The present study analyzed the cost-effectiveness of a mindfulness-based mental health prevention program provided by health coaches in a multi-site field setting on the basis of actually incurred costs. From a societal perspective, the program is likely to be cost-effective compared to usual care, with estimated incremental cost savings of €29 per unit improvement on the HADS for participants of the program (ICER = €-29). From a health care perspective, the intervention was associated with health benefits achieved at higher incremental costs of €91 per unit improvement on the HADS. Cost-effectiveness acceptability curves estimated willingness-to-pay levels of €225 from a societal and €192 from a health care perspective for a 95% probability that the intervention was more cost-effective than usual-care. Sensitivity analyses indicated differences in both, costs and effects, depending on the initial distress of participants.
The multifaceted, complex, and long-term nature of anticipated program benefits of universal prevention for healthy individuals have often been stated as a major reason for the relative shortage of offers and research in universal prevention (e.g. [
48]). In line with this debate, results of our sensitivity analyses suggest that one-unit improvement on the HADS requires a higher willingness-to-pay in the subgroup of symptom-free participants than in initially mildly or moderately distressed participants to achieve a 95% probability that the prevention program is cost-effective compared to usual-care. In other words, decision makers have to be willing to pay more for additional health improvements when offering the prevention program to healthy individuals. In the clinical effectiveness evaluation of this program, new cases of psychopathological symptoms were prevented in 1 of 16 participants [
28] – which might be an early indicator for potential longer-term benefits in monetary respects. Unfortunately, we did not acquire funding for a longer follow-up period. Large longitudinal studies are needed to appropriately show the effects for initially symptom-free populations.
From a health care perspective, the intervention was associated with incremental costs across subgroups of initial mental distress. Direct costs increased to an average of €181 per participant receiving the intervention. This points at a general issue in universal prevention: Many of the determinants and outcomes of poor mental health lie outside the health sector [
48]. The organization that funds a preventive program, in our case a statutory insurance fund, will most likely not profit directly from all its benefits [
49]. Consequently, our results underline the call for joint actions and mixed funding paradigms to cope with the challenge of preventing the onset of manifest mental disorders.
The fact that initially moderately and severely distressed participants show the most preferable ICERs from a societal perspective is in line with research on indicated prevention (e.g. [
8,
10,
12‐
14,
16]). Only one of these studies compared indicated and universal prevention. Hunter et al. [
11] conclude that identification of risk is likely to be more cost-effective than universal prevention. However, the researchers in this trial did not apply a specifically developed universal prevention program, but instead offered low-intensity depression prevention programs, such as bibliotherapy, online cognitive behavior therapy (CBT) or group therapy. It might, therefore, be possible that the applied programs were not optimally suited for universal prevention. Future research should identify approaches and specific components that are most effective in universal vs. indicative prevention programs [
50,
51].
Strengths and limitations
To our knowledge, this is the first cost-effectiveness analysis of a universal community-based health promotion program. The main strength of this study is the high ecological validity across several domains. The intervention had a large reach and availability for a large public by being offered via face-to-face group sessions on various weekdays and at various times in local health centers located throughout the state. Group sessions were provided by non-specialised health coaches, who continued to work in the program after completion of the evaluation study. Participants were included regardless of their initial psychological distress or other health-related issues, resulting in a quite heterogeneous sample. In future studies, it would be useful to test for unobserved heterogeneity in the intervention sample to identify possible subpopulations that are sensitive to the intervention, as well as who do not respond [
52].
A number of possible methodological biases has to be taken into account in the interpretation of our results, many of which relate to the selection and allocation of participants. Due to the complexity of this field trial, it was not feasible to randomize participants and offer an active control condition. This limitation was met by applying propensity score matching (PSM), a rigorous matching procedure which is recommended for the control of the treatment-outcome association in therapeutic studies where randomization is not possible or ethically acceptable [
34]. We included all available information as covariates and introduced two self-reported control items to account for potential differences in health-related activities and the willingness to participate in preventive services. Meta-analyses from other medical disciplines indicate that the treatment effects achieved in studies with PSM are comparable or differ only slightly from the effects of randomized-controlled trials: No significant differences were found in trials on surgical procedures [
53], a slight underestimation of effect sizes for interventions in critical care medicine [
54] and a slight overestimation in the treatment of acute coronary syndromes [
55].
