Mental health economics: A prospective study on psychological flourishing and associations with healthcare costs and sickness benefit transfers in Denmark

Background: Escalating healthcare expenditures highlight the need to identify modifiable predictors of the use and costs of healthcare and sickness benefit transfers. We conducted a prospective analysis on Danish data to determine the costs associated with flourishing as compared to the below threshold level of flourishing. Methods: We used data from a 2016 Danish survey of 3508 adults, which was linked to Danish register data. Flourishing was assessed with a validated psychological well-being scale. A two-part regression model was used to predict 2017 costs while adjusting for 2016 costs, demographic variables, and health status, including psychiatric morbidity and health behaviours. Costs are expressed in USD PPP. Results: Applying criteria from prior literature, the prevalence of flourishing in Denmark (measured in 2016) was 34.7%. Flourishing was associated with significantly lower healthcare costs ($-687.7, 95% CI = $-1295.0, $-80.4) and sickness benefit transfers ($-297.8, 95% CI = $-551.5, $-44.0) per person in 2017, as compared to the below threshold level of flourishing. Extrapolated to the Danish population (4.1 M people aged 16+ years), flourishing was associated with lower healthcare costs and sickness benefit transfers amounting to $-1.2bn (95% CI = $-2.3 bn, $-149.0 M). Conclusions: Flourishing is associated with considerably lower health-related government expenditure. Substantial reductions could potentially be achieved by increasing the number of people who are flourishing in the population.


Introduction
In Denmark, public healthcare expenditure, at constant prices, has increased 46% from 2000 to 2017 (Rasmussen & Kristensen, 2019). In 2017, healthcare accounted for 16.4% of all government expenditure (Dam, 2019). With escalating healthcare expenditures, there is a pressing need to identify factors that may reduce costs -in particular, modifiable factors that are predictive of health and disease, and by extension, healthcare utilization and costs. Well-being is one such modifiable predictor, with previous research suggesting that enhancing population levels of well-being is economically worthwhile in the short as well as the long term (Knapp et al., 2011;Nurse et al., 2014;Kim et al., 2021;Trudel-Fitzgerald et al., 2019). According to a number of reviews (Kim et al., 2021;Trudel-Fitzgerald et al., 2019;De Neve et al., 2013;Diener & Chan, 2011;Diener et al., 2017;Lyubomirsky et al., 2005;Walsh et al., 2018), states of elevated psychological well-being have been found to be protective of physical health, as well as mental health and longevity-for example, through improved cardiovascular, immune, and endocrine system functioning; reduced risk of heart disease, stroke, and infection; better health behaviours, as well as enhanced resilience and recovery from illness. All of these mediating factors may curb healthcare expenditure.
Researchers have become increasingly interested in the concept of psychological flourishing, a central concept in well-being science that represents optimal levels of psychological well-being. A person may be considered to be flourishing when well-being is maximized within a number of domains (Keyes, 2002;Huppert, 2009;Ryff & Singer, 1998). Although aspects of well-being have been studied individually with a focus on happiness (Lyubomirski et al., 2005), life satisfaction (Diener, 1989), or subjective well-being (Diener & Chan, 2011), psychological flourishing is a state that requires several aspects of one's psychological state to be good. More specifically, flourishing includes a combination of hedonic or "feeling good" facets (e.g. happiness and vitality) and eudaimonic or "functioning well" facets (e.g. purpose in life, competence, and social connectedness) of well-being (Keyes, 2002;. In a similar way that a diagnosis of clinical depression requires a combination of symptoms of anhedonia and malfunctioning, the determination of flourishing requires a combination of hedonia (i.e. feeling good) and functioning well in life. This combination is important because many individuals may have high levels of one componente.g., high hedonic well-being (positive feelings) -but lower levels of eudaimonic well-being (e.g., low purpose in life) or vice versa. Individuals who may be quite happy with life, but that are not functioning well have worse outcomes, such as higher rates of mental illness, compared to individuals who are flourishing (feeling good about life and functioning well) (Keyes & Annas, 2009). Flourishing has been found to be protective against premature mortality and morbidity, and is therefore a good candidate for reduced healthcare expenditure and other health-related social costs (e.g. costs relating to absence from work due to sick leave) (Keyes, 2015;Keyes & Grzywacz, 2005;Keyes & Simoes, 2012).
Eight studies have previously investigated the relationship between well-being measures and healthcare utilization/expenditure (Keyes & Grzywacz, 2005;Shi et al., 2013;Sears et al., 2013;, Harrison et al., 2012;Gandy et al., 2014;Roy et al., 2019;Riley et al., 2018). Seven of those were studies from the U.S. and based on self-reported healthcare utilization or health insurance expenditure. Across all these studies, well-being was negatively related to (1) healthcare utilization and expenditure (Keyes & Grzywacz, 2005;Shi et al., 2013;Sears et al., 2013;, Harrison et al., 2012;Gandy et al., 2014;Roy et al., 2019;Riley et al., 2018), (2) productivity loss (Keyes & Grzywacz, 2005;Shi et al., 2013;Sears et al., 2013), and (3) job turnover Sears et al., 2013). Most studies were non-prospective studies based on samples that were not randomized Gandy et al., 2014;Roy et al., 2019). Two studies included samples that were nationally-representative but were not prospective studies (Keyes & Grzywacz, 2005;Riley et al., 2018), and two were prospective studies but not based on randomized samples Harrison et al., 2012). More recently, we conducted a study on Danish population-based register data, in which we showed that a continuous measure of well-being (the Warwick-Edinburgh Mental Well-Being Scale -WEMWBS) in 2016 inversely predicted healthcare costs and sickness benefit transfers (note: the Danish government compensates long-term absence from work due to illness) in 2017 (Santini et al., 2021). The present study is similar in design, but it extends the previous work by estimating the potential reductions in costs associated with states of elevated well-being (i.e., psychological flourishing).
To the best of our knowledge, no studies to date have reported a comprehensive analysis of register-based data with flourishing as a predictor. Such studies are strongly needed to advance the field regarding the costs associated with flourishing and non-flourishing individuals in a population. This is important because whilst the European Commission and World Health Organization generally support the need for programs and policies to promote mental health and well-being (including flourishing), there is a research gap in terms of the potential return on investment that might be gained from successfully implementing such programs and policies (Forsman et al., 2015;EU, 2019;WHO, 2005). Moreover, it is important to analyze the associations between flourishing and economic outcomes from contexts other than the U.S., which have no federal statutory requirement for paid sick leave, and where even after the Affordable Care Act, 9.2% of adults still do not have health insurance coverage (based on data from 2019) (Keisler-Starkley & Bunch, 2019). In Denmark, all citizens have access to the publicly funded healthcare system, and long-term absence from work due to illness is compensated by the government. Denmark is therefore an ideal setting for investigating associations between flourishing and public healthcare expenditure.
In this study, using a large random sample of the adult Danish population, we set out to investigate to which extent flourishing is associated with lower government expenditure-specifically, healthcare costs and sickness benefit transfers. Based on the aforementioned evidence, we hypothesized that flourishing (as compared to the below threshold level of flourishing) would be associated with lower healthcare costs and sickness benefit transfers in the following year. A secondary objective of the study was to compare flourishing prevalence proportions in Denmark in 2016 with two previous time points (2006 and 2012). The rationale for this objective was to monitor the degree of change in the prevalence of psychological flourishing in Denmark over the period spanning the years 2006 to 2016. If flourishing is shown to be associated with lower expenditure, it is relevant to investigate prevalence rates of flourishing in previous years.

