Simulation models in a world of 'big data'
Recent research on the cost of complications based on a study involving resource use data from more than 300,000 people with diabetes from the Institute of Health Economics and Health Care Management, Helmholtz Zentrum, Munich, Germany
While the modeling landscape for diabetes is changing and growing, there is still much research to be done. For example there is no German model for type 2 diabetes based on patient level data. This is, however, much needed to impact the rising prevalence (~7% in 2011) and the resulting social and economic burden of the disease and its heterogeneous complications.
The aim of our recent study was to quantify the comprehensive short- and medium-term economic impact of typical type 2 diabetes-related complications based on nationwide data of the largest statutory health insurance provider in Germany. The overall motivation was to provide reliable and robust cost estimates for a German diabetes model that is related to UKPDS and CDC/RTI.
In the baseline year 2012, the population consisted of 316,220 patients with type 2 diabetes (63% male, mean age 65.9 years), which were selected based on ICD-10-GM codes E11 and E14, as well as the prescription of oral antidiabetic medications, and participation in a disease management program for type 2 diabetes. Costs for inpatient and outpatient care, pharmaceuticals, rehabilitation, and non-medical aids and appliances were assessed in the years 2013–2015. Quarterly observations were available for each year.
Total costs for complications were estimated (in 2015 Euro) using a generalized estimating equations (GEE) model with a normal distribution adjusted for age, sex, occurrence of different complications, and history of complications at baseline. Two- and three-fold interactions with age and sex were included in an extended model.
Figure 1* shows the annualized results of the extended model in comparison with the UKPDS model for the example of 70-79 years old male patients, who have total costs of 2,911 Euro in absence of complications.
* Comparing relative cost factors with the UKPDS Outcomes Model (Version 2) based on the example of 70-79 years old male patients. Cost factors were calculated by dividing the total costs for each complication (considering interactions with age and sex) by the total costs in absence of complications.
What we have learned so far:
The additive approach (using a GEE model with a normal distribution) showed a better model fit compared to a multiplicative approach with a gamma-based GEE model.
Complications are leading to increased costs (largely inpatient costs) not only in the quarter of first diagnosis, but also in subsequent years.
We could demonstrate a reasonable level of congruence with the UKPDS model, with greater deviations for IHD, diabetic foot, and fatal MI. In addition, the assumption of equal costs for the first and subsequent years of ESRD was not confirmed by the present data.
We found more differences between men and women, however, not all statistically significant.
Real-world data are one of the most powerful data sources (high coverage, large sample size, no recall bias), especially for cost information.
This study is the first of its kind for Germany. Apart from the highlighted strengths, some limitations must be considered when interpreting health insurance data, including restricted clinical data, unknown duration of diabetes, and the reliance on the diagnostic accuracy.
Hopefully, this study will draw modelers’ and politicians’ awareness on the importance of accurate cost estimations as a profound basis for the health economic evaluation of new diabetes interventions and programs.
[endif]-- Kähm K, Laxy M, Schneider U, Rogowski WH, Lhachimi SK, Holle R. Health Care Costs Associated With Incident Complications in Patients With Type 2 Diabetes in Germany. Diabetes Care. 2018 Jan 18 (online first).![endif]--