Introduction

Obesity is an increasing burden for most societies due to its adverse health effects and associated costs. According to WHO, 20.7% of the male adult population and 17.4% of the female adult population in Denmark were defined as obese (BMI >30) in 2014 [1]. Obesity leads to increased morbidity and mortality and is thus associated with higher health care costs [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21]. Bariatric surgery has been shown to cause sustained weight loss, which leads to a reduction in associated comorbidities as well as decreasing mortality rates [22,23,24,25,26,27,28]. Because bariatric surgery has positive implications for obesity-related risk factors and maintenance of weight loss, the question is whether this entails positive socio-economic implications as well. In this regard, previous research is not consistent [22, 28,29,30,31,32,33,34,35,36]. A widespread argument among proponents of bariatric surgery is future cost savings. However, most studies claiming this primarily focus on the patients’ health care costs after surgery [22, 28,29,30,31,32,33,34,35, 37,38,39,40,41,42,43] and the potential return of investment [44,45,46]. Some results on health care cost savings are based on statistical models, Markov models and assumptions used to forecast outcomes of interest [25, 36, 47,48,49,50], which often do not take into account adverse effects and complications of the surgery. In addition, health care costs merely constitute a part of the total socio-economic impact, which emphasizes the need for examining other costs as well, e.g. costs associated with the labour market. In Denmark, the Danish Civil Registration System assigns every citizen a personal identification number (central personal registration [CPR] number), which allows for the linking of information between national registries at the individual level. This enables identification of people undergoing bariatric surgery and calculations of both direct and indirect costs associated with these individuals. Based on this real-world data, the current study seeks to evaluate the total socio-economic impact of bariatric surgery by estimating both direct and indirect costs before and after surgery among individuals undergoing bariatric surgery, compared with a group of individuals eligible for but not undergoing bariatric surgery.

Methods

Predictions of health care costs, social transfer payments and income were made for an individual undergoing bariatric surgery relative to a similar individual eligible for but not undergoing bariatric surgery. In addition, two regression analyses were performed. Both analyses compare a group of individuals undergoing bariatric surgery with a group of individuals eligible for but not undergoing bariatric surgery. The first analysis estimates the relative risk (RR) of the surgical group’s health care costs relative to those of the non-surgical group before and after surgery. The second analysis estimates the RR of the surgical group’s social transfer payments and income relative to the non-surgical group before and after surgery.

Population and Primary Measure

The population of the surgical group included individuals undergoing bariatric surgery in 2010. These individuals were identified using the National Patient Registry (NPR), which includes information on diagnosis and treatment procedures and other types of hospital contacts. The population of the non-surgical group did not undergo bariatric surgery but did meet the criteria for bariatric surgery at the time. In 2010, bariatric surgery in Denmark was offered to people from 18 years of age with (1) BMI >35 and diagnosed with at least one comorbidity or (2) BMI >40 [51]. In December 2010, the guidelines were revised specifying the comorbidities required for eligibility for surgery: type 2 diabetes (T2D), hypertension, dyslipidemia, sleep apnea, polycystic ovarian syndrome or arthrosis in the lower extremities [52]. Thus, these criteria defined the non-surgical group, and information on these comorbidities was retrieved from NPR. The upper age limit in both groups was 64 years—reflecting the ceiling of the labour force—and individuals aged 18–24 were excluded due to skewed age distributions and a low number of individuals in the two groups. BMI for the non-surgical group was calculated (kg/m2) based on self-reported weight and height from the Danish National Health Profile, which is a national study of the Danish people’s health based on questionnaire surveys and data from national registries [53].

According to the Danish Health Data Authority, 4280 of 4360 bariatric surgeries of people over 24 years old in 2010 were done by gastric bypass, whereas gastric banding and gastric sleeve only accounted for 70 and 10, respectively [54]. According to the guidelines at the time, further criteria for bariatric surgery were completion of a prior treatment at the hospital in addition to a weight reduction of 8% of the current body weight within 3–6 months prior to surgery [51].

Outcome Variables

Outcome variables included health care costs, social transfer payments and income. Health care costs were retrieved as total costs and in subcategories: somatic inpatient services, somatic outpatient services, primary health care and drugs. Furthermore, health care costs were divided between selected disease groups defined by the International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10): diseases of the circulatory system, diseases of the digestive system, endocrine, nutritional and metabolic diseases, diseases of the musculoskeletal system and connective tissue and other diseases. Finally, drug costs were divided between selected drug codes defined according to the Anatomical Therapeutic Chemical (ATC) Classification System Index established by the WHO Collaborating Centre for Drug Statistics. Social transfer payments were retrieved as total costs and in subcategories: unemployment benefits, social security, disability pension, housing benefit, child benefit and sick pay. Income included annual earnings and other types of income not derived from the state. Because state education grants could not be separated from the data, this was included in income as well.

