We perform methodological research using Electronic Health Record data in the area of patient-level prediction and population level-effect estimation. We are highly involved in the Observational Health Data Sciences and Informatics (OHDSI) community that builds analytical pipelines on top of the OMOP Common Data Model (CDM) to generate reliable evidence to improve patient health. We are co-leading the patient-level prediction working group in OHDSI that develops a framework for patient-level prediction on top of the OMOP-CDM together with Jenna Reps from Janssen Research & Development:
Jenna M Reps, Martijn J Schuemie, Marc A Suchard, Patrick B Ryan, Peter R Rijnbeek; Design and implementation of a standardized framework to generate and evaluate patient-level prediction models using observational healthcare data, Journal of the American Medical Informatics Association, Volume 25, Issue 8, 1 August 2018, Pages 969–975, https://doi.org/10.1093/jamia/ocy032
Examples of the work done by the group are:
- Development and external validation of clinical prediction models, e.g. Heart Failure prediction in patients with Type 2 Diabetes Mellitus.
- Study of Heterogeneity of Treatment Effect (HTE) in population-level effect estimation.
- Deep Learning methods for prediction
- Learning Curve Analyses to assess bias and variance and sample size requirements
We collaborate with leading institutes in these areas, e.g. Columbia University, Stanford University.
Data standardization and federated analytics
Dr. Rijnbeek has been heavily involved in the standardization to the OMOP-CDM of several European Databases within the context of the European Medical Information Framework (EMIF) and will continue this work in the upcoming European Health Data and Evidence Network (EHDEN) project. The EHDEN project, a 5-year IMI project, will build an eco-system in Europe of a sustainable data network standardized to the OMOP-CDM. Through an open call process leveraging a large harmonization fund (17 million Euro), data sources can get support for their mapping effort. Small to medium enterprises (SMEs) will be trained and certified to support this important exercise.
Furthermore, the group is responsible for the development of the Jerboa software that enables federated analysis using standardized input files that are created on a study-by-study basis. Jerboa implements a distributed network design in which the analytical code is brought to the data source and is run locally. The tool produces analytical datasets with different levels of aggregation, i.e from patient level to fully aggregated summary statistic, that can be shared in a remote research environment for further processing. (see our GitHub site for more details).