Abstract
Background
The same information about the evidence on health effects can be framed either in positive words or in negative words. Some research suggests that positive versus negative framing can lead to different decisions, a phenomenon described as the framing effect. Attribute framing is the positive versus negative description of a specific attribute of a single item or a state, for example, "the chance of survival with cancer is 2/3" versus "the chance of mortality with cancer is 1/3". Goal framing is the description of the consequences of performing or not performing an act as a gain versus a loss, for example, "if you undergo a screening test for cancer, your survival will be prolonged" versus "if you don't undergo screening test for cancer, your survival will be shortened".
Objectives
To evaluate the effects of attribute (positive versus negative) framing and of goal (gain versus loss) framing of the same health information, on understanding, perception of effectiveness, persuasiveness, and behavior of health professionals, policy makers, and consumers.
Search methods
We searched the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library, issue 3 2007), MEDLINE (Ovid) (1966 to October 2007), EMBASE (Ovid) (1980 to October 2007), PsycINFO (Ovid) (1887 to October 2007). There were no language restrictions. We reviewed the reference lists of related systematic reviews, included studies and of excluded but closely related studies. We also contacted experts in the field.
Selection criteria
We included randomized controlled trials, quasi‐randomised controlled trials, and cross‐over studies with health professionals, policy makers, and consumers evaluating one of the two types of framing.
Data collection and analysis
Two review authors extracted data in duplicate and independently. We graded the quality of evidence for each outcome using the GRADE approach. We standardized the outcome effects using standardized mean difference (SMD). We stratified the analysis by the type of framing (attribute, goal) and conducted pre‐planned subgroup analyses based on the type of message (screening, prevention, and treatment). The primary outcome was behaviour. We did not assess any adverse outcomes.
Main results
We included 35 studies involving 16,342 participants (all health consumers) and reporting 51 comparisons.
In the context of attribute framing, participants in one included study understood the message better when it was framed negatively than when it was framed positively (1 study; SMD ‐0.58 (95% confidence interval (CI) ‐0.94 to ‐0.22); moderate effect size; low quality evidence). Although positively‐framed messages may have led to more positive perception of effectiveness than negatively‐framed messages (2 studies; SMD 0.36 (95% CI ‐0.13 to 0.85); small effect size; low quality evidence), there was little or no difference in persuasiveness (11 studies; SMD 0.07 (95% CI ‐0.23 to 0.37); low quality evidence) and behavior (1 study; SMD 0.09 (95% CI ‐0.14 to 0.31); moderate quality evidence).
In the context of goal framing, loss messages led to a more positive perception of effectiveness compared to gain messages for screening messages (5 studies; SMD ‐0.30 (95% CI ‐0.49 to ‐0.10); small effect size; moderate quality evidence) and may have been more persuasive for treatment messages (3 studies; SMD ‐0.50 (95% CI ‐1.04 to 0.04); moderate effect size; very low quality evidence). There was little or no difference in behavior (16 studies; SMD ‐0.06 (95% CI ‐0.15 to 0.03); low quality evidence). No study assessed the effect on understanding.
Authors' conclusions
Contrary to commonly held beliefs, the available low to moderate quality evidence suggests that both attribute and goal framing may have little if any consistent effect on health consumers' behaviour. The unexplained heterogeneity between studies suggests the possibility of a framing effect under specific conditions. Future research needs to investigate these conditions.
Keywords: Humans, Health Behavior, Persuasive Communication, Comprehension, Consumer Health Information, Consumer Health Information/methods, Health Communication, Health Communication/methods, Perception, Randomized Controlled Trials as Topic
Plain language summary
Framing of the health information message in either negative or positive words
Examples illustrating different types of framing
The same information about health effects can be framed either in positive words or in negative words. Attribute framing refers to the positive versus negative description of a specific attribute of a single item or a state, for example, "the chance of survival with cancer is 2/3" versus "the chance of mortality with cancer is 1/3". Goal framing is the description of the consequences of performing or not performing an act as a gain or a loss, for example, "if you undergo a screening test for cancer, your survival will be prolonged" versus "if you don't undergo screening test for cancer, your survival will be shortened".
Summary
There is a widely held belief that framing of health information messages can lead to different decisions and different health behaviours; this is described as the 'framing effect'.
This systematic review identified 35 studies of 16,342 people testing this hypothesis. It found that both attribute and goal framing may have little if any effect on health consumers' behavior.
In one study of attribute framing, participants understood the message better when it was framed negatively than when it was framed positively. Positively‐framed messages may have led to more positive perception of effectiveness than negatively‐framed messages. There was little or no difference in persuasiveness in the context of attribute framing.
In the context of goal framing, loss messages led to a more positive perception of effectiveness compared to gain messages for screening messages, and may have been more persuasive for treatment messages.
The unexplained differences in the results of the included studies suggests the framing effect may exist under specific but yet undetermined conditions. Future research needs to investigate these conditions.
Summary of findings
Summary of findings for the main comparison. Summary of findings table 1: Attribute framing.
| Attribute framing | ||||
| Patient or population: consumers Intervention: negatively‐framed health messages Comparison: positively‐framed health messages | ||||
| Outcomes | Average effect with positive message * | No of Participants (studies) | Quality of the evidence (GRADE) | Comments |
| Understanding | 0.58 SDs lower (0.94 to 0.22 lower) | 124 (1 study) | ⊕⊕⊕⊝ low1,2 | 1.2 point difference on a 10 point Likert scale. (moderate effect size |
| Perception (measured as rating on a scale of perceived effectiveness) | 0.36 SDs higher (0.13 lower to 0.85 higher) | 226 (2 studies) | ⊕⊕⊝⊝ low1,3 | 0.7 point difference on a 10 point Likert scale. (small effect size) |
| Persuasiveness (measured as a hypothetical decision or intention or willingness to adopt an intervention) | 0.07 SDs higher (0.23 lower to 0.37 higher)) | 1068 (8 studies) | ⊕⊕⊝⊝ low1,4,5 | 0.2 point difference on a 10 point Likert scale (no difference). |
| Behavior | 0.09 SDs higher (0.14 lower to 0.31 higher) | 282 (1 study) | ⊕⊕⊕⊝ moderate5,6 | 0.2 point difference on a 10 point Likert scale (no difference) |
| * SD: standard deviation We interpreted SMDs using the following rules suggested by the Cochrane Handbook: <0.40 represents a small effect size 0.40 to 0.70 represents a moderate effect size >0.70 represents a large effect size We back translated the point estimates of the results by multiplying SMD by the standard deviation from a representative study (SD of 2 on a 10 point Likert‐type scale). | ||||
| GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. | ||||
1 Indirectness: outcome is a surrogate for health behavior.
2 Imprecision: relatively small number of participants. 3 Imprecision: the CI includes both values in favour of negative framing and values in favour of positive framing. 4 Inconsistency: out of the 7 studies, 1 was in favour of negative framing, 4 were in favour of positive framing, and 3 were in favour of neither. I2=85%. Subgroup analysis based on the type of message found that I2 was low (0%) in only one subgroup including 2 studies (harm messages). 5 Did not downgrade for imprecision because CI included mostly values reflecting no to small effects. 6 Risk of bias: for the one included study, it was not reported whether allocation was concealed nor whether there were losses to follow‐up.
Summary of findings 2. Summary of findings table 2: Goal framing.
| Goal framing | ||||
| Patient or population: consumers Intervention: loss‐framed health messages Comparison: gain‐framed health messages | ||||
| Outcomes | Average effect with gain message * | No of Participants (studies) | Quality of the evidence (GRADE) | Comments |
| Perception ‐ screening message (measured as rating on a scale of perceived effectiveness) | 0.30 SDs lower (0.49 to 0.10 lower) | 513 (5 studies) | ⊕⊕⊕⊝ moderate1 | 0.6 point difference on a 10 point Likert scale (small effect size) |
| Perception ‐ prevention message (measured as rating on a scale of perceived effectiveness) | 0.11 SDs higher (0.12 lower to 0.33 higher) | 815 (9 studies) | ⊕⊕⊝⊝ low1,2,3 | 0.2 point difference on a 10 point Likert scale (no difference) |
| Persuasiveness ‐ screening message (measured as a hypothetical decision or intention or willingness to adopt an intervention) | 0.06 SDs higher (0.23 lower to 0.35 higher) | 931 (6 studies) | ⊕⊕⊝⊝ low1,2,3 | 0.1 point difference on a 10 point Likert scale (no difference) |
| Persuasiveness ‐ prevention message (measured as a hypothetical decision or intention or willingness to adopt an intervention) | 0.02 SDs higher (0.11 lower to 0.16 higher) | 1496 (13 studies) | ⊕⊕⊝⊝ low1,2 | 0 point difference on a 10 point Likert scale (no difference) |
| Persuasiveness ‐ treatment message (measured as a hypothetical decision or intention or willingness to adopt an intervention) | 0.50 SDs lower (1.04 lower to 0.04 higher) | 1788 (3 studies) | ⊕⊝⊝⊝ very low1,2,4 | 1 point difference on a 10 point Likert scale (moderate effect size) |
| Behavior | 0.06 SDs lower (0.15 lower to 0.03 higher) | 11629 (16 studies) | ⊕⊕⊝⊝ low2,5,6 | 0.1 point difference on a 10 point Likert scale (no difference) |
| * SD: standard deviation We interpreted SMDs using the following rules suggested by the Cochrane Handbook: < 0.40 represents a small effect size 0.40 to 0.70 represents a moderate effect size > 0.70 represents a large effect size We back translated the point estimates of the results by multiplying SMD by the standard deviation from a representative study (SD of 2 on a 10 point Likert‐type scale). | ||||
| GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate. | ||||
1 Indirectness: outcome is a surrogate for health behavior. 2 Inconsistency: Statistical test for heterogeneity shows a significant P value; large I2. 3 Did not downgrade for imprecision because confidence interval (CI) included mostly values reflecting no to small effects. 4 Imprecision: the CI includes both values in favour of loss framing and values in favour of gain framing. 5 Sensitivity analysis included 13 high quality comparisons and found the SMG at ‐0.07 (95% CI ‐0.18 to 0.04), consistent with the main analysis. 6 Risk of bias: the inverted funnel suggested a possible publication bias in favour of loss framing.
Background
Description of the condition
The clear and effective communication of research evidence is important for ensuring the success of evidence based practice and the participation of health consumers in healthcare decisions. Such communication should lead to decisions that are consistent with the values and preferences of the health consumer. The same information about health effects can be framed either in positive words or in negative words. The way information is framed can lead to different decisions, a phenomenon described as the framing effect.
Description of the intervention
Schneider 1995 proposed a taxonomy classifying the framing effect into three types :
'Risky choice framing' (see Appendix 1). We did not consider this type of framing in our review as it is not applicable to health messages communicated in public health or clinical settings.
'Attribute framing', also called 'statistical framing' and 'different consequences framing' (see Appendix 2 and the table below). Attribute framing is the positive versus negative description of a specific attribute of a single item or a state. For example, "the chance of survival with cancer is 2/3" versus "the chance of mortality with cancer is 1/3".
'Goal framing', also called 'action framing', 'behavior framing' and 'same consequences framing' (see Appendix 2 and the table below). Goal framing is the depiction of the consequences of performing or not performing an act as a gain versus a loss. For example, "if you undergo screening for cancer, your survival will be prolonged" versus " if you don't undergo screening for cancer, your survival will be shortened".
| Type of framing | Attribute Framing | Goal framing | ||||
| Type of message | Positive | Negative | Gain | Loss | ||
| Gain: attain a desirable outcome | Non‐loss: not attain an undesirable outcome | Loss: attain an undesirable outcome | Non‐gain: not attain a desirable outcome | |||
| Example | The chance of survival with cancer is 2/3 | The chance of mortality with cancer is 1/3 | If you undergo screening for cancer, your survival will be prolonged | If you undergo screening for cancer, your survival won't be shortened | If you don't undergo screening for cancer, your survival will be shortened | If you don't undergo screening for cancer, your survival won't be prolonged |
The Characteristics of included studies provides examples of these two types of framing by reproducing the framing messages used in each study.
How the intervention might work
According to the Prospect theory (Kahneman 1979; Tversky 1981), people make decisions by using a reference point to judge whether a particular outcome is a gain or a loss. Positively‐framed messages and negatively‐framed messages induce different shifts in the location of the reference point, leading to different effects on risk seeking.
Specifically, in goal framing, a negatively‐framed message emphasizing losses ("if you do this, you will suffer this negative consequence") may have a greater impact on a given behavior than a comparable positively‐framed message emphasizing gains ('if you do this, you will have this positive outcome"). Various theories (see Appendix 1) suggest that certain healthcare messages (for example, prevention, screening and therapeutic messages) are more or less prone to framing effects (Kahneman 1979; Tversky 1981; Rothman 1997).
The framing effect might vary in relation to screening, prevention and therapy due to the baseline risk of the condition. With screening decisions, the baseline risk of the condition that is being screened for is relatively low. With therapeutic decisions, the baseline risk is considered to be nearly 100%. Screening focuses on detecting disease with assumed consequences of later management, while therapeutic messages focus on change of the disease (that is, they deal with altering the course of a disease). These factors may influence health behavior (whether or not to undergo screening or therapy) and this behavior may be influenced by differentially framed messages.
Why it is important to do this review
Levin 1998 published a systematic review on the framing effect based on their proposed taxonomy, which is summarized above (Schneider 1995). They showed that within each type of framing, results show substantial consistency. In attribute framing, attributes are judged more favourably when labelled positively rather than negatively. In goal framing, a negatively‐framed message emphasizing losses tends to have a greater impact on a given behavior than a comparable positively‐framed message emphasizing gains. However, Levin's systematic review was not restricted to health messages. It also did not explore the Rothman and Salovey hypothesis that the framing effect varies by the type of health question (such as screening, prevention or therapy).
Objectives
To evaluate the effects of attribute (positive versus negative) framing and of goal (gain versus loss) framing of the same health information, on understanding, perception of effectiveness, persuasiveness and behaviour of health professionals, policy makers, and consumers.
Methods
Criteria for considering studies for this review
Types of studies
Randomized controlled trials, quasi‐randomised controlled trials, and cross‐over studies.
Types of participants
Health professionals, policy makers, and consumers.
Types of interventions
Interventions of interest consisted of positively versus negatively‐framed messages (attribute framing) or gain‐framed versus loss‐framed messages (goal framing). The two messages being compared should describe the same health information. As mentioned above, we excluded studies of risky choice framing.
We excluded interventions that compared different statistical presentations of the same heath information (Akl 2011), alternative graphical or verbal presentations of the same evidence, alternative orders of comparing risks or comparisons, or alternative media to present the same information.
We also excluded studies in which participants chose between different interventions with different benefits and harms using alternative presentation formats, since any effects of the presentation formats would be completely confounded with any effects of the differences in benefits and harms.
Types of outcome measures
The primary outcome was actual decisions or behaviours. We were also interested in the surrogates of that outcome including: understanding, perception of effectiveness, and persuasiveness. An additional outcome of interest was the consistency of the decision with values and preferences. We did not assess any adverse outcomes.
We only considered objective understanding (for example, correctly stating which treatment is more effective after being presented with statistical data) and not self‐reported understanding. Perception of effectiveness refers to how effective an intervention is perceived to be (for example, rating of the perceived effectiveness of vaccination). Persuasiveness refers to how likely participants are to make a hypothetical decision in favour of an intervention (for instance, a hypothetical decision to treat cholesterol). Appendix 3 provides examples of outcomes assessed in the studies classified according to this review typology of outcomes. Also, the outcome had to relate to the health behavior of interest (that is, in an experiment of the framing of the use of sunscreen for preventing skin cancer, an eligible perception of effectiveness outcome should relate to the perception of the efficacy of using sunscreen and not understanding how large the risk of getting skin cancer is in relation to sun exposure).
Search methods for identification of studies
Electronic searches
The search was part of a larger search for studies assessing alternative presentations of the same empirical evidence about health. We conducted the initial search in June 2002, and updated it in September 2004 and October 2007.
We first used OVID to search MEDLINE (1966 to 2007), EMBASE (1980 to 2007), and PsycINFO (1887 to 2007). We searched the Cochrane Central Register of Controlled Trials (CENTRAL, The Cochrane Library Issue 3 2007) using FRAM* and PRESENT* as text words. We searched MEDLINE, EMBASE, and PsycINFO by ANDing a search for study type with a search for intervention type. We used no language or date restriction (see Appendix 4).
Searching other resources
We used the 'Related Articles' feature of PubMed MEDLINE to find additional articles. We searched MEDLINE and PsycINFO databases for articles published by the first authors of included studies and of excluded but closely related studies. We reviewed the reference lists of related systematic reviews, included studies and excluded but closely related studies. Finally, we contacted experts in the field.
Data collection and analysis
Selection of studies
Two review authors independently screened the title and abstract of identified articles for relevance. We retrieved the full text of those articles judged potentially relevant by at least one review author. Two review authors then independently screened the full text article for inclusion or exclusion. The review authors resolved their disagreements by discussion or by consulting a third review author. Studies excluded after examination of the full text are listed in the table Characteristics of Excluded Studies, with the reason for exclusion given.
Data extraction and management
We developed a standardized data extraction form. Two review authors extracted data from each included study in duplicate and independently. They resolved their disagreements by discussion or by consulting a third review author. We extracted data that related to study methods, participants, intervention, assessed outcome, and study results. We extracted data for the longer follow‐up time when the authors reported more than one. We attempted to contact authors for incompletely reported data.
We noted the type of message (prevention, screening, treatment, other), the type of framing (attribute, goal), and the attainment and valence of the message.
Assessment of risk of bias in included studies
Two review authors independently assessed the methodological quality of each included study and resolved their disagreements by discussion or by consulting a third review author. Methodological data included:
Allocation concealment. We considered the following methods of concealment of allocation as adequate: (1) central randomization; (2) sequentially numbered drug containers; and (3) sequentially numbered, opaque, sealed envelopes. For cross‐over studies we assessed the randomization of the order of interventions.
Randomization. We considered the following methods of sequence generation as adequate: (1) repeated coin‐tossing; (2) throwing dice; (3) dealing previously shuffled cards; (4) using a published list of random numbers; and (5) using a list of random numbers generated by a computer. For cross‐over studies we assessed the randomization of the order of interventions.
Objectivity and directness of outcomes: yes (e.g., real understanding, behavior); no (e.g., hypothetical outcome, such as hypothetical decisions).
