The Role of Counterfactual Thinking in Narrative Persuasion: Its Impact on Patients’ Adherence to Treatment Regimen

ABSTRACT

The purpose of this study is twofold: (1) to explicate the underlying process of how narratives, accompanied with counterfactual thinking, exert cognitive and affective influence on audiences and (2) to examine how counterfactual thinking and regulatory focus, as story characteristics, enhance the persuasiveness of the narrative. Participants in the experiment were exposed to animated narratives in which the protagonist described her nonadherence to the peritoneal dialysis treatment regimen that resulted in her hospitalization. One hundred thirty-six patients undergoing peritoneal dialysis participated in a 2 (Goal failure framing: promotion-framed failure versus prevention-framed failure) by 2 (Counterfactual thinking: additive counterfactuals versus subtractive counterfactuals) between-subject factorial experiment. The analyses showed that narratives with additive counterfactuals, as opposed to those with subtractive counterfactuals, elicited greater anticipated regret and mental simulation, and, in turn, influenced the audience’s attitude toward and intention of adherence. More important, promotion-/prevention-framed failure and additive/subtractive counterfactuals jointly influenced the audience’s anticipated regret and mental simulation. Specifically, in the prevention-framed goal failure condition, narratives with additive counterfactuals elicited greater anticipated regret and mental simulation; however, in the promotion-framed goal failure condition, there was no significant difference on anticipated regret and mental simulation between narratives with subtractive counterfactuals and those with additive counterfactuals. The theoretical and practical implications were discussed.

Acknowledgments

The author expresses thanks to Kelvin Lee, PhD, for his assistance to develop the animated narratives for this experimental research, and to M. F. Lam, M.D., for his advice on the content related to peritoneal dialysis and peritonitis in the animated narratives.

Notes

¹ For the validity of this study, excluded from the recruitment process were patients (a) only on intermittent peritoneal dialysis; (b) switching to hemodialysis; and (c) physically, mentally, or psychologically incapable of verbally answering questions.

² These groups consist of patients with kidney related diseases who want to develop a stronger network and to receive peer support from other renal patients. An executive committee, formed by a group of volunteer patients, operates each self-help group. To increase the renal patient response rate, this study collaborated with the groups for participant recruitment because (1) these self-help group committee members have a wide personal network of peritoneal dialysis patients and are aware of which treatment their members are on, and thus, saved resources to identify and reach potential participants; and (2) there is a high level of trust between the committee and the patients as well as among the patients themselves. Therefore, PD patients responded more positively to the research invitation. Eight out of nine Hong Kong renal patients’ self-help groups agreed to assist in the participant recruitment.

³ The research assistant read the survey questions to the participants for two reasons. First, because a large portion of the end-stage renal disease patients in Hong Kong is over 60 years old (Leung, Cheung, & Li, 2015), a significant number of participants in this study were elderly (35.6%), and some of them have weak eyesight or limited reading and writing skills. Second, many patients undergoing long-term dialysis commonly experience fatigue (Jhamb, Weisbord, Steel, & Unruh, 2008). Thus, reading the survey items aloud to them was likely to lower the burden of filling out the questionnaire by themselves and was likely to increase their willingness to participate in the study. To avoid the interviewer’s bias, before the data collection, the assistant underwent a training session conducted by an experienced researcher.

⁴ When developing the measure of the mental simulation construct, during the searching process it was unable to identify a well-established multi-item scale in the literature of narrative persuasion. Although two related studies were found (Escalas, 2004; Janssen, Van Osch, De Vries, & Lechner, 2013), their measures were not useful for measuring mental simulation in the context of narratives accompanied with counterfactual thinking. However, a two-item measure in Escalas (2004) and a single-item measure in Janssen et al. (2013) were used in this current study as a background reference to construct the item of mental simulation.

⁵ The confirmatory factor analyses (CFA) showed that the factor loadings of the three statements of attitude toward adherence to the treatment regimen were (1) “To prevent peritonitis, carrying out the treatment procedures while exchanging fluid is necessary.” = .71, p < .001; (2) “For the sake of my health, it’s important to perform the treatment procedures.” = .70, p < .001; and (3) “Performing the treatment procedures while exchanging the fluid is helpful to prevent peritonitis.” = .73, p < .001. The factor loading of the three statements of behavioral intention were (1) “I will carry out the treatment procedures while exchanging fluid.” = .64, p < .001; (2) “I am going to practice the treatment procedures while exchanging fluid.” = .63, p < .001; and (3) “I intend to perform the treatment procedures as shown in the narrative.” = .81, p < .001. The factor loading of the three items of anticipated regret were (1) sorry = .61, p < .001; (2) regretful = .79, p < .001; and (3) remorseful = .75, p < .001.

⁶ According to guidelines for adequate and/or acceptable mode fit (Holbert & Stephenson, 2002; Hu & Bentler, 1995), a value of the root mean squared error of approximation (RMSEA) below .06 is considered a good fit, and a value less than or equal to .08 indicates an adequate fit with the upper bound of the 90% RMSEA confidence interval less than .10. A value of Comparative Fit Index (CFI) greater than .90 indicates an adequate fit, and a value greater than .95 suggests a good fit. The value of a standardized root mean squared residual (SRMR) of less than .08 is considered acceptable. A nonsignificant chi-square distributed test statistic (χ²) is considered a good fit; however, this statistic is sensitive to sample size. Therefore, the χ²/df was also reported where a value less than five suggested a good fit (Kline, 2005).

Additional information

Funding

This study was supported by the General Research Fund, Research Grant Council, Hong Kong (Grant No. HKBU12600115).