Socio-Economic Consequences of Mental Distress: Quantifying the Impact of Self-Reported Mental Distress on the Days of Incapacity to Work and Medical Costs in the Following 2 years: A Longitudinal Study in Germany

Background: Mental disorders are related to high individual suffering and signicant socio-economic burdens. However, it remains unclear to what extent self-reported mental distress is related to individuals’ days of incapacity to work and their medical costs. This study aims to investigate the impact of self-reported mental distress for specic and non-specic days of incapacity to work and specic and non-specic medical costs over a two-year span. Method: Within a longitudinal research design, 2,287 study participants' mental distress was assessed using the Hospital Anxiety and Depression Scale (HADS). Days of incapacity to work and medical costs were retrieved from a German health insurance’s routine data during the following two-year period. Results: Current mental distress was found to be signicantly related to the number of specic days absent from work and medical costs. Compared to participants classied as no cases by the HADS (2.6 days), severe case participants showed 27.3-times as many specic days of incapacity to work in the rst year (72 days) and 10.3-times as many days in the second year (44 days), and resulted in 11.4-times more medical costs in the rst year (2272 EUR) and 6.2-times more in the second year (1319 EUR). The relationship of mental distress to non-specic days of incapacity to work and non-specic medical costs is also signicant, but mainly driven from specic absent days and specic medical costs. Our results also indicate that the prevalence of presenteeism is considerably high: 42% of individuals continued to go to work despite severe mental distress. Conclusions: Our results show that self-reported mental distress, assessed by the HADS, is highly related to the days of incapacity to work and medical costs in the two-year period. Reducing mental distress by improving preventive structures for at-risk populations and increasing access to evidence-based treatments for individuals with mental disorders might, therefore, pay for itself and could help to reduce public costs. were obtained on non-specic costs in the rst year, but an increase was shown in non-specic medical costs for lower educated participants in the second year (χ 2 [3] = 43.12, p < .001). An opposite pattern was found concerning specic costs. Here, lower educated participants showed increased costs in the rst year (χ 2 [3] = 14.56, p = .002) but not in the second year. specic results demonstrate that mental distress impacts a person’s life for several years by predicting their sickness absence rates even two years later. This increased sickness absence rate might be, in turn, related to a generally reduced social and occupational functioning levels and reduced well-being of individuals Furthermore, mental distress appears to be a central challenge for employers in terms of productivity loss. The nancial consequences of specic DIW due to production loss can be calculated by multiplying the specic DIW by average income. Regarding average costs due to production loss in of 105 EUR per DIW [34], the averaged additional costs for an employee under severe mental distress due to absenteeism alone amount to 7,230 EUR in the rst years and 4,163 EUR in the second year, compared to an employee without mental distress. According to prior empirical ndings, the additional costs due to presenteeism can be estimated to be two to three times higher [31]. Our results revealed that 66% of participants classied as moderate cases and 42% of participants classied as severe cases, did not have any specic DIW in the two-year period that was analyzed. These results indicate that the percentage of people who go to work despite severe mental distress might be considerably high and illustrate the importance and spread of presenteeism. Given this high prevalence of presenteeism and the assumed adverse mental health outcomes, future studies should characterize this sub-sample's psychological and socio-demographic characteristics to better understand the risk factors for presenteeism. By doing so, a distinction should be made between whether work is perceived as a resource and thus contributes to the stabilization of mental health, or as a stressor that leads to the maintenance of high mental distress. Both the anxiety and depression subscales of the HADS were predictive for specic and unspecic DIW. This is not surprising since 80% of all specic DIW in our sample were caused by the diagnostic groups’ affective disorders (42%, e.g., depression) and neurotic,

A total of 34,207 policyholders of a large German health insurance company were contacted, 5,549 of whom declared their willingness to participate in the study. This corresponds to a response rate of 16%. The data could not be analyzed for 861 participants because (i) the questionnaire data was incomplete or DIW data was not available because of being insured with another health insurance company (n = 808), (ii) consent to participate in the study was withdrawn (n = 20), or (iii) the questionnaire was sent out twice (n = 23). The education variable "still in school" (n = 10) was excluded because the incomplete educational status could not be included in the ranking of the education factor. Of the remaining 4,688 insured persons, 1,329 insured persons were not included because they were not entitled to sickness bene ts (e.g., pensioners, family insured persons, rehabilitates, voluntarily insured persons not entitled to sickness bene t), and 1,072 because they belonged to the experimental group in the initial study. Finally, a total sample of 2,287 study participants were included in the analyses. For details, see Lyssenko et al. [18].