Sampling
Our data came from The Danish Mental Health and Well-Being Survey 2016 (DMHWBS2016) (Nielsen et al., 2017), which is a random population-based sample of Danes aged 16 years and above. The Danish government agency Statistics Denmark sent an electronic letter to 10,250 sampled individuals in October 2016 with information about the study and an invitation to participate. A total of 3508 people responded to the web-based survey (between October 18, 2016 andNovember 13, 2016), resulting in a response rate of 34%. Additionally, the survey was linked to the Danish Civil Registration System (Pedersen, 2011) via Statistics Denmark, which allows for the merging of data on employment status, household income, healthcare utilization, and social service use, among other data. Each citizen in Denmark has a personal registration number, enabling linkage between registers . All data are pseudonymized, so they cannot be traced back to specific participants. There is no formal agency for ethical approval of questionnaire-based survey studies in Denmark. The study complies with the Helsinki 2 Declaration on Ethics and is registered with the Danish Data Protection Authority; all confidentiality and privacy requirements were met. The participants' voluntary completion and return of the survey questionnaires constituted implied consent. To compare flourishing prevalence rates for Denmark between 2016Denmark between , 2012Denmark between , and 2006e, our secondary objective), we also used data from the European Social Survey (ESS) rounds 3 (2006) and 6 (2012). More information about these two ESS modules is reported in Appendix 1 (Table  A1.1) and elsewhere . Descriptives of the current sample and those reported previously for the ESS samples  indicate that the sample characteristics across these surveys were broadly matched .

