Evidence for strong modality-dependence of chronotype assessment from real world calendar app data

TL;DR

This study shows that chronotype assessments based on real-world calendar app data are highly dependent on the specific behavioral modality, indicating internal system-specific differences in circadian rhythms and challenging the assumption of a single unified chronotype.

Contribution

The paper provides evidence that different behavioral outputs yield nearly independent chronotypes, highlighting the modality-dependence of circadian assessments in real-world data.

Findings

Different behaviors produce nearly independent chronotypes.

Real-world data reflect modality-specific circadian rhythms.

Assuming a single chronotype may be misleading in behavioral studies.

Abstract

Chronotypes allow for comparisons of one individual's daily rhythms to that of others and the environment. Mismatch between an individual's chronotype and the timing constraints of their social environment create social jet lag, which is correlated with mental and physical health risks. The concept of chronotype implicitly supposes that a single phase applies to an individual, whereas the circadian rhythms of different internal systems entrain to or have their outputs masked by different environmental inputs. If the modern environment interferes with internal synchrony or generates different masking for different internal system's outputs, then real world data reflecting these outputs ought to reveal relatively low correlations, reflecting environmental interference. At the other extreme, if there is no behavior or tissue specific interference, then different internal system outputs should all be equally predicted from data modalities capturing their different outputs. Here we explore multimodal behavioral rhythm data from the Owaves calendaring app, focusing on behavioral outputs logged as: Sleep, Exercise, Eat, Work, Love, Play, Relax, Misc. We find that individuals show daily rhythms within each behavior type from which chronotypes can be assigned, but that chronotypes derived from different behaviors (or combinations of behaviors) lead to nearly independent sorting of individual's by phase. This suggests that if real world data are used to assign an individuals chronotype, then that chronotype may be specific to the internal system the output of which is related to each specific data modality assessed. Our findings suggest that researchers cannot confidently assume to account for outputs from other internal systems in real world settings without additional observations or controls.