Background A randomized study of the effect on people's health of improving their housing is underway in Torbay. The link between poor health, particularly respiratory health, and poor housing conditions has been recognized for a long time, but there have been few intervention studies to demonstrate that improving housing can improve health. In 1994, South and West Devon Health Authority set up a community development project in a deprived area of Torbay, in response to the concerns of local primary health-care workers. A community development worker helped local residents survey their homes for dampness and record their respiratory symptoms. The survey reported high levels of condensation/dampness and respiratory illness and the Council agreed to direct the majority of their housing improvement funds to the estate over the next 3 years. The Health Authority, University of Plymouth and Torbay Council were successful in obtaining funding to evaluate the housing improvements from the NHS R & D programme.
Participants and methods Of 119 houses eligible for the study, 50 were chosen at random and improved in the first year. The rest were improved the following year. Questionnaires screening for health problems were sent to all 580 residents and baseline surveys of the indoor environment were also carried out. More detailed health surveys were completed by community nurses visiting residents in their homes. All adults were asked to complete SF-36 and GHQ 12 questionnaires, as well as disease-specific questionnaires if appropriate.
Progress All houses in the study have now been improved, including insulation, double-glazing, re-roofing, heating, ventilation and electrical rewiring. Follow-up surveys are underway.
Links between poor housing and ill health have been described for many years, but few intervention studies have been attempted and the evidence supporting a causal link remains poor. We report here on the development of a study from its origins in local concerns about the poor health and service demands of the residents in an estate of council owned houses in Torquay. To our knowledge, only one study to date has randomly allocated housing.1 This is the first study to upgrade current accommodation at random.
Association between poor housing and poor health
Poor housing conditions have been linked to various aspects of poor health.2,3
Cold houses, costing a lot to heat, have been blamed for contributing to the excess winter deaths in the elderly, largely from cardiovascular and respiratory illness, seen more markedly in this country than elsewhere in Europe.4
Evidence for a link between housing conditions and asthma comes from surveys which have consistently reported an association between damp, mouldy housing and respiratory symptoms in adults3 and children.5–7
Poor mental health, such as stress and alcohol/drug problems, has also been linked to aspects of poor housing conditions such as poor design, infestation and noise.2
Despite these associations, few intervention studies have been undertaken. A recent systematic review8 searched the clinical, social science and grey literature extensively and identified studies as far back as 1887, but only identified 18 primary intervention studies. While several of the studies demonstrated health gain from housing improvements, methodological problems, such as small study size and lack of control groups, meant that the results were not generalizable. The authors concluded that large-scale studies of housing interventions, with good methodology, were required.
Development of the Watcombe project
Watcombe is an area to the north of Torquay in south Devon. In 1994, a local GP, whose practice serves Watcombe and the neighbouring areas of Barton and Hele, expressed concern about the high call-out and visiting rate to these areas from his practice. At that time, 50% of the practice population were on low incomes and receiving benefits. In addition, 45% of the children under 5 years were living in a household where the breadwinner was unemployed and 30% of children under 5 years were living in single parent households. The Jarman Index for the area was 22.7 compared with the Devon average of 12.75 and the out-of-hours visiting rate was the highest in Torquay, 15% above the average for the town.9
The locality manager for Torbay for South and West Devon Health Authority organized a discussion of these issues in a local pub with health professionals, local voluntary organizations, police, probation, teachers, social services and other Local Authority officers. As a result of this meeting, the Health Authority funded a 3 year project in the area: the Barton, Watcombe and Hele Health Gain Initiative.
The community development worker funded by the project became aware of the Watcombe residents' concerns about their housing. Many of the council-owned houses on the Watcombe estate were reported to be damp or had condensation problems, and many of the children reported respiratory illnesses. The community worker supported tenants in undertaking a questionnaire-based survey of 96 council houses and their occupants in 1997. The questionnaire included a pictorial guide to condensation, dampness and mould drawn by one of the tenants. Dampness in the house was reported by 64% of respondents. In reply to the question on health problems, 84/96 (87%) replied, with 47/84 (56%) reporting a health problem. The commonest health problem reported was asthma with 39 households reporting at least one sufferer, followed by arthritis, with 10 households reporting at least one sufferer. Of those houses reporting dampness/condensation problems, 60% also reported asthma or breathing problems, while only 10% of those without dampness/condensation reported respiratory illnesses.10
In December 1997, an interagency partnership steering group was formed (Torbay Healthy Housing Group) to put forward a research proposal to address the above problems. Subsequently, the proposal was put to the Housing Sub-Committee in May 1998 by the Head of Housing for Torbay Council and an agreement was gained from the members that the majority of the Council's housing maintenance budget could be directed at the Watcombe estate to address the identified dampness/condensation problems. The funds amounted to £600 000 over the following 3 years or however long it took to complete the relevant improvements. Both the Council and the Health Authority wished to evaluate the benefit to the tenants in health terms. An expanded project group was then successful in obtaining funding from the regional NHS R & D Committee to evaluate the improvements.