Outcome: healthcare costs and sickness benefit transfers
All costs were extracted from Statistics Denmark for years 2016 and 2017. This cost analysis utilized data from Danish national registers using each respondent's anonymized civil registration number linked to DMHWBS2016. Costs comprised (1) healthcare costs (general practitioners/specialists, hospitalizations, outpatient services, prescription medicines), and (2) sickness benefit transfers (compensations made by the Danish government for long-term absence from work due to illness). Unit costs for general practitioners and specialists are based on the current national health insurance rate . Charges based on Diagnostic Related Groups (DRG) were used as unit costs for both costs of hospitalizations and outpatient services. Costs were omitted for healthcare services that did not involve treatment for illnesses, such as health services for contraceptive management (ICD-10 codes Z30) and other circumstances (ICD-10 codes Z76). For the costs of prescription medicines, public expenditure was calculated by subtracting user payments from the retail price of the medicine.
All healthcare costs were aggregated for: (1) the full calendar year 2016 and (2) the full calendar year 2017. Sickness benefit transfers (long-term absence, 31+ days) were estimated based on the weekly number of hours absent from work and respective salaries . Sickness benefit transfers were also aggregated for: (1) the full year 2016 and (2) the full year 2017.
For a detailed description of cost components, see Appendix 2 (Table  A2). All costs used for analysis are in 2016/2017 prices (DKK) and results were subsequently converted to international dollars (United States Dollars -USD, Purchase Power Parity -PPP) using an online conversion tool (2017 rates for price and target year, PPP values from the International Monetary Fund, 1DKK=USD$0.13 PPP) (EPPI-Centre Cost Converter, 2021).

Predictor: flourishing
We used a psychological flourishing scale proposed by  as the predictor in this study. According to their conceptualization, flourishing involves features pertaining to three domains: (1) positive characteristics (5 items comprising emotional stability, vitality, optimism, resilience, and self-esteem), (2) positive functioning (4 items comprising engagement, competence, meaning, and positive relationships), and (3) positive emotion (1 item comprising happiness). Table 1 presents the items in the flourishing scale. The flourishing scale has previously been validated across different European regions (including Scandinavia) with acceptable model fit , and research performing cross-tabulation analysis has reported moderate agreement with other well-known operational definitions of flourishing (Hone et al., 2014).
Flourishing in this study was considered a state (as a binary variable) (Hone et al., 2014), where a person is flourishing when he or she meets the criteria of having the combination of (1) at least four out of five features pertaining to positive characteristics, (2) at least three out of four features pertaining to positive functioning, and (3) positive emotion. It was coded as follows: For the items pertaining to positive characteristics, we categorized a feature as present when the participant responded 'agree' or 'strongly agree' on a 5-point Likert scale with "strongly disagree" at the opposite end (for optimism, self-esteem, and resilience), or responded "most of the time" or "all or almost all of the time" on a 4-point Likert scale, with "none or almost none of the time" at the opposite end (for vitality and emotional stability). Note that the resilience item (under the positive characteristics domain) was negatively worded, and was therefore scored inversely. For the items pertaining to positive functioning, a feature was coded as present when a participant responded "agree" or "strongly agree" on a 5-point Likert scale, with "strongly disagree" at the opposite end (for competence, meaning, positive relationships, and engagement). For the items pertaining to positive emotion, we categorized the feature as present when a participant rated his or her happiness 8, 9, or 10 on a 10-point Likert scale, with 1 being "extremely unhappy" and 10 being "extremely happy." As a result of coding the scale according to the aforementioned criteria, we ended up with a binary variable-that is, either flourishing or below the threshold level of flourishing. Cronbach's alpha for the items included in the flourishing scale was 0.87. For the ESS data, the criteria for flourishing was the same, more information can be found in Appendix 1 and elsewhere .

