Investing in mental health and well-being: findings from the DataPrev project

A systematic review was conducted to determine the extent to which an economic case has been made in high-income countries for investment in interventions to promote mental health and well-being. We focused on areas of interest to the DataPrev project: early years and parenting interventions, actions set in schools and workplaces and measures targeted at older people. Economic evaluations had to have some focus on promotion of mental health and well-being and/or primary prevention of poor mental health through health-related means. Studies preventing exacerbations in existing mental health problems were excluded, with the exception of support for parents with mental health problems, which might indirectly affect the mental health of their children. Overall 47 studies were identified. There was considerable variability in their quality, with a variety of outcome measures and different perspectives: societal, public purse, employer or health system used, making policy comparisons difficult. Caution must therefore be exercised in interpreting results, but the case for investment in parenting and health visitor-related programmes appears most strong, especially when impacts beyond the health sector are taken into account. In the workplace an economic return on investment in a number of comprehensive workplace health promotion programmes and stress management projects (largely in the USA) was reported, while group-based exercise and psychosocial interventions are of potential benefit to older people. Many gaps remain; a key first step would be to make more use of the existence evidence base on effectiveness and model mid- to long-term costs and benefits of action in different contexts and settings.

withdrawal from the workforce on health grounds (McDaid, 2011).
While these are serious impacts, they are in themselves insufficient to justify investment in measures to promote mental health and wellbeing. For this, it is important not only to identify robust evidence-informed actions, but also to look at their costs and resource consequences, within and beyond the health system. Resources are always finite, with many potential alternative uses, and careful choices have to be made on investment and priority setting. It is perhaps even more critical to highlight whether investment in the promotion of mental health and well-being might represent good value for money and help avoid future costs of poor mental health during the current austere climate when health and other public sector budgets are under substantial pressure, and when mental health promotion may not be seen as a high priority for policy makers .
As part of the EC funded DataPrev project, a systematic review was conducted to identify the state of the evidence base on the use of economic evidence in helping to make the case for investment in mental health and well-being in the four areas of focus to the project: early years and parenting interventions, actions set in schools and workplaces and measures targeted at older people.

METHODS
Our objective was to identify economic evaluations, i.e. studies comparing the effectiveness and costs of two or more health-focused interventions, to promote mental health and wellbeing and/or prevent the onset of mental health problems.

Inclusion and exclusion criteria
Two distinct types of study were eligible for inclusion. First, economic evaluations conducted concurrently or retrospectively alongside a randomized controlled trial. An exception to this criterion was applied to workplace health promotion interventions where controlled trials are rare; in this case other empirical study designs alongside an economic analysis were also eligible. Economic evaluations conducted using a modelling approach, whereby effectiveness data were collected from one or more previous controlled studies and then combined with data on costs, were also included. Economic evaluations had to be consistent with different approaches commonly applied in health economics, including cost-effectiveness, cost -benefit, costconsequence, cost -utility and cost-offset analyses. While we cannot discuss the differences between these approaches here, the interested reader can refer to numerous guides, e.g. (Drummond et al., 2005;Shemilt et al., 2010).
To be eligible for inclusion studies also needed to include either a measure of positive mental health, e.g. use of the SF-36 mental health summary scale or other measures of quality of life, specific measures of well-being or alternatively quantify the prevention of psychosocial stress and/or mental disorders. We excluded studies relating to the prevention of dementia, as well as those focused on individuals with learning difficulties from our analyses. Interventions needed to have a primary objective of promoting health. This meant that we excluded some education and child care centred interventions that had subsequently been shown to have a positive impact on mental health (among other outcomes) (Barnett, 1998;Barnett and Masse, 2007).
Papers that focused on the treatment of individuals with existing mental health problems were excluded, with the exception of studies that looked at how the treatment of parents with mental health problems might promote/ protect the mental health of their children, as well as those reporting proxy outcomes, such as improvements in parent-child interaction and the prevention of child abuse. Children were assumed to be between the ages of 0 and 16, while studies in respect of older people focused on people aged 65 plus.

