Dynamics of Loneliness Among Older Adults During the COVID-19 Pandemic: Pilot Study of Ecological Momentary Assessment With Network Analysis

Abstract

Objective: The COVID-19 pandemic has had potentially severe psychological implications for older adults, including those in retirement communities, due to restricted social interactions, but the day-to-day experience of loneliness has received limited study. We sought to investigate sequential association, if any, between loneliness, activity, and affect.

Methods: We used ecological momentary assessment (EMA) with dynamic network analysis to investigate the affective and behavioral concomitants of loneliness in 22 residents of an independent living sector of a continuing care retirement community (mean age 80.2; range 68-93 years).

Results: Participants completed mean 83.9% of EMA surveys (SD = 16.1%). EMA ratings of loneliness were moderately correlated with UCLA loneliness scale scores. Network models showed that loneliness was contemporaneously associated with negative affect (worried, anxious, restless, irritable). Negative (but not happy or positive) mood tended to be followed by loneliness and then by exercise or outdoor physical activity. Negative affect had significant and high inertia (stability).

Conclusions: The data suggest that EMA is feasible and acceptable to older adults. EMA-assessed loneliness was moderately associated with scale-assessed loneliness. Network models in these independent living older adults indicated strong links between negative affect and loneliness, but feelings of loneliness were followed by outdoor activity, suggesting adaptive behavior among relatively healthy adults.

Keywords: aging; causal networks; dynamic networks; loneliness; negative affect; positive affect; social isolation.

Figure 1

Affect model describing how loneliness relates to affect: Undirected straight edges between variables indicate contemporaneous associations, and the directed labeled arcs represent lagged associations, with the label being the lag in multiples of sampling interval. The colors of the edges and the arcs represent negative (red) or positive (green) association. The variables with gray nodes indicate significant autocorrelation, or inertia, which can be interpreted as the values of these variables showing high resistance to change. If a variable measures polar quantity (happy-sad or relaxed-anxious), the variable is represented in the graph by the label to which higher values are assigned, and “how happy vs. sad were you” is simply “Happy.” Positive associations are in shades of green, and negative in shades of red. Lagged links are curved, have arrowheads and display lag in multiples of 8 h. Negative emotional states are associated with loneliness. Inverse relationship between happy and loneliness is also expected. Anxiety and worry display a positive feedback loop.

Figure 2

Behavior model describing how loneliness relates to behavior: Loneliness precedes being outdoor, which is associated with exercise and social interaction.

Figure 3

Affect, loneliness, and behavior: Loneliness is predictive of being outdoors. Loneliness is not contemporaneous to being outdoors, it precedes it and returns soon after.