Covariates
All sociodemographic variables were extracted from Statistics Denmark for the year 2016. The sociodemographic variables were as follows: age; sex (female, male); migration background (not immigrant; immigrant or descendent of an immigrant); marital status (married or registered partnership; divorced, terminated partnership or widowed; single); education (primary or unknown; high school; tertiary education); employment status (employed; unemployed; student; retired; social pension/early retirement; other -employment status not defined); and income (lowest quartile; second-lowest quartile; second-highest quartile; highest quartile). Because 6.8% of data on income were missing, a "missing" category was created for this variable.
Three variables pertaining to health status and health behaviours were included. The first twochronic and psychiatric conditionscame from the national registers. To classify the presence of chronic  1, 2, 3, 4, 5, 6; 7, 8, 9, and 10 (extremely happy). The criteria used for flourishing was having the combination of: (1) At least four out of five features pertaining to positive characteristics (present: 'agree/strongly agree' or 'most of the time/all or almost all of the time') (2) At least three out of four features pertaining to positive functioning (present: 'agree/strongly agree') (3) Positive emotion (present: 8, 9, or 10).
conditions, we used the Charlson Comorbidity Index (CCI). This is based on 19 different medical conditions, each weighted and assigned 1-6 points according to its potential impact on mortality, derived from relative risk estimates . The CCI score is then categorized into three comorbidity levels: low (CCI=0), medium (CCI=1-2), high (CCI≥3). To assess the presence of psychiatric conditions, a variable was created for any psychiatric or self-harm diagnosis based on the ICD-10 codes F00-F99 (mental and behavioural disorders, including disorders relating to psychoactive substance use) and X60-X84 (intentional self-harm), i.e. a binary variable was constructed (any psychiatric condition, no psychiatric condition). Finally, level of physical activity was assessed with a binary variable (weekly or more, less than weekly), obtained from the survey. The sociodemographic and health status/behaviour variables were included because they are associated with mental well-being and healthcare service utilization Harrison et al., 2012;Riley et al., 2018;Stewart--Brown et al., 2015;Diehr et al., 1999).

Statistical analysis
STATA version 14 was used to perform all analyses. Following recommendations regarding the use of healthcare cost data, we applied a two-part model (Mihaylova et al., 2011;Manning & Mullahy, 2001) with 2017 costs as the outcome. The two-part model made adjustment for the covariates described below and also for costs in 2016. In the first part of the model, the probability of incurring any expenditure was estimated by a probit model using the full sample. Then, a generalized linear model (GLM) with log link and a gamma distribution was fitted to the subset of people that had any expenditure in 2017. More formally, the model can be written as the product of expectations from the first and second parts of the model, as follows: Thus, the two-part model allows for a separate investigation of the effects of a predictor on the extensive margin (probit model, if any expenditures) and on the intensive margin (GLM, amount of expenditures, if any). Subsequently, the incremental effects of the predictor on the outcome for the combined probit and GLM were calculated (Deb & Norton, 2018).
The statistical analyses conducted were as follows. First, flourishing prevalence estimates were calculated for Denmark in 2016, and compared to prevalence estimates in 2012 and 2006 using two-sample tests of proportions. Subsequently, means (unadjusted) were computed to compare costs in 2017 between flourishing categories. Finally, we estimated the costs (2017 costs controlling for 2016 costs and other covariates) associated with flourishing as compared to the below threshold level of flourishing. Using 1-year time periods for the assessment of healthcare costs (total expenditures 12-months post baseline assessment as well as 12-month expenditure as a baseline covariate) is a common approach in economic modelling of longitudinal healthcare utilization and cost data Harrison et al., 2012;Ruhl et al., 2017;Curcio et al., 2019;Wooldridge, 2009), along with the use of two-step models in such analyses (Harrison et al., 2012;Ruhl et al., 2017).
For each analysis estimating costs, Model 1 adjusted for age, sex, migration background, marital status, education, income, and employment status (except not adjusting for employment status in models estimating sickness benefit transfers) and costs in 2016, while Model 2 adjusted for all the aforementioned variables as well as chronic conditions, the number of psychiatric conditions, and physical activity. Both of these two-part models were performed using the categorical flourishing variable as the predictor (reference category=below threshold level of flourishing). We performed the Model 2 analyses in order to avoid inflated results due to differing levels of physical or mental health problems. However, since we already adjusted for prior healthcare costs, this also ran the risk of overadjustment. Hence, we performed both models as a means to be able to compare the results and make sure that they were not markedly different from each other.
Finally, based on population data from Statistics Denmark (2021), statistically significant results were extrapolated to the entire Danish population (aged 16+ years) and expressed in international dollars (USD PPP). All extrapolated estimates were based on model 2 results. Based on 95% confidence intervals, we also generated extrapolations based on a lower bound estimate (based on the 95% CI lower bound limit), and an upper bound estimate (based on the 95% CI upper bound limit).