Search process
A search strategy designed to identify economic evaluations in bibliographic databases (Sassi et al., 2002) was combined with a range of mental health promotion/mental disorder terms and a set of population/setting specific keywords and phrases. Mental health-related terms and concepts included in the search included mental health, positive mental health, mental and emotional well-being, personal satisfaction, quality of life, happiness, resilience, energy and vitality. Health promotion and prevention-Investing in mental health and well-being i109 related keywords and phrases were also combined with terms related to poor mental health, including psychological stress, post-natal/postpartum depression, conduct disorder and child behavioural disorders.
We searched PubMed, PsycINFO, EMBASE, CINAHL, PAIS, Criminal Justice Abstracts, Web of Science, Scopus, EconLit and the National Health Service (NHS) Economic Evaluation Database at the University of York. Only results that reported abstracts (or chapter summaries) in English were included; geographical coverage was limited to the European Economic Area, plus EU Candidate Countries, Switzerland and other Organisation for Economic Co-operation and Development (OECD) members. Our review covered the period from January 1990 to December 2010. The electronic search was complemented by a limited search for key terms in Google Scholar, the general Google search engine and scrutiny of relevant websites, e.g. think tanks, universities, government departments and agencies. We also undertook a handsearch of a small number of journals and examined the reference lists of included studies, as well as citations of papers that met our inclusion criteria.
In addition, we also looked for any economic analyses of mental health promoting interventions previously shown in companion systematic reviews on effectiveness conducted as part of the DataPrev study to be effective in promoting mental health and well-being. Where these reviews identified evidence of the impact of an intervention on mental health and well-being, any studies that looked at the economic case for investment in those interventions, even if focused on non-health benefits, such as improved educational attainment, reduced crime and violence, were then eligible for inclusion.
References were initially screened independently by two reviewers (D.M. and A.P.) on the basis of their abstracts/summaries to determine whether they met study inclusion criteria. In the case of disagreement the two reviewers discussed the paper and came to a final decision on inclusion/exclusion, erring on the side of inclusion where no easy agreement could be reached. The full text of all references appearing to meet initial inclusion criteria was then retrieved and a final assessment made. Ultimately included studies were coded and stored in an Endnote database. An assessment of the quality of studies was also made, making use of two published economic evaluation checklists (Drummond and Jefferson, 1996;Evers et al., 2005). Overall this process meant that .3000 references were assessed (see Figure 1).

Parenting, early years and school-based interventions
There has been a considerable body of research into the effectiveness of interventions to promote/protect the mental health and wellbeing of children and their parents, both within and external to school settings (Adi et al., 2007a, b;Dretzke et al., 2009); there is also a small but growing number of studies looking at the economic case for taking action, albeit largely set in either a USA or UK context. We also identified one study protocol for an economic evaluation of an internet-based group intervention to prevent mental health problems in Dutch children whose parents have mental health or substance abuse problems (Woolderink et al., 2010). Overall the results are mixed, as the summary of findings from 26 papers and reports in Tables 1 and 2 indicate. Table 1 includes several studies looking at the impact of health visitors, including the wellcited Nurse Family Partnership programme developed in New York in the 1980s (Olds et al., 1993). Focusing on new mothers, but with a special emphasis on teenage, single-and lowincome mothers, the study followed 400 mothers and their children over a 15-year period. Looking at a broad range of outcomes going beyond positive maternal and child mental health outcomes, an initial analysis reported net costs per woman of $1582 (1980 prices) over the first 4 years for the whole population, but net savings of $180 per high-risk woman (Olds et al., 1993).