We then graded the quality of the underlying evidence for each outcome using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach (Guyatt 2008; Higgins 2011). As our main outcome of interest was behavior, we considered understanding, perception of effectiveness, and persuasiveness as surrogates for behavior.
Measures of treatment effect
We analyzed the results of included studies for each eligible comparison. Because outcomes in these studies are typically scaled responses to survey questions, we standardized the effects using Hedges adjusted standardized mean difference (SMD). We also calculated the adjusted standard error for the resulting SMD. The SMD ensures that the differences across the scales are standardized to a uniform scale before they are combined.
For comparisons with binary measures (with proportions p1 and p2) we calculated z value using the following formula: z = (p1 ‐ p2) / (p * (1 ‐ p ) * [(1/n1) + (1/n2)])1/2 where p = (p1*n1 + p2*n2) / (n1 + n2). For comparisons with continuous measures where we were not able to calculate the SMD directly, we calculated t‐values using the means and standard errors: t = (m1‐m2)/(se22/n1 + se22/n2)1/2. In both cases, we calculated the corresponding SMD using SMD = 2t/sqrt(N) (Cooper 1994) and adjusted it using the same adjustment factor.
We analyzed data as they related to the intended health behavior, for example, we transformed data reported for "willingness to have unprotected sex" into data relating to "willingness to have protected sex.
Assessment of heterogeneity
We tested results for homogeneity across studies using the I2 test (Higgins 2003) and used the following interpretation of the value of I2: 0 to 50 = low; 50 to 80 = moderate and worthy of investigation; 80 to 100 = severe and worthy of understanding; 95 to100 = aggregate with major caution (Julian Higgins, personal communication).
Assessment of reporting biases
We created inverted funnel plots of individual study results plotted against inverse of the variance in order to check for possible publication bias.
Data synthesis
We stratified the analysis by the type of framing (attribute or goal). We pooled multiple outcome measures for a single trial ‐ for example, three different questions about perception of effectiveness or responses to three different scenarios by the same participants ‐ using fixed‐effect models into a single SMD for that comparison. We pooled data from different studies when appropriate using random‐effects models with the inverse variance approach. We interpreted SMDs using the following rules (Higgins 2011):
< 0.41 represents a small effect size
0.40 to 0.70 represents a moderate effect size
> 0.70 represents a large effect size.
In addition, we back‐translated the results by multiplying SMD by the standard deviation from a representative study (e.g., SD of 2 on a 10 point Likert‐type scale).
If a paper reported the results of two or more separate comparisons enrolling different participants, we treated these as separate comparisons. If a study used a between‐subjects factorial design to compare an intervention of interest across another factor, we treated these as separate comparisons if there was no evidence of interaction.
Subgroup analysis and investigation of heterogeneity
We conducted pre‐planned subgroup analyses based on type of message, which we divided into screening, prevention, treatment, and other (including diagnostic, harm and public health messages). When the test for interaction was statistically significant, we reported the SMD for each subgroup.
Sensitivity analysis
We conducted pre‐planned sensitivity analyses excluding studies of lower methodological quality (i.e., did not meet at least two of the three methodological criteria. When the test for interaction was statistically significant, we planned to conduct the sensitivity analysis for each subgroup separately.
Results
Description of studies
Results of the search
Our electronic searches identified a total of 23,493 unique citations (10,732 unique citations in June 2002, 2,637 additional unique citations in September 2004, and 10,124 additional unique citations in October 2007). The title and abstract screening identified 99 citations as potentially eligible for this review, and we retrieved these studies in full text. Screening of these 99 full text studies identified 35 eligible studies.
Included studies
The 35 included studies recruited a total of 16,342 participants. All studies were conducted with health consumers. The response rate was high (> 50%) in 13 studies (26%), low (≤ 50%) in 2 studies (4%), and not reported in 35 (70%) studies.
The 35 studies reported a total of 51 comparisons: 13 comparisons related to attribute framing and 38 comparisons of goal framing. The messages used in these comparisons were about:
screening (n = 19),
prevention (n = 19),
treatment (n = 8), and
other (n = 5: 2 harm, 1 diagnosis, 1 public health, 1 on abortion); 1 study (Wilson 1987) used 2 types of messages.
Some of the 51 comparisons included more than 1 outcome; there was a total of 69 outcome measurements: 1 for understanding, 16 for perception of effectiveness, 34 for persuasiveness, 17 for behavior, and 1 for consistency of decision with people's values.
Excluded studies
We excluded 64 studies for the following reasons: not a comparison of interest (n = 11); not an original study (n = 11); not an appropriate study design (n = 3); use of different information in the comparison groups (n = 18); no outcome of interest evaluated (n = 8); and data for the comparison of interest not reported and not available from author (n = 13). These are reported in the table Characteristics of excluded studies,
Risk of bias in included studies
Figure 1 and Figure 2 respectively show the methodological quality graph and summary of included studies. Of the 51 comparisons, 42 were randomized controlled trials (82%), 8 were quasi‐randomised controlled trials (16%), and 1 was a cross‐over study (2%). Of the 51 comparisons, allocation was concealed in 8 (16%), not concealed in 5 (10%), and unclear whether concealed or not in 38 (75%). The design was randomized in 42 (82%), not randomized in 1 (2%), and unclear whether randomized or not in 8 (16%) studies. Of the 51 comparisons, 17 (33%) evaluated an objective outcome.
1.
Methodological quality graph: review authors' judgements about each methodological quality item presented as percentages across all included studies.
2.
Methodological quality summary: review authors' judgements about each methodological quality item for each included study.
Effects of interventions
Attribute framing
Understanding
One study reported one comparison using a screening message (Lewis 2003). Negative framing led to better understanding than positive framing; the SMD was ‐0.58 (95% confidence interval (CI) ‐0.94 to ‐0.22), a moderate effect size corresponding to a 1.2 point difference on a 10 point Likert‐type scale (Analysis 1.1). The quality of the evidence was low (Table 1).
1.1. Analysis.
Comparison 1 Attribute Framing, Outcome 1 Understanding.
Perception of effectiveness
Two studies reported two comparisons using one screening (Lewis 2003) and one other (harm) (Donovan 2000) message. Positively‐framed messages led to more positive perception of effectiveness than negatively‐framed messages; the pooled SMD was 0.36 (95% CI ‐0.13 to 0.85) (I2 = 70%), a small effect size corresponding to a 0.7 point difference on a 10 point Likert‐type scale. The quality of the evidence was low (Table 1).
In subgroup analysis, the test for interaction was not statistically significant (P = 0.18).
Persuasiveness
Eight studies reported 11 comparisons using 2 screening messages (Welkenhuysen 2001a; Welkenhuysen 2001b), 1 prevention message (Blanton 2001a), 5 treatment messages (Krishnamurthy 2001a; Krishnamurthy 2001c; Llewellyn‐Thomas 1995; Wilson 1987a; Zimmermann 2000) and 3 other types of messages (2 harm (Donovan 2000; Jasper 2001) messages and 1 message about abortion (Wilson 1987b)). The inverted funnel plot did not suggest publication bias. Positive and negative framing did not lead to different persuasiveness; the pooled SMD was 0.07 (95% CI ‐0.23 to 0.37) (I2 = 83%) (Analysis 1.3), corresponding to a 0.1 point difference on a 10 point Likert‐type scale. The quality of the evidence was low (Table 1).
1.3. Analysis.
Comparison 1 Attribute Framing, Outcome 3 Persuasiveness.
In subgroup analyses, the test for interaction was statistically significant (P < 0.0001). The pooled SMD was ‐0.20 (95% CI ‐0.93 to 0.53) (I2 = 90%) for the screening messages; ‐0.61 (95% CI ‐0.98 to ‐0.25) for the prevention message; 0.33 (95% CI ‐0.14 to 0.80) (I2 = 79%) for the treatment messages; and 0.10 (95% CI ‐0.30 to 0.51) (I2 = 69%) for the other type messages. None of the screening or prevention messages' comparisons was of high quality to conduct sensitivity analysis. Only one treatment message comparison was of high quality (Llewellyn‐Thomas 1995) and the associated SMD was ‐0.14 (95% CI ‐0.64 to 0.37). Two of the other type messages (the two using harm messages (Donovan 2000; Jasper 2001)) were of high quality and the associated SMD was 0.31 (95% CI 0.04 to 0.58) (I2 = 0%).
Behavior
One study reported one comparison using a prevention message (O'Connor 1996). Positive and negative framing did not lead to different behavior; the SMD was 0.09 (95% CI ‐0.14 to 0.31) (Analysis 1.4), corresponding to a 0.2 point difference on a 10 point Likert‐type scale. The quality of the evidence was moderate (Table 1).
1.4. Analysis.
Comparison 1 Attribute Framing, Outcome 4 Behavior.
Goal framing
Understanding
None of the included studies assessed this outcome in the context of goal framing.
Perception of effectiveness
Eight studies reported 14 comparisons using 5 screening messages (Cherubini 2005a; Cherubini 2005b; Hsiao 2002b; Ruiter 2003; Schmitt 2004) and 9 prevention messages (Block 1995a ; Block 1995b; Block 1995c; Block 1995d; Detweiler 1999; Hsiao 2002b; Kroll 2005; van Assema 2001a; van Assema 2001b). The inverted funnel plot did not suggest publication bias. Gain and loss framing did not lead to different perception of effectiveness; the pooled SMD was ‐0.03 (95% CI ‐0.22 to 0.16) (I2 = 61%) (Analysis 2.1), corresponding to no difference on a 10 point Likert‐type scale.
2.1. Analysis.
Comparison 2 Goal Framing, Outcome 1 Perception of effectiveness.
In subgroup analysis, the test for interaction was statistically significant (P = 0.0003). For screening messages, loss framing led to more positive perception of effectiveness compared to gain framing; the pooled SMD was ‐0.30 (95% CI ‐0.49 to ‐0.10) (I2 = 0%) , a small effect size corresponding to a 0.6 point difference on a 10 point Likert‐type scale. The quality of the evidence was moderate (Table 2). For prevention messages, gain and loss framing did not lead to different perception of effectiveness; the pooled SMD was 0.11 (95% CI ‐0.12 to 0.33) (I2 = 57%) corresponding to a 0.2 point difference on a 10 point Likert‐type scale. The quality of the evidence was low (Table 2).
The sensitivity analysis included three high quality comparisons for the screening messages (Cherubini 2005a; Cherubini 2005b; Schmitt 2004) and the pooled SMD was ‐0.37 (95% CI ‐0.61 to ‐0.12) (I2 = 0%). The sensitivity analysis included one high quality comparison for the prevention message (Detweiler 1999) and the SMD was 0.08 (95% CI ‐0.18 to 0.35).
Persuasiveness
Fourteen studies reported 23 comparisons using 6 screening messages (Banks 1995; Cherubini 2005a; Cherubini 2005b; Hsiao 2002b; Schmitt 2004; Scott 2006b), 13 prevention messages (Block 1995a; Block 1995b; Block 1995c; Block 1995d; Detweiler 1999; Hsiao 2002a; Kroll 2005; O'Connor 2005; Robberson 1988a; Robberson 1988b; Scott 2006a; van Assema 2001a; van Assema 2001b), 3 treatment messages (Carling 2010; Krishnamurthy 2001b; Krishnamurthy 2001d), and 1 other type (public health) message (Steward 2002). The inverted funnel plot did not suggest publication bias. Gain and loss framing did not lead to different persuasiveness; the pooled SMD was ‐0.06 (95% CI ‐0.18 to 0.06) (I2 = 73%) (Analysis 2.2), corresponding to a 0.1 point difference on a 10 point Likert‐type scale.
2.2. Analysis.
Comparison 2 Goal Framing, Outcome 2 Persuasiveness.
In subgroup analysis, the test for interaction was statistically significant (P = 0.0006). For screening messages, gain and loss framing did not lead to different persuasiveness; the pooled SMD was 0.06 (95% CI ‐0.23 to 0.35) (I2 = 75%) corresponding to a 0.1 point difference on a 10 point Likert‐type scale The quality of the evidence was low (Table 2). For prevention messages, gain and loss framing did not lead to different persuasiveness; the pooled SMD was 0.02 (95% CI ‐0.11 to 0.16) (I2 = 53%) corresponding to no difference on a 10 point Likert‐type scale The quality of the evidence was low (Table 2). For treatment messages, loss messages may have been more persuasive than gain messages; the pooled SMD was ‐0.50 (95% CI ‐1.04 to 0.04) (I2 = 89%) a moderate effect size corresponding to a 1 point difference on a 10 point Likert‐type scale. The quality of the evidence was very low (Table 2).
The sensitivity analysis included five high quality comparisons for the screening messages (Banks 1995; Cherubini 2005a; Cherubini 2005b; Schmitt 2004; Scott 2006b) and the pooled SMD was ‐0.03 (95% CI ‐0.15 to 0.08) (I2 = 0%). The sensitivity analysis included three high quality comparisons for the prevention messages (Scott 2006a; Detweiler 1999; O'Connor 2005) and the pooled SMD was ‐0.04 (95% CI ‐0.20 to 0.13) (I2 = 56%). Only one treatment message comparison was of high quality (Carling 2010) and the associated SMD was ‐0.21 (95% CI ‐0.36 to ‐0.07).
Behavior
Thirteen studies reported 16 comparisons using 10 screening messages (Apanovitch 2003 ; Banks 1995; Finney 2002a; Finney 2002b; Lalor 1990; Lerman 1992; Rivers 2005b; Schneider 2001b; Schneider 2001c; Williams 2001), 5 prevention messages (Detweiler 1999; Mann 2004; McCaul 2002; Rivers 2005a; Schneider 2001a), and 1 other type (diagnostic) message (Lauver 1990). The inverted funnel plot suggested a possible publication bias in favour of loss messages (Figure 3). Gain and loss framing did not lead to different behavior; the pooled SMD was ‐0.06 (95% CI ‐0.15 to 0.03) (I2 = 63%) (Analysis 2.3), corresponding to a 0.1 point difference on a 10 point Likert‐type scale. The quality of the evidence was low (Table 2).
3.
Funnel plot of comparison: 2 Goal Framing, outcome: 2.3 Behavior.
2.3. Analysis.
Comparison 2 Goal Framing, Outcome 3 Behavior.
In subgroup analysis, the test for interaction was not statistically significant (P = 0.62). The sensitivity analysis included 13 high quality comparisons (Apanovitch 2003 ; Banks 1995; Detweiler 1999; Lauver 1990; Lerman 1992; Mann 2004; McCaul 2002; Rivers 2005a; Rivers 2005b; Schneider 2001aSchneider 2001b; Schneider 2001c; Williams 2001) and the SMD was ‐0.07 (95% CI ‐0.18 to 0.04) (I2 = 67%).
Consistency of decision with values
Only one study (Carling 2010) assessed the effects of alternative framing of message on the consistency of decision (to take antihypertensive medication) with values. The group that was presented gain‐framed information appeared to make decisions that were most consistent with a second, more fully informed decision. However, the results did not clearly support gain framing as the "best" in terms of ensuring consistency.
Additional results
Detweiler 1999 compared the two forms of gain‐framed messages (gain and non‐loss) and the two forms of loss‐framed messages (loss and non‐gain) (see Appendix 2). They found no differences between the two gain‐framed messages or between the two loss‐framed messages. Although Apanovitch 2003 compared the two forms for gain and loss messages, they did not report analyses on differences.
Appendix 5 summarizes the results of comparisons and studies for which data are not reported but which are otherwise eligible. The results were not consistent with a discernable pattern.
Appendix 6 presents the results for the effect modification of the framing effects by 18 different factors. The results of the most studied factor (that is, level of involvement) were inconsistent.
Discussion
Summary of main results
In the context of attribute framing, participants in one included study understood the message better when it was framed negatively compared to positively (moderate effect size; low quality evidence). Although positively‐framed messages led to a more positive perception of effectiveness (small effect size; low quality evidence), there was little if any difference in persuasiveness (low quality evidence) and attribute framing had no effect on health consumers' behavior (moderate quality evidence).
In the context of goal framing, and compared to gain messages, loss messages led to more positive perception of effectiveness for screening messages (small effect size; moderate quality evidence) and tended to be more persuasive for treatment messages (moderate effect size; very low quality evidence). However, they had little or no effect on health consumers' behavior (low quality evidence).
These results do not unequivocally support framing effect theories or frameworks (Kahneman 1979; Tversky 1981; Schneider 1995; Rothman 1997) in relation to health messages. Indeed, these theories and frameworks were based on observations in laboratory settings as opposed to natural settings. They were also based on effects on surrogate outcomes as opposed to behavior. Understanding of information may not necessarily lead to a positive perception or persuade individuals to do or not to do something. Also, persuasion might not last long enough or might be so hampered by barriers (such as those related to the healthcare system) that it might not translate into actual behavior.
Overall completeness and applicability of evidence
The effectiveness of communication of health messages is directly linked to its effects on clinicians' evidence‐based practice and consumers' informed medical decision making. However, none of the included studies targeted health professionals or policy makers. The vast majority of included studies did not directly assess the effects on consumers' behavior. Indeed, most of the evidence is based on hypothetical scenarios and relates to surrogate outcomes of understanding, perception of effectiveness and persuasiveness. Another valuable outcome is the consistency between the consumer's behavior and his or her values (that is, the relative importance of the desirable and undesirable effects of the intervention) (Carling 2008). No study assessed this outcome; only one study (Carling 2010) assessed the effects of goal framing on the consistency of decisions with values. The results did not clearly support the superiority of gain or loss framing in terms of ensuring consistency, although gain framing led to decisions most consistent with more fully‐informed decisions.
Quality of the evidence
The vast majority of the comparisons were randomized controlled trials. However, 74% of studies did not report on the adequacy of concealment of allocation. The sensitivity analysis excluding studies at a higher risk of bias did not substantially alter the results for the outcomes of perception of effectiveness, persuasiveness and behavior for goal framing.
The assessment of quality by outcome using the GRADE approach showed that the quality of evidence varied from low to moderate for attribute framing, and from very low to moderate for goal framing. The quality of evidence was mainly affected by the use of surrogate outcomes and inconsistency. In spite of this inconsistency, in general, there does not seem to be a framing effect (particularly because the I2 test is very powerful for standardised mean difference (SMD)). While there may be specific conditions under which an effect might exist, we were not able to determine such attributes.
