which corresponds to a response rate of 74.2%. The participants completed an interview with questions on topics such as demographics, health and indoor climate. The data were collected in three rounds in February, May and September 2000 by trained interviewers in the respondents’ homes.

Details of questionnaire design and sampling are described elsewhere (14, 15). Linkage of participants from the DHMS 2000 and the Danish national health and administrative registers at the individual level was made possible by the unique and permanent identification number assigned to all Danish residents at birth or immigration (16). In Denmark, everyone has free and equal access to health care.

The study population consisted of all individuals aged 30 years or older who did not have COPD for 10 years before the interview date. COPD was defined based on registry information, see section “Chronic obstructive pulmonary disease” and Supplementary Table S1.

Assessment of housing conditions and indoor climate

Information on construction time, urbanization and population density was obtained through linkage with the Building and Dwelling Register (17) and the Danish Civil Register (16, 18), while information on dwelling type and self-reported indoor environment was obtained from the questionnaire.

The construction period was based on information on the year of construction and was categorised into < 1962, 1962–1982 and ≥ 1983. The categorisation reflected changes in Danish regulations on the construction of buildings.

Living in urban and rural areas was based on information from the Building and Dwelling Register (17) and the Danish Civil Register (16, 18) and was categorized into rural (< 200), 200–4,999, 5,000–49,999, and ≥ 50,000 inhabitants. For descriptive purposes only, the variable was dichotomized into ≥ 50,000 inhabitants or not.

Ownership status was determined from the Building and Housing Register (17), based on information on whether the apartment was occupied by the owner or by a tenant. Persons in cooperative housing were grouped together as tenants.

Population density was based on the size of the dwelling and the number of persons in the household. The latter was determined annually. Population density was divided into quartiles. For descriptive purposes only, the variable was set at the median value (less than 55 m² per person; yes/no).

The type of housing situation was specified by the interviewer in the questionnaire. The type of housing situation was divided into single-family houses, semi-detached and terraced houses, apartments, farms and others. The category “others” included institutions, universities, etc. For the descriptive analyses, the type of housing situation was divided into apartments and other housing situations.

The perceived indoor environment was based on 13 items: self-reported perceived annoyances within the last 14 days (12 items) and location of the apartment next to a busy road (no/yes) (1 item). Based on these items, three latent classes were identified using latent class analysis. The three classes were characterized by 1) very little annoyance (88.8%, N= 14,829 based on the most likely class affiliation), 2) moderate annoyances (5.9%, N= 980) and 3) many annoyances (5.3%, N= 879) (described in (19)).

Chronic obstructive pulmonary disease

Information on COPD was retrieved from the Danish National Patient Register (20) and the Danish Prescription Register (21) for the 10 years before inclusion/baseline and until the end of follow-up (prescriptions from 1995 onwards). An incident COPD was defined as first primary diagnosis J44 (COPD) or J96 (respiratory failure) as primary diagnosis combined with J44 as secondary diagnosis or J13-18 (pneumonia) as primary diagnosis combined with J44 or J96 as one of the secondary diagnoses and/or redemption of two prescriptions with the anatomical therapeutic chemical (ATC) code R03A or R03B or the indication code 464 or 379 within a 12-month period (22, 23) (see Supplementary Table S1).

Covariate

Age, gender, and socioeconomic status were identified a priori as confounders using a directed acyclic graph (DAG) (24, 25) (see Supplementary Figures S1a and S1b). When examining the association between housing type, home ownership, resident density, or perceived indoor climate and COPD risk, cohabitation was also identified as a confounder. For the association between perceived indoor climate and COPD, duration of residence in the dwelling was also identified as a confounder. Depending on the exposure studied, some of the housing conditions were also identified as confounders, e.g., year of construction and urbanization were identified as confounders for the association between housing type and COPD risk (see Supplementary Figures S1a–S1f). Smoking was identified as a confounder only in the association between perceived indoor climate and COPD risk. However, because smoking is a very strong risk factor for the development of COPD (26, 27), all analyses were adjusted for smoking status, except for the association between home ownership and COPD, where smoking was identified as a mediator and therefore not controlled. Home ownership is a known marker of socioeconomic status (28), and because smoking is more prevalent in lower socioeconomic status groups, smoking was considered a mediator. Furthermore, all analyses were adjusted for calendar years (2000–2003, 2004–2007, 2008–2011, 2012–2015, and 2016–2018) because the IR of COPD varied during the study period. All confounders were collected at baseline, except for age, population density, cohabitation, and household income, which were collected annually and included in the analysis as time-varying covariates.

Information on age (30–39, 40–49, 50–59, 60–69, 70–79 and ≥ 80 years), sex (male and female) and cohabitation (living alone and living in a cohabitation) was obtained from the Danish Civil Register (18). Socioeconomic position was assessed by educational level and equivalent household disposable income. The highest level of education achieved was obtained from the Danish Education Register (29) and categorised as primary school (International Standard Classification of Education (ISCED) 1–2), short-term education (ISCED 3) and intermediate/long-term education (ISCED 5–8). Information on equivalent household disposable income was obtained from the Income Statistics Register (30) and grouped into quintiles year by year. For some of the stratified analyses, income quartiles were used. Length of residence before participation was calculated using information on moving history from Statistics Denmark and categorized as follows: < 3 years, 3–10 years, 11–20 years, and ≥ 21 years.

Information on smoking status was obtained in the interview by asking the questions “Do you smoke?” and “Have you ever smoked?” These questions were combined into a single variable indicating the respondent’s smoking status: 1) never smoker; 2) former smoker; 3) current smoker.

For descriptive purposes, information from the questionnaire on body mass index (BMI) and passive smoking were included.

statistical methods

For descriptive analysis, median and interquartile range (IQR) were used for continuous variables and counts with proportions for categorical variables. Incidence rates (IR) were used to describe the COPD rate.

The association between housing conditions, indoor environment and COPD was examined using Poisson regression with COPD as the outcome and logarithmic transformation of follow-up time as adjustment. We adjusted the analyses for the potential confounders identified a priori (see the Covariates section for more information). Follow-up time was divided by calendar year. Results are presented as incidence rate ratios with corresponding 95% confidence intervals (IRR and 95% CIs). All participants without COPD were followed from the date of interview until onset of COPD, death, emigration, change of address or end of study (31 December 2018), whichever came first. Analyses were weighted for non-responses using weights estimated by Statistics Denmark based on information such as sex, age, education and income (31). This means that a weight value was calculated for each respondent in the survey and that this value indicates how much the respondent’s response counts in the Poisson regression.

To exclude the possible influence of smoking, all analyses were additionally stratified according to smoking status (non-smokers, former smokers and current smokers).

All analyses were performed using STATA software (Stata Statistical Software: Release 17.0. College Station, TX: StataCorp LLC).

Sensitivity analyses

According to the DAG, the association between population density and COPD as well as the association between perceived indoor climate and COPD should have been adjusted for housing type and home ownership. However, since there was a high degree of overlap between these two variables, both variables could not be included simultaneously. Therefore, the variable intended to control the most important information was included in the main analysis (e.g. housing type in the association between population density and COPD). The analyses were repeated with adjustment for the opposite variable.

Missing data

The number of missing data on covariates was low, ranging from 0 (e.g. age and gender) to 26 (e.g. smoking status), and all analyses were conducted as complete cases. Information on household income was not available for the last year of follow-up for 1,582 individuals. Since the income register is compiled at the end of the year, this can happen if individuals, for example, emigrate or die before the income register is compiled. For these individuals, information from the previous year was used. The same applied to the number of individuals living in the dwelling and cohabitation, for which the number of missing data was 1,581.