Empirical studies
Home visiting programmes have also been examined in England; some focused directly on child mental well-being, others on avoiding post-natal depression, a risk factor for poor child mental health (Murray, 2009). A controlled trial of an intensive home visiting programme and social support programme for i110 D. McDaid and A-La Park vulnerable families where children could be at risk of abuse or neglect reported a cost per unit improvement in maternal sensitivity and infant cooperativeness of £3246 (2004 prices) (Barlow et al., 2007;McIntosh et al., 2009). The challenge with such a finding, however, is judging whether this well-being improvement represents value for money, as it uses a clinical outcome measure which cannot be compared with other uses of resources within the health-care system. Both cost -utility analyses where outcomes are measured in a common metric, such as the Quality Adjusted Life Year (QALY) where a maximum cost per QALY deemed to be cost effective can be determined in different contexts, or cost -benefit analyses where both outcomes and costs are measured in monetary terms can be used to overcome this problem, although neither approach is without its own limitations (Kilian et al., 2010).
In England, a randomized controlled trial of health visitor delivered psychological therapies for women at high risk of post-natal depression improved outcomes at lower costs than health visitor usual care. There was a 90% chance that the cost per QALY gained would be ,£30 000; a level generally considered to be cost effective in an English context (Morrell et al., 2009). Another trial of women at high risk of post-natal depression compared health visitor delivered counselling and support for motherinfant relationships to routine primary care, finding that if society was willing to spend £1000 to prevent 1 month of post-natal depression then the intervention would have a 71% chance of being cost effective with mean net benefits of £384 (2000 prices) (Petrou et al., 2006). This contrasted with an earlier study on the use of post-natal support workers to reduce the risk of post-natal depression which did not appear cost effective (Morrell et al., 2000). However, the former study needs to be interpreted carefully as neither the change in costs or outcomes in the trial were significant and a comparable measure such as the QALY was not be used. Covering a longer time period and looking at additional benefits to children and mothers may have strengthened study findings.
Compared with standard health visitor care, no effectiveness or economic benefit was found in making use of supportive home visits to ethnically diverse mothers in London (Wiggins et al., 2004(Wiggins et al., , 2005. Home visiting was also compared with participation in a mother -child attachment group intervention in Canada. While no difference in effects was reported, costs were significantly lower in the attachment group (Niccols, 2008). We also found a recent Investing in mental health and well-being i111  (Foster, 2010)

Continued
Investing in mental health and well-being i115

Continued
Investing in mental health and well-being i119 The cost of implementing Triple P to one cohort of 2 year olds would be AUD 9.6 million. The average cost per child in the cohort would be AUD 51 RCT, randomized controlled trial; CBA, cost -benefit analysis; CEA, cost-effectiveness analysis; CCA, cost-consequences analysis; CUA, cost -utility analysis; COA, cost-offset analysis.
Australian study that reported that the provision of advice and materials within a maternal and child health centre to mothers of infants with sleep problems had similar costs but better mental health outcomes for mothers and improved sleep patterns for infants compared with standard clinic consultations (Hiscock et al., 2007). As Table 1 indicates, a number of economic evaluations of parenting studies conducted alongside randomized controlled trials have been published, some set in schools, others focused on pre-school age children. In addition we identified one published study protocol for an ongoing evaluation in Wales (Simkiss et al., 2010). An evaluation of the Webster-Stratton Incredible Years parenting programme in Wales, while finding the intervention to be costeffective for all 3-5-year-old children at risk of conduct disorder, suggested that the intervention would be most cost-effective for children with the highest risk of developing conduct disorder (Edwards et al., 2007). Analysis from a trial looking at 3 -8-year-old children in the USA also suggests that combining the parenting component of Incredible Years with child-based training and teacher training, even though more expensive, can be more cost-effective .
As with many health promotion interventions, benefits are only achieved if there is uptake and continued engagement with an intervention over a period of time. One Canadian study looked at community group versus clinic-based individual parenting programmes; while both approaches were effective in reducing the risk of conduct disorders the community group approach was six times more cost-effective because it reached a larger number of parents (Cunningham et al., 1995). A trial of the Incredible Years Programme, combined with a manualized intervention using reading to promote interaction between disadvantaged parents and their children in London, would however only be cost-effective if uptake and engagement rates could be improved (Scott et al., 2010).
The most negative studies were linked to empirical analysis of the Fast Track programme, a 10-year, multi-component prevention programme implemented in four areas in the USA and focused in part on the promotion of better mental well-being and the prevention of antisocial behaviour and violence. Although this included as one component a school curriculum approach based on PATHS (Promoting Alternative Thinking Strategies), it did not appear to be cost-effective. This may have been partly due to limitations in outcomes data in the study, but even if the intervention could be targeted solely at high-risk children it would only be cost-effective if society was willing to pay more than $750 000 (2004 prices) per case of conduct disorder averted (Foster andJones, 2006, 2007;Foster, 2010). In all of these Fast Track studies no specific monetary valuation was placed on the maintenance of better mental health and well-being, but rather on the longterm consequences to non-health sectors, such as criminal justice.