As a further measure to reduce the risk of a selection bias, an intent-to-treat-approach was selected for data analyses. Systematic monitoring of attendance rates and attrition from the program were too complex for the means of this study. Informal counting suggests a dropout rate of about 20% of all participants [
26], which would be within the range of drop-outs in reviews on health behavior interventions (e.g. [
56]). Although no conclusive judgement is possible, it seems reasonable to assume that the ITT analysis at least does not overestimate the effects of the program [
57]. Finally, the relatively high nonresponse rate among the population of program participants has to be taken into account in the assessment of a potential selection bias (see [
26]). Although it is unlikely that the small differences in age and education of responders compared to non-responders have yielded significant effects on outcome and costs, the results of this study can only be generalized to self-selected participants, who are willing to participate in mental health promotion and the corresponding research without further motivational incentives.
Another potential source of bias in single-study based economic evaluations concerns aspects of cost assessment and valuation [
58]. Due to the close cooperation with the health insurance fund, it was possible to obtain actually incurred direct costs and officially registered days of sick leave for the valuing of indirect cost. Any bias stemming from self-reported cost data or estimating health care costs can therefore be excluded [
59]. Accordingly, no cost-related bias is expected from the health care perspective. From a societal perspective, this approach can be considered as conservative and might underestimate cost benefits, because a variety of more distal costs are not included in our analyses – such as e.g. costs related to presenteeism [
60], productivity losses from unpaid work [
12,
13] or informal care [
10]. A recent study on indicated prevention of depression, for example, reports the largest indirect savings in terms of presenteeism [
12], a societal outcome that was not included in our analysis.
Our choice of the main outcome measure impedes the direct comparison with other economic evaluations reporting costs per quality adjusted life year (QALY) gained [
29]. There is an ongoing debate on the suitability, validity and responsiveness of generic preference-based measures for valuing mental health in economic evaluations (e.g. [
37]). One possible approach is the development of statistical mapping algorithms using the responses of condition-specific instruments for the estimation of QALYs [
61]. However, the development of these algorithm is still in its early stages and requires a strong database to acquire an adequate degree of accuracy. Empirical evidence on the transferability of the HADS shows varying correlations between samples [
62‐
64], and suggests a linear relationship with main deviances at the severe end of the scale [
65]. Preliminary analyses based on a linear transformation of the HADS-scores in our sample displayed a willingness-to-pay threshold of around € 9.500 per QALY in the cost-effectiveness acceptability curve to achieve a 95% probability of being regarded as cost-effective. Although this number seems to be within a reasonable range compared to cost-effectiveness-analyses of indicated prevention (e.g. [
12]), it can at most serve as a first impression of possible effects to stimulate future research on universal primary prevention.
Implications
Policy makers and insurance companies have to fill the gap between the pleas for the promotion of mental health and the prevention of mental disorders (e.g. [
6,
66]) and limited health care budgets. Informed decision making requires data on the cost-effectiveness of possible interventions. Economic evaluations of real-world prevention strategies are rare, largely because of the associated methodological challenges and high research costs (e.g. [
48]). The data presented in this article give a first impression that universal mental health prevention programs for adults in a population setting might be a cost-effective strategy to enhance well-being. However, further research is needed to enhance large-scale implementation of such programs. While this study evaluated a mindfulness-based intervention, most studies regarding the cost-effectiveness of selective and indicated interventions are based on CBT (e.g. [
7]). Large trials are necessary to compare different approaches and isolate the active components of interventions that might facilitate stronger outcomes and superior cost benefits [
50].
Longitudinal research on more distant parameters might further emphasize the societal relevance of easily accessible universal prevention. Better mental health has been associated with improved outcomes in a range of other domains, such as physical health, health behaviors, education, and earnings or crime reductions (e.g. [
67]), which are mostly not assessed in cost-effectiveness analyses due to the time lag of effects. The societal benefit of health promotion might thus be underestimated in shorter-term studies, such as ours.
Furthermore, this study adds to the growing evidence that trained laypersons with no prior professional mental health training can effectively be involved in health promotion and treatment of subsyndromal and mild mental disorders. The role of these non-specialist health workers is that of a coach, following structured intervention protocols, as opposed to a traditional therapist role [
68]. Non-specialist health workers are of particular importance in low- and middle-income countries where human resources for mental health are scarce (e.g. [
69]), but can also contribute to the affordability of mental health promotion in high-income countries by reducing costs for highly trained professionals. Recent research on the British “Improving the Access to Psychological Therapies” program shows a complex, non-linear relationship between non-specialist health workers’ competence and patient outcome [
70]. More research is needed to investigate the service of health workers in preventive interventions in industrialized countries.