We provide estimates of prevalence, incidence and concomitant co-morbidities of T2DM in a population from South Western Germany, which is comprised of over 10 million inhabitants. Our results are based on claims data of the AOK Baden-Wuerttemberg, a large statutory health insurance including about 4 million insurants in South Western Germany. The data show a clear increase of the lifetime diagnosis-based prevalence of T2DM from the year 2007 to 2010, which is still seen after excluding the demographic changes during this time period. Notably, in older people more than 25 % suffered from T2DM, associated with considerable comorbidity. Given the current demographic transition, the burden of T2DM and of diabetes associated co-morbidities will pose an increasing challenge to the health care system. Our data underline the need for early screening programmes as well as early treatment of co-morbidities to encounter this dramatic development. However, several limitations has to be considered concerning the analysis of claims data like non-uniform or incorrect coding, incomplete diagnosis of the physicians, lack of standardized criteria of diagnosis, missing validity checks of the diagnosis, and structural differences of the membership of health insurances.
Lifetime diagnosis-based prevalence of T2DM
The prevalence data of our study are comparable with recent investigations of health insured persons of the AOK Hessen, Germany, a comparable statutory health insurance. Those studies reported a diagnosis-based prevalence of 9.7 % for all types of diabetes together after standardization on the population of Germany in 2009 [
14,
25]. Furthermore, the recent nationwide representative DEGS1 survey found a slightly lower prevalence of known diabetes of 7.2 % in Germany in the year 2011. However in this survey the diagnosis of diabetes was established by self-report of diabetes or self-statement of antidiabetic drugs only, which can lead to underestimation. Interestingly, a special analysis of the data of the DEGS1 survey revealed a marked influence of the membership to different health insurances on the prevalence ranging for diabetes altogether from 3.8 % up to 9.0 % with the highest value for the health insurance we investigated in the present study [
1]. Taking into account that about 5 to 10 % of all patients with diabetes are considered to have type 1 diabetes the prevalences of our study revealed the same magnitude as estimated in the DEGS1 survey for the AOK health insurance data or the analysis of the data of the AOK Hessen.
A considerably lower diagnosis-based prevalence of T2DM was found in the recent analysis of another statutory health insurance in Germany. This analysis showed a prevalence of T2DM of only 4.7 % for Germany and 4.8 % for South Western Germany by cumulative evaluation of the insured persons of the years 2008 to 2010. However, this analysis is only partly comparable with our results as it cumulated the data of three years. In addition, the structure of the membership was different to our population, which might be in correspondence with the results of the DEGS1 survey the main reason for the different prevalences. Furthermore, their algorithm used for analysis considered the WHO-ICD-10 code E11 for the detection of persons with T2DM only. This excluded the substantial amount of insured persons with T2DM coded as E14 alone, which applies in our investigation about an eighth of all patients with a diabetes diagnosis [
22]. In contrast, we used an established algorithm first applied in insured persons of the AOK Hessen with a slight modification to focus on T2DM only [
24,
25].
In addition to the prevalence of known diabetes a considerable amount of undiagnosed diabetes may exist in the population. The prevalence of undiagnosed diabetes was estimated at 2.1 % in the DEGS1 survey for the age group of 18 to 79 years [
32]. Furthermore, the KORA survey 2000 in Southern Germany revealed for the age group of 55 to 74 years a prevalence of undiagnosed diabetes of 8.2 % after screening with an oral glucose tolerance test. This prevalence was nearly as high as the prevalence of known diabetes with 8.7 % [
33]. This is of particular interest due to the considerable amount of already existing co-morbidities like polyneuropathy, increased mortality and the increased requirement of drugs of persons with undiagnosed diabetes [
17,
34‐
36]. Therefore, the considerable amount of undetected diabetes might have led to an underestimation of the age group related prevalence of T2DM as well as of the prevalences of concomitant co-morbidities in our study.
The incidence rates of T2DM with 7.8 to 8.7/1000 in our study were about twice as high as those determined with 3.7/1000 person-years for South Western Germany in a recent analysis of the claims data of another health insurance. However, the prevalence of T2DM was about half as well in that investigation, which may explain the differences of the incidences found [
22]. A higher incidence rate of self-reported T2DM with 9.3/1000 person-years was recently shown for the South of Germany by the analysis of five regional population-based studies of the years 1997 to 2010. However, those studies investigated participants aged 45 years and over, a group with increased risk of developing a T2DM and the data are therefore not directly comparable with our study [
21]. The incidence rates found in our study are in the same range with those in other comparable countries. An incidence rate of 7.3/1000 person-years was determined for T2DM in England. Studies in the USA and Canada found incidence rates of diabetes of 8/1000 in 2008 (USA) and 8.2/1000 in 2003 (Canada) respectively [
37‐
39].