Modelling studies
As Table 2 indicates, economic models have been used to estimate some of the long-term potential costs and benefits associated with parenting, early years and school-based interventions. Further economic analysis, drawing on 15-year outcome data (Olds et al., 1997) suggested that the economic case for home visiting for all women was much stronger, given the impacts it had in terms of reducing abuse, violence, the need for social welfare benefits and improved employment prospects (Karoly et al., 1998(Karoly et al., , 2005. Benefits outweighed costs by a factor of 5.7 to 1 for high-risk women and 1.26 to 1 for low-risk women. As part of a wide-ranging economic analysis of early intervention programmes commissioned by the Washington State Legislature, several programmes relevant to DataPrev were modelled. It should be noted that the authors of these analyses acknowledged that a limitation of their modelling analysis was that it did not put a monetary value on the economic benefits associated with gains in social and emotional mental well-being or broad health benefits. This was due to the terms of the reference received from the Washington State Legislature, which limited the outcomes for all evaluations to crime, substance abuse, educational outcomes, teenage pregnancy, teenage suicide attempts, child abuse, neglect and domestic violence (Aos et al., 2004).
Nonetheless this Washington State review included further evidence of an economic case for action. Analysis of the Nurse Family Partnership, making use of further updated cost i122 D. McDaid and A-La Park data (Olds et al., 2002) reported a benefit to cost ratio of 2.88 to 1 when modelling benefits to child school leaving age, with major benefits due to crime avoided (Aos et al., 2004). Combining data from several similar home visiting programmes a benefit: cost ratio for programmes targeting high-risk mothers had a 2:1 return on investment, with net benefits per mother of $6077 (2003 prices). (Aos et al., 2004).
Turning to school-based interventions, the Caring School Community scheme developed in the USA (Battistich et al., 1996) and now being implemented in Europe, can be delivered at a cost of $16 per pupil over 2 years, and potentially generate a return on investment of 28:1, even when just looking only at benefits of reduced drug and alcohol problems alone (Aos et al., 2004). Using data from the Seattle Social Development Project, which implemented a teacher and parent intervention including child social and emotional development for 6 years and then followed up these children from age 12 to 21 (Hawkins et al., 2005), costs of $4590 (2003 prices) per child were outweighed by benefits that were three times as great. Again this analysis may be conservative, as no monetary value was placed on the significant improvements seen in mental and emotional health (Aos et al., 2004).
Another school-based intervention that has been modelled is the Good Behaviour Game (GBG), an approach which seeks to instil positive behaviours in children through participation in a game, with prizes given to winning teams who behave better. Potential net cost savings of between $15 and $20 million might be achieved for a hypothetical cohort of 5-and 6-year-old children if the programme could achieve a 5% reduction in special education placements, a 2% reduction in involvement with prison services and a 4% reduction in lifetime prevalence of tobacco use (Embry, 2002). Focusing solely on the economic benefits from evidence on a reduction in tobacco use rather than on any of its mental well-being benefits (Kellam and Anthony, 1998), another analysis of the GBG reported a return of investment of 25:1 (Aos et al., 2004).
As Table 2 shows several economic models have looked at the case for investing in the multi-component, manualized multi-level Triple P-Positive Parenting Programme in a number of different settings. Modelling the potential benefits of universal application of Triple P to the Queensland child population aged 2 -12, the average cost per child would be AUD 34 (2003 prices). It would appear to offer very good value for money when assumed to reduce the prevalence of conduct disorder by up to 4%, generating cost-savings of AUD 6 million. The intervention would have better outcomes and costs would be outweighed by conduct disorder averted as long as the prevalence of conduct disorder was at least 7% Mihalopoulos et al., 2007). In a USA context, an economic model predicted that the costs of Triple P could be recovered in 1 year through a modest 10% reduction in the rate of child abuse and neglect (Foster et al., 2008).
In England, modelling work for NICE (National Institute for Health and Clinical Excellence) looking at the universal use of a teacher delivered PATHS programme for children combined with parent training was reported to have a 66% chance of having a cost per QALY gained of ,£30 000. Combining emotional and cognitive benefits in the model's base case scenario the cost per QALY gained would be £5500 (McCabe, 2008). Other modelling work looking at universal use of social and emotional learning interventions for 11-16-year-old children, and drawing on a review of effectiveness evidence on its application to the prevention of bullying (Evers et al., 2007), suggested that if the intervention reduces victimization by 15% then it would have an 92% of having a cost per QALY ,£30 000 (Hummel et al., 2009).
Promoting mental health at the workplace A number of reviews have looked at evaluations of the effectiveness of interventions delivered in the workplace to promote better mental health and well-being (Kuoppala et al., 2008;Corbiere et al., 2009;Martin et al., 2009a). Actions can be implemented at both an organizational level within the workplace and targeted at specific individuals. The former includes measures to promote awareness of the importance of mental health and well-being at work for managers, risk management for stress and poor mental health, for instance looking at job content, working conditions, terms of employment, social relations at work, modifications to physical working environment, flexible working hours, improved employer -employee Investing in mental health and well-being i123 communication and opportunities for career progression. Actions targeted at individuals can include modifying workloads, providing cognitive behavioural therapy, relaxation and meditation training, time management training, exercise programmes, journaling, biofeedback and goal setting. Tables 3 and 4 summarize key findings on the economic case for investment in workplace mental health promotion from empirical and modelling-based studies. While the costs to business and to the economy in general of dealing poor mental health identified at work have been the focus of attention by policy makers in Europe and elsewhere in recent years (Dewa et al., 2007;McDaid, 2007), less attention has been given to evaluating the economic costs and benefits of promoting positive mental health in the workplace. A recent review for NICE found no economic studies looking specifically at mental well-being at work had been published since 1990 (National Institute for Health and Clinical Excellence, 2009a).
