Research design and methods
Heinz Nixdorf recall study
Individual characteristics/primary data of the participants were derived from the ongoing Heinz Nixdorf Recall (HNR; ‘Risk Factors, Evaluation of Coronary Calcium and Lifestyle’) study, a prospective, population-based cohort study evaluating modern risk stratification techniques for coronary outcomes. The study was approved by the responsible local ethics committee. In brief, 9484 inhabitants 45-75-years-old were randomly selected from the mandatory registries of residence of three German cities (Essen, Mülheim and Bochum) located in the northwest of Germany. Between 2000 and 2003, 4814 participants gave their informed consent and took part in the baseline study (recruitment efficacy proportion: 55.8%, response proportion: 53.3%).9 Further details have been described previously.10 Of the 4814 participants included in the baseline study, 4157 (86.4%) attended the second examination visit between 2006 and 2008 (5 year follow-up).
Data acquisition
Self-reported data were assessed by a standardized interview. Further clinical data were derived from laboratory and clinical investigations of the participants. In case of participants' approval, the main corresponding health insurances were asked to cooperate with the HNR study. For the current analyses, primary data from the HNR study were individually linked with routine data on healthcare utilization and associated costs from the main statutory health insurances in Germany.
Study population
Health insurance data of the main German statutory health insurances were available from 2184 participants. Of these, 1722 had health insurance data for at least 1 year before the 5 year follow-up and thus were eligible for the analyses. A further 13 participants were excluded because of missing data on their diabetes stage, resulting in 1709 participants for the analyses. Compared with the excluded participants (n=2448), the participants of the final study sample were on average 0.9 years older (pWilcoxon<0.001) and less frequently male (44.5% vs 53.0%, pχ2<0.001).
Definition of diabetes stages
According to the information provided by the participants and the results of a fasting plasma glucose (FPG) test at follow-up, all participants were divided into four groups of diabetes regulation. ‘Previously diagnosed diabetes’ was present if the participants reported a physician's diagnosis of diabetes or antihyperglycemic treatment at follow-up. Applying the criteria of the WHO, the International Diabetes Federation (IDF) and the European Diabetes Epidemiology Group (EDEG),11 ,12 ‘previously undetected diabetes’ was assumed if the FPG value was at least 7.0 mmol/L (corresponding to 126 mg/dL) and the participants reported no previous diagnosis of diabetes at follow-up. ‘Impaired fasting glycemia’ (IFG) was defined as an FPG value between 6.1 mmol/L and <7.0 mmol/L (corresponding to 110−<126 mg/dL) at follow-up. If the FPG value was lower than 6.1 mmol/L, participants were classified as having ‘normal fasting glycemia’ (NFG).
Sociodemographic and clinical variables
Sociodemographic data included age, sex, country of birth (Germany/other country), living with a partner (yes/no), education, status of employment and quartiles of equivalent income (net income adjusted for household size according to the Luxembourg Income Study13 ,14).
The following clinical variables from the follow-up were included: type of diabetes, diabetes treatment and diabetes duration among people with previously diagnosed diabetes, standardized analyzed glycated hemoglobin value (%, mmol/mol), prevalent hypertension (mean of the second and third systolic/diastolic blood pressure measurement >140/90 mm Hg or antihypertensive medication) since baseline ongoing self-reported previous medical diagnosis of stroke or myocardial infarction (validated by clinical experts) and current body mass index (BMI (kg/m2)) calculated from standardized measurements of height and weight.
Healthcare utilization and associated costs
Healthcare utilization, associated costs and diagnoses (documented as International Statistical Classification of Diseases and Related Health Problems, German Modification (ICD-10-GM)) or pharmaceutical coding of drugs (Anatomical Therapeutic Chemical (ATC) codes) were derived from the statutory health insurances' data. Numbers and percentages of healthcare utilization per capita as well as resulting costs referred to the four complete quarters before the individual follow-up examination (termed index year). The direct medical costs per capita were determined from the perspective of the statutory health insurance.
Inpatient data included the date of, length and diagnosis for each hospitalization (ICD-10-GM), diagnosis-related group (DRG), and corresponding costs reduced by patients' copayments. In case of missing net costs (10.5%), these costs were estimated using data reported by other statutory health insurances assuming sufficient transferability.
In Germany, costs of outpatient consultations are reimbursed according to the respective German compensation scheme (‘Einheitlicher Bewertungsmaßstab’ (EBM) reimbursement). Furthermore, extrabudgetary services, costs of dialysis equipment, and material expenses reported by the statutory health insurances were considered.
Information on prescribed medication comprised prescription date, ATC code, and costs of medication adjusted for mandatory drug discounts and copayments. Missing net medication costs (5.1%) were estimated using reported data by other health insurances. Costs of medication are reported as total medication costs, total costs without antihyperglycemic medication, and costs of cardiovascular medication.
Statistical analyses
The description of the data was presented for the total cohort and stratified by diabetes stage. The continuous variables were described using means and SDs. All categorical variables were described using the percentages with 95% CIs. All variables were age-standardized and sex-standardized in order to adjust for potential demographic differences between the diabetes stages (age strata: <65 years, ≥65 years) to the German population using data from the Federal Statistical Office (reference date: 31 December 2006).
Total costs and different cost categories are given as costs per capita in Euro for the last follow-up year 2008 (2008€). Costs of previous years were inflated to 2008 using the German Consumer Price Index.15 Owing to the right skewed distribution of healthcare utilization data including costs, 95% CIs for the mean values and percentages of health utilization were estimated using bootstrap procedures.16 ,17
The associations between mean total direct healthcare costs, costs of inpatient or outpatient treatment, or medication costs as dependent variables and diabetes stage as main independent variable were estimated using multiple regression analyses. For comparison of the costs between the different diabetes stages, cost ratios (CRs) were estimated adjusting stepwise for age and sex (model 1), additionally country of birth, living with a partner, and equivalent household income (model 2), hypertension, stroke, myocardial infarction, BMI (model 3), and finally healthcare utilization costs of the year before the index year (model 4). Since data contained persons with no healthcare utilization (8.1%), two-part models were used.18 ,19 By combining both parts of the model using generalized linear models (1st Poisson regression model with robust error variance,20 ,21 2nd γ regression model), expected CRs for the whole study population were estimated.