Information on all outcome variables was derived from national registries. Information on diagnoses and treatment was retrieved from NPR and linked with the Danish Case Mix System (diagnosis-related groups, DRG), which enabled calculation of the associated costs. Primary health care costs were derived from the Danish National Health Insurance Service Registry. Costs of drugs were calculated by multiplying the retail price for each drug with the prescribed quantity. This information is available from the Danish Medicines Agency. Information on income and social transfer payments was also derived from national registries from Statistics Denmark.

Data on all costs were traced retrospectively for the period 2007–2013, i.e. 3 years before and after surgery. The year of surgery (2010) is indicated as year 1, and the years in the preoperative and postoperative periods as years −3, −2 and −1 and years 2, 3 and 4, respectively. When performing the predictions, costs were adjusted to 2014 price levels and converted to EUR (1 EUR = 7.43 DKK).

Covariates

Baseline characteristics were retrieved from the Danish Civil Registration System and Statistics Denmark and were used for adjustment in the regression analyses. Baseline characteristics comprised sex (male or female), age (in categories: 25–34, 35–44, 45–54 or 55–64 years), marital status (married/civil partnership or single) and education (primary school, high school, vocational education, short education, medium education, higher education or education unknown).

Analytical Model

The following equation was used to predict health care costs, social transfer payments and income, respectively, from the baseline characteristics:

$$ {Y}_{\mathrm{i}}={\beta}_0+{\beta}_1{\times}_{1\mathrm{i}}+{\beta}_2{\times}_{2\mathrm{i}}+\dots +{\beta}_k{\times}_{k\mathrm{i}}+{\varepsilon}_{\mathrm{i}} $$

The independent variables in the regression were surgery/non-surgery, sex, marital status, education and age.

In the regression analyses, a “One model GLM with link = log and gamma distribution” was used for the estimation of health care costs, social transfer payments and income [55]. All cost variables are left skewed, and a large proportion of them have costs equaling 0. Therefore, a gamma-distributed two-step model was used. Analyses were performed using SAS 9.1.3, and the significance level was set to 0.05.

Results

Population Characteristics

The surgical group and the non-surgical group included 3904 and 1549 individuals, respectively, and the mean ages were 41.6 and 52.3 years, respectively. The age distributions in the two groups can be seen in Table 1. In the surgical group, 77.5% were female, whereas females constituted 52.2% of the non-surgical group. Vocational education was the most frequent education level in both groups (45.6 and 41.7%, respectively), and the majority of individuals in both groups were married or in civil partnerships (69.7 and 71.6%, respectively). However, this association was not significant.

Table 1 Population characteristics
Full size table

Predictions

Predictions of the yearly health care costs, social transfer payments and income were made for a 40-year-old, surgically treated male who is married or in a civil partnership and has a vocational education compared to those for a similar individual not undergoing bariatric surgery. The predictions are illustrated in Fig. 1 and the real numbers are presented in Table 2.

Fig. 1
figure 1

Predictions of health care costs, social transfer payments and income of a surgically treated individual and a similar but non-surgically treated individual

Full size image
Table 2 Estimates of health care costs, social transfer payments and income of a surgically treated man and a non-surgically treated man
Full size table

Three years prior to surgery, health care costs for the surgically treated man were approximately EUR 550 lower than those for the non-surgically treated man. One year before surgery, the costs for the surgically treated man reached the same level as for the non-surgically treated man, followed by a peak in year 1, reflecting the high costs associated with the surgery. In year 2, the surgically treated man had only slightly higher health care costs than the non-surgically treated man, and in the following 2 years, health care costs were similar for the two men. The costs of social transfer payments were slightly but non-significantly lower for the surgically treated man prior to surgery. Three years before surgery, the difference in costs was approximately EUR 400. For both men, social transfer payments increased up until surgery, after which the costs of the surgically treated man immediately started to decrease, whereas the costs for the non-surgically treated man continued to increase until year 2. Both men’s costs continued to decrease throughout the period, ultimately resulting in approximately EUR 1600 lower costs for the surgically treated man in year 4. With regard to income, the surgically treated man had an income approximately EUR 5000 higher than the non-surgically treated man in year −3. The men’s income followed roughly the same path of decrease up until surgery followed by a slight increase, resulting in an approximately EUR 6000 higher income for the surgically treated man in year 4.