Mental distress
Cross-sectional mental distress was assessed with the Hospital Anxiety and Depression Scale (HADS) [21]. The HADS is a self-report questionnaire measuring anxiety and depressive symptoms with good psychometric properties [22]. The questionnaire consists of seven items for each of the two subscales. Total scores can be calculated for each subscale, ranging from 0 to 21 or an overall score for both subscales ranging from 0-42, which can be interpreted as a global screener of mental distress [23]. Higher values in the subscales indicate more severe anxiety or depressive symptoms. Based on the values in one of the two subscales, the degree of mental distress can be differentiated as no distress (0-7), mild distress (8)(9)(10), moderate distress (11)(12)(13)(14)(15), and severe mental distress (≥ 16) [24]. In addition, cut-off values are provided to distinguish between inconspicuous values and values requiring therapy. For the HADS, cut-off values apply to one of the two subscales of ≥ 8 (values ≥ 11 are considered conspicuous) [25,26]. Therefore, a need for therapy should be clari ed by further procedures, even at a low level of mental distress. A meta-analytic consideration showed an averaged sensitivity of 0.82 and speci city of 0.74 when applying a cut-off point of 8 and an averaged sensitivity of 0.56 and speci city of 0.92 when applying a cut-off point of 11 across different clinical samples [27]. In our sample, the HADS showed good reliability, with a Cronbach's α of 0.91.

Days of incapacity to work
The number of speci c and non-speci c DIW was selected from all study participants' routine health insurance data. As speci c DIW, the days of incapacity to work due to mental illness were selected (i.e., ICD 10, F00-F99, "mental and behavioral disorders", and ICD 10, Z73, "problems with regard to di culties in coping with life", including burn-out). As non-speci c DIW, all days of incapacity to work due to any ICD 10 diagnosis were selected. Both speci c and nonspeci c DIW were retrieved cumulatively during the rst and second year, starting on the day after the HADS assessment.

Medical costs
Direct speci c and non-speci c medical costs were retrieved from routine health insurance data for all study participants. The direct speci c health care costs of the diagnostic main group "mental and behavioral disorders" (ICD 10, F00-F99) and "problems related to di culties in coping with life (Z73 including accentuation of personality traits, being burnt out)" were determined for the cost elds of outpatient treatment, hospital (main diagnosis), and rehabilitation (admission diagnosis). The drug costs were composed of the costs for antidepressants (N06A), psycholeptics, and psychoanaleptics in combination (N06C), anxiolytics (N05B), and hypnotics and sedatives (N05C). The averaged direct speci c and non-speci c medical costs are available in Euro and were retrieved cumulatively during the rst and second year, starting on the day after the HADS assessment.

Socio-demographic data
The socio-demographic characteristics of the sample age, gender, and employment status were retrieved from the routine health insurance data. The questionnaires also assessed education and marital status.

Statistical analyses
The DIW and the cost data represent a mixture of a discrete distribution with positive probability mass at zero (many participants do not have any DIW) and a continuous distribution for data > 0 and represent a left-skewed distribution. This form of mixed distribution is known as the Tweedie or Poisson-gamma distribution [28,29]. Accordingly, a generalized linear model with a Tweedie distribution with log link function was calculated with mental distress (HADS) as an independent variable and the speci c and non-speci c DIW in the rst and second year after the HADS assessment as dependent variables. Since prior studies have shown an effect of socio-demographic variables on DIW [30][31][32], we include our sample's socio-demographic variables as control variables in the model (age, gender, education, and relationship and employment status). The analyses were performed with SPSS 26.

