Using the Extended Parallel Process Model to Explain Physicians' Decisions to Test Their Patients for Kidney Disease

Abstract

A study was conducted to determine the effects of perceived threat and efficacy on physicians' intentions and behavior to test their patients' level of kidney functioning. Guided by the extended parallel process model, 151 physicians completed an initial survey measuring threat to patients (i.e., susceptibility and severity) and efficacy (i.e., response-efficacy and self-efficacy), as well as their current intentions and behavior. One-hundred and twelve of these physicians also completed and returned a follow-up survey sent approximately 4 months later using identical measures of intentions and behavior. As predicted, physicians who perceived greater threat to patients and greater efficacy demonstrated greater intentions and behavior to test their patients' level of kidney functioning. The theoretical and practical insights and implications of these findings are discussed.

Notes

¹This study was part of a larger study designed to determine physicians' perceptions of threat and efficacy regarding kidney disease and its detection, and also to determine how to best influence physicians' intentions and behaviors. In order to reduce complexity and provide a more focused report, we present just the results of our test of the EPPM in this article.

^a Using Bureau of Census (2002) classifications.

^a Mean item scores for behavior intentions are on a seven-point scale; mean item scores for behavior are on a six-point scale. Standard deviations are in parentheses.

^b Degrees of freedom are in parentheses.

∗p ≤ .01.

²The MANOVA procedure excludes cases listwise, meaning cases with missing values for any variable are excluded from all analysis. Thus, separate MANOVA were run for the initial and follow-up survey data to avoid the loss of participants who responded to the initial survey but not to the follow-up survey (n = 39), as well as the loss of participants who responded to both surveys but opted not to answer one of the dependent variables under investigation (n = 4). To be thorough, we still conducted the single MANOVA using all four dependent variables, and this analysis also yielded significant results, Wilks' λ = .72, F (12, 267.51) = 2.85, p <.001, η² = .10. Further, follow-up univariate analysis (i.e., one-way ANOVA with contrasts) for all four dependent variables were significant at the p < .05 level, and the pattern of means were consistent with the hypothesis for all four dependent variables (i.e., the mean for the high threat/high efficacy group was always greater than the means for the other three groups). The resulting 25% loss of power for the behavior measure on the initial survey, however, prevented many of the mean differences from reaching significance at the .05 level in the post hoc analysis with this variable. In sum, even under these most conservative of circumstances, the EPPM prediction still held in three of four instances. And, when a less conservative approach that minimizes the loss of research participants from the analyses is used, the EPPM predictions hold up in every instance.