STUDY OBJECTIVE To examine the association between damp housing and adult health, taking into account a wide range of other factors that may influence health and could confound this relation.
PARTICIPANTS AND SETTING A general population sample of adults, aged 18–64, from Oxfordshire, Buckinghamshire, Berkshire and Northamptonshire.
DESIGN Secondary analysis of responses to a postal questionnaire survey carried out in 1997 with a 64% response rate (8889 of 13 800). Housing dampness was assessed by self report. Health was measured by responses to a series of questions including presence of asthma and longstanding illness generally, use of health services and perceived health status (the SF-36). The effect of damp was examined using the χ2 test and one way analysis of variance. Significant associations with the various health outcomes were further explored taking into account 35 other housing, demographic, psychosocial and lifestyle variables using stepwise logistic and linear regression.
MAIN RESULTS Bivariate analyses indicated that damp was associated with the majority of health outcomes. Regression modelling however, found that being unable to keep the home warm enough in winter was a more important explanatory variable. Worry about pressure at work and to a lesser extent about money, showed an independent association with perceived health status equal to or greater than that of the housing environment, including cold housing, and that of health related lifestyles.
CONCLUSIONS This study shows that being unable to keep the home warm enough in winter is more strongly associated with health outcomes than is damp housing. However, as cold and damp housing are closely related, it is likely that their combined effects are shown in these results. The importance of worry as an independent predictor of health status needs testing in other studies. Its prevalence and relative importance suggest that it may be a significant determinant of public health.
- damp housing
- cold housing
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Several studies document a higher prevalence of ill health, particularly respiratory symptoms1 2 and longstanding illness3 in adults living in damp housing. In children, consistent associations have been shown between damp housing and ill health, in particular respiratory symptoms.1 4-7The relation between damp housing and ill health is complex as many factors associated with poor housing also affect health.8Teasing out specific associations from this “social complex”9 is difficult. A number of studies have succeeded in accounting for some potential confounding variables, by design or analysis, or both,1-3 6 10 and there is now evidence that the relation between damp housing and ill health is causal. For example, the association has been shown consistently in different populations.7 11-15 Several studies have demonstrated a dose response relation between severity of damp and prevalence of health problems.1 2 6 15 16 Also, the association between damp housing and some aspects of health is biologically plausible through the effects of house dust mites17 and moulds,18 which are associated with a variety of allergic diseases including asthma.17 19-24 Although the mechanism underlying the relation between other illnesses and damp is less clear, it has been suggested that fungal toxins may be involved in causing a variety of symptoms, including those that are non-respiratory1(quoting May et al 25). Alternatively, the wide range of health problems implicated, and their different aetiologies suggests that damp housing may create general susceptibility to poor health, by mechanisms that are not yet understood.
The third Oxford Healthy Life Survey (OHLS3) was commissioned by the Directors of Public Health for Berkshire, Buckinghamshire, Northamptonshire and Oxfordshire primarily to obtain information about key determinants of health. It provides a unique opportunity to examine the association between damp housing and health while accounting for a range of other variables not always available in housing and health studies. The survey measured health in a number of ways and collected data on a range of potential confounding factors.
The OHLS3 questionnaire was administered by post, in 1997, to a randomly selected sample of 14 868 adults aged 18 to 64 years living in Berkshire, Buckinghamshire, Northamptonshire and Oxfordshire. The methodology has been described elsewhere.26 Ethical approval was granted by the local research ethics committee. A secondary analysis was undertaken of the 8889 responses to this survey.
The health outcomes used in this analysis were: prevalence of longstanding illness, and asthma specifically; use of health services; and the anglicised version of the SF-3627 (see fig 1). Initially, bivariate associations between damp housing and each health outcome were examined using the χ2 test or one way analysis of variance. These associations were further examined in the context of 35 independent variables (see fig 2) using stepwise logistic or linear regression modelling. Number of visits to the GPs' surgery was transformed from a five category variable into a dichotomous variable comparing those who visited the surgery four or more times in the past year (that is, more often than the OHLS3 average) with those who visited less often or never. Six logistic regression models were undertaken, one for longstanding illness generally, one for asthma specifically, and four for aspects of health service use. Eight linear regression models were undertaken for health status (one for each dimension of the SF-36). For all regressions, a forward stepwise method was used, so each variable was added to the model in order of the significance of its score statistic. Variables were entered into the models at the 5% level of significance (p<0.05), however only those variables that were highly statistically significant (p<0.01) are shown in the tables. In the following sections, the term significant refers to p<0.01. All analyses were undertaken using SPSS version 7.5.
