Background
The management of patients with diabetes mellitus is a major challenge for health care systems. The International Diabetes Federation (IDF) estimates that about 366 million people suffered from diabetes worldwide in 2011, and the prevalence is expected to rise to 552 million by 2030[
1]. Diabetes is often associated with microvascular complications and macrovascular complications such as heart disease (heart attack, angina pectoris, heart failure) or hypertension[
2,
3], and leads to a significant burden of mortality[
4,
5]. Due to medical expenditures attributed to diabetes and its complications, as well as productivity losses of the patients, the economic cost of diabetes is immense[
6,
7]. The American Diabetes Association reported that the total estimated cost of diabetes was $174 billion in 2007 in the United States[
8].
In Switzerland, epidemiological data providing information on the burden of diabetes are scarce. There are a few studies, which used different approaches to estimate the prevalence[
9‐
11] or the direct costs of diabetes[
12,
13], and showed heterogeneous results. Furthermore, they were based on small study samples, not generalizable to the Swiss population, or are now outdated. Studies providing a current and comprehensive overview of the epidemiology and economic burden of diabetes are completely missing.
In absence of reliable population-based data on diabetes, health care claims data provide a useful basis for assessing the epidemiology and cost of diabetes. Claims data are reliable, longitudinal, cover a large population and are widely used in epidemiological, health services and outcome research e.g.[
14].
The aim of the study was to provide at first a comprehensive and up-to-date overview of the Swiss situation. Using longitudinal data from a large health insurance population allowed us to assess the current state as well as potential changes in diabetes prevalence, incidence and mortality between 6 years. Moreover, we also aimed to determine the economic burden of diabetes including the total health care costs from a social insurance payer perspective.
Methods
Study design and study population
Health insurance is mandatory in Switzerland and the basic benefit package is the same in the entire country. The database underlying this study included mandatory health insurance claims from approximately 1.2 million persons which lived all over Switzerland and were enrolled with the Helsana Group, consisting of four health and accident insurance plans. In a first step, we defined a sample which included all subjects which were at least 19 years old in 2011 (2011 sample; n = 1’015’106). These data were used to give information about the recent epidemiology and costs of diabetes in Switzerland. In order to additionally identify changes since 2006, we defined a further sample including all persons aged over 19 years in 2006 (2006 sample; n = 920’402). For prevalence and costs estimates, both the 2006 and 2011 samples included patients who were alive and continuously enrolled in the given year. For mortality estimates, we included all individuals. In order to determine the incidence of diabetes, we defined in a next step a retrospective cohort (2006–2011) which included adult persons who were continuously enrolled with Helsana Group during the 6-year period, from January 1, 2006, until December 31, 2011. Subjects who died within this period were also included. This cohort included 477’870 persons at risk in 2006 and 3’993 persons of those were new diabetes cases in 2007 (total: n = 491’863). Available population characteristics included gender, age, language area and type of insurance coverage (managed care model, deductible class). Analyzed data also contained information on patients’ treatment and its associated health care cost; they covered all settings of outpatient and inpatient care and nursing. Since the recorded insurance claims cover almost all health care invoices, the data achieve a high degree of completeness.
According to the national ethical and legal regulation, an ethical approval was not needed. Permission to access the study data was provided by the Helsana Group.
Identification of patients with diabetes mellitus
In absence of epidemiological data including medical diagnoses (e.g. International Classification System, ICD), the WHO Anatomical therapeutic chemical (ATC)[
15] classification system was used to identify patients with diabetes. Patients were classified as having diabetes when they were prescribed at least one drug for diabetes (hypoglycemic medication or insulin; ATC code A10A), oral blood glucose lowering drug (ATC code A10B) or other drug used in diabetes (ATC code A10X) anytime in a given calendar year. Patients with a prescription for diabetes in a given year were considered as still prevalent in the following year, irrespective of whether were additional prescriptions. This mapping approach allows a direct measure of treated diabetes and is frequently used as a reliable way to identify chronically ill persons in administrative data samples [e.g.16]. Based on previous studies, the use of at least one prescription for diabetes drugs can be used for an accurate identification of patients with diabetes across a large population e.g.[
17].
