Discussion

SM Carr; M Lhussier; N Forster; L Geddes; K Deane; M Pennington; S Visram; M White; S Michie; C Donaldson; A Hildreth

NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Carr SM, Lhussier M, Forster N, et al. An Evidence Synthesis of Qualitative and Quantitative Research on Component Intervention Techniques, Effectiveness, Cost-Effectiveness, Equity and Acceptability of Different Versions of Health-Related Lifestyle Advisor Role in Improving Health. Southampton (UK): NIHR Journals Library; 2011 Feb. (Health Technology Assessment, No. 15.9.)

Cover of An Evidence Synthesis of Qualitative and Quantitative Research on Component Intervention Techniques, Effectiveness, Cost-Effectiveness, Equity and Acceptability of Different Versions of Health-Related Lifestyle Advisor Role in Improving Health

An Evidence Synthesis of Qualitative and Quantitative Research on Component Intervention Techniques, Effectiveness, Cost-Effectiveness, Equity and Acceptability of Different Versions of Health-Related Lifestyle Advisor Role in Improving Health.

Show details

Contents

< Prev Next >

4Discussion

This research aimed to identify, describe, classify and analyse the range of models developed to date for delivering HRLA or training, for effectiveness, mechanism of effect, cost-effectiveness, equity and acceptability in improving the health and well-being of individuals and communities, with particular reference to the reduction of inequalities in the UK. One noticeable limitation of the review was the fact that only six of the included studies had taken place in the UK. While much can be learned from intervention components' details in other contexts, the extent of the transferability of the findings remains to be established. The initial phases of the project served to analyse the breadth of the research question, and to clarify the scope of this review. This is a key and complex step of the process,⁶⁹ which highlighted the importance of contextual issues and, later, guided the methodology adopted.

Including quantitative and qualitative research designs, the review identified 26 papers appropriate for inclusion in a systematic review. The wide variety of LA models, delivered in a wide variety of settings, and targeting a variety of populations, prevented the reviewers from engaging in establishing firm causal relationships between intervention mode and study outcomes. By convention, meta-analysis is designed to utilise results from several related studies by identifying a common measure of effect size that is modelled via meta-regression. With regard to the current review, however, neither the outcomes under investigation nor the methods used are constant. While most of the studies reviewed adopted a quantitative methodology, primary outcome measures were of either the parametric or frequency variety, thereby rendering direct comparisons impossible. The following section considers the issues surrounding the robustness of the review.

Analysis of the robustness of the review (sensitivity analysis)

The criteria for study inclusion in the review are provided in Chapter 2 (see Box 4). This section of the report provides further information on the quality assessment of the included studies. The purpose is to provide some critical commentary on the strength of evidence on which the review is based. The process has been guided by the principles of quality assessment⁶⁹^,³²⁰ and the work of Jackson and Waters.⁶⁴

To quote from the CRD guidelines,⁶⁹ ‘… the aim of assessing study quality is to establish how near the “truth” its findings are likely to be and whether the findings are of relevance in the particular setting or patient group of interest’ (p. 33). This is an issue worthy of comment, as the quality criteria required to achieve inclusion in this review limited data to a particular type of evidence. The consequence is that only a partial representation of the current practice of HRLA interventions may be reported in the review. This issue is further highlighted by drawing on anecdotal, early small-scale and formative evaluation evidence of the PAG with respect to one type of HRLA intervention: health trainers. Current practice activity is described as:

broader in focus (i.e. not limited to one health improvement issue)
possibly be more likely to reach disengaged populations (as, by nature, in research there is a process of consent to participate, which may alienate people)
possibly be more likely to link with other services (such as smoking cessation services, for example)
working on longer timescales, thus informally assessing knowledge acquisition and behaviour change
having a greater focus on barrier removal, which was addressed only in five of the studies included here
having more likely to engage in community development activities, which none of the studies included here did.

Appropriateness of study design

Some aspects of the study designs may have introduced bias. For example, some studies recruited participants from prior studies²⁶^,⁹⁷ or from existing health-care services.¹¹⁹

Some trials were unblinded and combined this with self-report data, for example Earp 2002¹⁶^,⁶³^,⁹³^,⁹⁴ and Paskett 2006¹¹⁶^,¹¹⁷ This may be difficult to avoid for some of the interventions studied, but there were few efforts to acknowledge or mitigate the problem. Longer follow-up would help to determine the duration of effect. This might also allow for the LA role/HRLA intervention to develop and mould itself to community-specific needs.

