The effect of housing on the mental health of older people: the impact of lifetime housing history in Whitehall II

Housing patterns and health inequalities

Studies of patterns of health inequalities in older age groups are primarily focused on 'lifestyle' rather than structural variables and largely ignore possible explanatory variables such as housing,[4] despite strong evidence linking housing tenure to adult health in various longitudinal [5, 6] and cohort studies [7, 8]. Housing costs, including fuel use, rent or mortgages, maintenance and repairs, are a significant component of the minimum income for healthy living required by older people [9] and those who are home-owners may be mortgage-free by the time they retire. Time-use surveys consistently show that older people spend more than 90% of their time indoors, mostly in their homes, [10] so that the indoor home environment is their most significant environmental exposure, as well as being the place that they have most contact with their families or friends.

Generally, housing affects people's health at several levels [11]. Housing tenure is a structural variable; houses are usually the largest capital asset owned by families and this wealth can be used to generate a stream of income, in addition to salary, wages and benefits. In Britain, wealth is highest for those close to retirement, but inequalities are pronounced and magnified by differential access to pensions [12, 13].

Those who rent, whether from a private landlord or a social housing agency, are likely to be poorer, although in some cases they may be trading off more income for less wealth [14]. There is also a possible cultural effect of tenure choices on mental health. In England, people generally aspire to home ownership and renting is seen as a temporary measure. Renting permanently is more unusual, or may be seen as a sign of failure (particularly renting in the public sector). However, in some European countries, renting is just another housing option. Nonetheless, there is cross-cultural evidence that people who own their own houses are in better health than people who rent their houses, even controlling for income [15].

Home-ownership seems to confer both psychological and material advantages on owner occupants, [16–18] although a recent systematic review concludes the evidence is not strong [19]. Psychologically, owning a home rather than renting seems to confer greater autonomy and social status [20]; what economists call 'positional goods'. Houses that are owned are generally in better condition than rented accommodation. Rental housing is generally of poorer quality and more insecure. Leases, though they vary from country to country, do not give the same security to tenants as a house title gives to an owner [21]. However, this is not a static situation, in part because the housing market is such a pivotal part of the general economy and in an economic recession, home-owners, who bought in a boom, may be left with negative equity in their houses [22, 23]. In this case, home-ownership may be less secure than rental housing, particularly if the home-owner is made unemployed or becomes chronically ill. Mortgage payment commitments and the costs of maintaining housing can be stressful and the quality of housing that can be afforded on reduced incomes may be less health promoting than rental housing that can be afforded for the same expenditure [17, 24].

Housing quality is also an intervening variable between SES and health in producing social inequalities in health. Cold, damp, mouldy housing affects people's health and well-being, as well as their use of health services [25, 26]. Housing conforms to the inverse care law first identified in health care in Britain [27]. Colder and windier parts of the UK have poorer housing, which is associated with reduced lung function, as well as raised diastolic and systolic blood pressure [28]. People living in cold homes are more likely to have poorer mental health [29]. Those in single-person households tend to have higher living costs and are more likely to suffer from fuel poverty, i.e. they need to spend more than 10% of their income on household energy to maintain indoor temperatures to an adequate level [30, 31]. Fuel poverty has been exacerbated by the retail price of domestic fuel increasing by 91% between 2002 and 2009 [13]. Experiencing financial difficulties in general may well capture fuel poverty in particular.

Le Grand has argued that housing is no different from any other good in a capitalist society and should be considered fundamentally in monetary terms [32]. The contrary view is that the ontological security provided by housing is high on the hierarchy of needs and confers a unique range of services. Analysis of the Joseph Rowntree Foundation Poverty and Social Exclusion Survey showed that poor renters were significantly more likely to be dissatisfied with their neighbourhoods (10% vs 4%), but poor home-owners were more likely to report a structural problem with their house, such as a leaky roof (13% vs 4%) [33]. Poor home-owners were more likely to report poor mental health than renters, but this result was not significant.

Variables

Outcome: The General Health Questionnaire (GHQ-30) is a 30-item questionnaire that measures minor psychiatric morbidity [39]. This was asked at all the phases of the study. The Likert scale scores were coded so that responses ranged from 1 (good health) to 4 (poor health). These scores were summed up to give a continuous score ranging from 4 to 120, with a normal distribution. A continuous GHQ score was chosen over a measure of GHQ caseness in order to describe changes and trajectories in mental health more accurately. The GHQ questionnaire was asked at all phases of the study.