Estimation of prevalence, incidence and mortality
Using our diabetes classification, we calculated diabetes prevalence for the 2006 and 2011 samples. The number of patients with diabetes was divided by all insured persons in the sample in the same year. Prevalence rates were stratified by age and sex. Furthermore, we used census data from the Swiss Federal Office of Statistics to extrapolate the prevalence of patients with diabetes in the general adult population for each year. The characteristics of the of the study population and the entire Swiss population are shown in Table
1. We assumed that our study sample is approximately representative. However, the sample included a slightly higher proportion of elderly persons and the market share of Helsana is not equally distributed in all 26 Swiss cantons. Therefore we used the method of extrapolation. We construct various levels in which the insured persons are represented and weighted the population by the levels which included the entire Swiss population (census data).The stratification for the extrapolation was carried out by age, gender and canton of residence. All 26 Swiss cantons and a total of 16 age classes, which are based on the classification of the risk equalization statistics, were taken into account[
18].
Table 1
Characteristics of the study population and the entire Swiss population
Overall | | | | |
All ages | 920402 | | 6’407’133 | |
19-39 | 282742 | 30.72 | 2’215’128 | 34.57 |
40-59 | 315446 | 34.27 | 2’369’500 | 36.98 |
>59 | 322214 | 35.01 | 1’822’505 | 28.44 |
Women | | | | |
All ages | 481616 | 52.33 | 3’279’256 | 51.18 |
19-39 | 138936 | 28.85 | 1’097’171 | 33.46 |
40-59 | 157528 | 32.71 | 1’175’379 | 35.84 |
>59 | 185152 | 38.44 | 1’006’706 | 30.70 |
Men | | | | |
All ages | 438786 | 47.67 | 3’127’877 | 48.82 |
19-39 | 143806 | 32.77 | 1’117’957 | 35.74 |
40-59 | 157918 | 35.99 | 1’194’121 | 38.18 |
>59 | 137062 | 31.24 | 815’799 | 26.08 |
The annual incidence of diabetes was defined as the number of newly identified cases with diabetes divided by the study population which is free of diabetes at the beginning of the given year. Our calculations excluded the number of persons with prevalent diabetes in the first study year. We classified patients as incident cases if they were identified as persons with diabetes in a given year and had a minimum diabetes-free period of one year in the previous years. Incidence estimates were based on the cohort data (2006–2011 cohort) and additionally standardized to the Swiss population using census data.
We included mortality as a further outcome variable. All deaths of the study sample in 2006 and 2011 were recorded but there was no information on cause of death. Therefore, we calculated the annual all-cause mortality in patients with diabetes, and additionally extrapolated to the general Swiss population by census data.
Estimation of costs
The analysis investigating the health care costs of diabetes was conducted from the perspective of the Swiss statutory health insurance system. Annual total health care costs (Swiss Francs (CHF) exchanged to Euro (EUR); rate of 0.81) of diabetes patients were obtained from providers’ claims and defined as the total payments made by the mandatory health insurance for outpatient and inpatient services per patient and year. Costs from the outpatient setting comprised payments for office-based physician visits, hospital outpatient visits, paramedical visits, nursing home, medications, laboratory tests and medical devices. Inpatient costs covered payments for hospital treatments, hospitalizations, rehabilitation, nursing home and emergency transport services. Inpatient costs also cover the cost of medications, laboratory and medical services during the inpatient episode.
Mean annualized costs of diabetes patients including standard deviation were calculated for each age and sex group. Furthermore, we estimated the costs attributable to diabetes by calculating the difference between the mean costs per capita/year in patients with diabetes and the mean cost per capita/year in patients without diabetes. Estimations were based on the 2006 and 2011 samples.
Additionally, we used census data to extrapolate the health care costs of patients with diabetes to the general adult population for each year.