Altogether, the prevalence of T2DM has risen considerably in Germany over the last decades. It was about 2 % in the year 1968 in Munich and 0.7 % in 1961 according to the former German Democratic Republic (GDR) diabetes register with an increase up to 4.0 % in the year 1987 [
40,
41]. The German National Health Interview and Examination Survey 1997/98 found a prevalence of known diabetes of 4.3 % for men and 3.8 % for women in the total population of Germany [
42]. Participants of this survey took also part in the DEGS1 survey. Comparing the results of both surveys the DEGS1 survey confirmed an increased prevalence of diabetes in Germany from 4.3 % to 7.2 % in about 10 years [
1]. Estimates of a sample of insured persons of the AOK Hessen have also shown an increasing diagnosis-based prevalence of diabetes in Germany over the years from 7.2 % in 1998 up to almost 10 % in recent years [
14,
24,
25]. In our study a significant increase of the prevalence of T2DM over the four years was still found after standardization of all data on the residential population of 2007 suggesting a demographic independent factor. Beside a real increase of the occurrence of T2DM in the last years an additional increase could partially be due to a rising awareness by the medical professions as well as by the population in general.
In addition, our study provides evidence of an increase of the prevalence of T2DM already starting at early adolescence and in relatively young age groups. This is of special interest due to a possible higher rate of complications in early-onset T2DM [
43]. The increase and the temporal course of the diagnosis-based prevalences of T2DM are comparable to current studies in the area of Germany and the steep increase with advancing age was found in other investigations also [
1,
24,
25]. In the retirement age over a quarter of the population suffered T2DM. In former studies indications were found for a decrease of prevalence in the age groups over 80 or 90 years, a pattern which was not seen in our study [
44]. However, we could find a decrease of the incidence of T2DM in this age group. The constant high prevalences with advancing age despite decreasing incidence may point to a possible improvement of medical care of these patients. In addition, as the occurrence of T2DM starts to increase earlier in life, these results underline the relevance of modification of established lifestyle factors in early age already to positively influence the development of T2DM in future [
45‐
48].
The prevalence of diabetes in our study may considerably differ from results estimated from claims data of other health insurance companies for various reasons. Altogether, insured persons of the AOK are known to show a relatively high prevalence of T2DM while insured persons of private health insurances show a lower one as recently determined in the DEGS1 survey. This observation fact is present despite standardization for age and sex. There are disparities in the structure of the membership of insured persons with respect to socio-demographic and life-style factors, which might be one explanation [
1,
14,
24,
49]. These underlying differences could partly be overcome in further investigations by the use of pooled data from different health insurances.
Several studies have shown differences in the prevalence and incidence of diabetes in different regions of Germany. A tendency towards a lower prevalence can be found in the South of Germany [
1,
22,
20,
21]. In addition, the investigation of distinct districts in our study showed a considerable range of the diagnosis-based prevalence of T2DM within South Western Germany. The differences were still evident after standardization for age and sex. These different regional prevalence patterns could be due to differences in the awareness and frequency of the diagnosis of T2DM, the non-uniform use of the WHO-ICD-10 codes by physicians in different regions, or regional differences in additional factors like social and economic status, regional deprivation, intensive traffic burden or different genetic disposition for T2DM as revealed e.g. in the DEGS1 survey or as pooled analysis of up to six population-based studies [
1,
20,
21,
50,
51]. Unfortunately, we had no suitable data to explore these differences further.
In the present study insurants with T2DM were determined according to the established algorithm of the analysis of the claims data of the AOK Hessen. This algorithm had successfully been used in these studies over years [
24,
25]. In order to analyze T2DM only the algorithm was slightly modified by excluding the code E10, which codes type 1 diabetes. Furthermore the code E13 was excluded as well due to the frequently use of this code for coding of pancreoprive diabetes. This form of diabetes is characterized by direct destruction of the insulin producing cells due to several reasons (e.g. toxic) and not by altered insulin-sensitivity and production. Therefore, pancreoprive diabetes shows pathophysiological similarities to type 1 diabetes. In addition, our additional analysis of the use of the different ICD-10 codes for diabetes by the physicians revealed considerable multiple coding by the physicians and showed that it is useful to consider more than one code. E.g. the code E14 concerning undefined diabetes was used alone for coding without other diabetic specific codes in 13.5 %. Most of these patients are usually regarded as T2DM and would have been missed out in the analysis by exclusion of E14 in the analysis. This may be one reason for the lower prevalence of T2DM found in the analysis of the claims data of another health insurance [
22].