Often comparisons were with control groups receiving ‘standard care’, but studies did not always report clearly what that entailed. Another issue relates to ‘background noises’ of health education campaigns, which was not acknowledged in any of the studies included here.

It was sometimes difficult to distinguish the effect of diverse intervention components. For example, Anand 2007⁸⁰ provided a very brief description of the intervention components, which makes it very difficult to assert what element (information giving, nurturing, removing of barriers, etc.) was the most important to intervention success. This has to be balanced with the fact that complex local needs may be best met with complex interventions, and that a detangling of discrete intervention components may come as a second phase to intervention trial for effectiveness. This is therefore not necessarily an issue of inappropriateness of study design, but a comment about where evidence research on this subject area tends to be placed on the MRC continuum of evidence development.⁶⁷

Choice of outcome measures

Few of the studies included explicitly measured outcomes taking into account socioeconomic profile, making it impossible to comment on equity of outcomes. Another issue for consideration is that many studies focused on rather homogeneous disadvantaged populations, so differential outcomes would be difficult or impossible to measure. Most interventions included aimed at changing behaviour. Assumptions thus appeared to be made about the linearity of the following chain of action: information provision – knowledge acquisition – behaviour change – physiological outcomes. There was very limited description detail of the information provision element (see above about ‘background noise’): four studies assessed knowledge acquisition as a result of the intervention,⁸⁴^–⁸⁷^,¹⁰⁸^,¹⁰⁹^,¹¹²^,¹¹⁶^,¹¹⁷ but none related this to subsequent likelihood of behaviour change. Those studies that assessed behaviour change did not relate this to changes in physiological outcomes. This may be because clear links have yet to be established in the health improvement literature, but reports did not acknowledge this issue or bring any methodological solution to it.

The benefits of lifestyle interventions are typically accrued decades into the future for younger people, hence capturing long-term outcomes (such as a reduction in mortality from smoking cessation) simply is not feasible. As such, the surrogate short-term outcomes reported are appropriate. However, there are concerns that lifestyle changes may not be maintained, and in this respect the very short duration of many studies (few were over 1 year – Gary 2003⁹⁸^–¹⁰⁰ are an example of this) raises concerns. There is also a lack of clarity as to whether maintenance requires continued input from LA – behavioural theory would suggest that a single input is unlikely to maintain behaviour change if the environment that the person is in does not also substantially change, for example drug users remaining in a drug-using community (unhelpful), smokers no longer being able to smoke in public venues (helpful).

Further concerns arise where self-reported data alone (i.e. not backed up by objective measures) are analysed from unblinded trials. There is evidence of inaccuracy in those data (particularly when one of the answers may be more socially acceptable), and a potential for bias in interventions where the trial (but not the control) arm have established a relationship with a LA (in which case the latter should not be involved in data collection). However, it has to be acknowledged that some outcomes that are very important can only be gained from self-report (e.g. QoL, attitude change, satisfaction). This issue might be mitigated by longer follow-up, and it is notable that one publication⁹⁵^,⁹⁶ found evidence of mitigation of the intervention effects after only 12 months. This might indicate either relapse from lifestyle changes or decreased motivation to report favourable outcomes (reduction in bias).

Statistical issues

The wide variety of LA models delivered in a wide variety of settings and targeting a variety of population groups and covering a range of health improvement aims prevented the reviewers from engaging in establishing firm causal relationships between intervention mode and study outcomes. Indeed, apart from the fact that the studies have all been designed to test the effect of intervention by LAs, neither the outcomes under investigation nor the methods used are constant. The disparate nature of the studies meant that no standardised method of estimating effect size was viable; hence the reporting of various effect sizes depending on, and restricted to, the topic under investigation. For example, the standardised mean difference is applied when reviewing studies assessing the same outcome but measured differently, for example via different instruments. In such circumstances results are standardised to a uniform scale prior to analysis. The resulting statistic communicates the size of the intervention effect in each study relative to the study's observed variability. However such an approach is clearly not applicable when dealing with different outcomes from dissimilar studies, albeit with a common, or perhaps similar, intervention philosophy.

Quality of reporting

Jackson and Waters⁶⁴ comment that ‘reviews have been criticised for their focus on individual health education interventions rather than complex environmental or structural interventions and the poor coverage of issues relating to the social determinants of health’ (p. 368). In this review, the evidence assessed did focus mostly on individual behaviour change interventions, and attempts were made to counter that by using a realist approach to reporting. This approach is also thought to have extracted the most meaning out of the data available.

