Abstract
This article examines how everyday media use and interpersonal communication for health information could influence health behaviors beyond intervention or campaign contexts. The authors argue that interpersonal communication works as an independent information channel and mediates the relation between media channels and health behaviors. In addition, the authors investigate whether interpersonal communication differently influences the relation between media connections and health behaviors for more and less educated individuals. Using data from the 2008 Annenberg National Health Communication Survey, the authors show that multiple communication channels for health information encourage health-enhancing behaviors but do not have significant relations with health-threatening behaviors. Interpersonal communication is directly linked to health-enhancing behaviors, but it also mediates the influence of individuals' multichannel media environment on health-enhancing behaviors. The mediating role of interpersonal health communication was only significant for less educated people. In addition, among media channels, television was a more important instigator of health-related conversations with family and friends for the less educated group. The theoretical and practical implications of these findings, as well as suggestions for future research directions, are discussed.
Notes
1The final sample size for the 2008 Annenberg National Health Communication Survey was 3,656. Listwise deletion was used to remove 16 cases that have missing values in key variables. The sample size of the final data analysis was 3,640. At the same time, using a combined weight variable in the Annenberg National Health Communication Survey data set, data was weighted to adjust for known deviations from an equal probability sample design as well as error resulting from nonresponse and noncoverage error.
2The final sample size for the 2008 Annenberg National Health Communication Survey was 3,656. Listwise deletion was used to remove 16 cases that have missing values in key variables. The sample size of the final data analysis was 3,640. At the same time, using a combined weight variable in the Annenberg National Health Communication Survey data set, data was weighted to adjust for known deviations from an equal probability sample design as well as error resulting from nonresponse and noncoverage error.
3We ran the measurement models to examine whether one factor structure (i.e., not distinguishing health-enhancing and heath-threatening behaviors and combining them into one health-related behaviors) would be better than dividing them into two separate variables. Although both models showed reasonably good fit statistics, two-factor model was favored over one-factor model in all model fit statistics: one factor model, χ2(5) = 49.775, CFI =.971, NFI =.968; IFI =.971, RMSEA =.050; two-factor model, χ2(4) = 37.780, CFI =.977, NFI =.975; IFI =.978, RMSEA =.048. A chi-square difference test did show a statistically significant difference between two measurement models, Δχ2(1) = 11.995; p > .01. Hence, the two-factor structure was favored over the one-factor model. At the same time, the correlation between health-enhancing behaviors and health-threatening behaviors was significant but weak (r = − .064, p < .001). On the basis of Lee's (Citation2010) study, self-rated health condition was served as a criterion to assess the validity of health-enhancing behaviors (r = − .13, p < .001) and health-threatening behaviors (r = .25, p < .001).
Note. TV refers to TV use for health information; Print refers to print media use for health information; Internet refers to Internet use for health information; Interpersonal refers to interpersonal communication for health information; HEB refers to health-enhancing behaviors; and HTB refers to health-threatening behaviors.
Cell entries are Pearson correlations between residual values of all six focal variables were obtained through regressing them with the seven control variables (i.e., age, gender, education, household income, ethnicity, personal and family health condition). Data were weighted before generating the residual values.
*p < .05; **p < .01; ***p < .001.
Note. TV refers to TV use for health information; Print refers to print media use for health information; Internet refers to Internet use for health information; Interpersonal refers to interpersonal communication for health information; HEB refers to health-enhancing behaviors; and HTB refers to health-threatening behaviors.
TV ← → Print, β = .497, p < .001; Print ← → Internet, β = .299, p < .001; TV ← → Internet, β = .321, p < .001.
*p < .05; **p < .01; ***p < .001.
Note. TV refers to TV use for health information; Print refers to print media use for health information; Internet refers to Internet use for health information; Interpersonal refers to interpersonal communication for health information; HEB refers to health-enhancing behaviors; and HTB refers to health-threatening behaviors.
TV ← → Print, β = .492, p < .001; Print ← → Internet, β = .299, p < .001; TV ← → Internet, β = .321, p < .001.
*p < .05; **p < .01; ***p < .001.
4We also examined the fit of all models (i.e., theoretical, alternative 1, alternative 2, and the final revised model) across the two education levels (groups). Overall, the final revised model was the best fitting model for the less educated group and the more educated group.
Note. TV refers to TV use for health information; Print refers to print media use for health information; Internet refers to Internet use for health information; Interpersonal refers to interpersonal communication for health information; HEB refers to health-enhancing behaviors; and HTB refers to health-threatening behaviors.
Bold values represents statistically significant differences (at p < .05 level) between parameters between the less educated subsample and the more educated subsample.
TV ← → Print, β = .517, p < .001; Print ← → Internet, β = .309, p < .001; TV ← → Internet, β = .330, p < .001 (less educated subsample); TV ← → Print, β = .449, p < .001; Print ← → Internet, β = .245, p < .001; TV ← → Internet, β = .257, p < .001 (more educated subsample).
*p < .05; **p < .01; ***p < .001.