Abstract
In the context of health campaigns, interpersonal communication can serve at least 2 functions: (a) to stimulate change through social interaction and (b) in a secondary diffusion process, to further disseminate message content. In a 3-wave prospective study of 1,079 smokers, the authors demonstrate that mass media messages (antismoking campaigns and news coverage relevant to smoking cessation) have an indirect effect on smoking cessation intention and behavior via interpersonal communication. Exposure to campaigns and news coverage prompts discussion about the campaigns, and, in turn, about smoking cessation. Interpersonal communication regarding smoking cessation then influences intention to quit smoking and attempts to quit smoking. The study finds evidence not only for the social interaction function of interpersonal communication, but also for the secondary diffusion function. A substantial number of smokers who are not directly exposed to the antismoking campaigns are nevertheless indirectly exposed via communication with people who have seen these campaigns. These results imply that encouragement of interpersonal communication can be an important campaign objective.
Acknowledgments
Collection of the data was supported by a grant from STIVORO, the Dutch Expert Center on Tobacco Control. Preparation of the article was supported by a personal grant to Bas van den Putte from the Dutch Science Foundation.
Notes
Note. The answers were selected from respondents who participated in Waves 1, 2, and 3, n = 1,079. For statistics that include intention at Wave 3, sample size is smaller (n = 1,027) because respondents who did not smoke at Wave 3 were not asked if they intended to quit. For statistics that include attempts to quit at Wave 2, n = 1,671.
a These coefficients were not significant. All other relations were significant at p < .05.
1Because this model combined linear regression with logistic regression, Mplus did not report fit statistics. In an additional analysis, we used linear regression (instead of logistic regression) to estimate the determinants of the binary smoking cessation behavior variable. In this model, the fit indices showed an adequate to good fit, χ2 = 185.34 (df = 23, p < .001), SRMR = .05, CFI = .93, RMSEA = .08.
Note. n = 1,079. Two unexplained variances were correlated in the model: (a) Correlated unexplained variance between campaign exposure W2–W3 and news media exposure W2–W3 was .27 (p < .001), and (b) correlated unexplained variance between campaign exposure W1–W2 and news media exposure W1–W2 was .38 (p < .001).
*p < .05. **p < .01. ***p < .001.
Note. For three-wave model that regressed intention to quit at Waves 2 and 3: n = 1,027, χ2 = 224.45, df = 23, p < .001, SRMR = .05, CFI = .92, RMSEA = .09. For three-wave model that regressed attempts to quit at Wave 3, n = 1,079. For two-wave model that regressed attempts to quit at Wave 2, n = 1,671. Relations among the interpersonal communication and exposure variables are not reported because the estimated parameters were almost identical to the results in Table 2.
*p < .05. **p < .01. ***p < .001.
2As in the previous model, no fit indices were reported by Mplus. In an additional analysis that included only linear regression, CFI and SRMR indicated a good fit, but RMSEA was too large, χ2 = 39.29 (df = 1, p < .001), SRMR = .03, CFI = .98, RMSEA = .15. When, in correspondence with the causal model in Figure , the direct paths from campaign exposure and communication about the campaigns to behavior were omitted as well as the direct paths from exposure to communication about smoking cessation, RMSEA became a satisfactory .08.
Note. For three-wave model that regressed intention to quit at Waves 2 and 3: n = 1,027, χ2 = 224.45, df = 23, p < .001, SRMR = .05, CFI = .92, RMSEA = .09. Model fit was identical to the fit of the three-wave model in Table 3 because the omitted causal relations among variables that were measured at the same wave were replaced by correlated unexplained variance between these variables. Relations among the interpersonal communication and exposure variables are not reported because the estimated parameters were almost identical to the results in Table 2. For three-wave models that regressed attempts to quit at Wave 3, n = 1,079. For two-wave models that regressed attempts to quit at Wave 2, n = 1,671.
*p < .05. **p < .01. ***p < .001.