Estimation of time trends
To determine time trends in epidemiology and costs, we additionally estimated logistic regression models using year as a binary variable for comparisons of prevalence, mortality and costs, based on insured persons in the 2006 and 2011 samples. Year was treated as a continuous variable for determining time trends in incidence in the 2006–2011 cohort sample. Estimates were reported as Odds Ratios (OR) including associated 95% confidence interval (95%-CI). All regression models were controlled for age and sex. Given the skewed nature of costs, we used a multiple regression model with log-transformed costs as outcome. Results were given as exponentiated coefficients with 95%-CI. We reported p-values <0.05 as significant. All statistical analyses performed using R, version 2.14.2. For performing the extrapolation the R-package “survey” was used.
Discussion
This study is the first national study to provide a comprehensive overview of the burden of diabetes mellitus in Switzerland. Based on a large data set, we estimated the current state as well as the development in diabetes prevalence, incidence, mortality and cost-of-illness between 2006 and 2011.
This study yielded an population prevalence of 4.9% in 2011. Our finding is lower than the prevalence rate of the Swiss CoLaus study (6%) of subjects aged 35–75 years and living in Lausanne in 2003[
9]. However, prevalence results originating from other studies performed in Switzerland were significantly lower than our result which might be explained by variations in study design and definition of diagnosed diabetes. For example, Bopp et al.[
10] analyzed diagnoses of hospital discharges and the death registry and reported a diabetes prevalence of about 3.5-4.3% in 2007/2008. This finding might be an underestimation, since diabetes is often not the primary reason for hospitalization and thus not stated in all medical records. Our estimates of prevalence and upward time trend are more in line with international studies reporting higher prevalence levels in the general population. For example, Wilke et al.[
20] reported a standardized prevalence of 5.5% for patients with Type 2 diabetes of all ages in Germany in 2008. Additionally, we observed a significant increase in diabetes prevalence between 2006 and 2011. This increasing trend is in line with estimates from Germany[
6] and Canada[
5] and from global estimates of the IDF Atlas (International Diabetes Federation), which predicted a major increase of 51% in worldwide cases with diabetes by 2030[
1].
While the diabetes prevalence in the Swiss population is increasing, our study suggests a decreasing incidence of treated diabetes over the 6-year time period. The extrapolated incidence of diabetes showed a relative decrease by 8% (percentage change from 2007 to 2011). Previous studies performed in Canada, U.S. and Italy, examining trends in diabetes incidence provide conflicting results[
5,
21,
22]. Differences in incidence trends of diabetes might be associated with several risk factors such as lifestyle habits (i.e. smoking), overweight and environmental patterns which vary between countries and cultures. Since obesity as a strong risk factor for diabetes is rising in Switzerland[
23], the observed decrease in diabetes incidence suggests a meaningful improvement in other diabetes related risk factors. Previous research has shown the positive impact of lifestyle interventions (exercise and/or diet) to reduce the diabetes incidence and indicates that the effectiveness of the intervention is independent from weight-changes[
24]. However, time-point incidence rates of most previous studies corroborate the results of our study, where the estimated diabetes incidence was 0.5-0.9%.
Our all-cause mortality estimates for diabetes patients are higher than those found in other studies[
5,
25,
26]. For example, findings from Canada reported a decrease of 3.3 to 1.6 in all-cause mortality from 1995 to 2005[
5]. However, comparability is limited since considered time intervals differ from each other.