Results
All results that include costs were converted (from DKK) and are presented in USD PPP in the main tables; the original results expressed in the Danish currency DKK are shown in Appendix 4 (Table A4.1, A4.2, A4.3, A4.4). Information regarding the sociodemographic distributions of the study sample is shown in Table 2. The mean age of the study population was 52.1 years, with 54.2% of the participants being female.
In 2016, 34.7% (95% CI = 33.1%, 36.3%) of the Danish population was estimated to be flourishing (See Table 2). The estimated prevalence was significantly lower than a prevalence of 55.8% (95% CI = 53.4%, 58.2%) in 2012 (p < 0.001), and a prevalence of 52.7% (95% CI = 50.2%, 55.3%) in 2006 (p < 0.001) (not shown in tables). The prevalence estimates for 2012 and 2006 were not significantly different from each other (p = 0.090). Unadjusted mean costs in 2017 by flourishing categories are shown in Table 3. As can be seen, healthcare costs and sickness benefit transfers were consistently lower as compared to the below threshold level of flourishing.
For the analytical statistics, only model 2 results are reported here in text (both model 1 and model 2 results are shown in Table 4). Table 4 shows the adjusted value of costs associated with the flourishing category in comparison to the below threshold level of flourishing. Flourishing was associated with significantly lower healthcare costs ($-687.7, 95% CI = $-1295.0, $-80.4) and sickness benefit transfers ($-297.8, 95% CI = $-551.1, $-44.0) per person in 2017, as compared to the below threshold level of flourishing. Table 5 shows the extrapolation of results to the Danish population in 2017. Extrapolated to the Danish population (population size of 4.1M people aged 16+ years) (see Appendix 3, Table A3), flourishing was associated with lower healthcare costs and sickness benefit transfers amounting to $-1.2bn (95% CI = $-2.3bn, $-149.0M) as compared to the below threshold level of flourishing.