Exposures

Housing tenure was phase varying: Respondents were asked if the accommodation they lived in was owned or rented at each phase in this analysis. Rental accommodation was further split into those renting from local authority or housing association vs. renting privately. However, the numbers renting privately were too small to analyse as a separate category.

Housing quality was measured by a single question on 'To what extent do you have problems with your housing, e.g. 'too small, repairs, damp', with responses ranging from 'very great problems' to 'very little'. These responses were categorised into a binary variable- 'some to very great problems' vs. 'very little to slight problems'. This question was asked at each phase in this analysis, except at phase 9. So responses at phase 8 of the study on housing quality were used to replace the missing phase 9 housing quality variable.

Household financial problems was measured by a single question 'To what extent do you have difficulty paying bills', with responses ranging from 'very great problems' to 'very little'. These responses were categorised into a binary variable- 'some to very great difficulty' vs 'very little to slight difficulty'. This question was asked at each phase in this analysis.

Confounders

The confounders analysed in this study were gender (phase-invariant) and other phase-varying covariates including age in years occupational grade, retirement status, smoking status, alcohol consumption above recommended limits and marital status. Occupational grade, marital status, smoking, alcohol consumption and gender are potential socioeconomic confounders of the association between housing and health, while retirement status is a potential confounder of any factor associated with changes in mental health.

Ethical approval was given by University College London Medical School Committee on the Ethics of Human Research.

Analysis

We first examined the bivariate cross-sectional associations of the housing exposures with GHQ-30 and other confounders at phases 1 and 9.

We then analysed GHQ-30 longitudinally using repeated measures (over four cross-sectional phases) multiple (linear) regression models, looking at the relative contribution of each of the housing variables-housing tenure, financial problems, and housing problems-on GHQ-30 mental health, while controlling for other confounding variables. Three nested multivariate models were compared. The most general model estimated the effect of each housing variable separately on GHQ-30 scores, adjusted for age and gender. The next model additionally included all the three housing variables together and their interactions with age. The final model additionally included all the confounding variables and their interaction with age. A quadratic function of age and interactions with the housing variables was initially included in all the models. However, this quadratic term of age and related interactions did not significantly reduce the deviance of the final model and so were excluded from the models shown in the results below.

A repeated measures multilevel model with phases (level 1) nested within respondents (level 2) was specified. A random effect of age (at level 2) was also specified in all the models to take account of heterogeneity in the effect of age on GHQ-30. For the final model, there are four units at level 1 (denoting the phase of the study) and 9846 units at level 2 (corresponding to most of the participants at baseline).

Bivariate analyses

Housing tenure in the Whitehall II Study is skewed towards home-ownership and became more so over the life-course. At phase one (1985-88), when the participants were aged between 35 and 55, 92% of these respondents owned, partly or fully, a house and 8% rented. At phase 9, home ownership increased to around 96% of the sample. The social gradient in home ownership remained over the life course. At phase one, only 16% of all owner occupiers had low grade civil service jobs. By phase 9, this had reduced to 8%.

At phase one, there was not much difference in GHQ scores between those who were renting and home owners. However, by phase nine, those in rented housing had higher (poorer) GHQ scores (55.48) than owner occupiers (51.98). In terms of difficulty paying bills and problems with housing, those who reported some problems in either domain had higher (poorer) GHQ scores at phase one (for difficulty paying bills 57.7 vs 54.34 and for problems with house 58.06 vs 53.99), and this difference increased by phase nine (for difficulty paying bills 59.64 vs 51.67 and for problems with house 56.68 vs 51.22). At phase one of the study, the questions on housing problems and financial difficulties were not asked in all versions of the questionnaire, resulting in significantly lower proportions of responses. However, there was no selection of participants in terms of who received a particular version of the baseline questionnaire, suggesting that such non-response is not likely to bias the results.

Multivariate analyses

In Model 2, also shown in Table 2, where all the housing variables are included at the same time, as well as their interaction with age, the results were similar, although the effect of housing tenure decreases. Model 3 includes Model 2 as well as possible confounders, cohabitation status, employment grade and retirement status and their interaction with age. Housing tenure no longer had a significant independent effect, whereas housing problems (increased GHQ score by 3.11) and financial difficulties (increased GHQ score by 3.40) retained their explanatory significance and increased over time.

These effects are shown in graph form in Figure 1. They show growing inequalities by age, by experiencing problems with housing and by difficulty paying bills, estimated from Model 3 in Table 2.