Abstract

Background: Housing conditions are recognised as an important determinant of health. In the UK, interventions to improve domestic heating are in place with the expectation that they will improve health. As a component of evaluating such policies, this study assesses whether specific health outcomes are significantly associated with the extent and duration of domestic heating use, either directly or via a possible mediating effect of internal environmental conditions. Methods: Baseline data from a prospective controlled study evaluating the health effects of a publicly-funded programme of heating improvements in Scotland were used to assess associations among heating use, internal conditions, and three specific health outcomes. Results: There were significant associations (P < 0.01) between measures of heating use and the presence of environmental problems in the home, such as mould and condensation. The presence of such problems was, in turn, found to be significantly predictive of two health outcomes derived from the SF-36 (P < 0.01) and of adult wheezing (P < 0.05). The direction of significant associations was highly consistent: greater levels of heating were associated with reduced likelihood of environmental problems, and the presence of environmental problems was linked to poorer health status. Heating use was not directly associated with the health outcomes considered. Conclusions: The study findings are consistent with a conceptual model in which health may be influenced by usage patterns of domestic heating, via the mediating effect of poor internal environmental conditions. Since these findings are based on cross-sectional data, interpretation must be carried out cautiously. However, if confirmed by planned future work they have important implications for policy initiatives relating to domestic heating and fuel poverty.

It has long been accepted that the physical conditions prevailing in houses exert an influence on the health of occupants.1 Within the UK, the contribution of housing conditions to health inequalities has been recognised in landmark publications such as the Black Report2 and the Acheson Report.3 The health/housing link is explicitly recognised in the recent Wanless report on public health policy in England,4 which states ‘inequalities in health may be due to … social and environmental factors such as housing and income’. A recent (2004) review of evidence conducted by the World Health Organisation5 acknowledges ‘there is considerable evidence that housing conditions do affect health status’. While the domestic may impact on residents' health via a number of routes, research on the physiological effects of cold environment suggests that low temperatures may be implicated in respiratory conditions69 and may be a risk factor for heart disease.1012 Low indoor temperatures may be caused by a number of factors, including poor thermal characteristics of the home13 and economic constraints on residents (low disposable household income and/or high fuel costs). Such constraints are now widely recognised and enshrined in the concept of fuel poverty.14 The effects of fuel poverty on health in the United Kingdom are accepted in the Government's fuel poverty strategy.15 Evidence for an effect of fuel poverty on health within a European context has emerged from the WHO LARES study, initial findings5 from this project specifically identify fuel poverty as a factor associated with cardiovascular problems.

In Scotland, the potential public health benefit associated with reducing levels of fuel poverty is one of the main drivers behind a major ongoing policy initiative—the Scottish Executive Central Heating Programme16—aimed at providing modern central heating systems in substantial numbers of public- and private-sector homes which lack such facilities. The intended public health impacts of the Central Heating Programme include improving the health of the elderly, reducing the number of winter deaths, lowering the incidence of cold-related illness, and reducing pressure on the National Health Service.17 Publicity material for the programme referred to a single study showing ‘a link between a low indoor ambient temperature and ill-health for the elderly’, and in general relatively few studies have directly considered the associations between cold housing and health.18 However, recent work indicates that inability to maintain sufficient warmth in the home in winter is associated with poor health.19,20 A national survey of UK homes carried out by the Department of the Environment in 197821 concluded that one major source of variance in domestic temperature was the type and operation of the heating system. Although the Eurowinter study22 included the duration of heating use in the living room and bedroom as predictors of mortality, detailed measures of the extent and duration of heating use have received little attention as possible determinants of health. Thus, there remains considerable uncertainty as to the extent to which the specific factor of heating behaviour—as distinct from general environmental conditions prevailing in the home—is associated with the health status of residents.

The present study used information on the duration for which individual rooms in dwellings are heated, collected during an evaluation of the health impacts of the Central Heating Programme. The evaluation employs a longitudinal design, intended to assess the extent of causal association between the provision of central heating and altered health status. Findings from the evaluation have the potential to inform similar health-related policy initiatives in the future, supporting the Wanless Report's recommendation that interventions should be evidence-based. Although the evaluation is longitudinal, only baseline data are available at present, and were used in cross-sectional analyses with the aim of investigating whether the measures of self-reported heating are associated with specific health outcomes.

To assess whether heating use exhibits meaningful associations with health, we considered two possible conceptual models (figure 1). Firstly, we considered a direct association between the extent and/or duration of heating use and health. Secondly, we considered that heating use may be associated with undesirable environmental conditions (such as mould or condensation) in the home, and that these are, in turn, associated with health.