Concomitant co-morbidities
The case control analysis of the present study shows dramatic excess risks for concomitant co-morbidities due to vascular alterations as well as adiposity for patients with T2DM. The respective excess risks were between twofold and threefold on average. These differences were even more pronounced when comparing patients with T2DM with the total population of South Western Germany. The excess risks found are comparable with results of other studies [
6,
8,
52,
53].
Concerning these concomitant co-morbidities it is of further particular importance, that most of them can be modified by therapeutic interventions. E.g. the occurrence of concomitant co-morbidities by T2DM is associated with increasing fasting blood glucose and can be reduced by a good adjustment of blood glucose levels [
6,
8,
9,
54]. Furthermore, a sufficient therapy of hypertension can markedly reduce the mortality risk related to vascular complications of diabetes. A reduction of the systolic blood pressure by 10 mm Hg can reduce the mortality of ischemic heart diseases by approximately 30 % and of strokes by 40 % [
55‐
57]. The increased risk of persons with adiposity to develop a T2DM and their about three-fold increased mortality should be an important reason for preventive activities. Teenagers with adiposity have an up to 8-fold increased risk to develop a T2DM [
58‐
61]. Furthermore, adiposity increases the risk of stroke as well as increased blood sugar and duration of diabetes [
62‐
64]. In addition, approximately a third of all patients with T2DM have a renal insufficiency 10 years after diagnosis with a high risk to develop end-stage renal disease and diabetic retinopathy [
65‐
68]. The risk of blindness is considerably increased for persons with T2DM. In current examinations it is about 2 ½-folds higher for T2DM than in the general population [
68‐
71]. Altogether, the occurrence of concomitant co-morbidities of patients with diabetes has declined in the last decades, but the burden of disease remained nearly unchanged due to the increased prevalence of diabetes [
72].
In this context it is of particular interest, that the concomitant co-morbidities in patients with T2DM do not only appear far more frequently but also far earlier in life compared to persons without T2DM. These factors contribute both to an increased burden of disease, and to a reduced average life expectancy of patients with T2DM [
3,
6‐
8,
18,
19]. Moreover, the treatment of concomitant co-morbidities cause on average three-quarters of all medical costs of patients with T2DM at present [
73].
Concerning the general population the determined prevalences of concomitant co-morbidities in our analysis were tendentiously lower than in some other studies. This was especially the case for adiposity and to a minor extent for coronary heart disease, stroke, and chronic renal insufficiency. The lifetime prevalence in the general population of Germany has been described in current surveys between 20 to 23 % for adiposity, at about 32 % for hypertension and between 7 to 9.3 % for coronary heart disease [
74‐
79]. These differences can be due to the use of claims data of a health insurance company in our study whereby only physician recorded diagnoses can be investigated missing the undiscovered ones in the general population. In contrast, systematic screening approaches in study populations can also discover asymptomatic disease. The use of routine claims data can lead to a considerable underestimation especially for diseases with an asymptomatic period, which can delay clinical manifestation and diagnosis. In addition, the ICD-coding of physicians for their claims can considerably differ depending on the presented complaints of their patients and on the extent of considering individual diagnoses. Furthermore, the ICD-coding in administrative data has only a limited sensitivity for identifying persons with diabetes [
80]. Finally, the social status of the insured persons can influence the prevalence of morbidities in the population as well. E.g. this was found for low social status and increase of hypertension or diabetes [
1,
79].
Strength and limitations
We analyzed the claims data of the total number of all insured persons of the AOK Baden-Wuerttemberg, a large statutory health insurance in South Western Germany. Thereby about two-fifth of the total population of South Western Germany was included. The completed claims data of the period from 2007 until 2010 were included in the analysis. The established algorithm for claims data of the analysis of insurants of the AOK Hessen was used for analysis with the slight modification of using the WHO-ICD-10 codes E11, E12 and E14 only. The relatively recent data and the large number should ensure a good external validity of our data.
Limitations of a study based on claims data should be considered. These are e.g. the non-uniform or incorrect use of the WHO-ICD-10 codes by the physicians, structural differences between the membership of the particular health insurances, the completeness of recording of diseases by the physicians, and the lack of detection of undiagnosed T2DM. Additional restrictions are the lack of the use of standardized criteria of diagnosis of diseases and the missing comparability and validity check of the physician’s diagnoses. All these limitations may have influenced the analysis and may have led to an under- or overestimation of disease as well as to an age shift towards later years in the present estimation.
A minor restriction of the present data analysis consists in the use of the complete data set of insured persons of the AOK Baden-Wuerttemberg for South Western Germany. These data contain a small proportion of persons with residence in Germany but residing outside of the State of Baden-Wurttemberg. It was assumed in this context that no essential alteration of the finally calculated prevalence arose from including these additional insured persons especially by residency in the border zone of South Western Germany.