Results

---

Table 1 provides the regression equations. In general, the findings support the model and the hypotheses presented here. Inspection of the table shows that the inclusion of only control variables does not have an impact on frequency of use (F = 0.98, p = 0.42), exploratory behavior (F = 0.94, p = 0.44), life benefits (F = 0.35, p = 0.85), search intensity (F = 1.59, p = 0.18), or communication effects (F = 0.29, p = .88). However, there are marginal relationships between income on frequency of use, and gender on search intensity. Taken as a whole, the covariates do not influence the model elements and are eliminated from further analysis.

Table 1: Regression Analysis Results

	Model Variables
Independent Variables	Frequency of Use	Exploratory Behavior	Life Benefits	Search Intensity	Communication Effects
Control Variables Only
Income	0.14[*]	0.04	0.02	0.04	0.05
Education	0.01	-0.06	0.04	0.07	0.04
Sex	0.01	0.06	0.01	0.13[*]	0.04
Age	-0.02	0.10	0.07	-0.09	0.02
Adjusted R2	0.02	0.00	-0.01	0.01	-0.01
ANOVA F	0.98	0.94	0.35	1.59	0.29
	Model 1		Model 2a		Model 2b
Paths
Situational Involvement	-.140[**]		-0.04		-0.15[**]
Frequency of Use	-		0.37 [***]		0.12[*]
Adjusted R2	0.15		0.13		0.03
ANOVA F	4.36[**]		17.88[***]		4.71[**]
			Model 3		Model 4
Test for Mediation
Frequency of Use			0.22[***]		0.59
Search intensity			-		0.54[***]
Exploratory Behavior			0.74[***]		-
Adjusted R2			0.90		0.34
ANOVA F			1027.73[***]		55.34[***]
[*]p < .10, [**]p <. 05, [***]p < .01

To examine Hypothesis 1, frequency of use was regressed onto situational involvement (Model 1 in Table 1). As expected, there is a relationship between an individual's current health situation and the frequency of Internet usage to obtain health information. As an individual's health condition deteriorates, her involvement with her health problem increases, and there is more frequent usage of the Internet for health information. The direct test of situational involvement on frequency of use indicates that the relationship is significant (F = 4.36, p = 0.04).

Hypotheses 2a and 2b were to determine if the frequency of use impacts the overall perceived benefits as well as an individual's communications with his or her health care providers. To examine these relationships, life benefits and communication effects were regressed onto situational involvement and frequency of use (Models 2a and 2b, respectively). Hypothesis 2a suggests that increased frequency of use will result in greater accrued life benefits as a long-term effect of Internet usage for health information. As suggested, if the Internet is used more frequently for health information, individuals feel that their lives have benefited from the information (F = 17.83, p < 0.01). The path from frequency of use is also significant (t = 5.8, p < 0.01). Additionally, for the short-term effects of an individual's frequency of use, a significant model to predict communication effects (F = 4.71, p = 0.01) is found. Interestingly, a significant path from frequency of use to communication effects (t = 1.78, p = 0.08), as well as an even stronger path from the individual's situational involvement (t = -0.152, p = 0.03), is also found. This significant path indicates that people that are considered a health risk are more likely to discuss Internet information with their health care provider. Though the path was not predicted, its significance is not unexpected. People in generally poor health may discuss more information with their doctor and the Internet is a source of information.

To test for mediation, we followed the four-step procedure suggested by Baron & Kenny (1986) as well as Judd & Kenny (1981):

Step 1: Establish that there is an effect that may be mediated. Separately, Life Benefits and Communication Effects served as dependent variables in regression equations with Frequency of Use as the independent variable. This step resulted in a significant standardized path coefficient of 0.9 (t(219) = 30.95, p< .001) to Life Benefits and a significant standardized path coefficient of 0.5 to Communication Effects (t(218) = 8.05, p< .001).

Step 2: Establish that the initial variable has an effect on the proposed mediator. Again in separate regressions, Frequency of Use was regressed onto the proposed mediators Exploratory Behavior and Search Intensity. This step resulted in a significant standardized path coefficient of 0.9 (t(216) = 33.13, p<.001) to Exploratory Behavior and a significant standardized path coefficient of 0.8 to Search Intensity (t(211) = 19.14, p<.001).

Step 3: Establish that the proposed mediator affects the outcome variable. Exploratory Behavior was regressed onto Life Benefits resulting in a significant standardized path coefficient of 0.9 (t(218) = 43.75, p<.001). Likewise, Search Intensity was regressed onto Communication Effects resulting in a significant standardized path coefficient of 0.6 (t(212) = 10.69, p<.001).

Step 4: Establish that the proposed mediator completely mediates the relationship by including both variables (from step 2 and 3) as predictors of the dependent variable. The direct path should not be significant indicating complete mediation. Model 3 provides this mediational regression for P3, Exploratory Behavior mediating the relationship from Frequency of Use onto Life Benefits. From Table 1, we can see that standardized path coefficients are both significant indicating partial mediation. The amount of mediation is defined as the reduction of the effect of the initial variable on the outcome variable (Baron & Kenny, 1986); in this case we can see that the direct path is reduced by 0.7, suggesting that approximately 78% of the effect of Frequency of Use on Life Benefits is mediated by Exploratory Behavior. Model 4 provides the mediational regression for P4, Search Intensity mediating the relationship from Frequency of Use onto Communication Effects. From Table 1, we can see that standardized path coefficients are not significant for Frequency of Use but are significant for Search Intensity, indicating complete mediation (t(209) = 5.81, p<.001).

---