Cold housing is more strongly associated with ill health than is damp housing when the two are examined together.
Cold and damp housing are closely related.
Cold housing is associated with a wide range of health outcomes.
The importance of reported cold over damp housing merits further consideration.
Cold housing is less strongly associated with health than are some other variables, notably “worries” about work and money.
A response rate of 64% (8889 of 13 800) was achieved for the OHLS3 after 1068 deletions were made mainly because of “unknown addressees”. A comparison with the 1991 census showed that respondents were broadly representative of the general population.26 Characteristics of respondents are shown in table 1.
ASTHMA, LONGSTANDING ILLNESS AND HEALTH SERVICE USE
In the bivariate analysis, damp housing was associated both with longstanding illness and with asthma specifically (see table 2). It was also associated with four or more visits to the GP's surgery in the past year, use of hospital casualty and of outpatient departments in the past three months. A linear dose response relation was demonstrated for all these associations (p<0.001) except for use of hospital inpatients, where p=0.048.
Despite these initial results, damp housing was not significant in any of the six logistic regression models (see table 3). However, being unable to keep the home warm enough in winter was associated with asthma, longstanding illness generally, visits to the GP's surgery and outpatient departments. Those who had difficulty keeping their home warm enough “most of the time” were nearly twice as likely to visit the surgery four or more times (odds ratio 1.8, 95% confidence intervals 1.2, 2.7), and twice as likely to use outpatient departments (odds ratio 2.1, 95% confidence intervals 1.4, 3.3) as those who never experienced this problem. The distribution of the regression coefficients provides evidence of a dose response relation.
Other variables consistently significant in these models were age, employment status and sex of respondent. Age was significant in five out of six models. Asthma, use of hospital inpatients and casualty were more common in younger people, while longstanding illness generally and use of outpatient departments were more common in older people. Sex of respondent was significant in five, and employment status in four of the six models. Women were more likely than men to report a longstanding illness, visit their GP's surgery four or more times, be admitted as an inpatient, or use outpatient departments, while men more commonly used casualty departments. Being unable to work because of illness/disability, or being retired, were the most important categories of employment status associated with negative health outcomes.
The “smoking” variable was significant in three models but the effect was attributable to being an ex-smoker rather than a current smoker. The former was associated with asthma, longstanding illness generally, and visits to the GP's surgery. Worry about pressure at work featured in three of the six models, and the regression coefficients are strongly suggestive of a dose response relation between increased worry and the prevalence of illness and greater health service use.
In the bivariate analysis, damp housing was associated with all eight dimensions of the SF-36 in a linear dose response relation (p<0.001). Those living in damp housing experienced poorer health status than those not in damp housing. The variables that were significant in the eight linear regression models varied according to SF-36 dimension (see table 4). Damp housing was only associated with “health perception”, and only when reported as “more of a nuisance than a problem”. The most important housing variable was cold housing, associated with poorer health status on all eight dimensions (p<0.001). Being unable to keep the home warm enough “quite often” and “most of the time” were associated with poorer health status than that of being unable to keep the home warm “only occasionally”.
Other variables associated with all eight dimensions were employment status, vigorous exercise, worry about money and about work pressure. The most important categories of employment status were being unable to work because of illness/disability, and retired, compared with being in full time employment. Never taking vigorous exercise was associated with poorer health status when compared with taking vigorous exercise five or more times per week. Where more than one category was significant, the distribution of the regression coefficients is strongly suggestive of a dose response relation. Similarly, there is strong evidence of dose response between poorer health status and more frequent worry about money and work pressure.
RELATIVE IMPORTANCE OF DIFFERENT VARIABLES
From the point of view of their consistent predictive value in the models, the most important factors relating to health were employment status, cold housing, and sex of respondent, each associated with 12 of the 14 outcomes. By far the largest effect on health was attributable to inability to work because of illness/disability. Because of the possibility that this category was determining the importance of employment status in the analyses, the modelling was repeated omitting this group (n=318). However, this resulted in no major changes to the models, suggesting that other occupational categories were also important predictors of ill health, namely those not involving paid employment. Being unemployed (seeking work), caring for home or family (not seeking work), or retired were associated with seven, nine and 12 health outcomes respectively, although the size of their effects on health were small compared with that of being unable to work.
In relation to cold housing, it is notable that the effect on health of being unable to keep the home warm enough “quite often” or “most of the time” was in almost all models greater than that of any of the health related lifestyles including vigorous exercise and smoking (either current or previously).
