Two-phase methodology for the combined analysis of secondary data sources
Description
- Administrative data of statutory health insurances are a valuable data source for pharmacoepidemiological studies and health services research. The validity of studies based on these data might be affected by the lack of information on some potentially important variables such as Body Mass Index (BMI) or smoking behaviour. If additional or more precise information can be obtained from other data sources for a subsample of persons, two-phase methodology can be used to analyze the complete information in the subsample (phase 2) in combination with the partial information available for all persons (phase 1). The estimation of unbiased and efficient risks relies on the method for inclusion of phase 1 information.
Aim of this project was to investigate the applicability of two-phase methods for the combined analysis of secondary data and to develop recommendations for the planning and conduct of such studies. Furthermore, software should be provided which will simplify the application of two-phase methodology.
A study using insurance claims data and additional data from the disease management program for diabetes mellitus served as an example. Methodological challenges arised from the multiplicity of the phase 1 data and from the selectivity of persons included in the disease management program.
The results of the study and a program to routinely analyze these types of studies were published in international journals. A simulation study on the comparison of up-to-date two-phase methods with other methods was conducted and shall be published in an international journal.
Funding period
- Begin: May 2012
End: August 2016
Sponsor
- German Research Foundation
Contact
Link
Selected project-related publications
- Enders D, Kollhorst B, Engel S, Linder R, Verheyen F, Pigeot I. Comparative risk for cardiovascular diseases of dipeptidyl peptidase-4 inhibitors vs. sulfonylureas in combination with metformin: Results of a two-phase study. Journal of Diabetes and its Complications. 2016;30(7):1339-1346.
https://doi.org/10.1016/j.jdiacomp.2016.05.015 - Schill W, Enders D, Drescher K. A SAS package for logistic two-phase studies. Journal of Statistical Software. 2014;57(9):1-13.
http://www.jstatsoft.org/v57/i09 - Enders D, Pigeot I. Extension of the pseudo likelihood method to analyze two-phase studies with selective phase 2 samples. XXVIIIth International Biometric Conference, 10-15 July 2016, Victoria, Canada.
- Behr S, Schill W, Pigeot I. Choosing efficient stratifications in logistic two-phase studies based on administrative data. 60. Biometrisches Kolloquium der Deutschen Region der Internationalen Biometrischen Gesellschaft (IBS-DR), 10.-13. März 2014, Bremen.
- Enders D, Schill W. Multiple Imputation als Methode zur Auswertung zwei-phasiger Kohortenstudien: Eine Anwendung. Workshop "Real world data" und Registerdaten in der klinischen und epidemiologischen Forschung: Chancen und Herausforderungen" der Arbeitsgruppen Statistische Methoden in der Medizin (IBS-DR), Statistische Methoden in der Epidemiologie (IBS-DR, DGEpi), Statistische Methoden in der klinischen Forschung (GMDS), Epidemiologische Methoden (DGEpi, GMDS, DGSMP), 20.-21. November 2014, Münster.
- Schill W, Wild P, Drescher K. Computing adjusted attributable fractions from two-phase case-control data. 60. Biometrisches Kolloquium der Deutschen Region der Internationalen Biometrischen Gesellschaft (IBS-DR), 10.-13. März 2014, Bremen.
- Enders D, Pigeot I. Impact of different stratification strategies in a two-phase study on the cardiovascular risk in type 2 diabetic patients. 4. gemeinsame Tagung der Deutschen Arbeitsgemeinschaft Statistik (DAGStat), 14.-18. März 2016, Göttingen.