Examining the Potential of Person-centered Temporal Network Models to Individualize Type 1 Diabetes Care

We aimed to characterize dynamic, individualized relationships among glucose, physical symptoms, emotions, and functioning (e.g., experiential variables) in adults with type 1 diabetes (T1D) using person-centered temporal network models, and assess whether network features relate to retrospective well-being and functioning.

We analyzed data from 158 adults with T1D who completed 14 days of blinded continuous glucose monitoring (CGM) and ecological momentary assessments of experiential variables. For each participant, we estimated a temporal network using the Group Iterative Multiple Model Estimation (GIMME) algorithm, which quantifies dynamic interconnections of variables within individuals. We assessed the variability of models across participants, and examined whether interconnections among variables (network density) were associated with retrospective measures of well-being and functioning.

GIMME models converged for 96.3% of participants, and showed excellent model fit for 77.2% of participants, establishing feasibility. Individual networks had a median of 20.5 (IQR 16-32.75) connections, of which 5 (IQR 3-8) were between glucose and experiential variables. Networks showed marked heterogeneity in the presence, direction, and strength of associations between variables. Individuals whose glucose was strongly predicted by experiential variables (i.e., those whose networks had higher "glucose-in" density), experienced greater depressive symptoms (β = 0.17, p = 0.03), perceived stress (β = 0.24, p < 0.001), anxiety symptoms (β = 0.19, p = 0.02), and negative affect (β = 0.16, p = 0.05) and poorer lifestyle/occupational balance (β = -0.16, p = 0.04).

GIMME models identify individualized patterns linking glucose with daily experiences, which may help generate hypotheses relevant to tailoring care. Network features are associated with retrospective measures of well-being and functioning, supporting the potential of network-informed approaches to T1D care.