Quality of intervention

Assessing the quality of interventions may be problematic ‘where there is no preliminary research suggesting that an intervention should be administered in a particular way… it is important to establish to what extent these are standardised, as this will affect how the results should be interpreted’⁶⁹ (p. 41). There is a methodological dilemma here, as high-quality research would require replicable interventions, and our synthesis shows that there was a tendency towards standardisation, but the nature of LA intervention may be more intuitive and resistant to the production of ‘one-size-fits-all’ model of delivery.

With respect to complex interventions typical of public health community-based programmes, ‘the quality can be conceptualised as having two main aspects: (i) whether the intervention has been appropriately defined, (ii) whether it has been delivered as planned’⁶⁹ (p. 42). As discussed above, the studies included in this review often lacked a detailed description of their intervention mechanisms, as planned and as delivered. Dickson-Gomez 2006⁹¹^,⁹² in adopting an ethnographic approach, is an exception to this. With this exception in mind, it is of note that this issue is linked to that of the quality of reporting, as detailed descriptive accounts of intervention components, while they would have been considered in a realist synthesis, could not be included in this review.

Interdependency issues

Some studies reported on interventions that would be difficult to implement or be too costly in real-life settings (Dickson-Gomez 2006⁹¹^,⁹² provides an example, in that the LAs would require salary to work in the longer term, and in that the background rates of HIV infection in the UK would mean that this intervention would not be cost-effective in this setting). Interventions are inevitably interdependent on their context, an issue that the realist approach used here has started to tease out, but this is rarely acknowledged in the published literature.

Generalisability

There is a particular issue around generalisability of the interventions described, in that (as described in Chapter 3, Section 2, Interventions context, mechanism and outcomes) there is no such thing as a ‘typical’ practice setting for LAs. The specificity of setting and intervention components may well prevent the success of some LA interventions to be generalisable. The strategy adopted in this review, which highlighted specificity, may offer service providers and funders with a ‘menu’ of intervention characteristics that is flexible enough to allow for local specificity and IK.

While not always granted statistical significance, small effect sizes may be important in public health setting. Indeed, Sorensen et al.³²¹ assert that ‘when risk is widely distributed in the population, small changes in behaviour observed across an entire population are likely to yield greater improvements on the population-attributable risk than larger changes among a smaller number of high-risk individuals’ (p. 380).

It is also worth bearing in mind that the achievement of a small, or even insignificant, effect size in a population, which would not otherwise be accessed by health improvement interventions, is not to be neglected. So while the generalisability of trial results could be statistically questionable, it may be that the consideration of issues of generalisability of interventions' contexts and components could play a key role in addressing health inequalities, for example.

Evaluation approaches and research designs

All research was conducted by professionals/academics, with no peer involvement in the research process; so, for example, when observed by ethnographers⁸⁹^–⁹² the peers ‘led the ethnographers around’, showed them ‘relatively’ safe environments, and people, and probably showed only behaviours regarded as ‘positive’ by the researchers. There may be issues of concern regarding the ‘not stated’/‘not seen’ behaviours. For example, Ungar et al.³⁷ showed peers wanting to be ‘invisible’ to fulfil their roles more effectively.

Evidence application and utilisation: processes and challenges

Given the caveats spelled out in the previous section, some caution has to be exercised in terms of the practice and service messages that can be drawn out in this discussion. The limitations that the review design placed on the type of data that could be included consequently means that there is an information base about HRLA provision outwith this review (see, for example, a newly developed database at www://piph.leedsmet.ac.uk/main/litreview.htm). Indeed, the quality assessment process that studies had to be submitted to prior to inclusion meant that the review favoured single-focus interventions, with defined and often standardised protocols, with a predominant focus on individual behaviour change rather than community development. This may have eliminated report of practices focusing on engagement or social capital or more overtly tackling health inequalities. Thus the series of continuums proposed in the initial phases of this review (Appendix 3), which was based on consultation with practice experts, needed development for a thorough description of the included studies. Equally, few included studies could be positioned on all of the continuums. This highlights the gap that still exists between HRLA practice and research. Mapping this review evidence against a model recently developed²⁰ allowed the location of this evidence base in the wider HRLA knowledge arena. While that model maps out practice foci, the model developed here (see Chapter 3, Figure 4) provides a detailing of models of practice within the individual/behaviour change quadrant of Visram et al.'s²⁰ model. This focus on intervention mechanisms, or intervention theories,⁷⁹ is a key feature of realist synthesis.