Potential biases in the review process
Our electronic search strategy was designed for the effects of alternative presentations of risk information and not specifically for message framing. We plan to design a specific strategy for our next update of this review. In addition to using more specific and more adequate search terms, the updated strategy would benefit from widening the scope of searching (to include, for instance, CINAHL, ERIC, some trial registers and some grey literature). Although our electronic search may not have captured all eligible studies, our additional search strategies were apparently effective. Still, the date of search for this systematic review will be three years old by the time of its publication.
Because the included studies reported their outcomes using different scales, we had to use SMD to analyze and present our results. We tried to make our results more interpretable by back translating the SMD into a scaled value. However, simulation studies have found SMD to be biased towards no effect, especially with smaller studies (Van Den Noortgate 2003).
As noted above, our statistical approach for dealing with binary data in cross‐over trials is limited. Since we designed our initial plan of analysis, new statistical methods have been developed, and we intend to use them with our next update (Curtin 2002; Elbourne 2002).
Agreements and disagreements with other studies or reviews
Earlier reviews of framing effects have tended to be narrative, did not conduct meta‐analyses or did not distinguish different types of framing (Edwards 2001; Rothman 1997; Wilson 1988; Moxey 2003). We discuss here the results of three systematic reviews that included meta‐analyses (Kuhberger 1998; O'Keefe 2007; O'Keefe 2009).
Kuhberger 1998 did not restrict their systematic review to health messages and the main analysis included all types of framing (including risky choice framing). The review found small to moderate differences between framing conditions (about a third of a standard deviation). However, the effect was mainly driven by studies of risky choice framing. The analysis found that both attribute framing (referred to as "framing by outcome salience") and goal framing (referred to as "message compliance design") were "ineffective".
O'Keefe 2007 conducted a systematic review assessing the persuasiveness of gain‐framed and loss‐framed messages for prevention behaviours. They found that gain‐framed appeals are statistically significantly more persuasive than loss‐framed appeals. However this superiority was driven by studies of dental hygiene behaviours and absent in studies of other preventive behaviours. Also, although statistically significant, the superiority was "quite small", corresponding to a correlation of 0.03. The authors concluded that the effect is so small that it should be considered negligible.
O'Keefe 2009 conducted another systematic review assessing the persuasiveness of gain‐framed and loss‐framed messages for screening behaviours. They found that loss‐framed appeals were more persuasive than gain‐framed appeals. However this superiority was driven by studies of breast cancer screening and absent in studies of other screening behaviours. Also, although statistically significant, the superiority was "quite small", corresponding to a correlation of ‐0.039. They concluded that loss‐framed messages are unlikely to substantially improve persuasiveness.
In both their reviews, O'Keefe 2007 and O'Keefe 2009 considered attitude change, post‐communication agreement, behavioral intention, and behavior as measures of persuasion, and averaged their measures to yield a single summary when more than one was reported.
Authors' conclusions
Implications for practice.
The main finding of this study is that attribute and goal framing has little if any consistent effect on behavior (low to moderate quality evidence). This finding is strengthened by its consistency with findings of other well‐conducted systematic reviews. Any effect (for example, in breast cancer screening) would possibly be further attenuated by the multiple complex factors (such as access to care and adequacy of follow‐up) that affect heath behavior, and eventually health outcomes. Moreover, there is evidence that the framing effects for both prevention and detection messages weaken with time (Rivers 2005a).
In the absence of evidence for the superiority of one frame over the other, a balanced presentation when producing patient information or decision aids is likely to be the safest approach.
Implications for research.
Future research in the field of health communication should use high quality randomised controlled trial design, be conducted in real life settings, and assess outcomes such as actual behavior and the consistency of choices and behaviours with preferences and values. Future updates of this systematic review should aim to explain the unexplained heterogeneity between studies; that is, through meta‐regression analyses exploring potential effect modifiers such as different perceived consequences (avoiding a bad health state as opposed to attaining a good health state) (See Appendix 2), type of heath message, level of baseline risk, level of involvement, and perceived susceptibility. Currently, such exploration is limited by the number of available studies, their quality, the large number of potential effect modifiers and potential confounding.
This review illustrates the danger of generalizing theory and laboratory research into clinical settings. Also, in the business field, a message framing intervention in a laboratory setting led to very different results in natural settings (Ganzach 1997). These facts highlight the need to test communication interventions empirically. While these studies would ideally take place in real life settings, they are complex and possibly not worth the cost. Carefully designed studies (with low risk of bias) using hypothetical scenarios may provide further evidence as to whether or not real life experiments are worth conducting.
Acknowledgements
The authors would like to thank Dominique Broclain for his contributions to the protocol for this review, the editors and members of the Cochrane Consumers and Communication Review Group, particularly Sophie Hill and Megan Prictor, for their careful review and editing, Ms. Ann Grifasi for her administrative assistance, and the authors of the primary studies who provided us with missing or supplementary information. They would also like to thank the State University of New York at Buffalo, the Italian National Cancer Institute Regina Elena and the National Institute of Public Health and the Norwegian Research Council for supporting their efforts on this review.
Appendices
Appendix 1. Theories of message framing
In their classic 'Asian disease' problem, Kahneman and Tversky asked study participants to choose between two alternative programs to combat an outbreak of an unusual Asian disease (Tversky 1981). With the first program, 200 people out of 600 will be saved, while with the second program there is a 1/3 probability that 600 people will be saved and a 2/3 probability that no people will be saved. Although the two alternatives have equivalent outcomes, the first is certain while the second is risky. When the outcomes of the scenario were framed positively (e.g. 200 will be saved) people were risk averse (i.e. selected the certain alternative over the risky alternative). When the outcomes of the scenario were framed negatively (e.g. 400 will die) people were risk seeking (i.e. selected the risky alternative over the certain alternative). Although the two alternatives have equivalent outcomes the first is certain while the second is risky in each of these two differently framed messages.
Kahneman and Tversky developed the prospect theory to explain the framing effect observed in their Asian disease experiment (Kahneman 1979; Tversky 1981). The theory argues that people make decisions by using a reference point to judge whether a particular outcome is a gain or a loss. Positively‐framed messages and negatively‐framed messages induce different shifts in the location of the reference point leading to different effects on risk seeking. This leads to the observed tendency of risk aversion with positively‐framed problems, and risk seeking with negatively‐framed problems.
Rothman and Salovey built on Kahneman and Tversky’s work to extend these concepts to the health domain. They hypothesized that gain frames would be more effective for disease prevention and recuperative (therapeutic) behaviors and that loss‐frames would be more effective for disease detection (screening) behaviors (Rothman 1997).
Indeed, prevention behaviors, such as the use of sunscreen or vaccination, are performed to prevent the onset of a health problem and are thus perceived as not risky. Similarly recuperative behaviors, such as chemotherapy or radiotherapy, may be performed to halt the progression of a health problem and/or to prevent its complications and are thus perceived as not risky. Based to the prospect theory, Rothman and colleagues suggested that gain frames should be more persuasive for these types of risk averse behaviors.
On the other hand, detection behaviors, such as colorectal endoscopy and mammography, are performed to discover early signs of a health problem. Although they are supposed to minimize long terms risks, they are perceived as risky in the short term. Based on the prospect theory, Rothman and colleagues suggested that loss frames should be more persuasive for these types of risk seeking behaviors.
Appendix 2. Definitions of attribute and goal framing
1. Attribute framing/ Statistical framing/ Different consequences framing
Framing relates to the description of a specific attribute of a single item or a state either positively or negatively: "Among women who get cancer two thirds/one third will survive/die from the disease"
2. Goal framing/ Behavior framing/ Same consequences framing
Framing highlights performing or not performing a behavior and the subsequent presence or absence of an outcome: "When you get/do not geta screening test, you will detect/will not detect cancer in early stages". The framing of the outcome part of the message is constructed along the 2 dimensions of attainment (attain versus not attain) and valence (desirable versus undesirable) (Brendl 1995; Detweiler 1999):
Gain message:
attain a desirable outcome (gain)
not attain an undesirable outcome (non‐loss)
Loss message:
attain an undesirable outcome (loss)
not attain a desirable outcome (non‐gain)
This differentiation has been referred to as the "kernel state of the consequence under discussion". The kernel state is the basic, root state mentioned in the message's description of the consequence.
| Type of framing | Attribute Framing | Goal framing | ||||
| Type of message | Positive | Negative | Gain | Loss | ||
| Gain: attain a desirable outcome | Non‐loss: not attain an undesirable outcome | Loss: attain an undesirable outcome | Non‐gain: not attain a desirable outcome | |||
| Example | The chance of survival with cancer is 2/3 | The chance of mortality with cancer is 1/3 | If you undergo screening for cancer, your survival will be prolonged | If you undergo screening for cancer, your survival won't be shortened | If you don't undergo screening for cancer, your survival will be shortened | If you don't undergo screening for cancer, your survival won't be prolonged |
Appendix 3. Examples of study outcomes classified according to the systematic review typology of outcomes
Understanding
Accuracy of responses to questions about the potential benefits and harms of mammography
Perception of effectiveness
Attitude towards the recommendation
Perceived utility of a prostate examination
Perceived efficacy of sun‐protective behavior
Attitude toward the immunization shot
Persuasiveness
Intention to get a screening test for HIV
Intention to get a mammogram
Likelihood of engaging in specific prevention behaviors within the next 30 days
Willingness to have protected sex
Intention to comply with recommendations
Decision whether to take antihypertensive medication
Motivation to perform a prostate examination
Self‐reported intention to use sunscreen at beach
Intention to immunize
Behavior
Getting an HIV test within 6 months of intervention
Mammography screening at 6 months
Redemption of a sunscreen coupon
Appendix 4. Electronic search strategies
The search used in MEDLINE (starting January 1966) and EMBASE (starting January 1980) were:
1‐randomized controlled trial.pt.
2‐controlled clinical trial.pt.
3‐((random$ or control$) adj5 trial$).mp.
4‐((random$ or control$) adj5 (trial$ or stud$)).mp.
5‐cross?section$.mp.
6‐(cross$ adj section$ adj3 (trial$ or stud$)).mp.
7‐(random$ adj allocat$).mp.
8‐randomized controlled trials/
9‐controlled clinical trials/
10‐cross‐sectional studies/
11‐random$.ti,ab.
12‐1 or 2 or 4 or 5 or 6 or 7 or 8 or 9 or 10 or 11
13‐*Risk/
14‐exp communication barriers/
15‐exp probability learning/
16‐(fram$ adj4 effect$).mp.
17‐(communicat$ adj5 risk$).mp.
18‐((quantit$ or amount) adj2 information).mp.
19‐((way$ or method$ or manner) adj2 (present$ or interpret$ or report$) adj3 (evidence or information or data or results)).mp.
20‐health education.mp.
21‐patient education.mp.
22‐graphic$.mp.
23‐(information$ adj5 display).mp.
24‐(risk adj5 presentation).mp.
25‐13 or 14 or 15 or 16 or 17 or 18 or 19 or 20 or 21or 22.mp. or 23 or 24
26‐12 and 25
In PsycLIT (starting January 1887), we used the same search for intervention type as in Medline and the following search for study type:
1‐randomi#ed controlled trial$.tw.
2‐((singl$ or doubl$ or trebl$ or tripl$) adj3 (blind$ or mask$)).tw.
3‐placebo/
4‐placebo$.tw.
5‐random$.tw.
6‐comparative studies$.tw.
7‐(clin$ adj3 trial$).tw.
8‐1 or 2 or 3 or 4 or 5 or 6 or 7
In addition, we searched MEDLINE, EMBASE and PscyINFO using “framing” as title word (framing.ti).
Appendix 5. Results of comparisons and studies for which data is not reported but otherwise eligible
Attribute framing
Levin 1988 assessed the framing of a treatment message (new cancer treatment). They found higher ratings of effectiveness, a higher likelihood of recommending the treatment to another person, and a higher likelihood of recommending the treatment to a family member when the message was framed positively.
Linville 2001 assessed the framing of a prevention message describing the effectiveness of condom use. They found that with a positive frame, participants were more likely to agree on advertising the condom as "effective" and to be willing to use the condom.
Marteau 1989 studied the framing of the risk of dying with a surgery for a terminal liver cancer (willingness to undergo surgery), risk of a fetus having spina bifida (willingness to undergo amniocentesis), and risk of a child of a haemophilia carrier mother of being affected (willingness to consider abortion). They found that patients were more likely to choose surgery when the message about perioperative survival was framed positively rather than negatively. There was a tendency for a framing effect with the spina bifida scenario (negative frame more persuasive in terms of performing an amniocentesis) but not with the haemophilia scenario.
Williams 2001 studied the framing of the percentage of women in different risk categories who are likely to get breast cancer. They found no effect of attribute framing.
Goal framing
Brug 2003 assessed the framing of a dietary prevention messages. They found no significant differences in attitudes or intentions between the gain‐frame and loss‐frame messages.
Collins 2006 assessed the framing of a prevention message relating to sexually transmitted diseases. There were no significant main effects of message framing on attitudes toward having sex or on intentions to have sex.
Steffen 1994 assessed the framing of a testicle self‐examination screening message. They found no framing effect on attitude, intention and behavior.
Lalor 1990 studied the framing of a breast self‐examination screening message. They found no significant effects of message type.
Trotto 2001 studied the framing the message about breast self‐examination in a group of female college students. The authors found no impact of framing on participants' perceived risk of breast cancer or intention to conduct a breast self‐examination as well.
Meyerowitz 1987 assessed the framing of a breast self‐examination screening message. Subjects presented with loss‐framed messages had more positive attitudes, intentions and behaviors related to screening.
Rothman 1993 found that exposure to loss‐framed information made women more likely but men less likely to intend adopting a screening behavior. No such effect was found for the prevention behavior (Rothman 1993). In another experiment, women exposed to gain‐framed messages were more likely than those who read loss‐framed messages to request sunscreen with an appropriate sun protection factor (prevention behavior) (Rothman 1993).
Maheswaran 1990 studied the framing of a cholesterol screening message. Millar 2000 studied goal framing of a safe driving behavior message. Miller 1999 assessed the goal framing of a message on the value of a diagnostic follow‐up colposcopy for precancerous cervical lesions in women. The three studies reported on interaction effects but not on main effects of framing.
Appendix 6. Results for the effect modification of the framing effects
Level of involvement
Levin 1988 studying the attribute framing of a treatment message found that the framing effect was independent of the level of involvement (recommending the treatment to others or to a family member).
Maheswaran 1990 studied goal framing of a cholesterol screening message in undergraduate students. They found that loss‐framed messages were more persuasive under high involvement (message describing risk of coronary heart disease in youth) while gain‐framed messages were more persuasive under low involvement (message describing risk of coronary heart disease in elderly).
Millar 2000 studying the goal framing of a safe driving behavior message found that when issue involvement was high, subjects receiving gain messages were more likely to agree with the message and to be willing to adopt the behavior. When issue involvement was low, there was no significant difference between the gain and loss messages in agreement or willingness.
Rothman 1993 found that exposure to loss‐framed information made women more likely but men less likely to intend adopting a screening behavior. No such effect was found for the prevention behavior (Rothman 1993). In another experiment, women exposed to gain‐framed messages were more likely than those who read loss‐framed messages to request sunscreen with an appropriate sun protection factor (prevention behavior) (Rothman 1993). The authors explained these findings by an interaction between framing and involvement based on their assumption that women are more "involved" in the issues of skin cancer and sun tanning compared with men.
Donovan 2000 evaluated the interaction between framing effect and level of involvement. There was no interaction for the outcomes of this review.
Krishnamurthy 2001a investigated both attribute framing and goal framing using the same treatment message. They found that positive framing was superior to negative framing in the context of attribute framing while negative framing was superior to positive framing in the context of goal framing when studying a population with low intrinsic self relevance (ie. students). They also found that the framing effect in the context of goal framing disappeared in a population with high levels of intrinsic self relevance (ie. patients).
Williams 2001 studied framing the message about breast cancer self‐examination. A loss frame led to greater positive behavioral change relative to the gain‐frame among women without a family history of breast cancer (low involvement), but not among women with family history of breast cancer (high involvement).
Finney 2002a found that a loss‐framed message increased compliance with mammography screening in women with a family history of breast cancer (considered to have a high level of issue involvement) at one month but not at 2 months. The investigators found no framing effect in women with no family history of breast cancer (considered to have a low level of issue involvement).
Scott 2006a found that the effects of goal framing of screening messages varied in women with or without parental history of high blood pressure or cholesterol: those with such a history were more likely to intend to 'check blood pressure or cholesterol in the next 30 days' when presented with a loss‐framed message; those without a history were more likely to express such an intention when presented with a gain‐framed message.
Wilson 1987a studied the effect of framing on a message relating to a hypothetical terminal liver disease and the probability of dying versus surviving of a surgery that would improve long term survival. The authors found that, when the choice of a medical option involved another individual's health outcome, subjects were still more likely to endorse riskier treatments when the outcomes were presented in a positive rather than negative context. However, when the medical decision involved a stranger's health outcomes, the contest effect did not occur.
Perceived susceptibility
Collins 2006 assessing the effect of goal framing of a prevention message relating to sexually transmitted diseases found that with a gain‐framed message, intentions to adopt the prevention behavior were marginally lower when the perceived risk of abstinence was high. With a loss‐framed message, intentions to adopt the prevention behavior were significantly higher when the perceived risk of abstinence was low. The authors measured sexual involvement and contextual risk as proxies for perceived risk of abstinence.
Lalor 1990 studying the goal framing of breast self‐examination screening message found no interaction between framing effect and high or low levels of feeling of susceptibility in terms of attitude about, intention to perform, and actually performing breast self‐examination.
Meyerowitz 1987 assessing goal framing of a breast self‐examination screening message found that framing effect did not interact with fear arousal, memory for the message content, perceived susceptibility to breast cancer or stronger beliefs in efficacy of screening.
Perceived efficacy/self‐efficacy
Meyerowitz 1987 assessing goal framing of a breast self‐examination screening message found that framing effect interacted with perceived self efficacy.
Block 1995a found that in the context of goal framing of prevention messages, a low efficacy condition (i.e., when it is uncertain that following the recommendations will lead to the desired outcome) motivates more in‐depth processing, in the context of which loss frames are more persuasive than gain ones. in contrast, a high efficacy condition generates less effortful message processing in which gain and loss frames are equally persuasive.
