Causal inference for planning randomised critical care trials: Protocol for a scoping review

Area of Science:

• Critical care medicine
• Clinical trial methodology
• Biostatistics

Background:

• Randomized clinical trials (RCTs) in critical care often face inconclusiveness due to optimistic effect size assumptions and inadequate handling of treatment effect heterogeneity.
• Secondary critical care data, rich in data types and temporal detail, combined with causal inference methods, offer potential solutions for planned predictive enrichment.
• A comprehensive overview of the application of these methods for trial enrichment is currently lacking.

Purpose of the Study:

• To conduct a scoping review assessing the use of causal inference from secondary data for planned predictive enrichment in critical care RCTs.
• To understand the extent and nature of causal inference methods applied to secondary data for this purpose.
• To identify methodological gaps and inform future trial design in critical care.

Main Methods:

• Systematic search of 10 general and specialty journals for RCT reports published after January 1, 2018, involving adult critically ill patients.
• Data collection on trial metadata, causal inference methods, estimation techniques, software, patient populations, data provenance, types, models, and data availability.
• Descriptive reporting of the collected results.

Main Results:

• The review will describe the application of causal inference methods using secondary data for predictive enrichment in critical care RCTs.
• It will detail the types of methods, data used, and patient populations involved.
• Findings will highlight current practices and potential areas for improvement.

Conclusions:

• This scoping review will provide insights into the current use of causal inference and secondary data for planned predictive enrichment in critical care RCTs.
• The findings will guide methodological advancements to enhance the utility and application of causal inference in planning future trials.
• Improved methods can lead to more conclusive and efficient critical care research.