Statistical methods to evaluate and interpret longitudinaldata have advanced substantially in recent years, especially for identifying relevant subgroups of patients and predicting trajectories of disease progression and responses to therapy. Many of these methods, developed by members of our Data Science Core team, are only just beginning to be applied to medicine. Hence, the work accomplished in the Division of Rheumatology to optimize standardized data collection across diseases can now be exploited more fully using new analytical methods.
The continually expanding and maturing patient cohorts developed through the Johns Hopkins Rheumatology Centers of Excellence provide an outstanding resource on which to focus evolving statistical approaches and complex computational methods designed specifically for the identification of subgroups by response and trajectory. Moreover, these models, once developed, integrate ongoing input for continuous refinement, enabling further improvement of individual patient response prediction, and can help to capture and account for heterogeneity of treatment response. These computational methods are only possible in sufficiently mature cohorts, and are designed specifically to account for multiple types of data input (e.g. clinical, laboratory, autoantibody, diagnostic, treatment).
Working together, clinician scientists and biomedical data scientists (biostatisticians) accelerate the pace of research discoveries by analyzing our rich datasets carefully collected over time with new advances in statistical modeling to identify multiple factors that influence the onset and course of rheumatic diseases and better predict individual responses to different treatments. This in turn can help guide evidence-based clinical practice.
Thus, the primary goal of the Data Science Core is to refine and apply modern statistical and computing methods to enable precision medicine in rheumatology. These methods will improve the clinical assessment of individual patients, identify disease and outcome trajectories, and translate results to bring additional benefits to people living with rheumatic diseases. They enable rheumatology investigators to more fully understand differences among patients and design and analyze data from studies that identify subsets for whom different interventions improve how patients feel, function or survive.
The Specific Aims of the Data Science Core are to:
- Provide data analytical support throughout the research process to enable investigators to generate, manage, analyze, and interpret data using modern statistical methods;
- Develop and apply Bayesian hierarchical models (BHMs) to longitudinal cohorts bringing together with diverse sources of data to identify patient subsets predictive of different trajectories of outcomes and responses to treatments;
- Apply modern analytic approaches to observational and experimental studies that rigorously address heterogeneity among individuals in their responses to treatments.
Leadership and Organization
The Data Science Core is led by Dr. Scott Zeger, Professor of Biostatistics and Epidemiology at the Hopkins Bloomberg School of Public Health, who has an extensive background in clinical and translational research. Dr. Zeger, who pioneered advances in longitudinal data analysis methods, leads the transformative Hopkins inHealth Initiative. The Data Science Core optimizes its impact and maximize productivity and efficiency by embedding biomedical data science faculty within each of the Centers of Excellence, thereby leading to reciprocal learning concerning the formulation of research questions and application of modern statistical methods to answer them. The Core interacts with the RDRCC Administrative Core, the Research Management and Patient Integrated Data (RAPID) Core, and the Sample Processing and Immunoassay Research (SPIRE) Core to synergistically evaluate and address the overall analytical needs of disease Centers and investigators.
Investigators have access to resources for basic through advanced statistical models, taking advantage of the range of expertise available through this Core and other institutional resources. The extensive interactions of the Core faculty and consultants across the institutional biostatistical community enable our investigators to readily exploit existing data, pilot new approaches, and facilitate new collaborations in rheumatic diseases.