Results
The sample of 2,287 participants consisted of 89% women and averaged 46.1 years of age (SD = 10.4). Most (72.4%) of the study participants were married. The percentage of participants holding an A-Level degree was 25.7%, whereas 2.5% had no school-leaving certi cate. A small group of participants (0.5%) became unemployed by losing their job during the study period. Regarding mental distress, 47.6% of study participants were classi ed as no case, 24.1% were classi ed as a mild case, 23.5% were classi ed as moderate cases, and 4.7% were classi ed as severe cases, by applying the proposed cut-off values to their HADS scores. Socio-demographics of the sample are depicted in Table 1.
However, 79.6% of our sample had no speci c DIW in the two years that followed the HADS testing. Based on the sample's HADS scores, the percentage of participants who did not have any speci c DIW was 89% for no cases, 82% for mild cases, 66% for moderate cases, and 42% for severe cases. Regarding nonspeci c DIW, 21% of our sample had no non-speci c DIW in the following two years. Based on the sample's HADS scores, the percentage of participants who did not have any non-speci c DIW was 27% for no cases, 20% for mild cases, 13% for moderate cases, and still 8% for severe cases.
Impact of socio-demographic variables Gender Gender revealed no signi cant differences in non-speci c and speci c DIW in the rst and second years (Tables 2 and 3). Accordingly, no differences in nonspeci c medical costs were obtained between male and female participants (Table 4). However, speci c medical costs were signi cantly increased for female participants in the second year (

Marital status
Marital status revealed no signi cant differences in non-speci c and speci c DIW in the rst or second years. Accordingly, no differences in non-speci c medical costs were obtained between married and unmarried participants. The speci c medical costs, however, were signi cantly increased in unmarried participants both in the rst year (χ 2 [1] = 30.56, p < .001, factor 1.5) and the second year (χ 2 [1]=12.51, p < .001, factor 1.3).

Education
Lower educated participants showed a signi cant increase in non-speci c DIW. Compared to participants holding an A-Level degree, lower educated participants showed 1.2 to 2.2 as many non-speci c DIW in the rst (χ 2 [3] = 33.59, p < .001) and in the second year (χ 2 [3] = 54.60, p < .001). The number of speci c DIW was also increased for lower educated participants, however, this only yielding signi cance in the second year. Lower educated participants showed in the rst year 1.2 to 2.0 as many speci c DIW (χ 2 [3] = 3.46, p = .326) and in the second year 1.4 to 3.3 as many speci c DIW (χ 2 [3] = 11.22, p = .011). Regarding the medical costs, no differences were obtained on non-speci c costs in the rst year, but an increase was shown in non-speci c medical costs for lower educated participants in the second year (χ 2 [3] = 43.12, p < .001). An opposite pattern was found concerning speci c costs. Here, lower educated participants showed increased costs in the rst year (χ 2 [3] = 14.56, p = .002) but not in the second year.

Employment Status
Employment status revealed no signi cant differences in non-speci c and speci c DIW in the rst and second years. However, the non-speci c costs of unemployed participants were signi cantly increased. Compared to employed participants, unemployed participants showed 2.0-times as many non-speci c medical costs (χ 2 [1] = 8.99, p < .003) in the rst year and 1.7-times as many non-speci c medical costs (χ 2 [1] = 4.07, p = .044) in the second year. Speci c medical costs of unemployed participants were descriptively increased (1.3 to 1.5), but this was not signi cant.

Impact of self-reported mental distress
The impact of HADS severity scores on speci c and non-speci c DIW and medical costs in the rst and second years after HADS assessment is depicted in Figure 1.

Days of Incapacity to Work
Mental distress, as measured with the HADS at t0, showed a signi cant effect on the number of non-speci c DIW in the rst year (χ 2 [3] = 320.78, p < .001; see Table 2) and the second year (χ 2 [3]  This increase of non-speci c DIW was mainly driven by an increase of speci c DIW in the rst year (χ 2 [3] = 196.98, p < .001) and the second year (χ 2 [3] = 106.21, p < .001; see Table 3): While participants, classi ed as no cases averaged 2.6 speci c DIW in the rst year (4.3 in the second year), mild case participants averaged 2.8 (1.5) times as many speci c DIW (M = 7.4 days in the rst and M = 6.2 days in the second year), moderate case participants The increases of DIW in the rst and second years were obtained on both the anxiety and depression subscales (all p values < .001).

Medical Costs
The increase of DIW depending on mental distress was also re ected in higher non-speci c medical costs in the rst year (χ 2 [3] = 164.41, p < .001; see Table 4) and the second year (χ 2 [3] = 110.45, p < .001). Again, the number of non-speci c medical costs increased continuously according to the degree of mental distress: While participants classi ed as no cases averaged 2031 EUR non-speci c medical costs in the rst year (2260 EUR in the second year), mild case participants had 1.2 (1.2) times as many non-speci c medical costs (M = 2330 EUR in the rst year and M = 2635 EUR in the second year); moderate case participants had 1.5 (1.5) times as many non-speci c medical costs (M = 3048 EUR in the rst and M = 3292 EUR in the second year); and severe case participants had 2.7 (1.9) times as many non-speci c medical costs (M = 5544 EUR in the rst and M = 5181 EUR in the second year).
The increase of non-speci c medical costs was again mainly driven by an increase of speci c medical costs in the rst year (χ 2 [3] = 496.56, p < 0.001; see Table 5) and the second year (χ 2 [3]  The increases in medical costs in the rst and second years were obtained on both the anxiety and depression subscales (all p values < .001).

