Health effects of housing improvement: systematic review of intervention studiesBMJ 2001; 323 doi: http://dx.doi.org/10.1136/bmj.323.7306.187 (Published 28 July 2001) Cite this as: BMJ 2001;323:187
- Hilary Thomson, higher scientific officer (, )
- Mark Petticrew, associate director,
- David Morrison, specialist registrar in public health medicine
- Correspondence to: H Thomson
- Accepted 25 April 2001
Objective: To review the evidence on the effects of interventions to improve housing on health.
Design: Systematic review of experimental and non-experimental housing intervention studies that measured quantitative health outcomes.
Data sources: Studies dating from 1887, in any language or format, identified from clinical, social science, and grey literature databases, personal collections, expert consultation, and reference lists.
Main outcome measures: Socioeconomic change and health, illness, and social measures.
Results: 18 completed primary intervention studies were identified. 11 studies were prospective, of which six had control groups. Three of the seven retrospective studies used a control group. The interventions included rehousing, refurbishment, and energy efficiency measures. Many studies showed health gains after the intervention, but the small study populations and lack of controlling for confounders limit the generalisability of these findings.
Conclusions: The lack of evidence linking housing and health may be attributable to pragmatic difficulties with housing studies as well as the political climate in the United Kingdom. A holistic approach is needed that recognises the multifactorial and complex nature of poor housing and deprivation. Large scale studies that investigate the wider social context of housing interventions are required.
What is already known on this topic
What is already known on this topic Many epidemiological studies have described associations between poor housing and health
What this study adds
What this study adds 18 studies were reviewed that studied the health effects of housing improvements
Most studies found some health gains
Small populations and lack of control for confounders limits the generalisability of the findings
More large scale, controlled studies of housing interventions are needed to give qualitative and quantitative data
Poor housing has been used both as an indicator of poverty and as a target for interventions to improve public health and reduce inequalities in health.1 Although housing still has a prime place on the health inequalities agenda, it also has wider importance because small health effects can have a large impact at the population level.
Policy makers are also increasingly interested in measuring the health effects of social interventions (such as social housing) and in gathering evidence to shape policy. 2 3 Much of the research investigating the links between housing and health has been cross sectional, and these studies have shown strong independent associations between housing conditions and health. However, results of studies in small areas are difficult to generalise to other contexts. Observational studies have also shown strong independent associations between poor housing and poor health, but their results remain open to debate and interpretation.4
Experimental studies of the health impacts of housing would provide stronger evidence. The randomised controlled trial has been regarded as the gold standard experimental model to show the effects of interventions in medicine. Such trials, however, are less common in housing research, where there is less of a history of experimentation.5 We carried out a systematic review of intervention studies of the health effects of housing improvement.
We searched the following databases: ASSIA (Applied Social Science Index and Abstracts, 1987-2000), CAB Health (1973-2000), DHSS-DATA (1983-2000), Embase (1974-2000), HealthSTAR (1975-2000), Medline (1966-2000), PAIS (Public Affairs Information Service, 1976-2000), PsycINFO (1887-2000), SIGLE (System for Information on Grey Literature in Europe, 1980-2000), Social SciSearch (1972-2000), Sociological Abstracts (1963-2000), Social Science Citation Index (1981-2000), Urbadisc, Cochrane Controlled Trials Database 2000 Issue 2, IBSS (International Bibliography of the Social Sciences), SPECTR (Social, Psychological, Educational, and Criminological Controlled Trials Register, searched December 2000), and the world wide web. Full details of the search strategy are available from the authors.
We hand searched the bibliographies of all reports, papers, and text books that we reviewed. We also requested information on unpublished and ongoing studies from subscribers to the Housing Studies Association newsletter and email list and the Health Action Zone discussion group. HT contacted health authority housing departments, academic departments in the United Kingdom, local authorities, and housing associations. We also asked delegates at an international housing conference for details of suitable studies, either completed or ongoing.
We sought primary studies in any language that used experimental or quasi-experimental approaches to examine the effects of housing improvements. These included randomised controlled trials and observational studies that used prospective or retrospective measures of health. Our outcome measures were based on a social model of health and included socioeconomic changes and illness based outcomes. Housing interventions were defined as rehousing, and all physical changes to housing were defined as infrastructure—for example installation of heating, insulation, double glazing, and general refurbishment.
