Initial Condition Decision to Ensure Reliable Circadian Phase Estimation With Shorter-Term Wearable Data

Area of Science:

• Chronobiology
• Medical Informatics

Background:

• Circadian medicine optimizes treatments by aligning them with patient circadian phases.
• Dim light melatonin onset (DLMO) is the gold standard for circadian phase measurement but is costly and time-consuming.
• Ordinary differential equation (ODE) models offer an alternative for circadian rhythm simulation using sleep data, but require accurate initial conditions (ICs).

Purpose of the Study:

• To investigate the impact of initial condition (IC) uncertainty on circadian phase estimation accuracy using ODE models.
• To develop novel IC estimation methods to reduce the time needed for reliable circadian phase assessment.

Main Methods:

• Collected sleep data from 28 shift workers over 34-75 days using ActiWatch.
• Analyzed the influence of subjective IC choices on circadian phase estimation over time.
• Developed and evaluated new period-based and work history-based IC estimation methods.

Main Results:

• Circadian phase estimation was found to be dependent on subjective IC choice for the first 17 days without accurate ICs.
• Previous IC estimation methods reduced this period to 11-14 days.
• The newly developed methods significantly reduced the required data duration to approximately 5 days for reliable estimation.

Conclusions:

• Accurate initial conditions are essential for reliable circadian phase estimation using ODE models.
• The novel period-based and work history-based IC estimation methods substantially decrease the data duration needed.
• These advancements enhance the practicality of chronotherapy by enabling more efficient and reliable circadian phase assessment.