Dispositional Motivations
Mann 2004 tested the framing effect in the context of a message promoting flossing. They found that loss‐framed message was associated with self‐reporting of flossing in avoidance‐oriented people while a gain‐framed message was associated with self‐reporting of flossing in approach‐oriented people. The congruency hypothesis states that individuals who are approach‐oriented are highly responsive to reward or incentive cues and those who are avoidance‐oriented are highly responsive to punishment or threat cues (Carver 2000).
Attitude toward affected group
Steward 2002 evaluated the framing effect in the context of advocating an HIV prevention program. He found that participants with more negative attitudes viewed the outcomes as more uncertain and, consequently, were more influenced by loss‐framed than gain‐framed messages, whereas those with more positive attitudes saw the outcomes as less uncertain and were unaffected by frame.
Prior intention
Detweiler 1999 found a gain‐frame advantage in promoting a prevention behavior (sunscreen use) among those who had not prior intention to use sunscreen but not among those who had a prior intention.
Emotional vividness
Cherubini 2005a found no interaction between framing and emotional vividness.
Motivation
Schmitt 2004 found that in individuals who are highly motivated to comply with a recommended health behavior, message framing may influence attitude towards the behavior but not have a direct effect on behavioral intentions or overt behavior.
Optimism
Lauver 1990 assessed the effect of goal framing of diagnostic message related to an abnormal Papanicolaou test and found no interaction between framing and optimism.
Fear
Ruiter 2003 assessed the impact of goal framing of a screening message on the perceived importance of breast self‐examination. They found loss framing to be more persuasive, particularly in the context of high fear arousal.
Need for cognition
Schmitt 2004 found no interaction between the effect of framing on persuasiveness and the "need for cognition".
Amount of information
In the study by Diamond 1992, the framing effects varied with the amount of information presented.
Certainty of the test outcome
Apanovitch 2003 tested the framing effect in the context of a message to motivate HIV testing among low‐income, ethnic minority women. Among participants who reported being certain of the test outcome, a gain‐framed message was associated with a higher rate of self‐reported testing compared with a loss‐framed message. Among women who perceived the outcome of HIV testing as relatively uncertain, the two types of framing led to similar rates of self‐reported testing and the trend was in the direction of greater effectiveness for the loss‐framed message.
Level of risk
Marteau 1989 studying attribute framing of a treatment message found that the framing effect varied with the level of risk presented in these messages. It was evident when the risk of survival from operation was 40% or less. They also found that only patients presented with the higher risk of being abnormal (spina bifida scenario) were more likely to consider amniocentesis when presented with the negative frame.
Time effect
Rivers 2005a showed that the framing effects for both prevention and detection messages weakened with time.
Culturally adapted message
Schneider 2001b;Schneider 2001c found a framing effect in detection message (mammography), with loss‐framed messages being more persuasive, when the message was multi‐cultural but not when it was targeted to a specific ethnic group.
Numerical versus verbal probabilities
Welkenhuysen 2001a;Welkenhuysen 2001b studied the effect of framing a message describing cystic fibrosis, in the context of a decision regarding prenatal diagnosis. They found no framing effect when using numerical probabilities but found that when using verbal probabilities, students exposed to negatively‐framed messages were more willing to decide on prenatal diagnosis. The verbal probabilities "moderate chance" and "high chance" were the translation of 25% and 75% numerical probabilities.
Other
Brug 2003 and van Assema 2001a;van Assema 2001b found no interaction with factors such as perceived personal relevance, credibility or novelty of the information or the perceived importance of the topic addressed.
Data and analyses
Comparison 1. Attribute Framing.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
|---|---|---|---|---|
| 1 Understanding | 1 | Mean ES (Random, 95% CI) | ‐0.58 [‐0.94, ‐0.22] | |
| 1.1 Screening Message | 1 | Mean ES (Random, 95% CI) | ‐0.58 [‐0.94, ‐0.22] | |
| 2 Perception of effectiveness | 2 | Mean ES (Random, 95% CI) | 0.36 [‐0.13, 0.85] | |
| 2.1 Screening Message | 1 | Mean ES (Random, 95% CI) | 0.12 [‐0.23, 0.47] | |
| 2.2 Other type of Message | 1 | Mean ES (Random, 95% CI) | 0.62 [0.22, 1.02] | |
| 3 Persuasiveness | 11 | Mean ES (Random, 95% CI) | 0.07 [‐0.23, 0.37] | |
| 3.1 Screening message | 2 | Mean ES (Random, 95% CI) | ‐0.20 [‐0.93, 0.53] | |
| 3.2 Prevention Message | 1 | Mean ES (Random, 95% CI) | ‐0.61 [‐0.98, ‐0.25] | |
| 3.3 Treatment Message | 5 | Mean ES (Random, 95% CI) | 0.33 [‐0.14, 0.80] | |
| 3.4 Other type of message | 3 | Mean ES (Random, 95% CI) | 0.10 [‐0.30, 0.51] | |
| 4 Behavior | 1 | Mean ES (Random, 95% CI) | 0.09 [‐0.14, 0.31] | |
| 4.1 Prevention message | 1 | Mean ES (Random, 95% CI) | 0.09 [‐0.14, 0.31] |
1.2. Analysis.
Comparison 1 Attribute Framing, Outcome 2 Perception of effectiveness.
Comparison 2. Goal Framing.
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
|---|---|---|---|---|
| 1 Perception of effectiveness | 14 | Mean ES (Random, 95% CI) | ‐0.03 [‐0.22, 0.16] | |
| 1.1 Screening Message | 5 | Mean ES (Random, 95% CI) | ‐0.30 [‐0.49, ‐0.10] | |
| 1.2 Prevention Message | 9 | Mean ES (Random, 95% CI) | 0.11 [‐0.12, 0.33] | |
| 2 Persuasiveness | 23 | Mean ES (Random, 95% CI) | ‐0.06 [‐0.18, 0.06] | |
| 2.1 Screening Message | 6 | Mean ES (Random, 95% CI) | 0.06 [‐0.23, 0.35] | |
| 2.2 Prevention Message | 13 | Mean ES (Random, 95% CI) | 0.02 [‐0.11, 0.16] | |
| 2.3 Treatment Message | 3 | Mean ES (Random, 95% CI) | ‐0.50 [‐1.04, 0.04] | |
| 2.4 Other type of message | 1 | Mean ES (Random, 95% CI) | ‐0.36 [‐0.65, ‐0.07] | |
| 3 Behavior | 16 | Mean ES (Random, 95% CI) | ‐0.06 [‐0.15, 0.03] | |
| 3.1 Screening Message | 10 | Mean ES (Random, 95% CI) | ‐0.07 [‐0.17, 0.04] | |
| 3.2 Prevention Message | 5 | Mean ES (Random, 95% CI) | ‐0.09 [‐0.37, 0.19] | |
| 3.3 Other type of message | 1 | Mean ES (Random, 95% CI) | ‐0.12 [‐0.48, 0.25] |
Characteristics of studies
Characteristics of included studies [ordered by study ID]
Apanovitch 2003.
| Methods | Randomized controlled trial; 2x2 factorial design: action framing (attain versus not attain) x valence (desirable versus undesirable). | |
| Participants | 531 women from low income neighbourhoods recruited from public housing developments or a community health center. 90% of recruited subjects included in the analysis. |
|
| Interventions | Videos with information about HIV screening framed either as gain or loss. | |
| Outcomes | Intention to get a screening test for HIV (persuasiveness) (data not available). Getting an HIV test within 6 months of intervention (behavior). |
|
| Notes | Screening message; goal framing; varying attainment and valence in a 2x2 factorial design. Examples of Messages: Gain frame (attain, desirable): There are many benefits, or good things, you may experience if you get tested for HIV. If you decide to get HIV tested you may feel the peace of mind that comes with knowing about health. (Accompanied by a photo of a couple embracing on a couch) Loss frame (not attain, desirable): There are many benefits, or good things, you may not experience if you don't get tested for HIV. If you decide not to get HIV tested you won't feel the peace of mind that comes with knowing about your health. (Accompanied by a photo of a couple sitting at opposite ends of a couch) Loss frame (attain, undesirable): There are many problems, or bad things, you may experience if you don't get tested for HIV. If you decide not to get HIV tested, you may feel more anxious because you may wonder if you're ill. (Accompanied by a photo of a couple sitting at opposite ends of a couch) Gain frame (not attain, undesirable): There are many problems, or bad things, you may not experience if you get tested for HIV. If you decide to get HIV tested, you may feel less anxious because you won't wonder if you're ill. (Accompanied by a photo of a couple embracing on a couch) |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | |
Banks 1995.
| Methods | Randomized controlled trial. | |
| Participants | 133 female utility employees who had never been diagnosed with breast cancer and were not adherent to national guidelines for obtaining mammography screening. No details provided regarding response rate. |
|
| Interventions | Educational video presentation about benefits of mammography screening using gain‐framed versus loss‐framed formats. | |
| Outcomes | Intention to get a mammogram (persuasiveness). Mammography screening (at 6 and 12 months) (behavior). | |
| Notes | Screening message; goal framing; varying attainment; using a desirable valence. Examples of Messages: Gain frame: We will show you that detecting breast cancer early will save your life. Loss frame: We will show that failing to detect breast cancer early can cost you your life. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Mammography screening at 6 and 12 months. |
Blanton 2001a.
| Methods | Quasi‐randomised controlled trial. Blanton 2001a relates to study 4. |
|
| Participants | 120 undergraduate students. No details provided regarding response rate. |
|
| Interventions | Message comparing those who use to those who don't use condoms framed either positively or negatively. | |
| Outcomes | Willingness to have unprotected sex (transformed into willingness to have protected sex) (persuasiveness). | |
| Notes | Prevention message; attribute framing. Messages: Positive frame: People who have sex with condoms are more responsible and less selfish than people who have sex without condoms. Negative frame: People who have sex without condoms are less responsible and more selfish than people who have sex with condoms. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Block 1995a.
| Methods | Quasi‐randomised controlled trial; 2 x 2 factorial design (framing) x (low versus high efficacy interventions by varying the certainty with which adherence to the recommendations will lead to the desired outcome). Block 1995a relates to study 1, low efficacy group. |
|
| Participants | 44 undergraduate students. No details provided regarding response rate. |
|
| Interventions | Messages about the effect of the adherence to precautions on contracting HPV and controlling the development of lesions using positive versus negative framing. | |
| Outcomes | Attitude towards the recommendation (perception of effectiveness) Intention to comply with recommendations (persuasiveness). |
|
| Notes | Prevention (and treatment) message, goal framing; varying attainment; using undesirable valence. Examples of Messages: Gain frame: If you use the following precautions, you will be able to avoid contracting HPV. If you do have HPV, using these same precautions may help control the development of lesions. Loss frame: If you don't use the following precautions, you will not be able to avoid contracting HPV. If you do have HPV, not using these same precautions may help speed the development of lesions. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Block 1995b.
| Methods | Quasi‐randomised controlled trial; 2 x 2 factorial design (framing) x (low versus high efficacy interventions by varying the certainty with which adherence to the recommendations will lead to the desired outcome). Block 1995b relates to study 1, high efficacy group. |
|
| Participants | 50 undergraduate students. No details provided regarding response rate. |
|
| Interventions | Messages about the effect of the adherence to precautions on contracting HPV and controlling the development of lesions using positive versus negative framing. | |
| Outcomes | Attitude towards the recommendation (perception of effectiveness) Intention to comply with recommendations (persuasiveness). |
|
| Notes | Prevention (and treatment) message, goal framing; varying attainment; using undesirable valence. Examples of Messages: Gain frame: If you use the following precautions, you will be able to avoid contracting HPV. If you do have HPV, using these same precautions may help control the development of lesions. Loss frame: If you don't use the following precautions, you will not be able to avoid contracting HPV. If you do have HPV, not using these same precautions may help speed the development of lesions. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Block 1995c.
| Methods | Quasi‐randomised controlled trial; 2 x 2 factorial design (framing) x (low versus high efficacy interventions by varying the certainty with which adherence to the recommendations will lead to the desired outcome). Block 1995c relates to study 2, low efficacy group. |
|
| Participants | 57 undergraduate students. No details provided regarding response rate. |
|
| Interventions | Messages about the effect of prevention behavior (sunscreen use and use of wide‐brimmed hats) on skin cancer. | |
| Outcomes | Attitude towards the recommendation (perception of effectiveness) Intention to comply with recommendations (persuasiveness). |
|
| Notes | Prevention message, goal framing; varying attainment; using undesirable valence. Examples of Messages: Gain frame: By using a sunscreen with a Sun Protection Factor (SPF) of 15 or greater, and wearing a wide‐brimmed hat...you will avoid UV rays. Loss frame: By not using a sunscreen with a Sun Protection Factor (SPF) of 15 or greater, and not wearing a wide‐brimmed hat...you will not avoid UV rays. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Block 1995d.
| Methods | Quasi‐randomised controlled trial; 2 x 2 factorial design (framing) x (low versus high efficacy interventions by varying the certainty with which adherence to the recommendations will lead to the desired outcome). Block 1995d relates to study 2, high efficacy group. |
|
| Participants | 58 undergraduate students. No details provided regarding response rate. |
|
| Interventions | Messages about the effect of prevention behavior (sunscreen use and use of wide‐brimmed hats) on skin cancer. | |
| Outcomes | Attitude towards the recommendation (perception of effectiveness) Intention to comply with recommendations (persuasiveness). |
|
| Notes | Prevention message, goal framing; varying attainment; using undesirable valence. Examples of Messages: Gain frame: By using a sunscreen with a Sun Protection Factor (SPF) of 15 or greater, and wearing a wide‐brimmed hat...you will avoid UV rays. Loss frame: By not using a sunscreen with a Sun Protection Factor (SPF) of 15 or greater, and not wearing a wide‐brimmed hat...you will not avoid UV rays. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Carling 2010.
| Methods | Internet‐based randomized controlled trial. | |
| Participants | 1528 adult volunteers in Norway recruited through a nationally televised weekly health program. No details provided regarding response rate. |
|
| Interventions | Information about the effects of antihypertensive medication displayed on a website framed either positively or negatively. | |
| Outcomes | Decision whether to take antihypertensive medication (persuasiveness). | |
| Notes | Treatment message, goal framing; varying attainment; using undesirable valence. Examples of messages: Gain frame: Of 1000 people who do not take medication to lower their blood pressure, 930 will escape having cardiovascular disease over the next 10 years. Of 1000 people who take medication to lower their blood pressure, 944 will escape having cardiovascular disease over the next 10 years. Loss frame: Of 1000 people who do not take medication to lower their blood pressure, 70 will have cardiovascular disease over the next 10 years. Of 1000 people who take medication to lower their blood pressure, 56 will have cardiovascular disease over the next 10 years. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Randomization part of the automatic process of logging on to the survey website. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | Hypothetical outcome. |
Cherubini 2005a.
| Methods | Randomized controlled trial; 2x2 factorial design: framing (positive versus negative) x emotional vividness (reassuring versus fear evoking). Cherubini 2005a relates to "reassuring" group. |
|
| Participants | 180 men from North Italy. 82% response rate. |
|
| Interventions | Six page pamphlet reporting information about prostate cancer, screening methodologies, and medical care recommendations framed either positively or negatively. | |
| Outcomes | Perceived utility of a prostate examination (perception of effectiveness). Motivation to perform a prostate examination (persuasiveness). Some of the data supplied by authors. |
|
| Notes | Screening message; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain frame: Only by having a prostate examination can you be sure that you do not have prostate cancer. Research shows that men who regularly undergo prostate examinations have an increased chance of finding a tumour in the early, more treatable stage of the disease. You can gain several potential health benefits by having an examination that's not at all distressful and requires just a little of your time. Take advantage of this opportunity. Loss frame: Not having a prostate examination, you cannot be sure that you do not have prostate cancer. Research shows that men who do not undergo prostate examinations have a decreased chance of finding a tumour in the early, more treatable stage of the disease. You can lose several potential health benefits by failing to have an examination that's not at all distressful and requires just a little of your time. Do not fail to take advantage of this opportunity. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Cherubini 2005b.
| Methods | Randomized controlled trial; 2x2 factorial design: framing (positive versus negative) x emotional vividness (reassuring versus fear evoking). Cherubini 2005b relates to "evoking" group. |
|
| Participants | 180 men from North Italy. 82% response rate. |
|
| Interventions | Six page pamphlet reporting information about prostate cancer, screening methodologies, and medical care recommendations framed either positively or negatively. | |
| Outcomes | Perceived utility of a prostate examination (perception of effectiveness). Motivation to perform a prostate examination (persuasiveness). Some of the data supplied by authors. |
|
| Notes | Screening message, goal framing. Examples of Messages: Gain frame: "If you are over 40, you should request a prostate examination from your physician. Only by having a prostate examination can you be sure that you do not have prostate cancer. Research shows that men who regularly undergo prostate examinations have an increased chance of finding a tumour in the early, more treatable stage of the disease. You can gain several potential health benefits by having an examination that's not at all distressful and requires just a little of your time. Take advantage of this opportunity." Loss frame: "Not having a prostate examination, you cannot be sure that you do not have prostate cancer. Research shows that men who do not undergo prostate examinations have a decreased chance of finding a tumour in the early, more treatable stage of the disease. You can lose several potential health benefits by failing to have an examination that's not at all distressful and requires just a little of your time. Do not fail to take advantage of the opportunity." |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Detweiler 1999.
| Methods | Randomized controlled trial; 2x2 factorial design: action framing (attain versus not attain) x valence (desirable versus undesirable). | |
| Participants | 217 beach goers, 18 or older. "Approximately 90% of those asked to participate agreed to do so, and of the brochures that were distributed, over 95% were completed and returned." |
|
| Interventions | Brochure with information about skin cancer and the use of sunscreen framed as gain versus loss. | |
| Outcomes | Perceived efficacy of sun‐protective behavior (perception of effectiveness). Self‐reported intention to use sunscreen at beach (persuasiveness). Redemption of a sunscreen coupon (behavior). |
|
| Notes | Prevention message; goal framing; varying attainment and valence in a 2x2 factorial design. Examples of Messages: Gain frame (attain, desirable): Protect yourself from the sun and you will help yourself stay healthy. Loss frame (not attain, desirable): Don't expose yourself to the sun and you won't risk becoming sick. Loss frame (attain, undesirable): Expose yourself to the sun and you will risk becoming sick. Gain frame (not attain, undesirable): Don't protect yourself from the sun and you won't help yourself stay healthy. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | |
Donovan 2000.