Figure 1

Conceptual models of postulated relationships between heating use and health

Figure 1

Conceptual models of postulated relationships between heating use and health

Investigation of the hypothesis expressed in the second conceptual model was justified by the studies reporting evidence of association between mould and dampness and poor health.23,24 Since the study reported here uses cross-sectional data, which do not permit robust identification of causal relationships, no element of causality is implied in either of the models.

Methods

Data source

The Central Heating Programme provides central heating and related thermal efficiency measures (e.g. loft insulation and pipe lagging) to all properties in the public rented sector in Scotland which lack central heating, and to properties in private ownership with the head of the household or spouse aged >60 years, and which lack central heating or possess a central heating system which is broken beyond repair. As part of the evaluation commissioned by the Scottish Executive, a sample of households scheduled to receive central heating under the programme were interviewed prior to installation. To assist in isolating any health effects associated with the provision of central heating, the study uses a quasi-experimental design in which the ‘intervention’ group (i.e. recipients of central heating) are matched to ‘comparison’ households with broadly similar social characteristics drawn from households not included in the programme. Both groups were presented with questions on a range of topics, including the length of time for which each room in the respondent's home is heated during winter. Data were collected by face-to-face interviews, conducted by trained interviewers between December 2002 and March 2004. There will be a postal follow-up survey (of both intervention and comparison group respondents) 1 year after the initial baseline interview, and then a final home interview 2 years after the initial interview.

For the purposes of this study, baseline data from both intervention and comparison households were combined on the grounds that no intervention had yet taken place. Responses were available for 3850 households with 1389 (36.1%) privately owned households, 2279 households (59.2%) in the public rented sector and 132 households (3.4%) renting from private landlords. The remaining 50 households conform to other tenure patterns, such as shared ownership or tied housing. The mean age of the respondents was 60.4 years (SD: 18.7 years) and 63.8% were females.

Measures of heating usage

For each room (to a maximum of 10) in the respondent's home (e.g. ‘kitchen’ and ‘bathroom’), interviewees were asked about the length of time, per day, for which the room is heated in winter. Separate responses were obtained for winter weekdays and for winter weekends. However, the two sets of values were almost identical and therefore only results for the weekday data are presented. Duration of heating use was recorded in time bands: 0 (no hours); 1–3 h; 4–6 h; 7–9 h; 10–12 h; 13–16 h; 17–19 h; 20–23 h; and 24 h. These were used to derive the summary quantities which form the basis of the analyses reported here (see table 1).

Table 1

Summary measures of domestic heating usage

Heating measure
 
Mean value (SD)
 
1. Proportion of rooms totally unheated i.e. duration of heating reported as 0 h per day (units of 10%) 3.6 (3.0) 
2. Proportion of rooms permanently heated i.e. duration of heating reported as 24 h per day (units of 10%) 1.2 (2.7) 
3. Proportion of rooms heated for 9 h per day or less (units of 10%) 6.6 (3.6) 
4. Duration (hours per day) of heating kitchen 5.8 (7.9) 
5. Duration (hours per day) of heating bathroom 5.2 (7.6) 
6. Duration (hours per day) of heating living room 12.8 (6.9) 
7. Duration (hours per day) of heating hall 8.2 (8.6) 
8. Duration (hours per day) of heating main bedroom 6.5 (7.7) 
9. Average duration of heating all ‘heated’ rooms i.e. excludes rooms which are totally unheated 11.4 (6.5) 
10. Average duration of heating all rooms in the dwelling 7.3 (6.0) 
Heating measure
 
Mean value (SD)
 
1. Proportion of rooms totally unheated i.e. duration of heating reported as 0 h per day (units of 10%) 3.6 (3.0) 
2. Proportion of rooms permanently heated i.e. duration of heating reported as 24 h per day (units of 10%) 1.2 (2.7) 
3. Proportion of rooms heated for 9 h per day or less (units of 10%) 6.6 (3.6) 
4. Duration (hours per day) of heating kitchen 5.8 (7.9) 
5. Duration (hours per day) of heating bathroom 5.2 (7.6) 
6. Duration (hours per day) of heating living room 12.8 (6.9) 
7. Duration (hours per day) of heating hall 8.2 (8.6) 
8. Duration (hours per day) of heating main bedroom 6.5 (7.7) 
9. Average duration of heating all ‘heated’ rooms i.e. excludes rooms which are totally unheated 11.4 (6.5) 
10. Average duration of heating all rooms in the dwelling 7.3 (6.0) 