Discussion
This study aimed to examine the impact of self-reported mental distress, assessed by the HADS, on the number of speci c and non-speci c DIW and medical costs in the two years following the testing. To address this aim, we conducted a longitudinal study, in which the HADS scores of 2,287 participants were used to predict their speci c and non-speci c DIW and medical costs in the rst and second years after HADS assessment.
Our results revealed that self-reported mental distress (HADS scores) was signi cantly related to the number of non-speci c DIW in the rst and second years.
Accordingly, the number of non-speci c DIW increased continuously based on the level of mental distress. Compared to the reference group classi ed as no cases, severe cases had 5.1-times as many non-speci c DIW in the rst year and 3.7-times as many non-speci c DIW in the second year. Not surprisingly, the increase of non-speci c DIW was mainly driven by a signi cant increase of speci c DIW. Compared to the no cases, severe cases showed 27.3-times as many speci c DIW in the rst year and 10.3-times as many speci c DIW in the second year.
These results demonstrate that mental distress impacts a person's life for several years by predicting their sickness absence rates even two years later. This increased sickness absence rate might be, in turn, related to a generally reduced social and occupational functioning levels and reduced well-being of individuals [33]. Furthermore, mental distress appears to be a central challenge for employers in terms of productivity loss. The nancial consequences of speci c DIW due to production loss can be calculated by multiplying the speci c DIW by average income. Regarding average costs due to production loss in 2014 of 105 EUR per DIW [34], the averaged additional costs for an employee under severe mental distress due to absenteeism alone amount to 7,230 EUR in the rst years and 4,163 EUR in the second year, compared to an employee without mental distress. According to prior empirical ndings, the additional costs due to presenteeism can be estimated to be two to three times higher [31]. Our results revealed that 66% of participants classi ed as moderate cases and 42% of participants classi ed as severe cases, did not have any speci c DIW in the two-year period that was analyzed. These results indicate that the percentage of people who go to work despite severe mental distress might be considerably high and illustrate the importance and spread of presenteeism. Given this high prevalence of presenteeism and the assumed adverse mental health outcomes, future studies should characterize this sub-sample's psychological and sociodemographic characteristics to better understand the risk factors for presenteeism. By doing so, a distinction should be made between whether work is perceived as a resource and thus contributes to the stabilization of mental health, or as a stressor that leads to the maintenance of high mental distress.
Both the anxiety and depression subscales of the HADS were predictive for speci c and unspeci c DIW. This is not surprising since 80% of all speci c DIW in our sample were caused by the diagnostic groups' affective disorders (42%, e.g., depression) and neurotic, stress and somatoform disorders (38%, e.g., anxiety disorders). This roughly corresponds to results from other studies in Germany, in which 88.6% of all speci c DIW resulted from affective (41.4%) or neurotic, stress, and somatoform disorders (47.2%) [32]. Accordingly, it seems plausible that both the anxiety and depression subscale of the HADS predicted the number of speci c DIW in our analyses. However, the impact of both HADS subscales for non-speci c DIW is in contrast to the results of Schneider et al. [30], in which only the anxiety symptoms, but not the depressive symptoms, were found to be a signi cant predictor of the duration of absences due to non-speci c DIW.
Beyond non-speci c and speci c DIW, our results demonstrated that mental distress is also signi cantly related to individuals' speci c and non-speci c medical costs in the rst and second year. Speci c costs in the rst year were 11.4-times higher for severe cases, compared to no cases. Even in the second year, severe cases showed 6.2-times as many speci c costs as no cases. This amounts to an additional average speci c cost of 2,073 EUR per person and year for severe cases in the rst year and 1,106 EUR per person and year in the second year for the public health care system. The predictive effect of nonspeci c costs was considerably smaller, but also signi cant. Compared to no cases, severe cases averaged 2.7-times the costs in the rst and 1.9-times the costs in the second year. This amounts to additional average non-speci c costs of 3,513 EUR per person and year for severe cases in the rst year and 2,922 EUR per person and year in the second year for the public health care system. These results underline the socio-economic burden of mental distress for public health care systems. However, they also show that this burden can be predicted by self-reported mental distress at an early stage, which opens the possibility for early interventions. Although these data represent costs from a German population, these results can be seen as an indicator for other industrialized countries, since both the prevalence of mental disorders (Germany 18%, EU 17.3%) and the percentage of direct and indirect medical costs due to mental illness in Germany (Germany 4.8%, EU 4.0%) are comparable to other EU countries [9].
Most demographic characteristics of our sample showed no consistent effects across the different dependent variables. However, these have been included mainly as control variables to control possible confounding variables. Future studies should speci cally focus on these variables to draw reliable conclusions about socio-demographic variables' in uence on absence days and medical costs. Only the participants' age showed a consistent pattern with increased nonspeci c DIW and non-speci c medical costs for both years, but no differences in speci c DIW and speci c medical costs. Lower education in our sample was signi cantly related to non-speci c DIW. However, on speci c DIW, the increase by lower education yielded signi cance only in the second year. These results are in line with prior studies showing that mental distress (anxiety symptoms), higher age, and lower education emerged as signi cant predictors of nonspeci c DIW [30]. Given these ndings, it seems likely that lower educational status and higher age can be considered a risk factor for non-speci c DIW. However, their effect on speci c DIW or medical costs remains uncertain. Future studies should include large and representative samples to investigate the differential effects of age and education on speci c and non-speci c DIW and speci c and non-speci c medical costs.
Contrary to prior studies, in which female gender was found to be a signi cant predictor of speci c DIW [31,32], our analyses showed no differences of speci c DIW between male and female participants. However, a closer look at the descriptive factors shows that the factors from our study (1.57) are comparable to those from previous studies (1.6) [31,32]. Therefore, the non-signi cant differences in DIW depending on the sample characteristics in our study could result from a too-small sample size in the different subgroups, thus limiting the power for individual comparisons. With 89%, the proportion of female participants was considerably high. Interestingly, female participants showed higher speci c medical costs in both years. This nding is in line with prior research, indicating a higher prevalence of anxiety and affective mental disorders in female populations [9,35].