In relation to sex of respondent, the finding that women's health was poorer than that of men in 11 of the 12 models where sex was significant (the exception being the use of hospital casualty departments) is consistent with published literature.28
The next most frequently significant variable was worry about pressure at work, associated with 11 outcomes. In terms of relative magnitude of effect on health outcomes (as indicated by the regression coefficients), worry about pressure at work and about money (associated with nine outcomes) were as important, or more important than either cold housing or health related lifestyles. The importance of these “worry” variables in relation to mental health is not surprising, but it is notable that their effect was greater than that of problems in the local environment, for example poor public transport, smells and fumes, noise, lack of open spaces, and vandalism. While respondents were asked whether they believed the area where they lived suffered from the latter problems, rather than how often they worried about them, as was the case with concern about money and pressure at work, in both cases respondents were offering a subjective impression. Given the greater prevalence in this survey of work and money worries compared with housing problems, these models suggest the former may be more important from a public health perspective.
Vigorous physical exercise was associated with 10 health outcomes, and respondent age with nine outcomes. As might be expected, the results show that regular vigorous exercise is associated with perceived good health and high energy levels. Both this and the apparent protective effect of any vigorous activity on longstanding illness may be an outcome rather than a cause. The relatively low impact on current health attributable to health related lifestyles other than physical activity is likely to be attributable to the fact that the detrimental impact of health related lifestyles may only be apparent in the long term.
COLD AND DAMP HOUSING
The most important housing variable seemed to be the perceived temperature of the home, not damp. Damp and cold housing are often related, and bivariate analyses of our data indicate a linear association between these two variables (χ2 test, p<0.001). It is therefore possible that reported cold housing reflects the combined effects of reduced temperatures and increased humidities on health. However, it is clear that when both damp and cold housing are entered into the multivariate models, it is the cold that predominates.
Despite a considerable literature on housing dampness, cold housing has been neglected, possibly because cold and damp are so interrelated. Although some housing and health studies have examined aspects of heating these variables have been included primarily as potential sources of pollutants (for example, Spengler et al 15). One study modelled the effects of perceived cold and dampness together.1 Associations between damp/mouldy housing and number of symptoms persisted when perceived cold and other housing factors were accounted for. Another study included “bedroom unheated” and “bedroom window left open at night” in a logistic regression model along with reported damp and mould in relation to health.12 Here, reported mould and “bedroom window left open at night” both contributed to parental reports of wheeze. These variables can only be considered as proxies for perceived temperature in the home. The cold housing variable in the OHLS3 is also only a proxy, and clearly contains an economic component as some people on low incomes may inadequately heat their homes to minimise heating costs. Although our cold housing variable was based on perceived rather than objectively measured cold, it may be an equally valid predictor of ill health as the latter could be associated with feeling cold rather than the actual temperature. The possible impact of feeling as compared with being cold does not seem to have received consideration.
Cold housing may affect health in its own right. The dangers of very low temperatures, in the form of hypothermia, are well established,29 but there is little information on less extreme temperatures. Evidence suggests that cooling is associated with changes in blood pressure and viscosity.30 Rises in blood pressure in elderly people are seen at 12°C.31 Also, low indoor temperatures may damage the lungs indirectly by reducing resistance to infections secondary to colds and influenza.32 Seasonal mortality has been discussed in relation to cold housing33 and Raw and Hamilton34 place “hygrothermal conditions” as one of the most important risk factors for health related to housing. The results of our study support this.
WORRY AS A CAUSE OF ASPECTS OF POOR HEALTH
The strength of association between physical and mental health and worry about work and money was a surprise finding and needs replicating in other datasets. The importance of worry relative to other factors of current public health concern suggests that such studies are urgent. Overall “emotional health problems” seem to be more common than physical health problems in this age group (18–64 years).35 While they seem to be as disabling as physical health problems35 they receive less attention from both public health and treatment services. Evidence is mounting that such problems may cause physical illness, in addition to being important in their own right. For example, studies exist that demonstrate that emotional distress from low job control,36examinations37 and life events38 39 can create susceptibility to physical illness and increase mortality. Animal studies provide evidence to support one possible causal mechanism via the immune response.40 41 The long term impact on health of health related lifestyles provides the basis for an alternative, potentially complementary, causal hypothesis. Smoking, drinking alcohol and consuming high fat foods are all valued by the public for their ability to relieve emotional distress.42
This survey did not collect data on whether cold or damp housing worried respondents. However, if worry about work or money creates general susceptibility to physical and mental health problems, worry about damp or cold housing might do so too. Research into the impact of housing on health has concentrated on mechanisms involving external biological agents and ignored the potential importance of psychosocial pathways, even though the latter offer a plausible explanation of the impact of housing on a range of very different health problems.