These measures represent either the extent of heating (e.g. the proportion of rooms in the dwelling which are totally unheated) or the duration of heating (e.g. the length of time for which the kitchen is heated per day). Measures of the extent of heating were expressed in units of 10%—for example, an absolute value of 25% was represented analytically as 2.5 units. This was done to yield parameter estimates with a useful practical interpretation, since single points of the total percentage of rooms heated in domestic dwellings have little meaning. In deriving measures of duration, the mid-point of each time band was used. For example, a response of 4–6 h was treated as an absolute value of 5 h. For individual rooms, duration values were calculated only for the five ‘core’ room types which existed in almost all homes (kitchen, bathroom, main living room, hall, and main bedroom). Of the 3850 properties in the sample, 653 (17.0%) possessed no rooms other than these.

Health outcomes

Three health outcomes were selected, on the grounds that they might plausibly be susceptible to influence by changes in heating or housing conditions: the reported presence (YES/NO) of wheezing in interview respondents in the past 12 months, the respondent's score on the SF-36 General Health scale,25 and score on the SF-36 Mental Health scale.

Presence of internal environmental problems

The interview recorded information about the presence of condensation on windows or walls; damp smell; mould growth on carpets/curtains/furniture; mould growth on walls, ceilings or floors; and mould or rot on window frames. The presence of any of these problems, in at least one of the five core rooms, was represented in the analyses by a binary indicator.

Statistical analysis (i): techniques

Binary logistic regression was used for analyses involving the first health measure (reported wheezing by respondents). However, although the two SF-36 derived outcomes are quasi-continuous quantities, their irregular distributions precluded modelling via conventional multiple regression. Consequently, scores for these two scales were reduced to a four-way ordered categorical scheme as follows: Level 1 (original scores from 0 to 25); Level 2 (>25 to 50); Level 3 (>50 to 75); and Level 4 (>75). This approach, while involving a loss of information, permitted the use of logistic regression with a polychotomous response model. The assumption of proportional odds was not supported for these data, so the categorised SF-36 measures were investigated via generalised logit models. In such models, levels of the outcome are contrasted individually with a preselected reference level and a separate set of parameter estimates produced for each contrast. For both measures, the lowest band of scores (Level 1, i.e. the band representing poorest health) was selected as the reference level. All analyses were carried out using the LOGISTIC procedure in SAS software, Version 8.02.

Statistical analysis (ii): selection of covariates

The models relating health outcomes to heating usage were adjusted for potential confounding factors which might reasonably be expected to influence any health/heating relationships: the respondent's age (years); gender (0 = male, 1 = female); exposure to smoking (1 = no exposure, 2 = no smoking by the respondent but exposure to passive smoking, 3 = the respondent smokes but no exposure to passive smoking, 4 = the respondent smokes and is also exposed to passive smoking); estimated annual household income (£1000s); the economic activity status of the household [four way classification: 1 = working, 2 = retired, 3 = sick/disabled, 4 = not working (including unemployed and in full-time education)]; and the presence of household pets (group of five binary indicators, denoting respectively the presence/absence of dogs, cats, birds, other furry pets, and other pets).

Statistical analysis (iii): analytical sequence

The analysis was carried out in three stages, as shown in figure 2.

Figure 2

Summary of the analytical sequence

Figure 2

Summary of the analytical sequence

Results

The models assessing the predictive effect of heating use on the health outcomes (Model Series 1, figure 2) yielded a total of 30 parameter estimates (from 10 heating measures, each predicting 3 outcomes). Of these 30 parameters, only 3 were significant at P ≤ 0.01. Parameter estimates for these effects are not reported here because, of these 3 significant effects, only 1 (out of the 30 parameters estimated) remained significant after the representation of environmental problems (EPs) was introduced in Model Series 2. In interpreting the results a significance level of 1% was applied throughout the study—in preference to the conventional 5%—to partially offset the risk of Type I errors associated with the large number of parameters estimated.

Associations between the presence of EPs (predictor) and health (outcome), estimated in Model Series 2, are shown in table 2 (for adult wheezing) and table 3 (for SF-36 general health and SF-36 mental health).