Strengths, Limitations, And Recommendations For Future Research
Our study's major strengths relate to its longitudinal research design and the analysis of real DIW and medical cost data from a health insurance company in conjunction with psychometrically assessed mental distress from individuals. By including DIW and medical costs in the rst and second year, we were able to show that self-reported mental distress was predictive for DIW and medical costs regardless of the DIW and medical costs occurring immediately after the HADS assessment, and this enabled us to show the long-term consequences of severe mental distress. By including speci c and non-speci c DIW and speci c and non-speci c costs as dependent variables, we were able to show the importance of mental health for general, occupational functioning and point to the consequences of mental distress for companies and the public health care systems. Furthermore, the available cut-off scores of the HADS to distinguish between no, mild, moderate, and severe cases allowed us to demonstrate clear, practical implications for the consequences of severe mental distress in applied settings. However, this study has some limitations, which should be considered when interpreting the results.
First, we only investigated the main effects of the sample characteristics and mental distress. However, more complex interaction effects between the independent variables are conceivable and should be investigated in future studies using larger sample sizes. Second, although our overall sample size was reasonably large, it does not signify a representative sample of the German population. Thus, some socio-demographic subgroups might be too small, resulting in limited power for examining the relationship between different sample characteristics and DIW and medical costs, respectively (e.g., n = 11 participants in the unemployment group). Third, in addition to the sample characteristics analyzed in this study, other variables might impact the relationship between mental distress and DIW, such as the quality of health management in organizations [36], subjectively perceived workplace characteristics (e.g., social support, leadership quality; [37,38]), or inter-individual differences in psychological conditions, such as self-e cacy or work attitude [39]). In addition, variables should be investigated, in uencing the relationship between mental distress and medical costs, such as access to psychotherapy or stigmatization. Finally, we analyzed DIW and medical costs independently of each other. Future studies should investigate how medical costs and DIW are related to each other over time (e.g., whether increased speci c medical costs help reduce DIW). Future studies should also systematically investigate how prevention programs for distressed individuals and evidence-based treatments for individuals with mental disorders contribute to saving money by restoring occupational and social functioning.