| Methods | Randomized controlled trial. | |
| Participants | 102 adult females (18‐45 years) in a shopping mall. 74% response rate. |
|
| Interventions | Participants were presented with information about a hypothetical new infant immunization with side effect information framed either positively or negatively. | |
| Outcomes | Attitude toward the immunization shot (perception of effectiveness). Intention to immunize (persuasiveness). |
|
| Notes | Harm message (in the context of a prevention message), attribute framing. Messages: 90% of children do not develop the side effect (negative frame; author categorized it as positive). 10% of children do develop a side effect (positive frame; author categorized it as negative). |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Finney 2002a.
| Methods | Quasi‐randomised controlled trial. | |
| Participants | 313 women, at least 40 years of age with a positive family history of breast cancer, and due for their annual mammogram. No details provided regarding response rate. |
|
| Interventions | Mammogram screening reminder letters framed either positively or negatively. Messages based on those used by Banks 1995. | |
| Outcomes | Compliance with mammogram screening (at 1 and 2 months) (behavior). | |
| Notes | Screening message; goal framing; varying attainment; using a desirable valence. Examples of messages: Gain frame: We will show you that detecting breast cancer early will save your life. Loss frame: We will show that failing to detect breast cancer early can cost you your life. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | Low risk | Compliance with mammogram screening. |
Finney 2002b.
| Methods | Quasi‐randomised controlled trial. | |
| Participants | 616 women, at least 40 years of age without a positive family history of breast cancer, and due for their annual mammogram. No details provided regarding response rate. |
|
| Interventions | Mammogram screening reminder letters framed either positively or negatively. Messages based on those used by Banks 1995. | |
| Outcomes | Compliance with mammogram screening (at 1 and 2 months) (behavior). | |
| Notes | Screening message; goal framing; varying attainment; using a desirable valence. Examples of messages: Gain frame: We will show you that detecting breast cancer early will save your life. Loss frame: We will show that failing to detect breast cancer early can cost you your life. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | Low risk | Compliance with mammogram screening. |
Hsiao 2002a.
| Methods | Randomized controlled trial; 2x2x2 factorial design: frame (positive versus negative) x type of exercise (regular exercise versus exercise testing) x emphasis (prevention versus detection). Hsiao 2002a relates to regular exercise with prevention focus and vary by frame. |
|
| Participants | 49 college students with "lack of regular physical activity and previous participation in exercise testing". No details provided regarding response rate. |
|
| Interventions | Pamphlet promoting regular exercise framed either positively or negatively. | |
| Outcomes | Attitude (perception of effectiveness). Behavioral intention (persuasiveness). |
|
| Notes | Prevention message; goal framing; varying attainment; using desirable valence. Data not available for self‐reported activity level (behavior). Messages: Gain frame: There are important health benefits to be gained from engaging in regular exercise. By engaging in regular exercise, you can effectively lower your LDL (the bad cholesterol that sticks to your artery walls) and increase your HDL (the good cholesterol that removes LDL from your artery walls). Additionally, by engaging in regular exercise, you can effectively lower your fat content and increase your muscle mass. Ultimately, when you engage in regular exercise, you are taking advantage of a safe and effective way to reduce your risk of heart disease and improve your cardiovascular fitness. With early prevention, you can avoid expensive medical treatment and unnecessary suffering caused by heart problems later on. Loss frame: There are important health benefits that can be lost from failure to engage in regular exercise. By not engaging in regular exercise, you cannot effectively lower your LDL (the bad cholesterol that sticks to your artery walls) and increase your HDL (the good cholesterol that removes LDL from your artery walls). Additionally, by not engaging in regular exercise, you cannot effectively lower you fat content and increase your muscle mass. Ultimately, when you do not engage in regular exercise, you fail to take advantage of a safe and effective way to reduce your risk of heart disease and improve your cardiovascular fitness. Without early prevention, you miss opportunities to avoid expensive medical treatment and unnecessary suffering caused by heart problems later on. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | Self‐reported activity level. |
Hsiao 2002b.
| Methods | Randomized controlled trial; 2x2x2 factorial design: frame (positive versus negative) x type of exercise (regular exercise versus exercise testing) x emphasis (prevention versus detection) Hsiao 2002b relates to exercise testing with detection focus and vary by frame. No details provided regarding response rate. |
|
| Participants | 46 college students with "lack of regular physical activity or previous participation in exercise testing." | |
| Interventions | Pamphlet promoting exercise testing framed either positively or negatively. | |
| Outcomes | Attitude (perception of effectiveness). Behavioral intention (persuasiveness). |
|
| Notes | Screening message; goal framing; varying attainment; using desirable valence. Messages: Gain frame: There are important health benefits to be gained from participating in an exercise test. By participating in an exercise test, you can effectively assess how your heart and blood vessels respond to physical exertion. As the body works harder during exercise, it requires more oxygen and the heart has to pump more blood. Thus by participating in an exercise test, you can effectively find out early if there is a lack of blood supply through the arteries that go to the heart. Ultimately, when you participate in an exercise test, you are taking advantage of a safe and effective way to detect your risk of heart disease and improve your cardiovascular fitness. With early detection, you can avoid expensive medical treatment and unnecessary suffering caused by heart problems later on. Loss frame: There are important health benefits that can be lost from failure to participate in an exercise test. By not participating in an exercise test, you cannot effectively assess how your heart and blood vessels respond to physical exertion. As the body works harder during exercise, it requires more oxygen and the heart has to pump more blood. Thus, by not participating in an exercise test, you cannot effectively find out early if there is a lack of blood supply through the arteries that go to the heart. Ultimately, when you do not participate in an exercise test, you fail to take advantage of a safe and effective way to detect your risk of heart disease and improve your cardiovascular fitness. Without early detection, you miss opportunities to avoid expensive medical treatment and unnecessary suffering caused by heart problems later on. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Jasper 2001.
| Methods | Randomized controlled trial, 84% response rate | |
| Participants | 105 female callers to a counselling service for pregnant women in Toronto. 84% response rate. |
|
| Interventions | Teratogenic‐risk information presented in either a positive frame or a negative frame. | |
| Outcomes | Willingness to take the drug (persuasiveness). | |
| Notes | Harm message (in the context of a treatment message), attribute framing. The Messages: Negative‐frame: In every pregnancy, there is a 1 to 3% chance that a woman will give birth to a child who has a major birth defect. This/these drug(s) [insert applicable drug name] has/have not been shown to change that. Positive‐frame: "In every pregnancy, there is a 97 to 99% chance that a woman will give birth to a child who does not have a major birth defect. This/these drug(s) [insert applicable drug name] has/have not been shown to change that. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Krishnamurthy 2001a.
| Methods | Randomized controlled trial; 2x2 factorial design: (goal versus attribute framing) x (positive versus negative framing). Each participant received 5 messages with 5 different levels of effectiveness (10%, 30%, 50%, 70%, 90%). Krishnamurthy 2001a relates to students who received attribute framing. No details provided regarding response rate. |
|
| Participants | 143 undergraduate college students (considered to have low intrinsic self relevance) participated in studies A and B. | |
| Interventions | Participants were presented with information about 5 possible new treatments for a health condition they chose ("that had most recently warranted a visit to a physician"). | |
| Outcomes | Likelihood of talking to the physician about the treatment (persuasiveness). | |
| Notes | Treatment message, attribute framing. The Messages: Positive attribute framing condition: Treatment A (B, C, D, E) provides better results for 50% (10%, 90%, 70% or 30%) of the patients with your health condition. Negative attribute framing condition: Treatment A (B, C, D, E) fails to provide better results for 50% (90%, 10%, 30% or 70%) of the patients with your health condition. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Krishnamurthy 2001b.
| Methods | Randomized controlled trial; 2x2 factorial design: (goal versus attribute framing) x (positive versus negative framing). Each participant received 5 messages with 5 different levels of effectiveness (10%, 30%, 50%, 70%, 90%). Krishnamurthy 2001b relates to students who received goal framing. No details provided regarding response rate. |
|
| Participants | 143 undergraduate college students (considered to have low intrinsic self relevance) participated in studies A and B. | |
| Interventions | Participants were presented with information about 5 possible new treatments for a health condition they chose ("that had most recently warranted a visit to a physician"). | |
| Outcomes | Likelihood of talking to the physician about the treatment (persuasiveness). | |
| Notes | Treatment message, goal framing; varying attainment; using desirable valence. The Messages: Positive goal framing condition: By taking Treatment A (B, C, D, E) you get a 50% (10%, 90%, 70% or 30%) chance of getting better results. Negative goal framing condition: By not taking Treatment A (B, C, D, E) you give up a 50% (10%, 90%, 70% or 30%) chance of getting better results. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Krishnamurthy 2001c.
| Methods | Randomized controlled trial; 2x2 factorial design: (goal versus attribute framing) x (positive versus negative framing). Each participant received 5 messages with 5 different levels of effectiveness (10%, 30%, 50%, 70%, 90%). Krishnamurthy 2001c relates to patients who received attribute framing. No details provided regarding response rate. |
|
| Participants | 117 patients of a health clinic (considered to have high intrinsic self relevance) participated in studies C and D. | |
| Interventions | Participants were presented with information about 5 possible new treatments for a health condition they chose ("that had most recently warranted a visit to a physician"). | |
| Outcomes | Likelihood of talking to the physician about the treatment (persuasiveness). | |
| Notes | Treatment message, attribute framing. The Messages: Positive attribute framing condition: Treatment A (B, C, D, E) provides better results for 50% (10%, 90%, 70% or 30%) of the patients with your health condition. Negative attribute framing condition: Treatment A (B, C, D, E) fails to provide better results for 50% (90%, 10%, 30% or 70%) of the patients with your health condition. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Krishnamurthy 2001d.
| Methods | Randomized controlled trial; 2x2 factorial design: (goal versus attribute framing) x (positive versus negative framing). Each participant received 5 messages with 5 different levels of effectiveness (10%, 30%, 50%, 70%, 90%). Krishnamurthy 2001d relates to patients who received goal framing. No details provided regarding response rate. |
|
| Participants | 117 patients of a health clinic (considered to have high intrinsic self relevance) participated in studies C and D. | |
| Interventions | Participants were presented with information about 5 possible new treatments for a health condition they chose ("that had most recently warranted a visit to a physician"). | |
| Outcomes | Likelihood of talking to the physician about the treatment (persuasiveness). | |
| Notes | Treatment message, goal framing; varying attainment; using desirable valence. The Messages: Positive goal framing condition: By taking Treatment A (B, C, D, E) you get a 50% (10%, 90%, 70% or 30%) chance of getting better results. Negative goal framing condition: By not taking Treatment A (B, C, D, E) you give up a 50% (10%, 90%, 70% or 30%) chance of getting better results. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Kroll 2005.
| Methods | Randomized controlled trial. | |
| Participants | 192 students from a metropolitan area private parochial high school "who met the criteria for the lowest two stages of exercise change". No details provided regarding response rate. |
|
| Interventions | Multimedia PowerPoint presentation about exercising with messages framed either positively or negatively. | |
| Outcomes | Perceived persuasiveness of the message (perception of effectiveness). Intention to exercise (on days 1, 7, 30) (persuasiveness). |
|
| Notes | Prevention message; goal framing; varying valence. Examples of messages: Gain frame: You'll be more healthy if you exercise. Loss framed: If you don't exercise, you'll get sick. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Lalor 1990.
| Methods | Randomized controlled trial. | |
| Participants | 55 undergraduate women, ages 19 to 42 who reported practicing breast self‐examination (BSE) less than 4 times per year. No details provided regarding response rate. |
|
| Interventions | Pamphlet on breast cancer examination framed either positively or negatively. | |
| Outcomes | Actual breast self‐examination (behavior). (Although study assessed perception of effectiveness and persuasiveness, it did not report the related raw data.) |
|
| Notes | Screening message, goal framing, varying attainment; using desirable valence. Examples of messages: Gain frame: By doing BSE now, you can learn what your normal, healthy breasts feel like so that you will be better prepared to notice any small, abnormal changes that might occur as you get older. Loss frame: By not doing BSE now, you will not learn what your normal, healthy breasts feel like so that you will be ill prepared to notice any small, abnormal changes that might occur as you get older. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | Cluster randomizations; "only one type of pamphlet was distributed at each training session so that subjects were unaware that there were two different pamphlets." |
| Objective outcome? | High risk | |
Lauver 1990.
| Methods | Randomized controlled trial. | |
| Participants | 116 women with abnormal Papanicolaou (pap) test. 85% response rate. |
|
| Interventions | When informed about the positive pap test, women were told about the gains with follow‐up (positive frame) or the losses without follow‐up (negative frame). | |
| Outcomes | Return to the clinic for a follow‐up colposcopy test within six weeks of contact (behavior). | |
| Notes | Mixed diagnostic and treatment message; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain Message: By going for colposcopy, you can have further tests to learn what the abnormal cells mean.... If (necessary), you would be in a good position to get early treatment. With early treatment you could raise your chances for living cancer‐free and symptom‐free. Loss Message: By not going for colposcopy, you cannot have further test to learn what the abnormal cells mean....If (necessary), you would not be in a good position to get early treatment. Without early treatment, you could lower your chances for living cancer‐free and symptom‐free. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Sealed envelopes. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Return to the clinic for a follow‐up colposcopy test. |
Lerman 1992.
| Methods | Randomized controlled trial. | |
| Participants | 446 women with abnormal mammograms during the previous year and eligible to receive an annual screening mammogram during the study period (data for about half included in meta‐analysis). 47% of eligible women included in the study. The response rate to telephone intervention was 71%. The response rate for checking the outcome 3 months later was 89%. |
|
| Interventions | Psychoeducational booklet, with information about abnormal mammograms and the necessity of continued screening, written in a positive frame or negative frame were mailed to the women followed one week later by a free mammogram referral. | |
| Outcomes | Participant self‐reported adherence to mammography screening three months after the intervention (behavior). | |
| Notes | Screening message, goal framing. Messages: not provided. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Adherence to mammography screening. |
Lewis 2003.
| Methods | Randomized controlled trial. | |
| Participants | 179 women aged 35 to 49, with no history of breast cancer, able to read and write in English, attending a General Internal Medicine practice. 67% response rate. |
|
| Interventions | Video presentations in which a female narrator conveys positively versus negatively‐framed messages about the benefits and harms of screening mammography. The verbal presentations were reinforced with diagrams. | |
| Outcomes | Change in the accuracy of the response to questions about the potential benefits and harms of mammography (understanding). Change in the perception of effectiveness of the importance of the benefits of mammography relative to its downsides (perception of effectiveness). |
|
| Notes | Screening message; attribute framing. Examples of Messages: Positive frame: Of 1000 women getting mammograms every year for 10 years, 1 woman's life is extended because she got yearly mammograms. Negative frame: Of 1000 women getting mammograms every year for 10 years, 999 women's lives are not extended because they got yearly mammograms. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Opaque, sealed and sequentially numbered envelopes. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | 1 out of 2: understanding. |
Llewellyn‐Thomas 1995.
| Methods | Randomized controlled trial. | |
| Participants | 60 patients with colorectal cancer. Of 264 eligible patients, 96 invited, 90 participated (60 included in the meta‐analysis). |
|
| Interventions | Booklets with information about a chemotherapeutic treatment presented in a positive or negative frame. | |
| Outcomes | Hypothetical decision to participate in a clinical trial incorporating the chemotherapeutic treatment (persuasiveness). | |
| Notes | Treatment (including benefit and harm) message; attribute framing. The Messages: Positive frame: 20 out of 100 patients will be free of these side effects, and will retain their ability to carry out their usual daily work and recreational activities. 5 years after treatment, 55 out of 100 patients will have survived. Negative frame: 80 out of 100 patients will have these side effects, which will interfere with their ability to carry out their usual daily work and recreational activities. 45 out of 100 patients will have died. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Sealed envelope. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Mann 2004.
| Methods | Randomized controlled trial. | |
| Participants | 70 undergraduate students enrolled in a research methods course (7 of them excluded from analysis). No details provided regarding response rate. |
|
| Interventions | Articles adapted from the American Medical Association Website presenting facts about gum disease and proper flossing technique, presented either in a gain frame or in a loss frame. | |
| Outcomes | Self‐reported use of flosses (behavior). | |
| Notes | Prevention message, goal framing; varying valence; using "attain" type of terms. Examples of Messages: Gain frame: Flossing your teeth daily removes particles of food in the mouth, avoiding bacteria, which promotes great breath. Loss frame: If you don't floss your teeth daily, particles of food remain in the mouth, collecting bacteria, which causes bad breath. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | |
McCaul 2002.
| Methods | Randomized controlled trial. | |
| Participants | 6522 men and women who lived in 29 North Dakota counties and had not submitted reimbursement requests to cover flu shots the previous year. "Subject loss was approximately 6%." |
|
| Interventions | A reminder letter from the state peer review organization about the consequences of the flu vaccine framed either positively or negatively. | |
| Outcomes | Vaccination rate (behavior). | |
| Notes | Prevention message; goal framing; varying attainment; undesirable valence. Examples of Messages: Gain frame: The gain insert featured the picture and testimonial of a North Dakota woman who had received a flu shot the previous year and had not gotten the flu. The testimonial stated, "I got a flu shot last year, and stayed healthy. I'm convinced that the flu shot protected me. I like to stay as healthy and active as I can, so I'm getting a flu shot again this year." In addition, the insert displayed information about three benefits of getting "your flu shot soon" ("You will be less likely to get the flu this fall"; "If you do get the flu, you will probably not be as sick" and "You will be less likely to enter the hospital because of the flu"). Loss frame: The loss insert featured the picture and testimonial of another North Dakota woman who had not received a flu shot last year and had spent several days in bed, sick with the flu. The testimonial stated, “I didn't get a flu shot last year, and caught the flu. I was sick and in bed for nine days. I want to avoid catching the flu again this year, so I’m getting a flu shot.” In addition, the insert displayed information about three costs “if you don’t get your flu shot soon” (“You will be more likely to get the flu this fall”; “If you do get the flu, you will probably be more sick” and “You will be more likely to enter the hospital because of the flu”). |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Vaccination. |
O'Connor 1996.
| Methods | Randomized control trial. | |
| Participants | 292 participants from outpatient respiratory and cardiac clinics and eligible for Influenza immunization. No details provided regarding response rate. |
|
| Interventions | Messages about the benefits and risks of influenza vaccine presented using positive or negative frames. | |
| Outcomes | Immunization rate (behavior). | |
| Notes | Prevention message, attribute framing. Examples of messages: The positive frame described the percentage of individuals who remain influenza‐free and free of vaccine side effects. The negative frame used the percentage who acquires influenza and vaccine side effects with and without immunization. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | |
O'Connor 2005.