Table 2

Parameter estimates for effect of EPs on adult wheezing (coding: 0 = NO, 1 = YES), when heating usage measures are also represented

Heating measure included
 
Estimate, effect of EPs: odds ratio (P)
 
% Unheated roomsa 1.48 (0.01) 
% Permanently heated rooms 1.49 (0.01) 
% Rooms heated for 9 h or less 1.47 (0.01) 
Heating duration: kitchen (h) 1.43 (0.02) 
Heating duration: bathroom (h) 1.52 (<0.01) 
Heating duration: living room (h) 1.39 (0.03) 
Heating duration: hall (h) 1.42 (0.02) 
Heating duration: main bedroom (h) 1.45 (0.02) 
Average duration excluding unheated rooms (h) 1.46 (0.01) 
Average duration, all rooms (h) 1.51 (<0.01) 
Heating measure included
 
Estimate, effect of EPs: odds ratio (P)
 
% Unheated roomsa 1.48 (0.01) 
% Permanently heated rooms 1.49 (0.01) 
% Rooms heated for 9 h or less 1.47 (0.01) 
Heating duration: kitchen (h) 1.43 (0.02) 
Heating duration: bathroom (h) 1.52 (<0.01) 
Heating duration: living room (h) 1.39 (0.03) 
Heating duration: hall (h) 1.42 (0.02) 
Heating duration: main bedroom (h) 1.45 (0.02) 
Average duration excluding unheated rooms (h) 1.46 (0.01) 
Average duration, all rooms (h) 1.51 (<0.01) 

Estimates are adjusted for the effects of the covariates listed in the statistical analysis (ii)

a: All measures which involve percentages of rooms are expressed in units of 10%, rather than individual percentage points

Table 3

Parameter estimates for effect of EPs on SF-36 general health and on SF-36 mental health when heating usage measures are also represented

Heating measure
 
General health estimate, effect of EPs: odds ratio (P)
 
Mental health estimate, effect of EPs: odds ratio (P)
 
% Unheated roomsa P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.52 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.86 (0.29) L2: 0.96 (0.85) 
% Permanently heated rooms P < 0.01 P < 0.01 
 L4: 0.48 (< 0.01) L4: 0.49 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.86 (0.28) L2: 0.94 (0.75) 
% Rooms heated for ≤9 h P < 0.01 P < 0.01 
 L4: 0.49 (< 0.01) L4: 0.50 (< 0.01) 
 L3: 0.63 (< 0.01) L3: 0.58 (< 0.01) 
 L2: 0.86 (0.31) L2: 0.96 (0.85) 
Heating duration: kitchen P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.52 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.60 (< 0.01) 
 L2: 0.87 (0.33) L2: 0.96 (0.82) 
Heating duration: bathroom P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.53 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.88 (0.37) L2: 0.98 (0.90) 
Heating duration: living room P < 0.01 P < 0.01 
 L4: 0.50 (< 0.01) L4: 0.51 (< 0.01) 
 L3: 0.64 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.87 (0.34) L2: 0.96 (0.84) 
Heating duration: hall P < 0.01 P < 0.01 
 L4: 0.50 (< 0.01) L4: 0.51 (< 0.01) 
 L3: 0.64 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.86 (0.28) L2: 0.97 (0.87) 
Heating duration: main bedroom P < 0.01 P < 0.01 
 L4: 0.51 (< 0.01) L4: 0.56 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.65 (0.02) 
 L2: 0.89 (0.42) L2: 1.07 (0.73) 
Average duration (excluding unheated rooms) P < 0.01 P < 0.01 
 L4: 0.48 (< 0.01) L4: 0.49 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.86 (0.31) L2: 0.93 (0.72) 
Average duration (all rooms) P < 0.01 P < 0.01 
 L4: 0.49 (< 0.01) L4: 0.50 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.85 (0.26) L2: 0.96 (0.82) 
Heating measure
 
General health estimate, effect of EPs: odds ratio (P)
 
Mental health estimate, effect of EPs: odds ratio (P)
 
