Poster #139 - LCA of Parent-Child Conversations about the COVID-19 Pandemic

• In Event: Poster Session 17

Abstract

Several studies (Rigney & Callanan, 2011; Geerdts et al., 2015) have argued that parent-child conversations provide a foundation for children’s developing understanding of the biological realm in general, and health and illness specifically. However, the majority of these studies have focused on the individual content of children's questions and the parent's responses. In the current study, we go beyond exploring individual parent-child conversations by examining whether children's questions and parents' responses about COVID-19 can be classified into different latent classes. To our knowledge, we are the first to use Latent Class Analysis (LCA) to classify children and parents into groups (i.e., latent classes) based on the overall pattern of children’s questions and parents’ responses.

Methods: We recruited 516 parents of 3- to 7-year-old children using Amazon’s Mechanical Turk from July 29th–August 10th, 2020. This was during the 2nd of the three major peaks in the COVID-19 pandemic. We asked parents to report up to three questions their children had asked them about the events surrounding COVID-19. Parents were also instructed to report how they responded to each of the questions reported.

Individual questions and responses were coded for content. Specifically, children’s questions were coded into the following mutually exclusive categories: (1) the virus, (2) safety of child, family, or friends, (3) preventive measures, (4) enforced lifestyle changes, or (5) other. The content of parents’ responses was coded into the following categories: (1) no explanations, (2) authority, (3) address COVID-19, (4) use of protection, (5) social responsibility, (6) reassurance, and (7) other. The reliability, measured through Cohen’s kappa, was deemed appropriate for all categories (see Table 1 and 2).

Results: On average, parents reported 2.7 child questions (n= 1,399) and 2.7 responses (n= 1,396). Children’s questions generally sought information about lifestyle changes (22%). While parental responses focused on providing information about the virus (23%) (see Table 1 and 2 for other categories).

We conducted LCA using the poLCA package in R to examine whether children’s questions and parents’ responses could be characterized by a few discrete latent classes. We fitted different LCAs, yielding latent class numbers ranging from one to six. As our criteria, Bayesian information criterion (BIC) and Akaike information criterion (AIC) were used to select the best models. The model with three classes had the lowest AIC and BIC for both children's questions (AIC=2433, BIC=2493) and parents' responses (AIC=3693, BIC=3778).

For children’s question, the three latent classes were described as: (1) primarily focusing on the virus [39%], (2) focusing on both the virus and lifestyle changes [21%], and (3) focusing on both lifestyle changes and preventive measures [40%]. For parents’ responses the three latent classes were characterized as: (1) focusing on providing information about the virus itself and self-protection [54%], (2) focusing on reassurance while providing some information about the virus [28%], and (3) providing no-explanation [17%].

These results suggest that the pattern of both children’s questions and parental responses can be captured in terms of a discrete number of latent classes.

Authors

• Graciela Trujillo-Hernandez, University of Rochester
  Presenting Author

• Seung Heon Yoo, University of Rochester

• David Menendez, University of Michigan - Ann Arbor

• Rebecca Klapper, University of Rochester

• Karl S Rosengren, University of Rochester