| Methods | Randomized controlled trial. | |
| Participants | 304 undergraduate students who had previously used some form of contraception. No details provided regarding response rate. |
|
| Interventions | Material with information about an imaginary male hormonal contraception framed either positively or negatively. | |
| Outcomes | Intention to use the male contraception (persuasiveness). | |
| Notes | Prevention message; goal framing; varying attainment; using a desirable valence. The Messages: Gain‐frame: "If you use the male contraceptive pill or injection, you will be able to take advantage of an alternative, convenient method, which would provide reliable contraception, does not carry a significant risk of side effects and would allow for a more equal sharing of the responsibility for contraception." Loss frame: "If you do not use the male contraceptive pill or injection, you will not be able to take advantage of an alternative, convenient method, which would provide reliable contraception, does not carry a significant risk of side effects and would allow for a more equal sharing of the responsibility for contraception." |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Rivers 2005a.
| Methods | Randomized controlled trial; 2x2 factorial design (gain versus loss) x (prevention versus detection). Rivers 2005a relates to the prevention comparison. |
|
| Participants | 441 women attending an urban community health clinic serving predominantly minority and lower‐income populations. 71% follow up rate at 6 months; 78% follow up rate at 12 months. |
|
| Interventions | 10 min video presentation about cervical cancer and Pap smear using either positive or negative framing. The video emphasized the prevention function of the test. | |
| Outcomes | Self‐reported Pap test utilization at 6 months (behavior). | |
| Notes | Prevention message; goal framing; varying attainment; using negative valence. Examples of Messages: Gain‐frame: If you get regular Pap smears, you can prevent cervical cancer from developing...and preventing cervical cancer can save your life. Loss frame: If you don't get regular Pap smears, you can't prevent cervical cancer from developing...and not preventing cervical cancer can cost your life. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | High risk | Computer‐generated table of randomly sorted combinations of conditions. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Self‐reported behavior. |
Rivers 2005b.
| Methods | Randomized controlled trial; 2x2 factorial design (gain versus loss) x (prevention versus detection). Rivers 2005b relates to the detection comparison. |
|
| Participants | 441 women attending an urban community health clinic serving predominantly minority and lower‐income populations. 71% follow up rate at 6 months; 78% follow up rate at 12 months. |
|
| Interventions | 10 min video presentation about cervical cancer and Pap smear using either positive or negative framing. The video emphasized the detection function of the test. | |
| Outcomes | Self‐reported Pap test utilization at 6 months (behavior). | |
| Notes | Screening message; goal framing; varying attainment. Examples of Messages: Gained‐framed: If you get regular Pap test, you can detect cervical cancer early...and detecting cervical cancer early can save your life. Loss framed: If you don't get regular Pap test, you can't detect cervical cancer early...and not detecting cervical cancer early can cost your life. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | High risk | Computer‐generated table of randomly sorted combinations of conditions. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Self‐reported behavior. |
Robberson 1988a.
| Methods | Randomized controlled trial; 3x2 factorial design: message framing (positive versus negative versus both) x type of appeal (health versus self esteem). Robberson 1988a refers to groups with "health" as type of appeal. |
|
| Participants | 84 non exercising females students enrolled in introductory psychology courses. No details provided regarding response rate. |
|
| Interventions | Information about the effects of exercise on bodily functioning and mental functioning using positive versus negative framing. | |
| Outcomes | Intention to exercise (persuasiveness). | |
| Notes | Prevention message; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain frame: One of the greatest advantages of becoming involved in a regular exercise program is the resulting increase in physical stamina and endurance. Loss frame: One of the greatest disadvantages of not being involved in a regular exercise program is the resulting decrease in physical stamina and endurance. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Robberson 1988b.
| Methods | Randomized controlled trial; 3x2 factorial design: message framing (positive versus negative versus both) x type of appeal (health versus self esteem). Robberson 1988b refers to groups with "esteem" as type of appeal. |
|
| Participants | 84 non exercising females students enrolled in introductory psychology courses. No details provided regarding response rate. |
|
| Interventions | Information about the effects of exercise on bodily functioning and mental functioning using positive versus negative framing. | |
| Outcomes | Intention to exercise (persuasiveness). | |
| Notes | Prevention message, goal framing. Examples of Messages: Gain frame: And what better results of exercise and fitness can there be than that general 'feeling good about yourself' that goes along with the improved appearance, personal style, confidence and control. There's a boost to self‐esteem that gives you greater comfort with yourself so that your pride and self‐acceptance grow. Loss frame: And then there's that general 'feeling bad about yourself' that often goes along with a poor appearance, lack of personal style. lowered confidence, and lessened control. Feeling of worthlessness and self‐degradation commonly increase. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Ruiter 2003.
| Methods | Quasi‐randomised controlled trial. | |
| Participants | 110 female Dutch undergraduate students. No details provided regarding response rate. |
|
| Interventions | Information about performing breast self‐examination in either gain frame or loss frame. | |
| Outcomes | Evaluation of the importance of breast self‐examination relative to five other health behaviours (perception of effectiveness). | |
| Notes | Screening behavior; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain frame: By performing breast self‐examination, your chances to detect breast cancer in an early and better treatable stage will be high. Loss frame: By not performing breast self‐examination, your chances to detect breast cancer in an early and better treatable stage will be low. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Unclear risk | No details provided. |
| Objective outcome? | High risk | |
Schmitt 2004.
| Methods | Randomized controlled trial; 2x2 factorial design: message framing (positive verus negative versus both) x screening function (illness‐detecting versus health affirming). | |
| Participants | 177 female community volunteers. 85% response rate. |
|
| Interventions | Pamphlets with information about cholesterol, heart disease and blood cholesterol screening framed either positively or negatively. | |
| Outcomes | Implementation Intention to obtain a blood cholesterol screening test (persuasiveness). | |
| Notes | Screening test; goal framing; varying attainment, using desirable valence. Examples of messages: Gain frame: If you obtain a blood cholesterol test, you will gain important information about your cardiovascular health Loss frame: If you do not obtain a blood cholesterol test, you will not gain important information about your cardiovascular health. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Schneider 2001a.
| Methods | Randomized controlled trial; 2x2 factorial design (gain versus loss visual) x (gain versus loss auditory). Schneider 2001a is a separate publication from Schneider 2001b and Schneider 2001c. |
|
| Participants | 437 undergraduates attending a public university in New England No details provided regarding response rate. |
|
| Interventions | Multimedia presentation using visual and auditory modalities randomizing patients in 2x2 (gain versus loss visual and gain versus loss auditory). | |
| Outcomes | Self‐reported smoking behavior after 6 weeks (behavior). | |
| Notes | Prevention behavior, goal framing. The factorial design (gain versus loss visual and gain versus loss auditory) weakness the validity of the results as half the participants received mixed framing. We used only data from groups receiving consistent messages (i.e. gain visual and gain auditory; and loss visual and loss auditory). Examples of Messages: Gain frame: "If you quit you'll look and feel better right away." Loss frame: "If you don't quit you won't look and smell better...." |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | |
Schneider 2001b.
| Methods | Randomized controlled trial. Schneider 2001b refers to multicultural video intervention. |
|
| Participants | 379 women (mostly minorities) over 40 recruited from inner‐city community health clinics or from public housing developments. Not clear what the response rate to the initial invitation is. Contact rates at 6 and 12 months follow up were 70 and 71%. |
|
| Interventions | 10 minutes multicultural video about breast cancer and screening mammography framed either positively or negatively. | |
| Outcomes | Actual mammography screening (at 6 and 12 months) (behavior). | |
| Notes | Screening behavior; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain frame: ...detecting breast cancer early can save a woman's life. When a woman gets regular mammograms, she is doing her best to detect breast cancer early. Loss frame: ...failing to detect breast cancer early can cost a woman her life. When a woman does not get regular mammograms, she is not doing her best to detect breast cancer early. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | High risk | Coin flip. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Actual mammography screening. |
Schneider 2001c.
| Methods | Randomized controlled trial. Schneider 2001c refers to targeted video intervention. |
|
| Participants | 373 women (mostly minorities) over 40 recruited from inner‐city community health clinics or from public housing developments. Not clear what the response rate to the initial invitation was. Contact rates at 6 and 12 months follow up were 70 and 71%. |
|
| Interventions | 10 minute long targeted video about breast cancer and screening mammography framed either positively or negatively. | |
| Outcomes | Actual mammography screening (at 6 and 12 months) (behavior). | |
| Notes | Screening behavior; goal framing; varying attainment; using desirable valence. Examples of Messages: Gain frame: ...detecting breast cancer early can save a woman's life. When a woman gets regular mammograms, she is doing her best to detect breast cancer early. Loss frame: ...failing to detect breast cancer early can cost a woman her life. When a woman does not get regular mammograms, she is not doing her best to detect breast cancer early. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | High risk | Coin flip. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Actual mammography screening. |
Scott 2006a.
| Methods | Randomized controlled trial; 2x2 factorial design (gain versus loss visual) x (present versus future time orientation). Scott 2006a refers to gain versus loss; prevention messages. |
|
| Participants | 395 college women. No details provided regarding response rate. |
|
| Interventions | Print messages about women and heart disease, main behavioral risk factors associated with heart disease, likely benefits of healthy activities and suggested actions. | |
| Outcomes | Likelihood of engaging in specific prevention behaviours within the next 30 days (persuasiveness). Data supplied by authors. Our analyses combine the variables ‘reduce salt’, ‘lower dietary cholesterol’, ‘increase physical activity’, and ‘reduced body fat’. | |
| Notes | Prevention messages; goal framing; varying attainment; using both desirable and undesirable valences. Prevention message analysis combined reduce salt, lower dietary cholesterol, increase physical activity, reduce body fat; smoking cessation not included per author recommendation because of data quality issues. Examples of Messages: Gain‐frame: The likely future benefits of healthy activities: achieve your goals for yourself as an active older adult; have the strength to go for long walks, garden, or play with future children or grandchildren; maintain a healthy weight; be less dependent on others to do things. Loss frame: The likely future consequences of unhealthy activities: fail to achieve your goals for yourself as an active older adult; lack the strength to go for long walks, garden, or play with future children or grandchildren; have difficulty maintaining a healthy weight; be more dependent on other to do things. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Participants pulled a folded piece of paper from a bag that had up to 400 pieces, each with a number that ranged from 1 to 400 and the number corresponded with message conditions to determine message assignment. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Scott 2006b.
| Methods | Randomized controlled trial; 2x2 factorial design (gain versus loss visual) x (present versus future time orientation). Scott 2006b refers to gain versus loss; screening messages. |
|
| Participants | 395 college women. No details provided regarding response rate. |
|
| Interventions | Print messages about women and heart disease, main behavioral risk factors associated with heart disease, likely benefits of healthy activities and suggested actions. | |
| Outcomes | Likelihood of engaging in specific detection behaviours within the next 30 days (persuasiveness). Data supplied by authors. Our analyses combine the variables 'check BP' and 'check cholesterol'. | |
| Notes | Screening messages; goal framing. Screening message analysis combined check BP and check cholesterol. Examples of Messages: No examples provided for screening messages. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Low risk | Participants pulled a folded piece of paper from a bag that had up to 400 pieces, each with a number that ranged from 1 to 400 and the number corresponded with message conditions to determine message assignment. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Steward 2002.
| Methods | Randomized controlled trial (study 1). | |
| Participants | 183 University undergraduate and graduate students. No details provided regarding response rate. |
|
| Interventions | Messages advocating an HIV prevention program emphasizing the benefits of funding (gain‐framed) or the costs of not funding (loss‐framed) the program. | |
| Outcomes | Support for the program (persuasiveness). | |
| Notes | Public health message; goal framing; varying attainment; using a desirable valence. Examples of Messages: Gain frame: By increasing their current budget, needle exchange programs could operate full time in neighbourhoods with high drug use. Loss frame: With their current budget, needle exchange programs cannot operate full‐time in neighbourhoods with high drug use". |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
van Assema 2001a.
| Methods | Randomized controlled trial; 2x2 factorial design: Frame (positive versus negative) x Dietary behavior (fat versus fruits and vegetables). van Assema 2001a relates to fat as dietary behavior. This study is also reported in Brug 2003 (study 1). No details provided regarding response rate. |
|
| Participants | 78 adults recruited via schools for adult education in three cities in the southern part of the Netherlands. | |
| Interventions | Booklet presenting a list of arguments for dietary behavioral change using positive versus negative frames. | |
| Outcomes | Cognitive attitude (perception of effectiveness). Intention to change the dietary behavior (persuasiveness). |
|
| Notes | Prevention message; goal framing; variation of both attainment and valence. Examples of Messages: Gain frame: People that eat a low fat diet: have more chance of staying healthy; probably have less chance of getting cancer; have less chance of getting a cardiovascular disease; etc. Loss frame: People that eat too much fat: have more chance of becoming ill; probably have more chance of getting cancer; have more chance of getting a cardiovascular disease; etc. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
van Assema 2001b.
| Methods | Randomized controlled trial; 2x2 factorial design: Frame (positive versus negative) x Dietary behavior (fat versus fruits and vegetables). van Assema 2001b relates to fruits and vegetables as dietary behavior. This study is also reported in Brug 2003 (study 1). |
|
| Participants | 70 adults recruited via schools for adult education in three cities in the southern part of the Netherlands. No details provided regarding response rate. |
|
| Interventions | Booklet presenting a list of arguments for dietary behavioral change using positive versus negative frames. | |
| Outcomes | Cognitive attitude (perception of effectiveness). Intention to change the dietary behavior (persuasiveness). |
|
| Notes | Prevention message; goal framing; variation of both attainment and valence. Examples of Messages: Gain frame: People that do eat enough F&V: have more resistance against diseases, such as flu and cold; have less chance of getting cancer; often have an adequate intake of dietary fibres and therefore more chance of healthy bowels. Loss frame: People that do not eat enough F&V: have less resistance against disease, such as flu and cold; have more chance of getting cancer; often have an inadequate intake of dietary fibres and therefore more chance of bowel problems. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Welkenhuysen 2001a.
| Methods | Randomized controlled trial; 2x2 factorial design (positive versus negative framing) x (numerical versus verbal presentation). Welkenhuysen 2001a relates to the numerical presentation. |
|
| Participants | 147 first year students in medical sciences. No details provided regarding response rate. |
|
| Interventions | Informative text describing the cystic fibrosis, in the context of a decision regarding prenatal diagnosis, using positive versus negative framing of the numerical probabilities of the disease. | |
| Outcomes | Decision whether to ask for a prenatal test (persuasiveness). | |
| Notes | Screening message, attribute framing. Examples of messages: Negative frame: 25% change of having a cystic fibrosis child. Positive frame: 75% chance of having a child without cystic fibrosis. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Welkenhuysen 2001b.
| Methods | Randomized controlled trial; 2x2 factorial design (positive versus negative framing) x (numerical versus verbal presentation). Welkenhuysen 2001b relates to the verbal presentation. |
|
| Participants | 153 first year students in medical sciences. No details provided regarding response rate. |
|
| Interventions | Informative text describing the cystic fibrosis, in the context of a decision regarding prenatal diagnosis, using positive versus negative framing of the verbal probabilities of the disease. | |
| Outcomes | Decision whether to ask for a prenatal test (persuasiveness). | |
| Notes | Screening message, attribute framing. "Moderate chance" and "high chance" were the translation of 25% and 75% probabilities. Examples of messages: Negative fame: moderate chance of having a cystic fibrosis child. Positive frame: high chance of having a child without cystic fibrosis. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | |
Williams 2001.
| Methods | Randomized controlled trial; 2x3 factorial design: statistical frame (positive versus negative) x behavior frame (positive versus negative versus no framing). | |
| Participants | 539 women aged 30 to 70. No details provided regarding response rate. |
|
| Interventions | Brochure with information regarding risk factors for breast cancer and the role of family history as a risk factor. Statistics conveying the percentage of women who are likely to get breast cancer was framed positively or negatively (statistical framing) and the message providing behavioral recommendation was framed positively, negatively or not framed (behavior framing). | |
| Outcomes | Breast self‐examination behavior (behavior). | |
| Notes | Screening message; goal framing. Examples of messages: "Three levels of framing were applied to the section in the brochure that provides behavioral recommendations: gain framing, loss framing, and no framing." No further details provided. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | Low risk | Breast self‐examination behavior. |
Wilson 1987a.
| Methods | Randomized controlled trial. Wilson 1987a is experiment 1. |
|
| Participants | 91 undergraduate students enrolled in Introductory Psychology at San Diego State University. No details provided regarding response rate. |
|
| Interventions | Subjects were presented with a hypothetical situation relating to a terminal liver disease and the probability of dying versus survival of a surgery that would improve long term survival. | |
| Outcomes | Choosing the alternative of surgery (persuasiveness). | |
| Notes | Treatment message; attribute framing. Examples of messages: Positive frame: the probability of surviving the operation was 20 per cent. Negative frame: while the probability of dying during the operation was 80 per cent. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | Hypothetical outcome. |
Wilson 1987b.
| Methods | Randomized controlled trial. Wilson 1987b is experiment 2. |
|
| Participants | 91 undergraduate students enrolled in Introductory Psychology at San Diego State University. No details provided regarding response rate. |
|
| Interventions | Scenario in which a genetic counsellor counsels an expectant woman who is a haemophilia carrier about the chances of having a problem birth. She is presented with the probability of having a "normal" versus an "affected" child. | |
| Outcomes | Opinion whether or not the woman should abort (persuasiveness). | |
| Notes | "Other" type of message; attribute framing. Examples of messages: Positive frame: the probability of having a 'normal' child was 50 per cent. Negative frame: the probability of having an 'affected' child was 50 per cent. |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | Unclear risk | No details provided. |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | Low risk | |
| Objective outcome? | High risk | Hypothetical outcome. |
Zimmermann 2000.