% Unheated roomsa P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.52 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.86 (0.29) L2: 0.96 (0.85) 
% Permanently heated rooms P < 0.01 P < 0.01 
 L4: 0.48 (< 0.01) L4: 0.49 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.86 (0.28) L2: 0.94 (0.75) 
% Rooms heated for ≤9 h P < 0.01 P < 0.01 
 L4: 0.49 (< 0.01) L4: 0.50 (< 0.01) 
 L3: 0.63 (< 0.01) L3: 0.58 (< 0.01) 
 L2: 0.86 (0.31) L2: 0.96 (0.85) 
Heating duration: kitchen P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.52 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.60 (< 0.01) 
 L2: 0.87 (0.33) L2: 0.96 (0.82) 
Heating duration: bathroom P < 0.01 P < 0.01 
 L4: 0.52 (< 0.01) L4: 0.53 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.88 (0.37) L2: 0.98 (0.90) 
Heating duration: living room P < 0.01 P < 0.01 
 L4: 0.50 (< 0.01) L4: 0.51 (< 0.01) 
 L3: 0.64 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.87 (0.34) L2: 0.96 (0.84) 
Heating duration: hall P < 0.01 P < 0.01 
 L4: 0.50 (< 0.01) L4: 0.51 (< 0.01) 
 L3: 0.64 (< 0.01) L3: 0.59 (< 0.01) 
 L2: 0.86 (0.28) L2: 0.97 (0.87) 
Heating duration: main bedroom P < 0.01 P < 0.01 
 L4: 0.51 (< 0.01) L4: 0.56 (< 0.01) 
 L3: 0.65 (< 0.01) L3: 0.65 (0.02) 
 L2: 0.89 (0.42) L2: 1.07 (0.73) 
Average duration (excluding unheated rooms) P < 0.01 P < 0.01 
 L4: 0.48 (< 0.01) L4: 0.49 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.86 (0.31) L2: 0.93 (0.72) 
Average duration (all rooms) P < 0.01 P < 0.01 
 L4: 0.49 (< 0.01) L4: 0.50 (< 0.01) 
 L3: 0.62 (< 0.01) L3: 0.57 (< 0.01) 
 L2: 0.85 (0.26) L2: 0.96 (0.82) 

Estimates are adjusted for the effects of the covariates listed in the statistical analysis (ii)

a: All measures which involve percentages of rooms are expressed in units of 10%, rather than individual percentage points

For both tables, each row shows the association between the presence of EPs and the health outcome, after controlling for a specific heating usage measure. As previously explained, table 3 shows results from generalised logit models, characterised by the production of a separate set of parameter estimates for each non-reference level of the outcome. In table 3, reporting of the overall significance level is followed by three individual parameter estimates, one for each non-reference outcome level. Outcome levels are identified by the prefix ‘L’, e.g. ‘L4’ indicates Level 4. It is stressed that Level 4 represents ‘best’ health.

An example may aid the interpretation of table 3. For the general health outcome, the effect of the presence of EPs when the first heating usage measure—proportion of unheated rooms—is also included as a predictor is shown in the topmost data cell in the centre column. The top line in the cell indicates the overall significance of the EPs indicator. The next line L4: 0.52 (< 0.01) indicates that the presence of EPs in the home is significantly associated with a reduced likelihood of the respondent being in the highest category of general health, relative to being in the lowest level. The remaining lines in the cell show the same information for other non-reference levels of the outcome.

Results for Model Series 3 are given in table 4. This shows the estimated effect of each heating measure (predictor) on the presence of EPs. For those heating measures which relate to the entire dwelling (e.g. the proportion of unheated rooms), the outcome is the presence of at least one EP in any of the five core rooms of the home. For heating measures which relate to a specific room type (e.g. the duration of heating use in the kitchen), the outcome is the presence of at least one EP in the corresponding room. Ratios >1.0 denote greater probability of EPs being present.

Table 4

Parameter estimates for effect of heating usage measures on the presence of EPs in the home

Heating usage measure
 
Estimate, effect of heating: odds ratio/95% CI/P
 
Proportion of unheated rooms (W)(units of 10%) 1.17 (1.14–1.21)P < 0.01 
Proportion of permanently heated rooms (W)(units of 10%) 0.93 (0.90–0.97)P < 0.01 
Proportion of rooms heated for 9 h/day or less (W)(units of 10%) 1.08 (1.05–1.11)P < 0.01 
Duration of heating: kitchen (h) 0.96 (0.95–0.97)P < 0.01 
Duration of heating: bathroom (h) 0.93 (0.92–0.95)P < 0.01 
Duration of heating: main living room (h) 0.99 (0.98–1.01)P = 0.31 
Duration of heating: hall (h) 0.97 (0.96–0.99)P < 0.01 
Duration of heating: main bedroom (h) 0.96 (0.94–0.97)P < 0.01 
Average duration of heating: heated rooms only (W) (h) 0.99 (0.97–1.00)P = 0.05 
Average duration of heating: all rooms (W) (h) 0.94 (0.93–0.96)P < 0.01 
Heating usage measure
 