We excluded cross sectional studies that did not investigate the effects of housing improvement before and after the intervention. We did not include interventions to improve the indoor environment through furniture or indoor equipment (such as vacuuming, mattresses, and air purifiers) unless the evaluation measured changes in residents' health and the measures were part of a package of interventions that included improvements to the house itself. Environmental studies of the adverse effects of lead, urea formaldehyde foam, poor air quality, allergens, or radon were not included. At least two reviewers independently screened all abstracts identified by the searches.
Assessment of validity assessment and data abstraction
Three reviewers critically appraised the included studies according to the criteria (box). Studies graded as C were not considered in the final assessment of the evidence. When reviewers' conclusions differed, the study was reviewed jointly by three reviewers. Data were abstracted by one reviewer (HT) and checked by a second reviewer (MP). When data on the group of interest were not given in the publication, we calculated them if possible—for example, new P values were calculated using the relevant sample sizes.
Criteria for assessing strength of evidence (adapted from NHS Centre for Reviews and Dissemination method)6
Level A: prospective study, follow up rate >80% and for ≥6 months
Randomised controlled trial, or controlled study with comparable control group
Objective assessment of health outcome(s)
Level B: prospective study with control group
Limited control of confounding
Appropriate assessment of health outcomes
Level C: prospective and retrospective studies that did not adjust for confounding factors
Studies with biased assessment of health outcomes
We identified 18 completed intervention studies (see BMJ ‘s website for details),7–29 the earliest dating from 1936.15 Six studies were identified from electronic databases (figure). Three studies examined the health impacts of rehousing based on medical need,7–10 11 examined the health effects of rehousing or refurbishment and renovation,11–23 and four assessed energy efficiency measures.24–29 Seven studies assessed housing improvement in the context of area or community regeneration. 11 12 15–17 19–21
We also identified 14 ongoing housing intervention studies based in the United Kingdom (see BMJ ‘s website for details). These are investigating similar interventions to the completed studies. Seven of these ongoing studies are prospective and controlled; one is using a randomised stepped wedge design.30
Medical priority rehousing
All three studies of rehousing on the basis of medical need found improvements in self reported physical and mental health. However, the only prospective study was small, 7 8 and no study controlled for the effects of possible confounding variables. One study examined the effects on use of health services and found no clear pattern.9
Rehousing, refurbishment and relocation or community regeneration
Two prospective controlled studies reported beneficial effects of rehousing or refurbishment on health outcomes, including improvements in mental health.11–14 Only one study had controlled for confounding. This study showed an initial increase in illness episodes in the intervention group at 9 months. At 18 months, however, the intervention group showed a larger reduction in illness episodes compared with the control group, although the absolute difference was small (29 episodes/1000 people) and the rate of follow up was not stated. 11 12 The other prospective controlled study reported improvements in mental and physical health, but the study was small and the comparability of the control group is unclear. 13 14
Energy efficiency measures
Although the four studies that we identified all found that energy efficiency measures improve respiratory and other symptoms, only one study adjusted for potential confounding variables.27 High rates of attrition in this and most other studies limit the generalisability of these findings.
Use of health services and social effects
Some studies assessed the effects of improving housing on use of health services; decreased visits to the general practitioner, reduced likelihood of inpatient and outpatient use of health services, and reduced prescribing of hypnotic and respiratory drugs were reported. None of the evidence for these effects came from methodologically robust prospective controlled studies.
Broader social impacts of housing improvement were reported in some studies, including improvements in social outcomes, such as perceptions of safety and social and community participation. 16 17 19 23 One small study reported a small increase in social support after the intervention.19 Two studies that examined the effects of housing improvement in the context of area regeneration, reported that residents' concerns about local crime were reduced. 17 23 Another small study reported that fewer days were lost from school because of asthma after heating improvements.25
We found few studies examining the effects of housing improvements on health, and the quality of the studies identified was generally poor. Improvements have been reported in overall self reported physical and mental health, as well as reductions in symptoms and use of health services. There is also some evidence of improvements in broad indicators of social inclusion such as neighbourliness and fear of crime. However, because of the methodological limitations of the studies, it is impossible to specify the nature and size of health gain that may result from a specific housing improvement. In particular, there are few large prospective controlled studies, and many studies are now quite old.