So that this discussion may achieve maximum utility to policy-makers and service providers, Figure 6 serves as an anchor for the following paragraphs, where each aspect is covered in turn. The ambition for this approach is that it will allow readers to locate the evidence synthesis and the issues arising from that in their particular cultural and organisational context.

FIGURE 6

Lifestyle advisor intervention.

Contextual issues

Evidence

Overall, previous reviews suggest that LAs may be of use in improving access to health care, and may reduce health disparities. However, the evidence is variable and can give only limited support to LAs having a positive impact on health knowledge, health behaviours and health outcomes. All of the previous reviews identified the need for future research that was of high methodological quality and high reporting quality. This should clearly identify and describe the character and role of the LA, and the character of the population to whom they delivered the intervention. More research is needed to understand the health effects of HRLA in combination with other interventions. The research should use valid, reliable and sensitive outcomes of importance to the participants and increase community involvement. There is a need for longitudinal research to evaluate the duration of effect of the interventions and more research into the social and health costs of providing such services.

Policy priorities

Although most included studies stated that their target population was underprivileged and lacked access to services, none referred to tackling health inequalities as a study aim. Some, however, such as the screening interventions, tackled inequalities in that they made the screening more available to otherwise mostly disengaged populations. However, this was not an explicit aim of the study. Maybe more obviously, the HIV infection prevention studies (Dickson-Gomez 2006⁹¹^,⁹² in particular) sought to engage drug users in the delivery of health care. As such, they fulfilled a dual purpose of (1) engaging these hard-to-reach groups and (2) making safe practice advice and materials more available to them, thus reducing barriers to health. The ethnographic design of the study also meant that the message delivered was not solely from on high, but also took local practices and microcultural dimensions into consideration. Dickson-Gomez 2006⁹¹^,⁹² put a particular emphasis on highlighting the benefits of HRLA work on the LAs themselves and described how, for many of them, undertaking the LA role was the first step to employment and a possible end to homelessness and addiction.

Models of health care

In the introduction to this review the LA role was located within a general movement in the public health field away from a paternalistic to a partnership approach. The development of LA roles, most of which are rooted in some way in the target community, is an example of this policy shift. However, in practice this was only minimally presented in this review, with the premise of correcting inappropriate behaviours being at the root of many interventions. Included studies thus illustrate a partnership approach that was operationalised through a change in workforce rather than a change in message focus. This, however, had an obvious impact on social capital in the case of Dickson-Gomez 2006⁹¹^,⁹² for example.

Understanding need

Few of the studies included in this review make reference to accessing or capitalising on IK as a key component of the intervention. As described above, LAs in the included studies acted as translational agents, who sometimes removed barriers to the prescribed behaviour or helped to create a social environment facilitative or supportive to behaviour change. LAs clearly used their IK in Dickson-Gomez 2006⁹¹^,⁹² to access hard-to-reach individuals, and did report, to an extent, on other, unforeseen, local needs and issues. In this case, the use of LAs who were peers with a common experience and who had lived in the community for some time was crucial to intervention success.

However, the ways in which this capitalisation on IK through LA is realised remained unclear in most studies. Questions remain as to what knowledge was lacking to require LA intervention in the first place, how it was sought and how the message was delivered. This issue relates to an operationalisation of an understanding of local needs, for which techniques such as social marketing could offer potential.

Population focus

The original population continuum (see Appendix 3) did not allow for detailed description of the complexity of intervention target groups. The studies included showed that multiple characteristics can be used simultaneously to describe intervention target groups (Table 10). However, the rationale for selecting a particular combination of characteristics was not made explicit. In particular, although some studies described the local population as hard to reach, the ‘hard-to-reachness’ of study participants was not always asserted. The quality assessment process that all studies were submitted to allocated lower strength to studies when a low proportion of the population agreed to take part. This suggests that some studies that did target disengaged populations had to be excluded from this review.

Intervention location

Intervention location was a key element of intervention delivery mode and approach, in that, for example, people with chronic conditions were often part of interventions that took place in health-care settings. While both types of interventions were classified as community based, there is a key distinction to be made between screening interventions, which had elements of mass education campaigns, and HIV infection prevention interventions, in which location was crucial to engagement, message delivery and acceptability.

