Comparison with literature
Our findings are in contrast with previous studies that found gain framing to be most effective in the promotion of physical activity (see Gallagher and Updegraff (19) for an extensive review). However, these studies mostly involved healthy participants (primarily students). This discrepancy in framing effects could be explained by the fact that OA patients often experience pain whilst being physically active. Our study showed that loss-framing was more effective in the promotion of pain medication use in support of being physically activite. Our findings are in line with Janke, Spring (20) who found that individuals with chronic pain were more motivated by a loss-framed message in the promotion of pain self-management.
Several other aspects might explain the preference for negatively framed messages in the present study. For instance, several studies have found support for the hypothesis that people with a high personal involvement with a subject favour negatively framed messages, and those with a low personal involvement favour positively framed messages (47–49). People with OA are likely highly involved in their own disease as self-management is an essential part of treatment. In addition, loss-framed messages are more effective for individuals with higher self-efficacy (50–52). As people with OA are likely highly involved in their own disease, they might score relatively high on self-efficacy. Although self-efficacy is not measured in the current study, high self-efficacy for self-management of OA was found in several other studies (53–56). Furthermore, the present findings are in line with previous research showing that loss framing is more effective than gain framing when highlighting the delayed or long-term behavioral consequences of adherence to treatment (57). As such individuals ponder on the delayed or long-term consequences, losses will likely be more salient than gains (58, 59). Our message focused on these delayed behavioral consequences, as OA is a chronic condition which requires long-term commitment to treatment. This might have enhanced the effect of loss framing.
Strengths and limitations
Our study also had several strengths and limitations. Firstly, one of the strengths of the present study was that we conducted our experiment with patients who experienced actual health problems. Previous research on message framing and testimonials were often conducted in a laboratory setting with college students (sometimes using hypothetical health problems) rather than in a patient setting with a more salient target population. In the present study, the Dutch Knee Panel was used for patient recruitment, consisting of people with OA symptoms in the knee from all over the country.
The involvement of a panel could also bring some limitations, such as the possibility of volunteer bias, and participants of the panel might possess different psychosocial and clinical characteristics from the target population (60). For instance, research shows that women are more likely to volunteer for studies than males (61), which might result in underrepresentation of males in studies. In addition, participants in our study were overall highly educated (55%). However, other sociodemographic and disease-related characteristics corresponded well to other studies that included Dutch primary care patients with OA (33, 62).
A further limitation may be that the intervention and survey of the present study was internet based. Certain populations are less likely to have internet access and to respond to online questionnaires. In addition, within the online survey system that was used, there was no way to control for the fact whether the participant had watched the video or not. This means participant who did not watch the video might have given a false statement in order to continue to the rest of the questionnaire. Furthermore, there was an average time difference of 21 months between the enrolment questionnaire and the present survey. As such, some of the information in Table 1 might not be entirely accurate .
In addition, the intended sample size of 286 participants was not met, because the response rate was lower than expected. Thus, our study might have been underpowered to detect relevant effects of framing. As such, results of this study are exploratory. Although we did detect a statistically significant effect of framing on pain medication attitudes, a larger sample size might have resulted in more statistically significant effects of framing on other variables.
The present study tested the effects of message framing and the use of a patient testimonial on several dependent variables, including attitudes, behavioral beliefs and intentions. This multiple testing increases the risk of drawing a false-positive conclusion (63, 64). However, whilst we only found significant effects for pain medication beliefs, associations with other variables showed similar, consistent, but not statistically significant, patterns.
Furthermore, we only measured intentions to change behaviour and did not assess actual behavior. However, findings of several studies suggest that forming an intention to change is a critical factor to changing actual behaviors (65). In addition, intentions were measured by two single items. As there is no multi-item counterpart available, no reliability test could be performed. However, we believe there is good theoretical reason to suspect that these measures provide an adequate assessment of the construct of interest and that they have good face validity (66)(67). As we wanted to measure the readiness to change instead of a simple agree/disagree answer option, we chose to use a clearly defined single item for both intentions with the use of Likert scales originating from the validated Multidimensional Pain Readiness to Change Questionnaire. Single-item measures are acceptable when constructs are unidimensional, clearly defined, and narrow in scope (68).
Messages were developed through an iterative process consisting of focus groups with two patient representatives (JM, SA) and several specialists in the field (JO, ED, CE), and explicitly used antonyms for the key messages in otherwise identical texts. As such, the assumption that both framing and the testimonial had matching counterparts is quite justifiable and it was decided not to pre-test. Nonetheless, results showed that participants perceived the messages to emphasize more on the negative consequences when the testimonial was present within the gain-frame than when the testimonial was absent within the gain-frame. This might be because the messages in which a testimonial was embedded started with the patient’s mother, whose symptoms only worsened as she did not adhere to self-management behaviors for OA. This additional sentence was not present in the condition where the testimonial was absent. In addition, the texts had an emphasis on pain, complaints and symptoms. This could have led to an underestimation of physical activity items. Nevertheless, participants still scored high on physical activity attitudes, beliefs and intentions regardless of which conditions participants were in. Such inconsistencies could have been avoided by means of pretesting.
Recommendations
First of all, as the present study was underpowered, results of this study are likely preliminary and further research with a larger sample size is advised. Respondents showed positive beliefs on the benefits of physical activity and high intentions to be physically active regardless of the type of message. This might be explained by a ceiling effect of beliefs and intentions for physical activity.. These findings implicate that it may be better to focus on diminishing barriers for engaging in physical activity, of which pain is an important one (69). Interventions concerning health behaviors in OA may thus benefit from embedding loss-framed messages on the importance of pain medication use to support physical activity to influence beliefs and intentions.
Results of this study are a generalization and reflect differences on group level. For one-on-one interactions such as consultations, effects will likely be dependent on the specific characteristics of the patient in question. Nowadays, healthcare providers often adapt their narrative intuitively. Ideally, they should have clues as to which technique to use for who. However, further research is needed into the underlying determinants of framing effects, such as the patient characteristics within our OA population.
In addition, the present study consisted of a message at a single time point. A few studies have used a “multi-dose” approach, where participants read a framed message at multiple points in time. Framing-manipulation has been shown to be more effective in promoting healthy behaviors when a “multi-dose” approach was used compared to a “single-dose” (e.g., Latimer, Rench (70)). In the case of our experiment, responses were measured directly after the message. Furthermore, our single-dose makes it unclear how long effects would have lasted. Nevertheless, repetition of the message over time might have enhanced the effect.