In part this may be due to a lack of incentives for business to undertake such evaluations, as well as issues of commercial sensitivity. There have been few controlled trials of organizational workplace health promoting interventions, let alone interventions where mental health components can be identified and even fewer where information on the costs and consequences of the intervention are provided (Corbiere et al., 2009). Moreover, many actions within the corporate world do not tend to be published in academic journals or books but rather in company literature. This makes studies more difficult to find and a full search of company literature was beyond the scope of our review. Most workplace health promotion evaluations related to mental health have focused on helping individuals already identified as having a mental health problem remain, enter or return to employment (Lo Sasso et al., 2006;Wang et al., 2006;Brouwers et al., 2007;McDaid, 2007;Zechmeister et al., 2008).
In fact, we were able to identify several economic analyses with some focus on mental health promotion (Table 3), largely from a US context where employers have had an not inconsiderable incentive to invest in workplace health promotion programmes, given that they typically have to pay health-care insurance premiums for their employees (Dewa et al., 2007). At an organizational level, modelling work undertaken as part of the UK Foresight study on Mental Capital and Well-being suggests that substantial economic benefits that could arise from investment in stress and well-being audits, better integration of occupational and primary health-care systems and an extension in flexible working hours arrangements (Foresight Mental Capital and Wellbeing Project, 2008). Modelling analysis of a comprehensive approach to promote mental well-being at work, quantifying some of the business case benefits of improved productivity and reduced absenteeism was also produced as part of guidance developed by NICE (Table 4). It suggested that productivity losses to employers as a result of undue stress and poor mental health could fall by 30%; for a 1000 employee company there would be a net reduction in costs in excess of E300 000 (National Institute for Health and Clinical Excellence, 2009b). Another analysis looking at the English NHS workforce reported potential economic gains from reducing absence levels down to levels seen in the private sector that would be equivalent to .15 000 additional staff being available every day to treat patients. This would amount to an annual cost saving to the English NHS of £500 million per annum (Boorman, 2009).
Most analyses have focused on actions targeted at individuals, such as stress management programmes, which are less complex to evaluate. There have been a number of economic assessments of general health promotion and wellness programmes (Pelletier, 1996(Pelletier, , 2001(Pelletier, , 2005(Pelletier, , 2009Chapman, 2005), but few have specifically mentioned mental well-being orientated components, and even when they do include these components they may not report mental health or even stress-specific outcomes. The Johnson and Johnson wellness programme, which includes stress management, has been associated with a reduction in health-care costs of $225 per employee per annum (Ozminkowski et al., 2002), while a 4-year analysis of the Highmark company wellness programme, including stress management classes and online stress management advice, reported a return on every $1 invested of $1.65 when looking at the impact on health-care costs (Naydeck et al., 2008). Neither analysis reported specific impacts on mental well-being or stress. Another study of an intervention to help cope with stress in the computer industry did not find any significant difference in stress levels, but it was associated i124 D. McDaid and A-La Park  No ratio reported, as no significant difference in stress, anxiety and coping C: Self-help groups with e-mail personal feedback (partial intervention) and waiting list control 12 months CCA Costs would be lower at $47.50 if delivered by in house medical professionals There was a nearly significant difference in self-reported days of illness for the intervention group But significant 34% reduction in health-care utilization by intervention participants compared with the control groups (p ¼ 0.04) Concluded that this reduction in costs would more than cover the costs of delivering the intervention if delivered by in-house professionals (Renaud et al., 2008), Canada I: Comprehensive health promotion programmes to provide employees with information and support for risk factor reduction, using a personalized approach and involving the organization's management as both programme participants and promoters. Programme includes modules on stress management, healthy eating and physical activity 270 company employees Before and after study. No controls. COA Cost of the intervention not reported Costs avoided not directly reported in monetary terms, but in terms of absenteeism and staff turnover Significant reduction in stress levels away from work as reported using Global Health Profile Score over 3 years falling from 27 to 17% (p , 0.0001). There was also a reduction in feelings of depression with 54.8% of participants stating that they rarely felt depressed after 3 years compared with 38.5% at baseline (p , 0.0001). There was also a reduction in the number of people experiencing signs of stress (p , 0.0001)