Estimate, effect of heating: odds ratio/95% CI/P
 
Proportion of unheated rooms (W)(units of 10%) 1.17 (1.14–1.21)P < 0.01 
Proportion of permanently heated rooms (W)(units of 10%) 0.93 (0.90–0.97)P < 0.01 
Proportion of rooms heated for 9 h/day or less (W)(units of 10%) 1.08 (1.05–1.11)P < 0.01 
Duration of heating: kitchen (h) 0.96 (0.95–0.97)P < 0.01 
Duration of heating: bathroom (h) 0.93 (0.92–0.95)P < 0.01 
Duration of heating: main living room (h) 0.99 (0.98–1.01)P = 0.31 
Duration of heating: hall (h) 0.97 (0.96–0.99)P < 0.01 
Duration of heating: main bedroom (h) 0.96 (0.94–0.97)P < 0.01 
Average duration of heating: heated rooms only (W) (h) 0.99 (0.97–1.00)P = 0.05 
Average duration of heating: all rooms (W) (h) 0.94 (0.93–0.96)P < 0.01 

Whole-dwelling heating measures are identified by ‘W’. Estimates are adjusted for the effects of the covariates listed in figure 2

All measures which involve percentages of rooms are expressed in units of 10%, rather than individual percentage points

Discussion

Although housing and the wider domestic environment are now generally accepted as important determinants of health, this study is among the first to consider the specific relationship of health with the extent and duration of domestic heating use. While the study is subject to certain limitations (discussed below), it is believed that the measures of heating use that are described here provide a potentially useful representation of how completely, and for how long, a domestic property is heated. In view of the cost, complexity and general difficulty (both practical and ethical) of carrying out detailed recording of heating operation, room temperature, and room use, it may be that measures of this type will be of use in a range of investigations in the field of housing conditions and health.

As stated earlier, a significance level of 1% has been applied when considering results, in preference to the conventional 5%. However, given the large number of parameters estimated in the study, this strategy is in itself inadequate to guard against Type I errors. Recognising this, the interpretation and discussion which follow are based on general patterns of results rather than estimates for specific individual variables.

Results derived in Models Series 1 and 2 suggest an almost complete lack of direct association between the heating usage measures and the specific health outcomes selected. Of the 30 parameters estimated in Series 1, only 3 were significant, and only 1 remained so when control for EPs was introduced. On this basis, our conceptual Model A—which postulates a direct link between the extent and/or duration of heating use and health—is not supported.

However, tables 2 and 3 indicate significant associations between the presence of EPs and the health outcomes. For the two SF-36 outcomes (table 3), this association is consistently highly significant (P < 0.01). For the adult wheezing outcome (table 2), the result is less clear, but still persuasive. For all three outcomes the direction of relationships is as expected: the presence of EPs is associated with a greater likelihood of reported wheezing, and with a reduced likelihood of higher (i.e. better) scores on the SF-36 scales. It would have been desirable—in view of the body of research identifying cold as a risk factor for heart disease1012—to supplement the three health outcomes used with a further one relating specifically to cardiovascular health. However, no reliable indicator of this class of disease was included in the data collected.

The finding that internal EPs, such as mould, are associated with poor health is not new.23,24 However, confirmation of this association acquires additional meaning when findings are viewed in the context of the conceptual Model B (heating use ←→ environmental conditions ←→ health) because the results are fully consistent with the second (rightmost) relation in this model. Table 4 indicates that, for the most part, measures of heating usage are significant predictors (P ≤ 0.01) of the presence of EPs and that the association is overwhelmingly in the expected direction: ‘more heat’ (i.e. greater extent or duration) is associated with a reduced probability of EPs being reported.

That heating usage is significantly associated, in the expected direction, with the presence of conditions such as mould and condensation invites two inferences. The first is that the measures themselves may be held to possess a degree of validity. It might intuitively be anticipated that less heating would be linked with an increased prevalence of environmental manifestations, and this indeed appears to be the case. More rigorous formal validation of the proposed measures is required before they can be considered as reliable tools for more general research use. A second inference may be drawn when the associations between heating use and EPs are regarded in conjunction with the associations between EPs and the health outcomes considered. Taken together, the two sets of findings are consistent with the conceptual Model B, and hint at a unified framework of relations among the three factors involved, i.e. heating use, EPs, and health. While many previous studies have suggested links between housing conditions (including cold and damp) and health, little published work has postulated an association between health and the specific use which is made of the household's domestic heating apparatus. The work described here suggests that such an association may in fact exist, via the mediating influence of domestic environmental conditions. It is accepted that some of the estimated effect sizes are modest: indeed, one of the strengths of this study is its large sample size, which provides the ability to detect small effects. While the practical implications of some of the estimated effects are open to debate, it should be remembered that many of the parameters reported represent small unit effects (e.g. the effect of a single hour's increased heating duration), and the practical result of larger increases will be appreciably greater. This may be illustrated briefly using an example. The bottom line of table 4 indicates that a 1 h increase in average heating duration is associated with an odds ratio of 0.94 for the presence of EPs. However, for a notional increase in average heating duration from 0 to 7.3 h (the mean value for the data), the corresponding odds ratio is 0.63.