Mechanisms

Intervention aim

No studies described the aim of the intervention in terms of placing themselves on a health maintenance–health promotion–primary, secondary or tertiary disease prevention continuum. Of note is the fact that no intervention tackled health maintenance in any population. That is to say, most studies were narrowly focused on one issue or behaviour pattern, and measured outcomes directly related to this. While this is understandable from the methodological point of view of study design, it comes at odds with the potential of local or IK as operationalised by the use of LAs. Indeed, the problematisation of hard-to-reach communities, whether it is in terms of lack of access to services, high prevalence of risky behaviours or diseases, is unlikely to be linked to single causal elements that can be addressed by a single intervention foci.

Intervention delivery mode

This was particularly relevant to the included studies, in that by necessity of producing and recording evidence (often by means of activity logs), few studies reported on ad hoc informal, but yet informative, conversations. Dickson-Gomez 2006⁹¹^,⁹² was an exception to this, but used an ethnographic study design to observe interactions between drug users and trained LAs.

Figure 5 highlights a gap in evidence of informal interventions targeted at groups of the general population. The relationship of LA with other service provision was not well articulated. This is an important deficit with respect to being a bridge between communities and service providers.

The CDSM and associated programmes deserve particular attention. They have been assessed in this review as potentially cost-effective and have been widely replicated throughout the world. In contrast with the acknowledgement in the Choosing health⁷ document that one size fits all might not be appropriate; Lorig 1999¹¹⁰ have developed a highly formulaic intervention, adaptable to a large number of disease groups. There are a few notable differences between this and other HRLA interventions. While groups are defined by behavioural characteristics in the case of interventions targeting healthy eating or screening, for example, they are defined by physiological or physical characteristics in the case of CDSM programmes. The aims of the interventions also differ, in that people with chronic conditions are helped to live with their condition; whereas in other areas, participants are not learning to live with lack of exercise or physical activity, but try to change – thus a change in engagement with risk is a key differential factor.

This diversity is in keeping with the understanding described in UK policy that ‘different neighbourhoods will need different types of health trainers’.⁷ However, what is not clear from the data is why a particular model was selected and also which model achieves the best results in which environment, and that different models of provision will be required to achieve best outcomes. In other words, the links between contexts, mechanisms and outcomes are not explicitly established.

Intervention approach

Most studies included in this review focused on providing information by an alternative message giver (as was predominantly the case in the diet and physical activities study group, with the exception of Anand 2007⁸⁰ and Elder 2006⁹⁵^,⁹⁶); seven studies used this approach only. The assumption is that the message is thus translated in a more acceptable and effective manner. Less than half of the studies described the creation of supportive social environment to help behaviour change. It thus appears that most interventions included here were in support of standard advice (chronic care, the buddy schemes in smoking cessation or breastfeeding, for example). In a few cases, as in the HIV infection prevention studies or in mental health, they were presented as an alternative and more effective approach to standard care. The screening studies focused on reaching out to populations to bring them to standard screening practices, so fulfilled more the role of a bridge between disengaged populations and standard models of care.

Evidence suggests that few studies use one approach only, but, equally, few studies are explicit about approach components, and their effect in isolation and/or in combination. This suggests that intervention approach may be even less explicit in practice, and left to develop from the IK held by the LAs. The categories of information delivery, nurturing for behaviour change, creation of supportive social conditions and barrier removal were created inductively from the included studies, but may thus be insufficient to describe the complexity of interventions in practice. However, crucially, these categories need further unpicking, as some studies appeared to create favourable social conditions, for example, how this was achieved remains unclear. While it is in keeping with the philosophy of HRLA to capitalise in an informal manner on the knowledge held by LA, this is also preventing an articulation of what approach works in what context.

Training

The relationship of the amount and area of LA training to intervention effectiveness remains unclear. Equally, the effect of training on the lay and/or peer status of LAs remains unexplored. Earp 2002¹⁶^,⁶³^,⁹³^,⁹⁴ presents the impact of training on intervention acceptability and credibility, as participants nominated the fact that LA had taken a course as one of the reasons why they would feel comfortable talking to them. Some studies followed a pseudoprofessional approach with respect to recruitment, training and remuneration, and the LAs were rarely selected by the community they were intended to serve.

In practice, the LA role is represented by a range of titles that obscure its key characteristics. For example, distinction between peerness and layness was not made in any of the studies included in this review. This is also true in the UK, where HRLA is often delivered by ‘health trainers’, a title that, in itself, does not assume any degree of peerness or layness. This lack of clarity may have major implications on the mechanisms of action and intervention outcomes.