Discussion
Our results suggest that 34.7% of the Danish population was flourishing in 2016, a prevalence that appears to have declined significantly when compared to ESS prevalence estimates in 2016 and 2012. The objective of this study was to estimate the costs associated with flourishing as compared to the below-threshold level of flourishing. In line Note. Data are unweighted n (weighted %) unless otherwise specified. a All costs are in $PPP. Zero-costs were omitted   with previous findings, our results confirmed our expectations that flourishing would be significantly associated with lower costs in the year following the baseline survey. Four prior studies are worth considering in the context of our findings. Keyes & Grzywacz (2005) conducted a cross-sectional study on American data to assess the association between complete health (including flourishing based on Keyes' flourishing scale) and healthcare utilization. They found that those not having complete health were significantly more likely (OR = 2-4) to utilize healthcare services (work injury, overnight hospitalization, medical visits due to physical or mental health problems) as compared to those fulfilling criteria for complete health. This study was the only one of the four to use a measure of flourishing. Both Harrison et al. (2012) and Sears et al. (2013) used American insurance data to assess the association between baseline well-being scores (a continuous measure ranging from 0 to 100) and healthcare expenditure 12-months post baseline well-being assessment. They found that each increase in well-being was associated with significantly lower future healthcare costs (prescription and medical costs). More recently, Santini et al. (2021) conducted a similar study on Danish data, assessing the association between baseline well-being (a continuous measure ranging from 14 to 70) and healthcare costs/sickness benefit transfers in the following year. The results showed that each point increase was significantly associated with lower healthcare costs (US$ 42.5) and lower sickness benefit transfers (US$ 23.1). In the Present study, after adjusting for a wide range of covariates (including 2016 costs, psychiatric conditions, chronic illnesses and health behaviour), we found that flourishing was significantly associated with lower healthcare costs ($-687.7, 95% CI = $-1,295.0, $-80.4) and lower sickness benefit transfers ($-297.8, 95% CI = $-551.1, $-44.0) per person in 2017, as compared to the below threshold level of flourishing. These are particularly strong findings, given that all models adjusted for costs in the previous year, since these are known to be highly correlated with future costs Harrison et al., 2012). Extrapolated to the Danish population (population size of 4.1M people aged 16+ years), flourishing was estimated at $-1.2bn (95% CI = $-2.3bn, $-149.0M) in healthcare costs and sickness benefit transfers. Additionally, we follow recommendations to assess the implications of each limit within our confidence intervals (Amrhein et al., 2019). According to these estimates, the lower costs associated with flourishing may range from $149.0M to $2.3bn. Both of these values and all values inside this interval are reasonably compatible with our data. Notably, even the lowest value within this interval is substantial and warrants attention. In 2017, the Danish government spent a total of $24.5bn on the healthcare sector (Dam, 2019). If the estimated $1.2bn in reduced costs could have been achieved by moving a large segment of the population to a flourishing state (this scenario being strictly hypothetical), the Danish government would have had the freedom to potentially allocate these financial resources to other priorities within or beyond the healthcare system. Similarly, funds that would otherwise have been transferred to workers due to sickness absence could also have been used for different welfare budgets or other purposes.
Considering the large differences in expenditure between those who flourish and those below the threshold level of flourishing, the potential return on investment for programs and policies that promote mental health and well-being would be sizable and financially well worth the effort. Policy and research priorities formed by the European Commission and the World Health Organization support the view that a focus on promoting mental health and well-being is crucial for long-term growth and sustainable development (Forsman et al., 2015;EU, 2019;WHO, 2005). Our results add to a growing evidence base (Knapp et al., 2011;Nurse et al., 2014) suggesting that increasing the population prevalence of flourishing individuals could have the potential added benefit of curbing care costs for physical as well as mental health problems. That said, it should be noted that the results of the present study show that flourishing predicts future healthcare expenditure, but we cannot make firm inferences regarding directions of causality. For example, we cannot exclude the possibility that in the years prior to the survey, healthier lifestyles could have predicted both flourishing and a reduced need for future healthcare services. Although there is evidence from the U.S. that change in well-being (as a result of intervention) is related to reductions in healthcare expenditure , future intervention research, especially within a European setting, is strongly warranted to provide causal evidence.
The prevalence of flourishing in 2016 was 34.7%, which is substantially lower than 55.8% in 2012 and 52.7% 2006. It is possible that the decline in flourishing prevalence over this period could have contributed to escalating healthcare expenditure over the same period. It may also be noted that the prevalence does not appear to have declined due to the global financial crisis in 2007-2008; in fact, our data revealed a slight, albeit non-significant, increase in the prevalence of flourishing from 2006 to 2012. This trend is in line with previous research suggesting that while low levels of well-being and mental health problems are strongly associated with socioeconomic factors, the same pattern is not observed for high levels of well-being (Stewart--Brown et al., 2015;Kahneman & Deaton, 2010;Santini et al., 2020;Van Lente et al., 2012;Nielsen et al., 2016;Solin et al., 2019).
Promoting mental health and well-being is not the same as preventing poor mental health or mental illness (although the two are related and overlap). Whereas the prevention of poor mental health and illness generally focuses on reducing risk factors for these conditions (e. g., combating poverty or bullying), promoting mental health and wellbeing focuses on strengthening protective factors and resources for thriving and flourishing (e.g., boosting self-efficacy or encouraging gratitude, meaning, and purpose in life). To exemplify the difference: Being poor or bullied increases the risk of mental health problems or mental illness, and hence ensuring social safety and implementing antibullying initiatives are important preventative measures. But not being poor or bullied will not in itself result in good mental health and wellbeing, let alone flourishing. Therefore, initiatives with a specific focus on promoting mental health and well-being are also needed. In light of our results, such initiatives may not only promote mental health and well-being, but may also be beneficial from a financial perspective (Santini et al., 2021). Protective factors and resources for mental health and well-being should be promoted at several levels-namely, individual, community, and societal-as there will always be an interplay between these levels. Various research studies have explored social, recreational and lifestyle factors that are predictive of high levels of well-being and flourishing (Lyubomirsky et al., 2005;Santini et al., 2020;VanderWeele, 2020VanderWeele, , 2017Barry, 2009;Kalra et al., 2012;Regan et al., 2016;Lyubomirsky & Layous, 2013;Sin & Lyubomirsky, 2009;Bolier et al., 2013). Societal and policy efforts are important in promoting flourishing, such as strengthening social capital and community coherence. Individual activities can also make important contributions, such as keeping active in various ways, connecting with others, and engaging in meaningful causes or challenges VanderWeele, 2020;Lyubomirsky & Layous, 2013;Sin & Lyubomirsky, 2009;Bolier et al., 2013). From a health economic perspective, taking the step further would imply considering the cost-benefit of any action that could promote flourishing (presumably alongside other relevant outcomes) (Knapp et al., 2011;Nurse et al., 2014).
Some strengths and limitations should be kept in mind when interpreting the results. Major strengths include the prospective design, the use of a validated scale for measuring flourishing, and the use of a population-based survey linked with national registers. This approach made it possible to make direct links between flourishing in one year and cost outcomes expressed in monetary terms in the subsequent year, as well as a range of register-based covariates. Some limitations to consider include: first, the relatively low (34%) response rate to the survey; although this is not unusual for web-based surveys, selection bias cannot be ruled out. Second, response rates were higher for women, individuals aged 45 years old and above, individuals with higher (tertiary) education, individuals who were married or in a registered partnership, employed individuals, individuals with a non-immigrant background, and individuals with higher incomes (Nielsen et al., 2017). We applied weights in all analyses, which to some extent reduced the statistical uncertainty related to selection bias. Third, in terms of the criteria used for flourishing, it may be considered that the distinction between hedonic and eudaimonic well-being is not universally agreed upon (Kashdan et al., 2008), and some scholars have argued that eudaimonic well-being tends to predict hedonic well-being, rather than the two simply reflecting different facets of well-being (Sheldon, 2018). Fourth, we were not able to restrict healthcare cost outcomes to those pertaining to mental health care only, both due to data restrictions and due to the fact that some mental health care in Denmark is provided through primary care or other care services outside the psychiatry. Hence, we were not able to tease out the differences between associations with general health care and mental health care specifically, and future research is needed to explore this. Last, it is worth noting that the results are not necessarily generalisable to other countries, especially those outside Scandinavia. Denmark is different from most countries in the world with regard to its healthcare system, economy, and social structure, which may affect how flourishing is associated with healthcare expenditure as compared to other country settings. As mentioned earlier, only one American study reported an inverse association between flourishing (as a component of complete health) and healthcare utilization (Keyes & Grzywacz, 2005), which is in line with our findings. However, healthcare utilization in the U.S. is heavily dependent on private or company health insurance, which may affect the relationship between flourishing/non-flourishing and healthcare utilization or expenditure (e. g. individuals may not be flourishing but may also not utilize healthcare services due to lack of insurance).
Because some relevant variables, such as alcohol consumption, smoking, and nutrition, were not available in our dataset, we could not adjust for these factors, although they may also be mediators. Furthermore, the flourishing assessment was made based on one particular conceptualization of flourishing, and conceptualizations that include other components, such as a greater emphasis on character (e.g., ability to delay gratification) and virtue (e.g., doing good for others) (Van-derWeele, 2017) may have produced different results. Importantly, flourishing here was specifically psychological flourishing. Some flourishing assessments include physical health (VanderWeele, 2017), which would have been problematic in this study given its focus on assessing the effects of flourishing on healthcare costs; thus, we restricted the assessment to psychological flourishing and adjusted for physical health in the analyses. Finally, there was some overlap between the time of the survey (October 2016) and the data on costs (2016 and 2017). In this study, our final results are based on analytical models that adjusted for health status and health behaviour (apart from demographics and socioeconomic factors). We did this to minimize the confounding of health factors; however, overadjustment is a possibility, as we already adjusted for past healthcare utilization, hence, we conducted two models with and without the adjustment for health factors. Finally, our results pertain to the included outcomes, but flourishing may also be related to numerous other costs (for more information, see Santini et al., 2021), for example, our results do not capture loss to productivity (both in regard to paid and unpaid work) due to long-term sick leave. Also, our results capture reductions in costs associated with flourishing over the short-term (one year from assessment), but not flourishing over the longer term (e.g. continuously flourishing over repeated assessments/years). Taking these things into account, our results may be considered conservative from a public health and economic standpoint.

