Inter- and intra-physician variability in insulin injection adjustments compared with Bayesian algorithm...

Area of Science:

• Endocrinology
• Artificial Intelligence in Medicine
• Diabetes Management

Background:

• Type 1 diabetes requires precise insulin management, often involving multiple daily injections.
• Automated insulin adjustment systems are needed for unsupervised use to improve patient outcomes.
• Current insulin adjustment protocols can be subjective and vary among clinicians.

Purpose of the Study:

• To evaluate the McGill decision support system (DSS), a Bayesian algorithm, for automated insulin adjustments.
• To compare the algorithm's recommendations with those made by endocrinologists for type 1 diabetes management.
• To assess intra-physician variability in insulin adjustment decisions.

Main Methods:

• A survey of 13 Canadian endocrinologists was conducted using retrospective participant data.
• Physicians made mock insulin adjustments based on weekly (Part A) and biweekly (Part C) data, comparing them to the McGill DSS algorithm.
• Intra-physician variability was assessed by comparing a physician's recommendations on identical datasets (Part B vs. Part A).

Main Results:

• The agreement rate for the direction of weekly insulin adjustments between the algorithm and physicians was non-inferior to inter-physician agreement for both prandial bolus and basal insulin.
• Full disagreement rates were comparable between the algorithm and physicians.
• Physicians showed significant intra-variability, agreeing with their own previous decisions only 66-67% of the time.

Conclusions:

• The McGill DSS algorithm demonstrated comparable insulin adjustment directions to endocrinologists.
• The study underscores the potential of automated decision support systems in diabetes care.
• Significant intra-physician variability highlights the subjective nature of insulin management in type 1 diabetes.