Outcomes

Although papers included in the review often discuss the content of the intervention, what they do that leads to positive change is very rarely described. So, although it is possible to say that LAs are effective in improving health and well-being, the outcome–causality chain is not clear. Disappointingly, the situation reported by the WHO 20 years ago with respect to CHWs, i.e. a lack of understanding of how to realise the potential of the role¹² continues to plague the LA role. The dominant mechanisms of action appears to balance on the assumption that a change in knowledge leads to change in beliefs, which leads to change in health behaviours, which leads to improvements in health, QoL, activity, participation, etc. Three key issues emerge for comment. First, the time scale of many of the reported studies is too short to allow demonstration of movement along a knowledge to improved health trajectory. The general assumption is that the movement is linear and not dependent on continued or evolving and cumulative interventions. Second, there is a clear need to identify and measure intermediate outcomes to demonstrate progress on such an outcomes continuum. Third, because of the contextual sterility of intervention descriptions, it remains unclear to what extent the LA intervention was a contributor to other programmatic interventions, as is often the case with respect to public health practice. Thus the partnership or cumulative impact or potential is therefore not clear.

Acceptability

Levels of acceptability appear to be high. However, this is often reported as a generic statement with respect to a HRLA service, rather than providing clarity on what aspect of the LA influenced acceptability. Earp 2002¹⁶^,⁶³^,⁹³^,⁹⁴ presented an exception to this, as participants explained how important it was to them that the LA was someone local who they knew well and trusted. Other important elements were that the LA had professional or personal experience of breast cancer and had undertaken training. So the key element in here is that participants wanted the health improvement message translated. In Dickson-Gomez 2006⁹¹^,⁹² the delivery setting was a particularly important acceptability factor, as outreach workers were able to deliver messages in settings not usually targeted, such as disused buildings and other drug injection sites.

Equitability

There are clear gaps in HRLA provision, covering both target groups, such as men, older people or homeless people, for example. There was indeed a clear dominance of interventions targeted at women, but the rationale for this was unclear (i.e. women might be clear change agents in some communities, but this was not made explicit). Interventions were always focused and no evidence could be found of holistic interventions (i.e. tackling health promotion, maintenance, primary, secondary or tertiary prevention).

Cost-effectiveness

The economic analysis suggests that lay-delivered smoking cessation interventions are highly cost-effective. Neither promotion of screening nor exercise/healthy eating is cost-effective. Programmes directed towards improved disease management have the potential to be cost-effective. The conclusions on physical activity and healthy eating flow from a lack of evidence of effectiveness in these areas. Where there is evidence of effectiveness, LAs are not always cost-effective. The key driver is the size of the potential health gain from the behaviour promoted. This is large for smoking cessation, and justifies a relatively intensive intervention. The gain from mammography is simply insufficient to justify even a low-intensity promotion programme. The benefits from improved management of diabetes are potentially large, and may justify a low-intensity call centre-based intervention to encourage healthier lifestyles. While the benefits of averting HIV infection are large, the background rate is too low to justify intensive peer-promoted risk reduction programmes for injecting drug users in the UK.

A considerable amount of uncertainty pervades much of this analysis. Estimates of the health gains are likely to be robust in mammography and diabetes management, as they are based on extensive trial data modelled by experienced groups. Less attention has been paid to modelling the health gains from smoking cessation, but extensive epidemiological data suggest that the estimate used here is conservative. Consequently, the conclusions on smoking cessation in this study are likely to be robust, and they are similar to many published studies of cessation services. The analysis of mammography used generous estimates of the benefits of the programme and the results are likely to be robust, although other authors have come to different conclusions. The greatest uncertainty exists over the benefits of breastfeeding. Few would doubt that benefits exist, but the evidence of improvements in cognitive ability and reductions in obesity and type 1 diabetes is controversial given the inevitable environmental confounders. Without an estimate of the health gains from breastfeeding it is very difficult to judge whether promotion is cost-effective. It should be noted that the small number of studies reviewed in each area raises the possibility of publication bias, leading to an overestimate of the effectiveness of LA programmes.

The greatest uncertainty arises with respect to maintenance of behaviour changes. Data from the smoking literature are encouraging, in as much as they suggests that a proportion of quitters remain abstinent. Few data exist on whether changes in diet or physical activity are maintained but the evidence from the weight loss literature is not encouraging.³²² It is quite possible that long-term abstinence from smoking is easier to maintain than dietary improvements and physical activity routines given the financial incentives to abstain. Data on long-term maintenance are essential if judgements on the viability of diet and physical activity promotion programmes are to be made.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK109529

Contents