Understanding Online Health Risk Information Seeking and Avoiding during the COVID-19 Pandemic

ABSTRACT

This study tested the utility of the Risk Information Seeking and Processing (RISP) model in understanding why people seek or avoid online information about COVID-19. Data collected at three different time points (i.e., February, April, and May 2020) showed the measured RISP model constructs explained between 70–78.8% of the variance for information seeking, and between 36.9–62.5% of the variance for information avoiding. Specifically, fear, information insufficiency, and relevant channel beliefs consistently predicted information seeking. Further, information insufficiency and relevant channel beliefs consistently predicted information avoidance. However, fear had no association with information avoidance. Longitudinally, the study found that within individuals, there were larger increases in most RISP model constructs between Time 1 and Time 2, and smaller changes occurred from Time 2 to Time 3. However, there was no significant change in information seeking over time.

Notes

1. The following item, which was adapted from Spence et al. (2011), was included on the Time 1 survey to check the assumption that the internet was an important source of information about COVID-19 for our sample: “Thinking about the past 30 days, how much have you actively looked for information about the coronavirus from each of the following sources?” Participants were asked to rate five different sources [i.e., television, internet, friends and family, doctor or other healthcare providers, other (newspapers, magazines, radio, etc.)] using a five-point scale ranging from “never” to “very often.” Paired-sample t-tests indicated that participants were mostly likely to get information about COVID-19 from the Internet (M = 3.56; SD = 1.22), followed by friends and family (M = 3.04, SD = 1.16), followed by other sources (M= 2.20, SD = 1.16) and television (M = 2.05, SD = 1.22), followed by a healthcare provider (M = 1.83, SD = 1.08) (i.e., Internet > friends and family > other = television > health care provider).

2. Originally, we intended to also include perceived information gathering capacity in our model. Unfortunately, even though our items were adapted from established scales (Brinker et al., 2020; Lu, 2015; Yang, 2012), Cronbach’s alphas were in the very low to low range (i.e., .48, .68, and .61 for T1, T2, and T3 respectively), and deleting items did not improve measurement reliability. Thus, this construct was ultimately excluded from this study.

3. We ran the Intraclass Correlations (ICC) for all eight variables (i.e., susceptibility, severity, affective response, current knowledge, sufficiency threshold, relevant channel beliefs, information seeking, and information avoiding) and for three different samples (i.e., those who participated at any one time, two times, or all three times) to determine whether or not multilevel modeling is needed to account for time as a level-2 variable (Hox, 2013; Peugh, 2010). All ICCs were small (M = .031, SD = .038) and insignificant (p < .05) and are therefore consistent with the fact that multilevel modeling with time as the level-2 factor was not necessary for this data.