Company perspective
No ratio. Significant reduction in high levels of stress, signs of stress and feelings of depression

Continued
Investing in mental health and well-being i127  with a significant reduction in overall reported illness and a one-third decrease in the use of health-care services which would more than cover the costs of the intervention (Rahe et al., 2002). One study that did report mental health outcomes looked at the economic case for investing in multi-component workplace-based health promotion programme ( personalized health and well-being information and advice; health-risk appraisal questionnaire, access to a tailored health improvement web portal, wellness literature, and seminars and workshops focused on identified wellness issues). Using a pre-post test study design, participants were found to have significantly reduced health risks, including work-related stress and depression, reduced absenteeism and improved workplace performance. The cost of the intervention to the company was £70 per employee; there was a 6-fold return on investment due to a reduction in absenteeism and improvements in workplace productivity (Mills et al., 2007).
The experience of employees in another health promotion scheme over 3 years was compared with matched controls. Overall levels of risk to health were significantly reduced, while there was also a significant reduction in the prevalence of depression, although rates of anxiety significantly increased. There were net cost savings from a health-care payer perspective, although the costs of participation in the health promotion programme were not reported (Loeppke et al., 2008). In Canada, an uncontrolled evaluation of a comprehensive workplace health promotion programme, including information for stress management reported a significant reduction in stress levels, signs of stress and feelings of depression at the end of a 3-year study period. While costs of the programme were not reported, staff turnover and absenteeism decreased substantially (Renaud et al., 2008). A small controlled study looking at a programme to prevent stress and poor health in correctional officers working in a youth detention facility in the USA, reported incremental cost savings of more than $1000 over 3 months, although the sample size was too small to be significant. However, the study did not monetize the value of reported productivity gains, while there were positive changes in outlook, attitudes, anger and fatigue (McCraty et al., 2009).
Studies can also be identified where no impacts on absenteeism rates of stress management interventions were identified (van Rhenen et al., 2007). In other cases analyses of a combination of organizational and individual stress management measures did report improvements in emotional well-being, as well as in productivity and reduced absenteeism, but no cost data were provided (Munz et al., 2001). We also identified an ongoing cost -benefit analysis currently being conducted alongside a randomized controlled trial of a mental health promotion intervention to prevent depression targeted at managers in small and medium size companies involving cognitive behavioural therapy and delivered by DVD in Australia (Martin et al., 2009b).