Implications For Practice
This study shows the extent to which self-reported mental distress is related to the subsequent inability to work and to medical costs. On an individual level, our results indicate that mental distress affects a person's life after a span of two years by reducing occupational and social functioning. Our results demonstrate the high socio-economic costs of mental distress through productivity losses due to reduced functional levels at the societal level. The results, therefore, suggest that joint efforts should be made to effectively reduce mental distress. Individuals with mild and moderate mental distress who do not yet suffer from a manifested mental illness should be given access to preventive services. Preventive structures should be established within peoples' everyday lives (e.g., at the workplace) to enable low-threshold access [15][16]. Not recognizing mental distress, ignoring it, or not taking effective countermeasures might exacerbate the problem and result in signi cant negative nancial impact. A preventive commitment from employers to the workforce's mental health should ultimately lead to a better working atmosphere, a better quality of life for employees, and an increase in productivity [11].
Individuals with severe mental distress or those with manifested mental disorders should be given rapid access to specialized help in the form of evidencebased psychotherapeutic or psychiatric treatments. Prior studies from the UK have shown that increasing access to psychological therapies would largely pay for itself by reducing other depression and anxiety-related public costs (e.g., medical costs and productivity loss) and increasing revenues (e.g., paying taxes [16]). Rapid access to mental health services should be enabled, since the time spent waiting to start psychological treatments was negatively associated with treatment outcome [15]. One best-practice example of how access to psychotherapy can be improved is the English Improving Access to Psychological Therapies (IAPT) service, which delivers psychological therapies recommended by the National Institute for Health and Care Excellence for depression and anxiety disorders to more than 537,000 patients in the UK each year [15]. Evaluation by the IAPT approach has shown that 40.3% of patients showed reliable recovery, 63.7% showed a reliable improvement, and 6.6% showed reliable deterioration [40].
The short processing time of the HADS and the simultaneously good predictive validity regarding the DIW allow for an e cient, cost-effective, and early assessment of the mental distress across different settings. Therefore, it could be used as a suitable screening tool for mental distress to effectively allocate prevention measures in the context of selected or indicated prevention. To improve access to specialized treatments, general practitioners in private practice could use the HADS as a risk-assessment parameter for mental disorders. Since the HADS is suitable for identifying speci c mental disorders in physical health settings, it might be advisable to consult an expert (i.e., a psychiatrist or psychotherapist) if conspicuous values of the HADS appear [27].

Conclusion
In summary, our study demonstrates the extent to which mental distress is associated with reduced occupational and social functioning. Accordingly, mental distress signi cantly impacts the number of DIW and medical costs for a span of two years following the initial HADS assessment. These results indicate that improving preventive structures for at-risk populations and increasing access to specialized treatments for individuals with mental disorders might reduce individual suffering as well as public costs. This study was approved by the ethical review committee at the University of Heidelberg (2013620NMA). Written informed consent to participate in the study was obtained from participants.

Consent for publication
Not applicable, no individual data.

Availability of data and material
The datasets generated and/or analyzed during the current study are not publicly available due to the data protection policy from the cooperating insurance fund but are available from the corresponding author on reasonable request (in anonymized form).

Competing interests
G. Mueller and M. Bombana are employees of the sponsor. The authors declare that the research was conducted in the absence of any commercial or nancial relationships that could be construed as a potential con ict of interest.

Funding
This study was nanced by the health insurance fund AOK Baden-Württemberg. The compilation of health care costs directly from the insurance fund's records required a close cooperation throughout all phases of the project. In order to ensure scienti c objectivity and simultaneously comply with the data protection policy, the sponsor commissioned an internal research o cer (GM) and signed a contractual agreement with the research facility (CIMH Mannheim) that all data may be published regardless of possible positive or negative results.
Authors' contributions GM, LL, MBoh and NK designed the study. LL, GM and RV collected the data. GM analyzed the data and discussed the results with RV. MBom and MHG provided statistical advice. GM and RV prepared the rst draft of the manuscript. All authors reviewed the manuscript.   Note. N=2.287. t0 = assessment of predictor variables. M = Mean non-speci c direct costs per person and year. CI = 95% con dence interval. HADS = Hospital Anxiety and Depression Scale. a = reference category. Method = log-link function, Tweedie-distribution of residuals. Note. N=2.287. t0 = assessment of predictor variables. M = Mean speci c direct costs per person and year. CI = 95% con dence interval. HADS = Hospital Anxiety and Depression Scale. a = reference category. Method = log-link function, Tweedie-distribution of residuals.