This paper reports on the design of the evaluation, the choice of outcome measures for the houses and their occupants and associated problems and issues. Future papers will report on the health and environmental outcomes before and after the improvements.
The funds allocated by Torbay Council were sufficient to improve around 50 of the houses in Watcombe each year. Following consultation with the local residents and the housing department, it was agreed that houses could be randomized to receive their improvements either in the first or second year of the project.
The study design is thus one of randomization to waiting list. All occupants of houses eligible for improvements were asked to take part in the evaluation study. A baseline survey of the indoor environment in each house was carried out at the same time as a brief self-completed questionnaire which included health of the occupants, number of people in the house, pets and smokers. A more detailed health survey was carried out by community nurses.
Randomization was carried out at a public meeting on the estate. The eligible houses were stratified by street and those for improvement in the first year were drawn out of a bucket by a local councillor at a public meeting.
The first round of housing improvements took place after all baseline surveys were completed. The same surveys were repeated on all houses and all occupants a year later, after the first and before the second round of improvements. The final surveys will be conducted after all the improvements have been completed. The surveys include the outcome measures listed below. Changes in these measures will be compared in improved and unimproved houses after year 1 and after each phase of improvement.
This design permits intergroup comparison of houses (environmental parameters) and their occupants (health and economic parameters) after the first round of improvements when only the 50 houses in Phase I have been improved. Before and after comparisons on the same parameters can also be done for both Phase I and Phase II.
A combination of general and disease-specific outcome measures has been chosen, in order to capture both change in health from as many people as possible and to reflect the commonest disease complaints identified in the residents' original survey. Change in general health is being measured using the SF3611 and GHQ12 questionnaires. Change in respiratory health is being measured using a symptom-based outcome measure,13 lung function tests and symptom diaries for children. The presence of ischaemic heart disease is being assessed using the Rose Angina Questionnaire14 and change in morbidity from arthritis/rheumatism is being assessed using the AIMS/Rand questionnaire.15 Use of prescribed and over the counter medication for the three specific disease categories is also recorded and will be used to estimate severity of disease.
The environmental parameters have been chosen both to give an overall picture of the indoor environment of each house and to assess the levels of specific parameters known to affect the chosen diseases. Changes to the following parameters will be assessed: temperature, humidity, dampness, large (3–7 μm) and small (0.3–3 μm) particles, house dust mite in mattresses, moulds and CO2.
The effect of the housing improvements will be summarized by measuring the change in energy efficiency of the houses using the SAP,16 a widely used energy rating. The change in heating costs for occupants will be assessed and the change in costs for the NHS will be assessed through change in use of prescribed drugs and primary and secondary care contacts.
Organization and process
An interagency group, the Torbay Healthy Housing Group, was formed as a steering group for the project, with representation from the Health Authority, Torbay Primary Care Group, the Research and Development Support Unit for South and West Devon, Torbay Voluntary Council, local general practices, South Devon Healthcare Trust, Watcombe residents and Torbay Council. It was chaired by the Head of Housing for Torbay Council. The community development worker from the Barton, Watcombe and Hele Health Gain Initiative was appointed as the researcher for the evaluation study. The steering group now meets regularly and has established a subgroup structure to oversee the various aspects of the project.
The researcher produces regular newsletters for the local residents to keep them up-to-date and is available to discuss progress with all those involved in the study at regular times.
The Housing Department surveyed all houses on the estate before randomization to estimate the amount of work needed. After randomization, detailed assessments of the 50 houses in the first phase of improvements were undertaken to determine exactly what work would be done in each house in order to bring them all up to a common standard.