| Methods | Non‐randomized cross‐over study. | |
| Participants | 40 breast cancer patients, 25 to 70 years old, with history of breast surgery and past adjuvant chemotherapy experience without recurrence of cancer at the time of the study; residing in Verona, Italy. No details provided regarding response rate. |
|
| Interventions | Cancer treatment information presented in a positive (probability that the cancer is cured) or a negative (probability that the cancer will return) frame. For each frame 3 messages were presented each with different level of risk of recurrence (10%, 20% and 30%). | |
| Outcomes | Hypothetical treatment decision about adjuvant chemotherapy (yes, no) (persuasiveness). | |
| Notes | Treatment message, attribute framing. Examples of Messages: Mrs. A (B) (high probability of cure) (low risk of recurrence) In case Mrs. A (B) chooses chemotherapy, the probability of cure (recurrence) in the next five years increases from 85% to 90% (decreases from 15% to 10%). No chemotherapy probability of cure 85% (probability of recurrence 15%). Chemotherapy probability of cure 90% (probability of recurrence 10%). |
|
| Risk of bias | ||
| Bias | Authors' judgement | Support for judgement |
| Allocation concealment (selection bias) | High risk | |
| Randomized design? If cross over design, order of presentation of the different formats randomized? | High risk | Patients presented first with 3 negatively‐framed scenarios and 2 months later with 3 positively‐framed scenarios (fixed order). |
| Objective outcome? | High risk | |
Characteristics of excluded studies [ordered by study ID]
| Study | Reason for exclusion |
|---|---|
| Aarts 2007 | Not a comparison of interest (examined the role of negative affect in moderating goal priming effects) |
| Abood 2005 | Not a comparison of interest (compared loss‐framed message to “usual communication”) |
| Armstrong 2002 | Not a comparison of interest (compared two types of graphs and not verbal health messages) |
| Arora 2000 | Not an appropriate study design (study design not clear) |
| Blanton 2001b | Not a comparison of interest (framing is not of the health information itself) |
| Blumenschein 1998 | Not a comparison of interest (not a health message; part of utility elicitation tool and outcome is the utility of an outcome) |
| Brug 2003 | Data for the comparison of interest not reported and not available from author (study 1 reported in van Assema 2001a; studies 2 and 3 not reported elsewhere) |
| Brunton 2007 | Not an appropriate study design (not RCT; all patients received the same information) |
| Carlson 2005 | Not an original study (review) |
| Christensen 1995 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B (or test versus no test)) |
| Collins 2005 | Use of different information in the comparison groups (differences in the presentation besides framing) |
| Collins 2006 | Data for the comparison of interest not reported and not available from author |
| Cox 2001 | Use of different information in the comparison groups |
| Cunningham 2006 | Use of different information in the comparison groups |
| Diamond 1992 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B) |
| Druckman 2001 | Use of different information in the comparison groups (the risky choice problem) |
| Fagley 1987 | Use of different information in the comparison groups (the risky choice problem) |
| Gallagher 2007 | Not a comparison of interest |
| Gurm 2000 | Use of different information in the comparison groups (different presentation format besides framing: percentage for positive frame and frequency for negative frame) |
| Hein 1997 | Not an original study |
| Kahneman 1979 | Use of different information in the comparison groups (the risky choice problem) |
| Kahneman 2007 | Not an original study |
| Kalichman 1995 | Use of different information in the comparison groups (different amount of information) |
| Kiene 2005 | Use of different information in the comparison groups |
| King 2005 | Not an original study (review) |
| Kuhberger 1998 | Not an original study (review) |
| Lalor 1989 | Data for the comparison of interest not reported and not available from author |
| Levin 1988 | Data for the comparison of interest not reported and not available from author |
| Levin 1990 | Use of different information in the comparison groups (the risky choice problem) |
| Levin 1993 | Use of different information in the comparison groups (the risky choice problem) |
| Levin 1998 | Not an original study (review) |
| Linville 2001 | Data for the comparison of interest not reported and not available from author |
| Maheswaran 1990 | Data for the comparison of interest not reported and not available from author |
| Marteau 1989 | Data for the comparison of interest not reported and not available from author |
| Mayhorn 2002 | Use of different information in the comparison groups (the risky choice problem) |
| McKee 2001 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B) |
| McNeil 1982 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B) |
| Meyerowitz 1987 | Data for the comparison of interest not reported and not available from author |
| Millar 2000 | Data for the comparison of interest not reported and not available from author |
| Miller 1999 | Data for the comparison of interest not reported and not available from author |
| O'Connor 1985 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B) |
| O'Connor 1989 | No outcome of interest evaluated (study outcome was the preference for treatment A versus treatment B) |
| Oliver 2004 | No outcome of interest evaluated (standard gamble) |
| Orbell 2006 | Not a comparison of interest |
| Ortendahl 2006 | Not an original study |
| Percy 1995 | Not a comparison of interest (framing is not of the health information itself but of the shifting frame in a probability trade off task) |
| Plank 1994 | Not an original study |
| Rettig MckKee 2001 | No outcome of interest evaluated (study outcome was the preference for surgery versus radiation therapy) |
| Ronnlund 2005 | Use of different information in the comparison groups (the risky choice problem) |
| Rothman 1993 | Data for the comparison of interest not reported and not available from author |
| Rothman 1997 | Not an original study (review) |
| Rybash 1989 | Not a comparison of interest (messages delivered as positively framed are in fact partly negatively‐framed and vice versa) |
| Salovey 2004 | Not an original study (review) |
| Schneider 1992 | Use of different information in the comparison groups |
| Schneider 1995 | Not an original study |
| Shiloh 2001 | Use of different information in the comparison groups |
| Stalmeier 1999 | Not a comparison of interest (about health utilities) |
| Steffen 1994 | Data for the comparison of interest not reported and not available from author |
| Stuart 2003 | Use of different information in the comparison groups |
| Trotto 2001 | Data for the comparison of interest not reported and not available from author |
| Tversky 1981 | Use of different information in the comparison groups (the risky choice problem) |
| Wang 2001 | Use of different information in the comparison groups (the risky choice problem) |
| Wilson 1990 | Not a comparison of interest (framing is not of health information itself) |
| Wong 2002 | Not an appropriate study design ("Inappropriate number of subjects at follow‐up" for the outcome of interest) |
Differences between protocol and review
The protocol title was: 'Negative versus positive framing of health information'.
The protocol inclusion criteria for type of studies include RCTS, quasi‐RCTs and controlled before and after studies (CBAs). We omitted CBAs from the review because its categorization reflects mainly the outcome assessment method which could apply to different types of study designs.
The protocol inclusion criteria for type of outcome included any measure (including self‐reported) of the different outcomes. In the review, and for the outcome of understanding, we considered only objective measurements after carefully considering the nature of that outcome.
Contributions of authors
EAA: study conception and design, screening, data extraction, data analysis and interpretation. ADO: study conception and design, data analysis and interpretation. JH: study conception and design, data extraction data analysis and interpretation. GV: screening, data extraction, data analysis and interpretation. CC: screening, data extraction. DB: screening, data extraction. HJS: study conception and design, screening, data extraction, data analysis and interpretation.
Sources of support
Internal sources
-
State University of New York at Buffalo, NY, USA.
Salary support to EA.
-
Italian National Cancer Institute, Regina Elena, Rome, Italy.
Salary support to HJS
External sources
-
Norwegian Research Council, Norway.
Funding to AO.
-
European Commission: The human factor, mobility and Marie Curie Actions. Scientist Reintegration Grant: IGR 42194 ‐ GRADE, Not specified.
Funding to HJS.
Declarations of interest
Some of the review authors were also authors of one included study: Carling 2010. A review author who was not a study author (IT), as well as EA, were involved in data abstraction and analysis for these studies.
New
References
References to studies included in this review
Apanovitch 2003 {published data only}
- Apanovitch AM, McCarthy D, Salovey P. Using message framing to motivate HIV testing among low‐income, ethnic minority women. Health Psychology 2003;22(1):60‐7. [DOI] [PubMed] [Google Scholar]
Banks 1995 {published data only}
- Banks SM, Salovey P, Greener S, Rothman AJ, Moyer A, Beauvais J, et al. The effects of message framing on mammography utilization. Health Psychology 1995;14(2):178‐84. [DOI] [PubMed] [Google Scholar]
Blanton 2001a {published data only}
- Blanton H, Vanden Eijnden RJ, Buunk BP, Gibbons FX, Gerrard M, Bakker A. Accentuate the negative: social images in the prediction and promotion of condom use. Journal of Applied Social Psychology 2001;31(2):274‐95. [Google Scholar]
Block 1995a {published data only}
- Block LG, Keller PA. When to accentuate the negative: the effects of perceived efficacy and message framing on intentions to perform a health‐related behavior. Journal of Marketing Research 1995;32(2):192‐203. [Google Scholar]
Block 1995b {published data only}
- Block LG, Keller PA. When to accentuate the negative: the effects of perceived efficacy and message framing on intentions to perform a health‐related behavior. Journal of Marketing Research 1995;32(2):192‐203. [Google Scholar]
Block 1995c {published data only}
- Block LG, Keller PA. When to accentuate the negative: the effects of perceived efficacy and message framing on intentions to perform a health‐related behavior. Journal of Marketing Research 1995;32(2):192‐203. [Google Scholar]
Block 1995d {published data only}
- Block LG, Keller PA. When to accentuate the negative: the effects of perceived efficacy and message framing on intentions to perform a health‐related behavior. Journal of Marketing Research 1995;32(2):192‐203. [Google Scholar]
Carling 2010 {published data only}
- Carling C, Tove Kristoffersen D, Oxman AD, Flottorp S, Fretheim A, Schünemann HJ, et al. The effect of how outcomes are framed on decisions about whether to take antihypertensive medication: a randomized trial. PLoS One 2010;5(3):e9469. [DOI] [PMC free article] [PubMed] [Google Scholar]
Cherubini 2005a {published data only}
- Cherubini P, Rumiati R, Rossi D, Nigro F, Calabro A. Improving attitudes toward prostate examinations by loss‐framed appeals. Journal of Applied Social Psychology 2005; Vol. 35, issue 4:732‐44. [0021‐9029]
Cherubini 2005b {published data only}
- Cherubini P, Rumiati R, Rossi D, Nigro F, Calabro A. Improving attitudes toward prostate examinations by loss‐framed appeals. Journal of Applied Social Psychology 2005; Vol. 35, issue 4:732‐44. [0021‐9029]
Detweiler 1999 {published data only}
- Detweiler JB, Bedell BT, Salovey P, Pronin E, Rothman AJ. Message framing and sunscreen use: gain‐framed messages motivate beach‐goers. Health Psychology 1999;18(2):189‐96. [DOI] [PubMed] [Google Scholar]
Donovan 2000 {published data only}
- Donovan RJ, Jalleh G. Positive versus negative framing of a hypothetical infant immunization: the influence of involvement. Health Education & Behavior 2000;27(1):82‐95. [DOI] [PubMed] [Google Scholar]
Finney 2002a {published data only}
- Finney LJ, Iannotti RJ. Message framing and mammography screening: a theory‐driven intervention. Behavioral Medicine 2002;28(1):5‐14. [DOI] [PubMed] [Google Scholar]
Finney 2002b {published data only}
- Finney LJ, Iannotti RJ. Message framing and mammography screening: a theory‐driven intervention. Behavioral Medicine 2002;28(1):5‐14. [DOI] [PubMed] [Google Scholar]
Hsiao 2002a {published data only}
- Hsiao ET. Using message framing to promote regular physical activity in college‐age women. Dissertation. Ohio State University 2002.
Hsiao 2002b {published data only}
- Hsiao ET. Using message framing to promote regular physical activity in college‐age women. Dissertation. Ohio State University 2002.
Jasper 2001 {published data only}
- Jasper JD, Goel R, Einarson A, Gallo M, Koen G. Effects of framing teratogenic risk perception in pregnant women. Lancet 2001;358:1237‐8. [DOI] [PubMed] [Google Scholar]
Krishnamurthy 2001a {published data only}
- Krishnamurthy P, Carter P, Blair E. Attribute framing and goal framing effects in health decisions. Organizational Behavior and Human Decision Processes 2001;85(2):382‐99. [DOI] [PubMed] [Google Scholar]
Krishnamurthy 2001b {published data only}
- Krishnamurthy P, Carter P, Blair E. Attribute framing and goal framing effects in health decisions. Organizational Behavior and Human Decision Processes 2001;85(2):382‐99. [DOI] [PubMed] [Google Scholar]
Krishnamurthy 2001c {published data only}
- Krishnamurthy P, Carter P, Blair E. Attribute framing and goal framing effects in health decisions. Organizational Behavior and Human Decision Processes 2001;85(2):382‐99. [DOI] [PubMed] [Google Scholar]
Krishnamurthy 2001d {published data only}
- Krishnamurthy P, Carter P, Blair E. Attribute framing and goal framing effects in health decisions. Organizational Behavior and Human Decision Processes 2001;85(2):382‐99. [DOI] [PubMed] [Google Scholar]
Kroll 2005 {published data only}
- Kroll E. The effects of message framing and gender on physical exercise communications to high school students. Dissertation. Hofstra University 2005.
Lalor 1990 {published data only}
- Lalor KM. The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast examination. Applied Research and Evaluation 1990;10:183‐92. [DOI] [PubMed] [Google Scholar]
Lauver 1990 {published data only}
- Lauver D, Rubin M. Message framing, dispositional optimism, and follow‐up for abnormal Papanicolaou tests. Research in Nursing & Health 1990;13(3):199‐207. [DOI] [PubMed] [Google Scholar]
Lerman 1992 {published data only}
- Lerman C, Ross E, Boyce A, Gorchov PM, McLaughlin R, Rimer B, et al. The impact of mailing psychoeducational materials to women with abnormal mammograms. American Journal of Public Health 1992;82(5):729‐30. [DOI] [PMC free article] [PubMed] [Google Scholar]
Lewis 2003 {published data only}
- Lewis CL, Pignone MP, Sheridan SL, Downs SM, Kinsinger LS. A randomized trial of three videos that differ in the framing of information about mammography in women 40 to 49 years old. Journal of General Internal Medicine 2003;18(11):875‐83. [DOI] [PMC free article] [PubMed] [Google Scholar]
Llewellyn‐Thomas 1995 {published data only}
- Llewellyn‐Thomas HA, McGreal MJ, Thiel EC. Cancer patients' decision making and trial‐entry preferences: the effects of "framing" information about short‐term toxicity and long‐term survival. Medical Decision Making 1995;15(1):4‐12. [DOI] [PubMed] [Google Scholar]
Mann 2004 {published data only}
- Mann T, Sherman D, Updegraff J. Dispositional motivations and message framing: a test of the congruency hypothesis in college students. Health Psychology 2004;23(3):330‐4. [DOI] [PMC free article] [PubMed] [Google Scholar]
McCaul 2002 {published data only}
- McCaul KD, Johnson RJ, Rothman AJ. The effects of framing and action instructions on whether older adults obtain flu shots. Health Psychology 2002;21(6):624‐8. [DOI] [PubMed] [Google Scholar]
O'Connor 1996 {published data only}
- O'Connor AM, Pennie RA, Dales RE. Framing effects on expectations, decisions, and side effects experienced: the case of influenza immunization [published erratum appears in Journal of Clinical Epidemiology 1997 Jun;50(6):747‐8]. Journal of Clinical Epidemiology 1996;49(11):1271‐6. [DOI] [PubMed] [Google Scholar]
O'Connor 2005 {published data only}
Rivers 2005a {published data only}
Rivers 2005b {published data only}
Robberson 1988a {published data only}
- Robberson MR, Rogers RW. Beyond fear appeals: negative and positive appeals to health and self‐esteem. Journal of Applied Social Psychology 1988;18(3):277‐287. [Google Scholar]
Robberson 1988b {published data only}
- Robberson MR, Rogers RW. Beyond fear appeals: negative and positive appeals to health and self‐esteem. Journal of Applied Social Psychology 1988;18(3):277‐87. [Google Scholar]
Ruiter 2003 {published data only}
- Ruiter RAC, Kok G, Verplanken B, Eersel G. Strengthening the persuasive impact of fear appeals: the role of action framing. The Journal of Social Psychology 2003;143(3):397‐400. [DOI] [PubMed] [Google Scholar]
Schmitt 2004 {published data only}
- Schmitt SK. The effects of message framing on women's goal and implementation intentions to obtain a blood cholesterol screening test. Dissertation. Walden University. Schmitt, Susan K : , US, 2004.