In this study, we also determined the mean annual direct medical costs of diabetes. Patients with diabetes caused health care costs of EUR 8‘239 per year; comparison with persons without diabetes lead to a total of EUR 5’331 per patient attributable to diabetes in 2011. Our cost estimates from 2011 suggest a meaningful increase in direct costs (excluding nursing home), when comparing with previous national estimates in patients with type 2 diabetes from 1998/1999 (EUR 2’232)[
12]. Comparison with international results is difficult as health systems, medical practice and unit cost differ greatly. Köster et al.[
6] showed that the average health care costs (including nursing home) of patients with diabetes in Germany were EUR 5’726 in 2007 and that the costs attributable to diabetes were EUR 2’605. Another study calculating the medical costs of diabetes from the U.S. reported an annual cost (including nursing care) per diabetes case of EUR 5’120 in 2007[
27], which is meaningful lower than findings from our study. Additionally, our results show an increase of the cost attributable to diabetes in 2006 to 2011. This finding is in line with internationally reported data showing an increase of health care costs over the past years[
28,
29]. According to our results, about 387’100 persons (4.9% of 8 million Swiss residents) were diagnosed with diabetes and treated in 2011. This implies total health care costs attributable for diabetes in the compulsory health insurance of CHF 2.5 billion, corresponding to 9.5% of total Swiss statutory health insurance expenditures in 2011. This finding represents a significant burden of disease in Switzerland, and is expected to rise in the future, due to demographic changes as increasing life expectancy and increasing number of elderly patients[
30]. In addition, with the current increase in overweight and obesity observed in Switzerland[
23], it stresses the need for public health strategies to manage patients with diabetes and optimize resource allocations in health service delivery systems. The promotion of chronic care management might be an effective approach to manage and improve health care for patients with diabetes[
31] and adequately prepare for the increasing burden of diabetes on health care resources. Moreover, our results highlight the importance of prevention and early intervention of diabetes mellitus, especially in the elderly population. There is a meaningful impact of age on the occurrence of diabetes mellitus. The percentage of elderly persons continues to rise and older people are faced with chronic diseases and multimorbidity. Previous research has shown that lifestyle interventions are very effective in preventing or delaying the occurrence of chronic diseases such as diabetes mellitus[
23,
32]. Physical activity and healthy diet helps to prevent diabetes (type 2). Public health efforts should notably address the prevention of diabetes in the elderly population and give it a priority in health policy, decision-making and resource allocation debates.
Several strengths and limitations of our study have to be taken into account. The main strength is that the study is based on a very comprehensive health care claims database which covers a large and geographically diverse Swiss population and allows for a good and low cost monitoring of diabetes occurrence. Since we additionally performed an extrapolation of our estimates to the general adult population, we regard our results as being representative for the prevalence, incidence, mortality and cost of diabetes for Switzerland. The study also has several limitations. First, the number of patients with diabetes may be biased, because clinical diagnoses (e.g., ICD-10) were not available. Additionally, our prevalence might be underestimated. Undiagnosed persons with diabetes, patients with lifestyle-treatment and patients with diabetes who were non-adherent to their medication could not be included in the study. However, diagnoses based on dispended medications are a valid proxy for medical diagnoses and widely used in epidemiological and outcomes research to assess prevalence[
14,
16]. Second, cost estimates may be slightly too low as about 3% of the used claims were not paid by the health insurer but directly by the patient (out-of-pocket). Third, our cost estimates might be slightly overestimated, since the data do not allow a distinction between the costs of the treatment of diabetes solely and the costs of its major comorbidities. Further research is needed to estimate the costs of diabetes considering the occurrence of diabetes specific complications such as cardiovascular diseases. A fourth limitation of this study is that incidence trend estimations are not completely representative of the general population, because our cohort only included continuously enrolled patients; persons newly insured during the observation period were not covered. Since we have more or less high fluctuations in our stock of insured persons over the years, we assumed to have a switching bias in calculating the incidence. However, to define a consistent cohort of persons and to follow them over several years it is a very common and frequently used method to determine trends in diseases. Cohort studies are very important in giving us a lot of epidemiological information. Using this study design allows us to see what happens to them with diabetes as well as to see who will develop a new disease. We are able to follow the patients and thus to assess the incidence and its trend.
Competing interests
Financial support for this study was provided by Merck, Sharp and Dohme & Co (MSD), Switzerland. The sponsor had no role in collection, analysis, and interpretation of data; writing of the paper; or in the decision to submit the paper for publication. MS received no funding.
Authors’ contributions
CAH, MS and OR performed the conceptual development and the study design. CAH and RR analyzed the data and CAH drafted the manuscript. OR and MS revised the manuscript. All authors participate in the interpretation of data, critically reviewed for important intellectual contents and gave the final approval of the version to be published.