Conclusion
The results of the present study lend support to and expand prior findings that flourishing predicts lower costs in terms of future healthcare expenditure and sickness benefit transfers. Flourishing was associated with lower healthcare costs ($-687.7, 95% CI = $-1295.0, $-80.4) and lower sickness benefit transfers ($-297.8, 95% CI = $-551.1, $-44.0) per person in 2017, as compared to the below threshold level of flourishing. Extrapolated to the Danish population (population size of 4.1M people aged 16+ years), flourishing was estimated at $-1.2bn (95% CI = $-2.3bn, $-149.0 M) in healthcare costs and sickness benefit transfers. Because these estimates cover the cost outcomes included in this study and refer to the potential reductions in short-term costs associated with flourishing, they must be considered conservative from a societal perspective. Even so, the lower costs associated with flourishing are substantial and warrant attention. Increasing the number of individuals who flourish in the population could potentially free up resources and reduce costs in the short term, as well as generate cost savings for society in the longer term.

Contributor statement
All authors have contributed to the work submitted.

Declaration of Competing Interest
No conflicts of interest declared.

Acknowledgments
None declared.

Ethics
This study is a secondary data analysis with no human subject issues. Ethics statement is included in the paper.

Data availability
We do not have permission to share data.

Funding
Nordea-fonden. No support from any organization for the submitted work; no financial relationships with any organizations that might have an interest in the submitted work in the previous three years, no other relationships or activities that could appear to have influenced the submitted work.

Transparency declaration
The manuscript is an honest, accurate, and transparent account of the study being reported. No important aspects of the study have been omitted. Any discrepancies from the study as planned (and, if relevant, registered) have been explained.