Given that this study has produced some evidence in support of the indirect relationships involved in Model B, the almost complete absence of direct association between heating and health—as postulated in Model A—is surprising. However, analysis of unadjusted bivariate associations between the heating measures and the health outcomes (not reported here) indicates that some degree of direct association may in fact exist. Of the 20 rank correlation coefficients (Spearman's) estimating the association between the heating measures and the SF-36 outcomes, 9 are significant at P ≤ 0.01. However, the associations are modest (maximum rs = 0.12). It may be that a tenuous pattern of direct association is obscured in the regression approach reported here due to the distributional characteristics of the heating measures (which exhibit marked irregularity), but becomes manifest when simple monotonic associations are considered.

The limitations of this study should be noted. Firstly, because results are derived from cross-sectional data, causation cannot be assumed: we have not conclusively shown that restricted use of domestic heating leads directly to EPs, and that these in turn lead to poorer health. A variety of interpretations are plausible. For example, it may be argued that, even if limited heating use does cause EPs, other factors such as the thermal properties and ventilation characteristics of the building are likely to be involved in the development of EPs. Furthermore, it seems probable that the relationship between heating use and EPs is not unidirectional. Restricted heating use might cause the problems, but presence of the problems might also generate increased heating use (in an attempt to eradicate mould, condensation etc.). Secondly, the representation in the models of EPs is itself questionable in that self-reports of residential mould have been shown to be subject to systematic reporting bias when compared with objective environmental measurements.26 Similar concerns might be expressed about the possible bias in the self-reporting of health status, but these are largely allayed by the acknowledged stature of the SF-36, which is one of the most extensively used, tested, and validated instruments of its kind, featuring in many hundreds of studies.27 Thirdly, the models include only a limited number of the most obvious potential confounders and the possibility that these results are influenced by one or more factors not represented in the models cannot be discounted. Fourthly, because data were collected over a 15-month period, the possibility of recall bias cannot be discounted. Although the heating measures are based on reported use of heating in winter, respondents interviewed at other times of the year may vary in the accuracy of their recollections of winter heating behaviour. To investigate this possibility, the models of Series 1 were re-run with an added term representing the season of the year. It was found that the original parameter estimates for associations between heating and health were almost unchanged, suggesting that recall bias may be minimal. Finally, the heating usage measures describe only the extent and duration of heating use, without any consideration of the temperature levels achieved, or the extent to which rooms are actually ‘lived in’. Thus, they do not fully gauge the extent to which homes (and their occupants) are exposed to adequate levels of heating, which is really the factor of fundamental interest.

Despite limitations, this study potentially extends the body of knowledge relating to housing and health by introducing a largely unexplored factor: the effect of heating use. As Howden-Chapman comments ‘surveys to explore new associations … between housing and health are important’.28 This point is echoed in the Wanless Report, which states that ‘the major constraint to further progress on the implementation of public health interventions is the weakness of the evidence base regarding their effectiveness’. The findings reported here indicate that the possible links between heating use and health are worthy of further exploration via intervention studies, to provide the highest possible quality of evidence.

This study was funded by the Scottish Executive. Mark Petticrew is funded by the Chief Scientist's Office of the Scottish Executive Health Department. Stephen Platt and Richard Mitchell are funded by the Chief Scientist's Office of the Scottish Executive Health Department. The views expressed in this paper are those of the authors, and do not necessarily reflect those of the funders.

Key points

  • This study sought to determine whether the extent and duration of domestic heating use were associated with the residents' health.

  • Self-reported heating use was found not to be directly associated with a number of health outcomes examined.

  • However, heating use was significantly associated with the presence of EPs, such as mould and condensation.

  • Such problems were in turn significantly predictive of two health outcomes derived from the SF-36, and of wheezing among adults.

  • Evidence of indirect links between heating use and health further enhances the imperative to reduce fuel poverty and improve domestic heating.

Funding: Chief Scientist Office, The Scottish Executive.

