Analytical Methods for a Learning Health System: 1. Framing the Research Question

Area of Science:

• Health Services Research
• Comparative Effectiveness Research
• Patient-Centered Outcomes Research

Background:

• Learning health systems leverage electronic health data (EHD) for continuous improvement.
• Observational studies using EHD are increasingly important for healthcare research.
• Challenges exist in inferring causality from EHD due to data volume and analytical limitations.

Purpose of the Study:

• To explore the utility of EHD for various research questions.
• To discuss the role of observational studies alongside randomized clinical trials (RCTs).
• To guide the design and application of EHD in health research.

Main Methods:

• Review of research questions addressable by EHD.
• Discussion of causal inference in observational studies versus RCTs.
• Exploration of study designs for delivery system science.

Main Results:

• EHD is suitable for descriptive research where causal inference is not required.
• Observational studies, when carefully designed and analyzed, can yield valid results for intervention effectiveness, balancing internal and external validity.
• Delivery system science requires diverse methods beyond causal inference, including qualitative and mixed approaches.

Conclusions:

• The appropriate use of EHD depends on the research question.
• Well-designed observational studies using EHD can be valuable complements to RCTs.
• Rigorous observational study designs can be modeled after RCT principles to build a robust evidence base.