In conclusion, our predictions indicate that bariatric surgery does not affect health care costs, that it has a slightly positive effect on social transfer payments and that it does not affect income notably.

Regression Analyses

Health Care Costs

Total health care costs were significantly lower in the surgical group in the preoperative period but higher in the postoperative period; see Table 3. However, estimates for the postoperative period were insignificant. The surgical group had lower costs for somatic inpatient services before surgery, but the postoperative period reflected higher costs among the surgically treated individuals, ending with 38% higher costs for somatic inpatient services in year 4. Costs of somatic outpatient services were higher in the surgical group 1 year before surgery, and the difference between the two groups decreased subsequently, resulting in 14% lower costs of outpatient services in the surgical group compared with the non-surgical group in year 4. Estimates of costs of primary health care were insignificant in the preoperative period but significantly lower in the surgical group after surgery, compared with the non-surgical group. Costs of drugs were lower in the surgical group during the whole period but significantly lower in the postoperative period (RR 0.49, RR 0.46 and RR 0.49, respectively).

Table 3 RR of health care costs in the surgical group compared with the non-surgical group
Full size table

Table 4 shows health care costs of specific disease groups. Costs of diseases of the circulatory system were lower in the surgical group during the whole period but notably lower in the postoperative period (RR 0.47, RR 0.47 and RR 0.37, respectively). The surgical group had 64% higher costs of diseases of the musculoskeletal system etc. 3 years before surgery and 31% lower costs in year 4. However, these costs already started to decrease in year −2, which complicates interpretation of the impact of the surgery on these costs. Costs of endocrine, nutritional and metabolic diseases where higher in the surgical group before surgery and the incremental costs increased additionally after surgery resulting in a RR of 7.76 in year 4. A similar tendency was seen in costs of diseases of the digestive system, which were lower in the surgical group prior to operation but subsequently exceeded the costs in the non-surgical group in year 4 (RR 2.71). Except for years 1 and 2, costs of other diseases were lower in the surgical group.

Table 4 RR of costs of selected disease groups in the surgical group compared with the non-surgical group
Full size table

A decrease in costs was seen for the majority of ATC codes after surgery; see Table 5. Costs of code A, alimentary tract and metabolism; code A10, anti-diabetic medication; code M, musculoskeletal system and code R, respiratory system, were all higher for the surgical group in the preoperative period, but the costs in this group became significantly lower after surgery. Especially affected were costs of anti-diabetic medication; in year −3, the RR was 1.12 whereas in year 4, it was 0.07. Costs of code C, cardiovascular system, were lower for the surgical group in the preoperative period, but after the surgery, the RR decreased, resulting in 64% lower costs for this group in year 4. Except for year 2, costs of code H, systemic hormonal preparations etc., were also lower for the surgical group, both before and after surgery. Costs of other drugs were significantly lower for the surgical group in the preoperative period, but insignificantly higher in year 4.

Table 5 RR of costs of selected ATC codes in the surgical group compared with the non-surgical group
Full size table

All estimates in the analysis of health care costs were adjusted for age, sex, municipality and education.

Social Transfer Payments and Income

Costs of social transfer payments and income are shown in Table 6. The surgical group incurred lower total costs in the form of social transfer payments both before and after surgery, but the difference between the two groups was the greatest in the postoperative period (RR 0.79, 0.79 and 0.81, respectively). This was primarily due to a decrease in the difference in costs of unemployment benefits between the two groups resulting in significantly lower costs for the surgical group compared with the non-surgical group in year 4 (RR 0.65). In contrast, the surgical group had lower costs of social security in the preoperative period, but the difference between the two groups increased after surgery, resulting in significantly higher costs for the surgical group compared with the non-surgical group in year 4 (RR 1.34). Costs of disability pension were lower for the surgical group for the whole period but the lowest prior to surgery. Costs of housing benefit were higher in the surgical group both before and after surgery, but the incremental costs were lower after surgery (RR 1.36 in year 4). Costs of child benefit were also higher in the surgical group for the whole period, but the incremental costs were the highest in the postoperative period (RR 1.26, RR 1.27 and RR 1.26, respectively). The surgical group had higher costs of sick pay than the non-surgical group before surgery but lower costs in years 2 and 3 (2 and 9%, respectively). However, the difference in costs between the two groups returned to the same level as 3 years before surgery (RR 1.33 and RR 1.35 respectively). Finally, the surgical group had a significantly higher income than the non-surgical group both before and after surgery. The excess income was slightly higher in year 4 (RR 1.26), but the increase in RR from year −3 to year 4 was not significant. In addition, the primary increase commenced before surgery, which renders the picture of bariatric surgery’s impact on this parameter unclear.