References

1
Lowry S. Housing and health. London: British Medical Journal,
1991
.
2
Black D. Inequalities in health: report of a research working group. London: DHSS,
1980
.
3
Acheson D. Independent inquiry into inequalities in health report. London: The Stationery Office,
1998
.
4
Wanless D. Securing good health for the whole population: Final Report. London: HMSO,
2004
.
5
Bonnefoy XR, Annesi-Maesano I, Aznar LM, et al. Review of evidence on housing and health: background document for the Fourth Ministerial Conference on Environment and Health, Budapest 2004. Apr. Report No.: Eur/04/5046267/BD/1. Geneva: World Health Organisation,
2004
.
6
Strachan DP, Sanders CH. Damp housing and childhood asthma: respiratory effects of indoor air temperature and relative humidity.
J Epidemiol Community Health
 
1989
;
43
:
7
–14.
7
Rasmussen F, Borchsenius L, Winslow J, Ostergard E. Associations between housing conditions, smoking habits and ventilatory function in men with clean jobs.
Scand J Respir Dis
 
1978
;
59
:
264
–76.
8
Ross A, Collins M, Sanders C. Upper respiratory tract infection in children, domestic temperatures, and humidity.
J Epidemiol Community Health
 
1990
;
44
:
142
–6.
9
Hyndman SJ. Housing dampness and health amongst British Bengalis in east London.
Soc Sci Med
 
1990
;
30
:
131
–41.
10
Tochihara Y, Ohnaka T, Nagai Y, et al. Physiological responses and thermal sensations of the elderly in cold and hot environments.
J Therm Biol
 
1993
;
18
:
355
–61.
11
Woodhouse PR, Khaw KT, Plummer M. Seasonal variation of blood pressure and its relationship to ambient temperature in an elderly population.
J Hypertens
 
1993
;
11
:
1267
–74.
12
Keatinge W, Coleshaw S, Cotter F. Increases in platelet and red cell counts, blood viscosity and arterial pressure during mild surface cooling: factors in mortality from coronary and cerebral thrombosis in winter.
BMJ
 
1984
;
291
:
1405
–8.
13
Howieson SG. Housing: raising the Scottish standard. Glasgow: Technical Services Agency Ltd. (undated).
14
Boardman B. Fuel poverty: from cold homes to affordable warmth. London: Belhaven Press,
1991
.
15
Inter-Ministerial Group on Fuel Poverty. The UK fuel poverty strategy. London: Department of Trade and Industry,
2001
.
16
The Scottish Executive. The Scottish Executive's Central Heating Programme and the Warm Deal: Annual Report 2001/02. Edinburgh: The Scottish Executive,
2002
.
17
The Scottish Executive. Free central heating for pensioners and families (Scottish Executive Press Release SE2497/2000). Edinburgh: The Scottish Executive,
2000
.
18
Hopton J, Platt S, Macleod L. Housing conditions and health consequences. In: Gill P, de Wildt G, editors. Housing and health: the role of primary care. Oxford: Radcliffe Medical Press Ltd.,
2003
: 17–45.
19
Evans J, Hyndman S, Stewart-Brown S, et al. An epidemiological study of the relative importance of damp housing in relation to adult health.
J Epidemiol Community Health
 
2000
;
54
:
677
–86.
20
Gemmell I. Indoor heating, house conditions, and health.
J Epidemiol Community Health
 
2001
;
55
:
928
–9.
21
Hunt DRG, Gidman MI. A national field survey of house temperatures.
Building and environment
 
1982
;
17
:
107
–24.
22
Keatinge WR, Donaldson GC, Bucher K, et al. Cold exposure and winter mortality from ischaemic heart disease, cerebrovascular disease, respiratory disease, and all causes in warm and cold regions of Europe. Eurowinter Group.
Lancet
 
1997
;
349
:
1341
–6.
23
Platt SD, Martin CJ, Hunt SM, Lewis CW. Damp housing, mould growth and symptomatic health state.
BMJ
 
1989
;
298
:
1673
–8.
24
Martin CJ, Platt SD, Hunt SM. Housing conditions and ill health.
BMJ
 
1987
;
294
:
1125
–7.
25
Ware JE, Snow KK, Kosinski M. SF-36® Health survey: manual and interpretation guide. Lincoln, RI: QualityMetric Incorporated,
1993
.
26
Dales RE, Miller D, McMullen E. Indoor air quality and health: validity and determinants of reported home dampness and moulds.
Int J Epidemiol
 
1997
;
26
:
120
–5.
27
Ware JE, Gandek B. Overview of the SF-36 Health Survey and the International Quality of Life Assessment (IQOL) Project.
J Clin Epidemiol
 
1998
;
51
:
903
–12.
28
Howden-Chapman P. Housing and inequalities in health [editorial].
J Epidemiol Community Health
 
2002
;
56
:
645
–6.

Author notes

*
Work described in this paper has previously been presented orally at the ENHR International Housing Conference (Cambridge; 2–6 July 2004) and at the WHO Second International Housing and Health Symposium (Vilnius; 29 September–1 October 2004).

Supplementary data

Comments

0 Comments