LIMITATIONS OF THE STUDY
The response rate of 64% limits the extent to which definitive conclusions can be drawn. However, as no major biases among respondents were found and as the relations between variables found in the responding group should not have been affected by the response rate, our conclusions about the relative importance of the different variables should be robust.
Other limitations arise from the fact that the dataset was collected primarily for other purposes. For example, no data were collected on some potentially important variables in the analysis concerning asthma, such as pet ownership.
In addition, the survey was restricted to adults aged 18 to 64. As the elderly experience a high degree of ill health and many live in poor housing, their omission could have biased our results. However, it could be argued that their inclusion would strengthen rather than weaken associations between poor housing and health. Within our sample of working age adults, it is younger, rather than older people who tend to report the worst housing problems. Those aged 18–33 were most likely to report either damp housing as “a serious problem” or cold housing “most of the time”, while those aged 50–64 were least likely to report these problems (linear by linear association, p<0.01 for both). It is therefore important to evaluate the possible impact of housing problems in younger, as well as older, age groups.
Cross sectional surveys cannot determine the direction of relations between variables, although evidence of dose response is suggestive of causality. For some associations in our data, ill health may be a cause rather than, or as well as, an outcome of the characteristic; for example the association between ill health and employment status.43 There is evidence both that unemployment causes avoidable morbidity and premature mortality,44 and that people with psychological ill health are more likely to remain unemployed.45 Similarly, the association between ill health and retirement in our study could reflect either an effect of prior ill health on the decision to retire, or a result of retirement. There is some evidence that women who care for children or for sick family members experience more symptoms of minor illness than others.46 47 If poor housing is a cause of ill health it could be argued that those who stay at home during the day are more likely to notice and be affected by poor housing conditions such as damp or cold.
Other associations of this type are those with ill health and being an ex-smoker, and worrying about work. It is well recognised that many smokers quit because they have become ill and therefore that illness tends to be associated with being an ex-smoker more than with being a current smoker. Similarly, being unwell can prevent physical activity and exacerbate worry about pressure at work.
Studies of damp housing and health have tried to overcome the possibility of reporting bias by including both objective and subjective measures of health and housing (see review by Hyndman48), but for logistical reasons no objective measures were used in this survey. However, self reported health and housing have been shown to be associated with objectively measured health and housing respectively.10 General health perception has also been shown in longitudinal surveys to be predictive of mortality.49 In our survey, the questions on housing and health were widely separated in the questionnaire with those on health being reported before those on housing. Nevertheless, it would be advantageous to confirm our findings in datasets that include both objective and subjective measures of damp and cold housing.
POTENTIAL COST OF POOR HOUSING TO THE NHS
The cost to the NHS of treating conditions caused by poor housing has been estimated as £2 billion per year. This was based on a study that found respondents in poor housing used health services 50% more than average.50 51 Our finding that cold housing is associated with a doubling of the use of outpatient departments and GP surgeries suggests that the cost to the NHS could be even higher.
Epidemiological studies tend to focus on one particular area, and collect data accordingly. The strength of this study has been its ability to examine the relation between damp housing and health in the context of a range of environmental, demographic, psychosocial and lifestyle variables. Taking these variables into account, those most consistently related to health are: employment status, cold housing, worry about pressure at work, sex of respondent, and vigorous physical exercise. Particularly interesting are the apparent predominance of “cold housing” over damp when the two are entered into models together, and the significance of worry about work and money. The results concerning cold and damp do not mean that damp housing is unimportant. It is possible that it is the combined effects of cold, damp housing that is being demonstrated. The relative prevalence and importance of “worry” in relation to both physical and mental health in this study is notable. This raises important issues concerning the possible direct effects of emotional distress on health in general, as well as its role as a possible mediator in relations between housing and health.
The authors would like to acknowledge with thanks: the Directors of Public Health for Buckinghamshire, Berkshire, Northamptonshire and Oxfordshire, who funded the OHLS3 itself; specialists in health promotion and public health in these four counties who supported the survey through the Advisory Group; Viv Peto and other members of the Health Services Research Unit for their help in administering the survey; all the members of the public who took time to complete the OHLS3 questionnaire; and the Department of Health/Department of Environment, Transport and the Regions who funded the specific analyses reported in this paper.
Funding: this work was funded by a Department of Health/Department of Environment, Transport and the Regions initiative, together with the Directors of Public Health in Oxfordshire, Buckinghamshire, Berkshire and Northamptonshire.
Conflicts of interest: none.