Table 6 RR for change in social transfer payments and income in the surgical group compared with the non-surgical group
Full size table

All estimates were adjusted for age, sex, municipality and education.

Discussion

In this study, we have generated real-world evidence on the direct and indirect costs associated with obese individuals undergoing bariatric surgery compared with a group of individuals eligible for but not undergoing bariatric surgery. To our knowledge, we are the first to conduct a registry-based study evaluating the impact of bariatric surgery on both direct and indirect costs over a 7-year period. Our predictions indicated that bariatric surgery did not affect health care costs, induced slightly lower costs of social transfer payments and led to no notable change in income. These findings were partly confirmed in the regression analyses, which indicated higher health care costs, slightly lower costs of social transfer payments and no significant impact on income. In conclusion, our regression analyses did not demonstrate any net cost savings of bariatric surgery during the period of analysis. However, we identified some notable effects of bariatric surgery on some of the subcategories.

The use of national registries for retrieving real-world data from the preoperative and postoperative periods allowed us to look back and ahead from the time of surgery and thereby evaluate the impact of the surgery on health care costs, social transfer payments and income at the individual level. By including all types of costs and dividing them into subcategories, we were able to identify the parameters on which bariatric surgery had or did not have an effect. With regard to health care costs, especially costs of somatic inpatient services and costs of drugs were affected. Costs of somatic inpatient services were significantly higher in the surgical group, whereas costs of drugs were significantly lower in the surgical group after surgery. Particularly code A10, anti-diabetic medication, was affected, the costs for this going from being 12% higher in the surgical group in year −3 to being 93% lower in year 4. In year −1, somatic inpatient services have decreased and somatic outpatient services decreased, which may be a result of the criteria for surgery: Somatic inpatient services may have decreased due to the 8% weight reduction and somatic outpatient services may have increased due to the treatment prior to surgery including frequent medical examinations.

Although our results on health care costs in the postoperative period are insignificant, increased health care costs after bariatric surgery have been identified in other studies, especially increases in costs of inpatient services and medical expenses, which is in line with our findings [22, 30, 31, 33,34,35, 37, 38, 42, 44]. For instance, Lopes et al. [56] conclude in their meta-analysis that bariatric surgery was associated with an average reduction in drug costs of 49.8% during a postoperative period of 6–72 months. Our finding of a significant decrease in drug costs is primarily due to the notable reduction in drug code A10, anti-diabetic medication. This supports previous findings of bariatric surgery resulting in improvement or remission of T2D [22,23,24, 26, 27, 56, 57]. The decrease in costs of diseases of the circulatory system is due to reductions in the prevalence of comorbidities, e.g. hypertension, ischemic heart disease and stroke [3, 4, 6]. The higher costs of endocrine, nutritional and metabolic diseases and diseases of the digestive system are probably indications of complications of the surgery, e.g. nutritional deficiencies and other gastrointestinal complications [24, 58,59,60,61]. Already in year −1, diseases of the circulatory system, diseases of the musculoskeletal system and other diseases have decreased in the surgical group, which may be explained by the required weight reduction prior to surgery.

Our findings underline the notable positive effect of bariatric surgery on T2D in addition to some circulatory diseases. However, in a socio-economic perspective, these beneficial effects are outweighed by other complications indicated by the increase of endocrine, nutritional and metabolic diseases and in diseases of the digestive system.

In the analysis of social transfer payments, we identified significantly lower costs of unemployment benefits and significantly higher costs of social security. These results can be explained by a displacement of costs due to expiry of the unemployment benefit period. Income was affected by a small increase, though the primary increase was identified prior to surgery, which makes the picture of income unclear. The weight reduction prior to operation might also contribute to this finding, i.e. improved ability to work supported by the decrease in sick pay from year −3 to year −1. Obesity has been associated with reproductive disorders and complications [6, 62], which supports our finding of significantly higher costs of child benefit in the postoperative period, indicating that the weight loss achieved by the surgery induces an increased fertility rate among female patients.