Investing in the mental health and well-being of older people
The final area we reviewed concerned the mental health and well-being of older people. Sixteen per cent of older people may have depression and related disorders; potentially the prevention of such depression, particularly among high-risk groups such as the bereaved, might help avoid significant costs to families, and health and social care systems (Smit et al., 2006). Evaluations from a wider range of countries were identified, most notably from the Netherlands (Table 5). In addition to published studies discussed below, we also were able to identify some ongoing cost-effectiveness studies where protocols had already been published in open access journals (Joling et al., 2008;Pot et al., 2008).
Several studies looked at different types of home visiting interventions to promote wellbeing and reduce the risk of depression, with mixed results. Neither a home visit programme by nurses in the Netherlands nor a programme to promote the befriending of older people in England was found to be effective or costeffective (Bouman et al., 2008a, b;Charlesworth et al., 2008;Wilson et al., 2009). We did identify a cost -utility analysis from the Netherlands conducted alongside a randomized controlled trial comparing a home visiting service provided by trained volunteers with a brochure providing information on depression . It targeted older people who had been widowed for between 6 and 9 months and who were experiencing some degree of loneliness.
Investing in mental health and well-being i129

Continued
Investing in mental health and well-being i131 Although improvements in quality of life were marginal, because of health service costs avoided the intervention had a 70% chance of being cost-effective, with a baseline cost per QALY gained of E6827 (2003 prices). In Canada a home nursing programme used to bolster personal resources and environmental supports of older people was also associated with a reduction in the risk of depression at no additional cost (Markle-Reid et al., 2006). Recently a controlled trial of a stepped care approach for the prevention of depression in older people in the Netherlands was also found to be highly cost-effective at E4367 per depression/anxiety-free year gained (2007 prices) ( Van't Veer-Tazelaar et al., 2010). Economic analyses also supported investment in some different types of group activities. Regular participation in exercise classes by older people was found to have some mental health benefits and be cost-effective from a health system perspective in England with a cost per QALY gained of E17 172 (2004 prices) (Munro et al., 2004). Several studies also reported the beneficial effects to mental health of Tai Chi (MacFarlane et al., 2005), but no formal cost-effectiveness analysis appears to have been undertaken. A study of 166 people randomized to participation in a choral singing group or no action was associated with a reduction in loneliness and lower health-care costs in the USA, albeit that the costs of the intervention were not estimated (Cohen et al., 2006).Weekly group activity sessions led by occupational therapists in Canada significantly improved mental and physical health outcomes compared with participation in regular group social activities only. The incremental cost per QALY gained from a health and social care perspective was also considered to be costeffective (Hay et al., 2002). In Finland, a trial of psychosocial group therapy for older people identified to be lonely was also reported to be effective with a net mean reduction in healthcare costs per participant of E943 (Pitkala et al., 2009).