The effect of publication bias on our study also needs considering. Given the small positive effect sizes and small sample sizes, any summary of the published studies may overestimate the effects of housing improvements. In addition, the fact that we identified only six out of the 18 included studies using electronic databases suggests that systematic reviews of non-clinical interventions need to develop specially tailored search strategies.
Difficulties in studying housing and health
Reasons for the lack of studies into the effects of housing on health may include methodological difficulties and political obstacles. There are many methodological difficulties inherent in assessing the health effects of housing interventions. Poor housing conditions often exist alongside other forms of deprivation, and housing interventions rarely occur in isolation. This may affect the sociodemographics of an area and make before and after comparisons problematic.20 Moreover, response and follow up rates in studies of deprived areas are often low.
More generally, the experimental approach to housing research has been criticised for being reductionist and ignoring the multifactorial nature of causality in housing, deprivation, and health.31 Nevertheless, broad generalisations about the link between deprivation and ill health can have only a limited role in informing specific policy decisions.32 Evidence of the effectiveness and cost effectiveness of specific interventions is therefore particularly important. Assembling such evidence requires a holistic approach, combining quantitative and qualitative methods and taking into account a range of possible influences and mechanisms.33 Although there is a long tradition of this type of evaluation in the United States, it has rarely been attempted in the United Kingdom.34 In 1989, Smith recommended that housing research should embrace a public health approach and include more multidisciplinary studies.35 This is now starting to happen. Of the ongoing studies we identified, eight are collaborations between housing and health agencies and academics.
Political obstacles to conducting housing research may also exist. Traditionally, policy makers in the United Kingdom have not had access to much evidence on the health effects of social interventions. This lack of evidence and the methodological limitations of existing studies may be used by governments to absolve themselves of responsibility for improving housing. However, the current Labour government's interest in identifying “what works,” and its emphasis on joined up decision making, may facilitate a less fragmented approach to tackling deprivation. The number of current collaborative housing studies suggests a greater willingness to use such joined up approaches. What is also needed is robust evidence of their effects.
Sources of evidence other than experimental studies are also important. Longitudinal studies have been recommended as a useful, if expensive, study design in evaluating complex interventions such as housing.35 For example, recent results from the 33 year follow up from the longitudinal national childhood developmental study show that poor housing adversely affects health in later life. The study found a dose-response relation, with multiple housing deprivation leading to greater risk of disability or severe ill health in later life.36 Data from the Boyd-Orr cohort also show that childhood housing conditions have an effect on adult health independent of the effects of socioeconomic deprivation.37
Further research is also needed into the direction of the relation between health and housing. Previous work has suggested that poor health can negatively affect housing opportunities.38
Although our review focused on major housing improvements, good evidence exists from systematic reviews that other interventions to improve health inside the home may be effective. Among these are interventions to reduce house dust mites and to reduce accidents among children and elderly people.39–41
The basic human need for shelter makes the relation between poor housing and poor health seem self evident.42 Despite, or perhaps because of, this intuitive relation, good research evidence is lacking on the health gains that result from investment in housing. We know little about the mechanisms of interaction of social factors and the effects of poor housing over the lifecourse. There is also a lack of comparative information on the costs and effects of specific housing improvements, such as central heating or major refurbishment. It is this type of evidence that is likely to be most valuable to policy makers and housing providers. Large scale studies that investigate the wider social context of housing improvements and their comparative effectiveness and cost effectiveness are now required.
We thank Professor Sally Macintyre for helpful comments on the manuscript, Julie Glanville and Su Golder for help with the searches, and Mary Robins for librarian support.
Contributors: HT contributed to the search strategy of the review, collection, and analysis of the data, and writing the paper and is the guarantor. MP contributed to the collection and analysis of the data and writing the paper. DM contributed to reviewing the studies and writing the paper.
Funding The authors are funded by the Chief Scientist Office of the Scottish Executive Department of Health.
Competing interests None declared.
Tables giving full details of the included and ongoing studies are available on the BMJ's website