Appendix 1. The European social survey (ESS)
The European Social Survey (ESS) is a biennially repeated crosssectional investigation conducted in a wide range of European countries. We used data specifically from the well-being modules (Rasmussen & Kristensen, 2019;Dam, 2019), which were only included in the third and sixth round of the survey conducted in 2006 and 2012, respectively. The ESS selected participants using strict probability samples of the resident national population aged 15 or older living in private households. Data were gathered via face-to-face interviews with standardized questionnaires. Statistical data and comprehensive methodological documentation are freely available on the website of the ESS (www. europeansocialsurvey.org). The ESS subscribes to the Declaration on Professional Ethics of the International Statistics Institute (Knapp et al., 2011). According to this declaration, participants must be protected against potentially harmful effects of taking part in the survey. Hence, participation was based on participants' voluntarily given informed consent. For this analysis, we included only one of the available ESS countries at both rounds -namely, Denmark. The sample size for Denmark was 1,505 in 2006 (response rate 50.8%) and 1650 in 2012 (response rate 49.4). In the 2006 data, 4.8% of the flourishing data were missing, leaving a final sample size of 1433, while 3.9% of the flourishing data were missing in the 2012 data, leaving a final sample size of 1585. Because some items in the flourishing scale were changed from round three to round six of the ESS, items on positive relationships and engagement were not identical in both rounds. However, it is possible to arrive at almost identical flourishing scales for both rounds (Dam, 2019;Nurse et al., 2014). The flourishing scale has been validated using the ESS data, including measurement invariance testing across European regions (including Scandinavia) (Kim et al., 2021).

Table A2
Information on healthcare costs components.

Register-based cost studies
The tracking of individual contacts within the healthcare system over time and across institutions is made feasible by an anonymized unique personal identification number (CPR-number). A resident's CPR number is granted at birth or upon immigration and is included in all national registers. Statistics Denmark encrypts the CPR number before releasing data for research [1,2].

General Practitioners (GPs) and specialists
Unit costs for contacts with the primary care sector were estimated based on the fee paid by the government to the healthcare professionals (General Practitioners and specialists). The data on contacts with the primary care sector was stratified from the Danish National Health Service Register (NHSR)). The NHSR contains information on doctor and patient centered data. Data on services for GPs and specialists comprises more than 200 individually priced services. The prices are agreed upon between the Organization of General Practitioners and the Danish regions. Data on contacts with patients and type of service delivered are reported to the NHSR as all GPs are linked to a uniform computer system [1,3].

Hospitalizations and outpatient services
The National Patient Register (Danish: Landspatientregisteret (LPR)) was used to obtain the costs for hospitalization and outpatient treatment [4]. The LPR includes both administrative and clinical data. The administrative data are patient centered. As soon as a person has been examined or hospitalized, the hospital records a series of information about the patient: The person's CPR number, background information on causes leading to hospital contact, etc. Clinical data relates to diagnosis and treatment procedures. Here, the LPR adapts the International Classification on Diseases, version (ICD-10). The National Patient Register includes all full-time admissions, emergency room contacts, and outpatient contacts for each CPR number respectively. Each treatment for a similar condition is linked to a rate that represents the average (continued on next page) How happy are you a Response options were as follows: strongly agree, agree, neither agree nor disagree, disagree, and strongly disagree. b Response options were as follows: all or almost all of the time, most of the time, some of the time, and none or almost none of the time. c Response options were as follows: 0 (extremely unhappy), 1, 2, 3, 4, 5, 6; 7, 8, 9, and 10 (extremely happy).   cost of a treatment course. We used these Diagnosis Related Groups charges (DRG) for admissions and Danish outpatient charges (DAGS) for outpatient contacts as unit cost estimates [5]. Admissions related to normal births, sterilization or healthy companions (the system registers when a patient needs a companion during appointments) were not included.

Prescription medicines
The Danish National Prescription Registry (DNPR) was used to estimate the unit costs of prescription medicines. The DNPR holds information on prescription drugs acquired at Danish pharmacies. We applied the market price as a unit cost estimate [9]. Out-of-pocket payments were deducted from the total cost, thereby generating a cost outcome that pertains strictly to government healthcare expenditure.

Sickness benefit transfers (longterm)
The LPR was used to extract relevant information on sickness benefit transfers. The Danish register includes information on sickness benefit transfers (long-term) when individuals are absent from work due to sickness for at least 31 days, i.e. absence is not registered for periods less than 31 days (short-term). Sickness benefit transfers (long-term) are estimated based on the weekly number of hours absent from work and respective salary [6][7][8]. Analyses on sickness benefit transfers were only applied to the subgroup of the study sample who fall in the working age population (range of 16 to 64 years).  Note. DKK = Danish Krone (official currency of Denmark), 1DKK = USD$0.13 PPP. All extrapolations (see Appendix 3) are based on model 2 results that were statistically significant (p < 0.05). a The result was extrapolated to the Danish population of employed individuals aged 16-64 years old.