All properties lacked cavity wall insulation and the roof space insulation varied, leading to cold bridging at the first floor ceiling level, promoting the formation of condensation mould growth. Some properties had partial central heating, some had none. Bathrooms and kitchens had virtually no ventilation, making it impossible to remove excessive moist air. Electrical upgrading was also required, although it was recognized that this work did not affect the energy efficiency of the house.
All occupants of improved houses received a leaflet explaining the correct use of the new heating and ventilation systems installed in their home, in order to help them make the most efficient use of the equipment.
After all the improvements have been completed, occupants will also receive individual guidance on the effect of lifestyle and domestic activities on the indoor environment. Such advice has been requested by the occupants, but has been delayed until completion of all data collection in order not to bias the main study results.
Following the local residents' survey and before the study started, there were 142 houses owned by Torbay Council on the Watcombe estate (this figure was variable owing to the rights to buy). As the study was developing, 15 houses were bought by their occupants, leaving 127 eligible for the Council's housing improvements. Progress through the study is shown in Fig. 1.
Randomization has produced a reasonably balanced distribution of relevant characteristics between the two phases of the study. There were no statistically significant differences between the two groups (Table 1).
Table 1. Comparison between Phase I and Phase II households (119 houses were randomized)
The 50 houses in Phase I were improved during the second half of 1999. The work included installation of full gas-fired central heating, upgrading of partial heating and renewal of undersized boilers (at a total cost of £120 000). In order to obtain feedback on the use of ventilation units, data gathering extract fans were installed, controlled by the ambient temperature and humidity. The cost of the fans and the electrical upgrade was £50 000. Roofs were fitted with breathable roofing felt and had 50 mm insulation installed between the rafters and the wallplate/ceiling level (£180 000). All cavities were insulated with rockwool fibres blown in through holes drilled into the outer skin. The over-ceiling insulation was topped up to 200 mm thickness with glass fibre quilting (£11 000). Front and back doors and French windows were replaced with PVCu doors to achieve a virtually draught-free property (£27 000).
The average cost of the improvements for each house in the first year was £7760.
The first round of follow-up surveys were completed in May 2000, with response rates comparable to the baseline surveys the previous year (Table 2).
Table 2. Response rates
The second round of improvements started in July 2000 and was completed in December 2000. The third round of follow-up surveys will complete in May/June 2001.
This study only became possible because of the community development work undertaken by the Health Authority, with the co-operation of the Council, at the request of the local residents and health workers. Because of this background, the project has so far met with enthusiastic co-operation from local people, who have been asked to answer lengthy questionnaires on their health and to have their homes assessed by an environmental engineer. It would not be surprising if the response rate to these intrusive surveys dwindled rapidly after the first year, along with general interest in the project, but so far that has not happened. It is worth identifying the specific local circumstances and issues of study design which have contributed to the success of the study so far, in order to develop a more generalizable methodology for future studies.
Issues of study design
The first challenge to the evaluation team was to superimpose a rigorous design on the renovation work of the council. A straightforward before and after assessment would not advance our knowledge because of its methodological drawbacks.8 Without a control group we have problems of attributability and without randomization confounding would prevent valid inferences. On the other hand, we were concerned that the widely accepted gold standard of a randomized controlled trial might prove ethically unacceptable, as well as difficult to implement fully in this setting. Despite the latter concerns, we have been able to set up a randomized study, the second we are aware of in this field. In gaining acceptance of this design, it has been essential to explain the reasons for it to both the Council and the local residents. As the Council were unable to improve all the houses on the estate in 1 year, it was straightforward enough to explain that we needed some way of allocating houses to be improved in either the first or second phase of the work. Normally, allocation to each phase would be done by the Council, on the basis of (a) the need of the occupants, (b) current state of repair of the houses, (c) nature of the improvements to be undertaken or (d) convenience of the Council's building contractors. Randomization was accepted easily by the Council's Housing Department and, in principle, by the residents, on the understanding that a lottery was the fairest way of allocation. It did not prevent some lively discussion at the public meeting when randomization was carried out, including offers from some people to swap their allocation for another's on the grounds of the latter's needs being greater.
While acceptance of the principle of randomization was the first major hurdle, there followed considerable discussion of the unit of randomization. The Council initially wished to improve houses in blocks, so that we first planned to randomize by street, which would have given us a weaker design; in particular intracluster correlations would reduce our statistical power. We succeeded, however, in getting agreement that houses could be improved individually, resulting in our ability to randomize at the level of the house. We nevertheless decided to stratify by street in order to increase acceptability to residents.