DISCUSSION
Our review indicates that there is an economic evidence base in all of the areas examined by DataPrev for some interventions to promote mental health and well-being in some very specific contexts and settings. In addition, we were able to identify published protocols of additional economic studies now underway. However, much of the existing economic literature that is available was beyond the scope of this review as it focused on actions targeted at the prevention of further deterioration, as well as the alleviation of problems in people already identified as having clinical threshold levels of mental disorder. This is consistent with the findings of previous reviews (Zechmeister et al., 2008).
One important limitation of our review was the restriction to English language only materials, although papers in other languages that had abstracts in English were included in the review. Certainly the overwhelming majority of material that we found came from English-speaking countries, but this is consistent with previous reviews of economic evaluations of public health interventions where no language restrictions were applied (McDaid and Needle, 2009).
We will have missed relevant studies concerning workplace interventions that have been published in diverse corporate literature with apparent positive returns on investment, but with insufficient information to be included in this review (Price Waterhouse Coopers, 2008). This includes case studies on the UK Health, Work and Wellbeing website looking at four large and small companies in the pharmaceutical, hotel and leisure, transport and manufacturing sectors. All report some positive impacts on absenteeism and/or staff retention rates. In the case of London Underground, for example, a return of 8:1 on investment in a stress management programme was reported (http://www.dwp.gov.uk/health-work-and-wellbeing/case-studies/).
Great caution must be exercised in drawing any firm conclusions on the economic case for investment, but the case for action in childhood or targeted at mothers appears strong. The economic consequences of poor mental health across different sectors and persisting into adulthood mean that effective health visiting and parenting programmes can have very favourable cost -benefit ratios; all economic analyses reported here from a societal perspective were cost-effective. Narrower perspectives adopted in some other child focused studies where evidence of effect was found, for instance from a health or education i134 D. McDaid and A-La Park perspective alone, may undervalue the potential case for action.
Nine of the ten economic analysis set in the workplace reported favourable outcomes. Most of these studies looked solely at the impacts for employers, either in terms of paying for the health care of their employees or dealing with absenteeism and poor performance at work. No studies looking solely at the benefits of organizational level actions to promote well-being and mental health were found. Given that there is a literature on the effectiveness of some of these measures, there is scope for modelling work to look at the potential economic costs and benefits of these measures. Of the 10 studies looking at programmes for older people, 3 were found to have little chance of being costeffective, but reasonable cost-effectiveness was reported for some group activities and home visiting activities.
In all areas we were able to identify published studies where no evidence of effect was found; these are also critical in helping to ensure resources are not used inappropriately. It is also the case that there has been little incentive to undertake formal economic evaluations of very low cost, but effective interventions, especially where costs are largely not borne by the public purse. One example of this are initiatives, often initially evaluated in low-and middle-income country contexts to promote skin-to-skin touch between mothers and their new born, where the principal cost is the time that the mother spends with her infant (Moore et al., 2007;Maulik and Darmstadt, 2009).
Going forward our analysis of the methodological quality of studies suggests much room for improvement. While high-quality analyses were identified, most studies failed to separate presentation of data on resources used to deliver interventions from the costs of these resources. Few studies undertook more than a very cursory sensitivity analysis to account for uncertainty around estimates of effect and cost. There was little discussion of the distributional impacts of interventions, an issue that is of particular relevance in the context of public health and health promotion interventions, where engagement and uptake can be critical to effectiveness (McDaid and Sassi, 2010).
There is also a need for more common and consistent endpoints to improve comparability across different interventions and country settings. Reliance solely on topic-specific outcomes such the cost per unit improvement of maternal sensitivity or a reduction in loneliness mean that it is difficult to compare the case for different potential areas of intervention. One key challenge in economic analysis going forward is to develop measures that can adequately capture the benefits of improved mental wellbeing. The principal quality of life measure reported in studies here, the QALY, was designed to identify the benefits of the absence of illness rather than well-being. Work on other approaches to well-being is underway; but in the meantime making use of validated well-being instruments such as the Warwick-Edinburgh Mental Wellbeing Scale (Tennant et al., 2007), alongside instruments used to value QALYs, such as the EQ-5D or SF-36, is merited. None of the cost -benefit analysis reported in this paper has elicited direct values for positive mental health: indeed the difficulty in putting a monetary value on well-being for cost -benefit analyses has been noted (Aos et al., 2004). Another issue is that despite the links between poor physical and poor mental health, little economic analysis has focused on the economic case for preventing co-morbidity, for instance on the prevention of depression to promote cardiovascular health. This is another area that economists might explore further.
More use can also be made of economic modelling in the short term to help strengthen the evidence base for investing in mental health and well-being. Such an approach has recently been used to help inform policy making on the case for prevention of various mental health problems in both England and Australia (Knapp et al., 2011;Mihalopoulos et al., 2011). The DataPrev project has demonstrated that there is a substantial evidence base on effective interventions; most of these have not been subject to economic evaluation. Working with programme implementers to determine resource requirements, costs of delivery and any necessary local adaptations, economic models could be used to determine the likelihood that interventions are likely to be cost-effective in different contexts, and over different time periods. They can also be used to look at the case for investing in multi-level approaches to promotion and prevention, with some interventions targeted at the general population and others targeted solely at high-risk groups. Published examples of this approach include the Triple P programme for children (Mihalopoulos et al., Investing in mental health and well-being i135 2007;Foster et al., 2008;Van't Veer-Tazelaar et al., 2010) and stepped care for older people. Such models could also factor in key critical factors such as probability of uptake and continued engagement by different population groups.

This work was supported by the European Commission
Sixth Framework Research Programme. Contract SP5A-CT-2007-044145. Funding to pay the Open Access publication charges for this article was provided by the LSE Institutional Publication Fund.