The fact that we did not identify any net savings does not mean that bariatric surgery should be considered ineffective. The increase in health care costs, primarily due to complications and adverse effects of the surgery, should be weighed against the beneficial impact of bariatric surgery, including positive clinical effects, as indicated by our results (e.g. reductions in the prevalence of T2D and circulatory diseases, and reduced drug use) in addition to considerable and sustained weight loss [27, 43, 57, 60, 63], enhanced quality of life [25, 64,65,66,67,68] and a small decrease in total social transfer payments, as found in our study. In line with the statement made by Agren et al., expressing that increased hospital care is a moderate price to pay considering the well-documented benefits of bariatric surgery [32], the present argument for bariatric surgery as applied weight loss intervention should not be based on possible cost savings, but rather on the potential improved quality of life and clinical benefits. According to the Danish health legislation, the argument for (carrying out) health care services are clinical benefits but in the public debate, clinical benefit is often considered equal to economic benefit, though this is not statutory. Likewise, the argument for bariatric surgery is often cost savings and the endpoint of improving health and quality of life fades into the background. However, these clinical benefits may entail economic benefit in the longer run if improved health of the surgically treated individuals leads to improved ability to work, fewer hospital contacts etc. The period of analysis in the current study is obviously not long enough to reveal whether bariatric surgery induces cost savings in the longer run. The identified increase in costs of somatic inpatient services actually seemed to ease off slightly during the postoperative period, which indicates that these costs might be counterbalanced over a longer period. However, studies on the economic impact of bariatric surgery over longer periods of time are limited. Most studies with long time horizons are based on different types of statistical modelling and forecasting methods, which often do not capture the dynamics and complexities in terms of, for instance, complications and comorbidities, which is possible when using real-world data. Weiner et al. conducted a study of health care costs of surgically treated individuals relative to a matched non-surgical group, in which they found an increase in total health care costs for the surgical group, even 6 years after operation [34]. In contrast, the Swedish Obese Subjects (SOS) study on health care use over 20 years after bariatric surgery has demonstrated that costs of inpatient and non-primary outpatient services among surgically treated obese individuals only exceeded the costs for conventionally treated control subjects in the first 6 years after surgery [41]. However, it should be noted that the surgical population of the SOS study was primarily treated with vertical-banded gastroplasty, whereas gastric bypass was the customary procedure in Denmark in 2010, which complicates a proper comparison. Nevertheless, the results from the SOS study indicate that bariatric surgery has the potential to induce reductions in health care costs in the long run. With regard to social transfer payments, the direction of these costs is more unclear. We tend to assume that the correlation between obesity and higher social transfer payments is evidence of a causal effect, i.e. when the obese individuals lose weight, they will return to the labour market. However, this may not be the case and, thus, public resources will not necessarily be as strongly affected as expected.

Taken together, these results emphasize the need for further research using real-world evidence on both direct and indirect costs over longer periods of time instead of modelled estimates, in order to examine the actual socio-economic impact of bariatric surgery.

Limitations

Even though the non-surgical group was defined by criteria that also applied to the surgical group, it should be noted that the two groups are not identical and this shortcoming should be acknowledged. It appears from Table 1 that the individuals in the surgical group were younger than those in the non-surgical group. Also, the surgical group mainly consisted of women, whereas this was not the case in the non-surgical group. However, the discrepancies between the two groups have been taken into account by adjusting all estimates in the regression analyses. In addition, our results elucidate that the surgical group initially had more sick days than the non-surgical group and higher costs of drugs in some ATC codes. This indicates that this group might already be in worse condition than the non-surgical group. Furthermore, bariatric surgery is followed by some recommendations concerning diet and exercise but we have no data on compliance in this matter. Neither do we know the magnitude of BMI reduction due to surgery and whether this reduction is sustained in the years following operation. Taken together, the difference between the surgical group and non-surgical group and the lack of information on BMI reduction and postoperative lifestyle could explain some of the estimates in the regression analyses, e.g. some of the results on comorbidities after surgery.

Conclusion

In this study, we used real-world data to conclude that bariatric surgery had a slight positive effect on social transfer payments, had no significant effect on income and did not induce lower health care costs 3 years after surgery. Taken together, no net saving is gained from bariatric surgery in the short run. The present argument for bariatric surgery should be based on the beneficial clinical effects, sustained weight loss and improved quality of life rather than on the potential of saving public resources. Further real-world evidence on direct and indirect costs over a long period of time is needed to examine if costs eventually will be counterbalanced and thereby making bariatric surgery a socio-economically beneficial intervention.