In transferring the model of a randomized trial to this setting, we were aware that blinding of the residents to the intervention was never an option. The study will thus always be subject to bias in the respondents. It would, nevertheless, be possible to blind the health and environmental assessments. We have not formally done so, however, as we felt that the disadvantages outweighed the advantages of trying to do so. The environmental assessments would always have to be carried out in the house, but in the second year, the engineer was not formally told which houses had been improved in advance of his visit. He was, of course, unable to remain blind to newly installed central heating once there. More seriously, we considered asking the residents to attend a neutral setting where the nurses could carry out the health surveys without being aware of the state of the house. This possibility was rejected on local advice, as we thought it very probable that we would have more difficulty in maintaining high response rates if the nurses did not visit people in their houses. Previous experience17 also suggested that people were unlikely to respond to invitations for health interviews in a setting away from home. We have therefore gone for the option of obtaining good data in terms of responses at the expense of design rigour.
As very few rigorous studies in this area have been completed, we are uncertain of the timescale over which changes in health are likely to occur. We will have data on houses improved in the first year for 12–18 months after completion of those improvements. We can only speculate at the moment whether we see an early improvement in health, which may or may not be sustained. Some dimensions of health may be more likely to change (for better or worse) than others and over different timescales.
In addition to the issues of study design, there was the choice of outcome measures. Poor housing is linked to several aspects of poor health. By improving a house, the benefits are presumably available to everyone who lives in it, not just those with an illness recognized by the health services. If we then concentrate our attention on a single disease, or a single age group, we will not be capturing the full effect of housing improvement on the health of all the residents. The Cornwall study focused on identifying children with asthma and then improving the houses they lived in. The Watcombe project has started with a group of houses in need of improvement and is trying to capture both general and disease-specific outcomes for health. Our chosen diseases were based on the reports from the early survey by the residents and we have added general health measures as well – if we had not, then some of the houses included in the improvement scheme would not have had the health of any residents evaluated. This policy has resulted in a substantial burden of information for our nurses to collect and the residents to supply, yet we are aware that we will still not capture some aspects of health gain or other benefit (or harm) of the housing improvements. We would like to explore issues of mental health at a future date and also gain a better understanding of how improved housing influences domestic and social behaviour, e.g. by enabling school children to do homework in their bedrooms rather than in the one heated room with the rest of family present.
As indicated above, we have been aware of the need to keep the local community involved in the project. Much of the success in doing this has come from the origins of the study in the Barton, Watcombe and Hele Health Gain Initiative, which has helped to build community awareness on the estate. The appointment of the community worker as the research assistant on the project has provided a local and trusted advocate for the study. Her local knowledge has proved invaluable in order to get the involvement of residents through attending meetings or completing surveys. Timing visits and meetings is also important, e.g. with regard to television schedules. She has also helped with gathering other information: questionnaires may not be completed without her help as illiteracy is a recognized problem for a few local people. She has also provided added incentives for participation in the evaluation surveys, by getting local businesses to donate prizes for draws and persuading the Council to fund a children's Christmas party. Her role includes distributing regular project newsletters, and organising displays about the project in the local community office and in the local school.
There are nine independent organizations working together in this study and the `joined up thinking' promoted by the present Government has not been easy to achieve. Ensuring that the study's objectives are met by agencies with differing functions, timescales and cultures requires excellent communications and diplomatic skills. A logical management structure with clear lines of responsibility is a minimum requirement for sustainable collaboration. Understanding each other's priorities and pressures has taken time, but the collaboration is improving with time and familiarity.
It is also necessary to adapt to changing circumstance. The local elections in May 2000 changed the political makeup of Torbay Council, although we understand that the now dominant Conservative Party will continue to support the project. In addition, the Council is in the process of devolving the management of all its public sector housing to a Housing Trust. Although the funding for improvements in the current year is secure, we are currently engaging the Trust board in discussions so that the collaboration may continue beyond this year and this particular project.
The Torbay Healthy Housing Group acknowledge the financial assistance of the South and West Devon Health Authority and NHS Executive for pilot and main studies, respectively. Support in kind has been provided by the Torbay Council, Riviera Housing Trust and local residents.