Schneider 2001a {published data only}
- Schneider TR, Salovey P, Pallonen U, Mundorf N, Smith NF, Steward WT. Visual and auditory message framing effects on tobacco smoking. Journal of Applied Social Psychology 2001;31(4):667‐82. [Google Scholar]
Schneider 2001b {published data only}
- Schneider TR, Salovey P, Apanovitch AM, Pizarro J, McCarthy D, Zullo J, et al. The effects of message framing and ethnic targeting mammography use among low income women. Health Psychology 2001;20:256‐66. [DOI] [PubMed] [Google Scholar]
Schneider 2001c {published data only}
- Schneider TR, Salovey P, Apanovitch AM, Pizarro J, McCarthy D, Zullo J, et al. The effects of message framing and ethnic targeting mammography use among low income women. Health Psychology 2001;20:256‐66. [DOI] [PubMed] [Google Scholar]
Scott 2006a {published data only}
Scott 2006b {published data only}
Steward 2002 {published data only}
- Steward WT. Effects of perceived outcome and attitude toward affected group. Dissertation. Yale University 2002.
van Assema 2001a {published data only}
- Assema P, Martens M, Ruiter RA, Brug J. Framing of nutrition education messages in persuading consumers of the advantages of a healthy diet. Journal of Human Nutrition and Dietetics 14;6:435‐42. [DOI] [PubMed] [Google Scholar]
van Assema 2001b {published data only}
- Assema P, Martens M, Ruiter RA, Brug J. Framing of nutrition education messages in persuading consumers of the advantages of a healthy diet. Journal of Human Nutrition and Dietetics 14;6:435‐42. [DOI] [PubMed] [Google Scholar]
Welkenhuysen 2001a {published data only}
- Welkenhuysen M, Evers‐Kiebooms G, d'Ydewalle G. The language of uncertainty in genetic risk communication: framing and verbal versus numerical information. Patient Education and Counseling 2001;43(2):179‐87. [DOI] [PubMed] [Google Scholar]
Welkenhuysen 2001b {published data only}
- Welkenhuysen M, Evers‐Kiebooms G, d'Ydewalle G. The language of uncertainty in genetic risk communication: framing and verbal versus numerical information. Patient Education and Counseling 2001;43(2):179‐87. [DOI] [PubMed] [Google Scholar]
Williams 2001 {published data only}
- Williams TL, Clarke V, Borland R. Effects of message framing on breast‐cancer‐related beliefs and behaviors: the role of mediating factors. Journal of Applied Social Psychology 2001;31(5):925‐50. [Google Scholar]
Wilson 1987a {published data only}
- Wilson D, Kaplan R, Schneiderman L. Framing of decisions and selections of alternatives in health care. Social Behavior 1987;2:51‐9. [Google Scholar]
Wilson 1987b {published data only}
- Wilson D, Kaplan R, Schneiderman L. Framing of decisions and selections of alternatives in health care. Social Behavior 1987;2:51‐9. [Google Scholar]
Zimmermann 2000 {published data only}
- Zimmermann C, Baldo C, Molino A. Framing of outcome and probability of recurrence: breast cancer patients' choice of adjuvant chemotherapy (ACT) in hypothetical patient scenarios. Breast Cancer Research and Treatment 2000;60(1):9‐14. [DOI] [PubMed] [Google Scholar]
References to studies excluded from this review
Aarts 2007 {published data only}
Abood 2005 {published data only}
Armstrong 2002 {published data only}
- Armstrong K, Schwartz JS, Fitzgerald G, Putt M, Ubel PA. Effect of framing as gain versus loss on understanding and hypothetical treatment choices: survival and mortality curves. Medical Decision Making 2002;22(1):76‐83. [DOI] [PubMed] [Google Scholar]
Arora 2000 {published data only}
- Arora R. Message framing and credibility: application in dental services. Health Marketing Quarterly 2000;18(1‐2):29‐44. [DOI] [PubMed] [Google Scholar]
Blanton 2001b {published data only}
- Blanton H, Vanden Eijnden RJ, Buunk BP, Gibbons FX, Gerrard M, Bakker A. An introduction to deviance‐regulation theory: the effect of behavioral norms on message framing. Personality and Social Psychology Bulletin 2001;27(7):848‐58. [Google Scholar]
Blumenschein 1998 {published data only}
- Blumenschein K, Johannesson M. An experimental test of question framing in health state utility assessment. Health Policy 1998;45(3):187‐93. [DOI] [PubMed] [Google Scholar]
Brug 2003 {published data only}
- Brug J, Ruiter RAC, Assema P. The (ir)relevance of framing nutrition education messages. Nutrition and Health 2003;17(1):9‐20. [DOI] [PubMed] [Google Scholar]
Brunton 2007 {published data only}
- Brunton MA. One message for all? Framing public health messages to recognize diversity. International Journal of Intercultural Relations 2007; Vol. 31, issue 1:127‐32.
Carlson 2005 {published data only}
- Carlson LE, Feldman‐Stewart D, Tishelman C, Brundage MD, SCRN Communication Team. Patient‐professional communication research in cancer: an integrative review of research methods in the context of a conceptual framework. Psycho‐Oncology 2005;14(10):812‐28; discussion 829‐30. [1057‐9249] [DOI] [PubMed] [Google Scholar]
Christensen 1995 {published data only}
- Christensen C, Heckerling P, Mackesy‐Amiti ME, Bernstein LM. Pervasiveness of framing effects among physicians and medical students. Journal of Behavioral Decision Making 1995;8(3):169‐80. [Google Scholar]
Collins 2005 {published data only}
- Collins CA. The influence of message framing and previous alcohol consumption on attitudes, perceived norms, intentions, and behavior related to alcohol use in a college sample. Dissertation. University of Kansas 2005.
Collins 2006 {published data only}
- Collins PJ. Message framing and sexual abstinence messages. Dissertation. Purdue University. Collins, Philip John: , US, 2006.
Cox 2001 {published data only}
- Cox D, Cox A. Communicating the consequences of early detection: the role of evidence and framing. Journal of Marketing 2001;65:91‐103. [Google Scholar]
Cunningham 2006 {published data only}
Diamond 1992 {published data only}
- Diamond L, Lerch FJ. Fading frames: data presentation and framing effects. Decision Sciences 1992;23:1050‐71. [Google Scholar]
Druckman 2001 {published data only}
- Druckman JN. Evaluating framing effects. Journal of Economic Psychology 2001;20:91‐101. [Google Scholar]
Fagley 1987 {published data only}
- Fagley NS, Miller PM. The effects of decision framing on choice of risky vs certain options. Organizational Behavior and Human Decision Processes 1987;39:264‐77. [Google Scholar]
Gallagher 2007 {published data only}
Gurm 2000 {published data only}
- Gurm HS, Litaker DG. Framing procedural risks to patients: is 99% safe the same as a risk of 1 in 100?. Academic Medicine 2000;75(8):840‐2. [DOI] [PubMed] [Google Scholar]
Hein 1997 {published data only}
- Hein K. Framing and reframing. Jornal of Adolescent Health 1997;21(4):215‐7. [DOI] [PubMed] [Google Scholar]
Kahneman 1979 {published data only}
- Kahneman D, Tversky A. Prospect theory: an analysis of decision making under risk. Econometrica 1979;6:621‐30. [Google Scholar]
Kahneman 2007 {published data only}
Kalichman 1995 {published data only}
- Kalichman SC, Coley B. Context framing to enhance HIV‐antibody‐testing messages targeted to African American women. Health Psychology 1995;14(3):247‐54. [DOI] [PubMed] [Google Scholar]
Kiene 2005 {published data only}
- Kiene SM, Barta WD, Zelenski JM, Cothran DL. Why are you bringing up condoms now? the effect of message content on framing effects of condom use messages. Health Psychology 2005;24(3):321‐6. [DOI] [PubMed] [Google Scholar]
King 2005 {published data only}
- King M, Nazareth I, Lampe F, Bower P, Chandler M, Morou M, et al. Conceptual framework and systematic review of the effects of participants' and professionals' preferences in randomised controlled trials. Health Technology Assessment (Winchester, England) 2005; Vol. 9, issue 35:1‐186. [DOI] [PubMed]
Kuhberger 1998 {published data only}
- Kuhberger A. The influence of framing on risky decisions: a meta‐analysis. Organizational Behavior and Human Decision Processes 1998;75(1):23‐54. [DOI] [PubMed] [Google Scholar]
Lalor 1989 {published data only}
- Lalor KM, Hailey BJ. The effects of message framing and feelings of susceptibility to breast cancer on reported frequency of breast self‐examination. International Quarterly of Community Health Education 1989;10(3):183‐92. [DOI] [PubMed] [Google Scholar]
Levin 1988 {published data only}
- Levin IP, Schnittjer SK, Thee SL. Information framing effects in social and personal decisions. Journal of Experimental Social Psychology 1988;24(5):520‐9. [Google Scholar]
Levin 1990 {published data only}
- Levin IP, Chapman DP. Risk taking, frame of reference, and characterization of victim groups in AIDS treatment decisions. Journal of Experimental Social Psychology 1990;26(5):421‐34. [Google Scholar]
Levin 1993 {published data only}
- Levin IP, Chapman DP. Risky decision making and allocation of resources for leukemia and AIDS programs. Health Psychology 1993;12(2):110‐7. [DOI] [PubMed] [Google Scholar]
Levin 1998 {published data only}
- Levin IP, Schneider SL, Gaeth GJ. All frames are not created equal: a typology and critical analysis of framing effects. Organizational Behavior and Human Decision Processes 1998;76(2):149‐88. [DOI] [PubMed] [Google Scholar]
Linville 2001 {published data only}
- Linville PW, Fischer GW, Fischhoff B. AIDS risk perceptions and decision biases. In: Pryor JB, Reeder GD editor(s). The Social Psychology of HIV Infection. Hillsdale, NJ: Lawrence Erbaum, 2001. [Google Scholar]
Maheswaran 1990 {published data only}
- Maheswaran D, Meyers‐Levy J. The influence of message framing and issue involvement. Journal of Marketing Research 1990;27:361‐7. [Google Scholar]
Marteau 1989 {published data only}
- Marteau TM. Framing of information: its influence upon decisions of doctors and patients. British Journal of Social Psychology 1989;28:89‐95. [DOI] [PubMed] [Google Scholar]
Mayhorn 2002 {published data only}
- Mayhorn CB, Fisk AD, Whittle JD. Decisions, decisions: analysis of age, cohort, and time of testing on framing of risky decision options. Human Factors: The Journal of Human Factors and Ergonomics Society 2002;44(4):515‐21. [DOI] [PubMed] [Google Scholar]
McKee 2001 {published data only}
- McKee DR. The effects of framing on younger and older adults' medical decision making. Dissertation. Eberly College of Arts and Sciences at the West Virginia University 2001.
McNeil 1982 {published data only}
- McNeil BJ, Pauker SG, Sox HC Jr, Tversky A. On the elicitation of preferences for alternative therapies. NEJM 1982;306(21):1259‐62. [DOI] [PubMed] [Google Scholar]
Meyerowitz 1987 {published data only}
- Meyerowitz BE, Chaiken S. The effect of message framing on breast self‐examination attitudes, intentions, and behavior. Journal of Personality and Social Psychology 1987;52(3):500‐10. [DOI] [PubMed] [Google Scholar]
Millar 2000 {published data only}
- Millar MG, Karen MU. Promoting safe driving behaviors: the influence of message framing and issue involvement. Journal of Applied Social Psychology 2000;30(4):853‐66. [Google Scholar]
Miller 1999 {published data only}
- Miller SM, Buzaglo JS, Simms SL, Green V, Bales C, Mangan CE, et al. Monitoring styles in women at risk for cervical cancer: implications for the framing of health‐relevant messages. Annals of Behavioral Medicine 1999;21(1):27‐34. [DOI] [PubMed] [Google Scholar]
O'Connor 1985 {published data only}
- O'Connor AM, Boyd NF, Tritchler DL, Kriukov Y, Sutherland H, Till JE. Eliciting preferences for alternative cancer drug treatments: the influence of framing, medium, and rater variables. Medical Decision Making 1985;5(4):453‐63. [DOI] [PubMed] [Google Scholar]
O'Connor 1989 {published data only}
- O'Connor AM. Effects of framing and level of probability on patients' preferences for cancer chemotherapy. Journal of Clinical Epidemiology 1989;42(2):119‐26. [DOI] [PubMed] [Google Scholar]
Oliver 2004 {published data only}
- Oliver A. Testing the internal consistency of the standard gamble in 'success' and 'failure' frames. Social Science and Medicine 2004;58(11):2219‐29. [DOI] [PubMed] [Google Scholar]
Orbell 2006 {published data only}
- Orbell S, Hagger M. Temporal framing and the decision to take part in type 2 diabetes screening: effects of individual differences in consideration of future consequences on persuasion. Health Psychology 2006;25(4):537‐48. [DOI] [PubMed] [Google Scholar]
Ortendahl 2006 {published data only}
- Ortendahl M, Fries JF, Ortendahl M, Fries JF. Discounting and risk characteristics in clinical decision‐making. Medical Science Monitor 2006;12(3):RA41‐5. [PubMed] [Google Scholar]
Percy 1995 {published data only}
- Percy ME, Llewellyn‐Thomas H. Assessing preferences about the DNR order: does it depend on how you ask?. Medical Decision Making 1995;15(3):209‐16. [DOI] [PubMed] [Google Scholar]
Plank 1994 {published data only}
- Plank DM. Framing treatment options: a method to enhance informed consent. Clinical Nurse Specialist 1994;8(4):174‐9. [PubMed] [Google Scholar]
Rettig MckKee 2001 {unpublished data only}
- Rettg McKee, D. The effects of framing on younger and older adults' medical decision making. Dissertation 2001.
Ronnlund 2005 {published data only}
- Ronnlund M, Karlsson E, Laggnas E, Larsson L, Lindstrom T, Ronnlund M, et al. Risky decision making across three arenas of choice: are younger and older adults differently susceptible to framing effects?. Journal of General Psychology 2005;132(1):81‐92. [DOI] [PubMed] [Google Scholar]
Rothman 1993 {published data only}
- Rothman AJ, Salovey P, Antone C, Keough K, Martin CG. The influence of message framing on intentions to perform health behaviors. Journal of Experimental Social Psychology 1993;29:408‐33. [Google Scholar]
Rothman 1997 {published data only}
- Rothman AJ, Salovey P. Shaping perceptions to motivate healthy behavior: the role of message framing. Psychological Bulletin 1997;121(1):3‐19. [DOI] [PubMed] [Google Scholar]
Rybash 1989 {published data only}
- Rybash JM, Roodin PA. The framing heuristic influences judgements about younger and older adults' decision to refuse medical treatment. Applied Cognitive Psychology 1989;3(2):171‐80. [DOI] [PubMed] [Google Scholar]
Salovey 2004 {published data only}
- Salovey P, Williams‐Piehota P. Field experiments in social psychology: message framing and the promotion of health protective behaviors. American Behavioral Scientist 2004;47(5):488‐505. [Google Scholar]
Schneider 1992 {published data only}
- Schneider SL. Framing and conflict: aspiration level contingency, the status quo, and current theories of risky choice. Journal of Experimental Psychology: Learning, Memory, and Cognition 1992;18(5):1040‐57. [DOI] [PubMed] [Google Scholar]
Schneider 1995 {published data only}
- Schneider SL, Levin IP, Gaeth G. The three faces of framing: a typology of framing effects. 36th Annual Meeting of the Psychonomic Society, Los Angeles, California. 1995.
Shiloh 2001 {published data only}
- Shiloh S, Eini NJ, Ben‐Neria Z, Sagi M. Framing of prenatal screening test results and women's health‐illness orientations as determinants of perceptions of fetal health and approval of amniocentesis. Psychology & Health. European Journal of Social Psychology 2001;16(3):313‐25. [Google Scholar]
Stalmeier 1999 {published data only}
- Stalmeier PF, Bezembinder TG. The discrepancy between risky and riskless utilities: a matter of framing?. Medical Decision Making 1999;19(4):435‐47. [DOI] [PubMed] [Google Scholar]
Steffen 1994 {published data only}
- Steffen VJ, Sternberg L, Teegarden LA. Practice and persuasive frame: effects on beliefs, intention, and performance of a cancer self‐examination. Journal of Applied Social Psychology 1994;24(10):897‐925. [Google Scholar]
Stuart 2003 {published data only}
- Stuart AE, Blanton H. The effects of message framing on behavioral prevalence assumptions. European Journal of Social Psychology 2003;33(1):93‐102. [Google Scholar]
Trotto 2001 {published data only}
- Trotto PA. The effect of message framing and imagination on breast self‐examination compliance. Dissertation. College of Liberal Arts and Sciences, DePaul University 2001.
Tversky 1981 {published data only}
- Tversky A, Kahneman D. The framing of decisions and the psychology of choice. Science 1981;211(4481):453‐8. [DOI] [PubMed] [Google Scholar]
Wang 2001 {published data only}
- Wang X, Simons F, Bredart S. Social cues and verbal framing in risky choice. Journal of Behavioral Decision Making 2001;14(1):1‐15. [Google Scholar]
Wilson 1990 {published data only}
- Wilson DK, Wallston KA, King JE. Effects of contract framing, motivation to quit, and self‐efficacy on smoking reduction. Journal of Applied Social Psychology 1990;20(7):531‐47. [Google Scholar]
Wong 2002 {published data only}
- Wong CO, McMurray NE. Framing communication: Communicating the antismoking message effectively to all smokers. Journal of Community Psychology 2002;30(4):433‐47. [Google Scholar]
Additional references
Akl 2011
- Akl EA, Oxman AD, Herrin J, Vist GE, Terrenato I, Sperati F, et al. Using alternative statistical formats for presenting risks and risk reductions. Cochrane Database of Systematic Reviews 2011, Issue 3. [DOI: 10.1002/14651858.CD006776.pub2] [DOI] [PMC free article] [PubMed] [Google Scholar]
Brendl 1995
- Brendl CM, Higgins ET, Lemm KM , Brendl CM, Higgins ET, Lemm KM. Sensitivity to varying gains and losses: the role of self‐discrepancies and event framing. Journal of Personality and Social Psychology 1995;69(6):1028‐51. [DOI] [PubMed] [Google Scholar]
Carling 2008
- Carling C, Tove Kristoffersen D, Herrin J, Treweek S, Oxman AD, Schünemann HJ, et al. How should the impact of different presentations of treatment effects on patient choice be evaluated? a pilot randomized trial. PLoS One 2008;3(11):e3693. [DOI] [PMC free article] [PubMed] [Google Scholar]
Carver 2000
- Carver, CS, Sutton SK, Scheier MF. Action, emotion, and personality: emerging conceptual integration. Personality and Social Psychology Bulletin 2000;26:741‐51. [Google Scholar]
Cooper 1994
- Cooper H, Hedges LV. The Handbook of Research Synthesis. New York: Russell Sage Foundation, 1994. [Google Scholar]
Curtin 2002
- Curtin F, Elbourne D, Altman DG. Meta‐analysis combining parallel and cross‐over clinical trials. II: Binary outcomes. Statistics in Medicine 2002;21:2145‐59. [DOI] [PubMed] [Google Scholar]
Edwards 2001
- Edwards A, Elwyn G, Covey J, Matthew E, Pill Roison. Presenting risk information? a review of the effects of framing and other manipulations on patient outcomes. Journal of Health Communication 2001;6(1):61‐82. [DOI] [PubMed] [Google Scholar]
Elbourne 2002
- Elbourne DR, Altman DG, Higgins JP, Curtin F, Worthington HV, Vail A. Meta‐analyses involving cross‐over trials: methodological issues. International Journal of Epidemiology 2002;31:140‐9. [DOI] [PubMed] [Google Scholar]
Ganzach 1997
- Ganzach Y, Weber Y, Or PB. Message framing and buying behavior: on the difference between artificial and natural environment. Journal of Business Research 1997;40:91‐5. [Google Scholar]
Guyatt 2008
- Guyatt G, Oxman AD, Vist GE, Kunz R, Falck‐Ytter Y, Alonso‐Coello P, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ 2008;336:924‐6. [DOI] [PMC free article] [PubMed] [Google Scholar]
Higgins 2003
- Higgins JPT, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta‐analysis. BMJ 2003;327:557‐60. [DOI] [PMC free article] [PubMed] [Google Scholar]
Higgins 2011
- Higgins JPT, Green S (editors). Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0. The Cochrane Collaboration, March 2011. [Google Scholar]
Moxey 2003
- Moxey A, O'Connell D, McGettigan P, Henry D. Describing treatment effects to patients: how they are expressed makes a difference. Journal of General Internal Medicine 2003;18(11):948‐59. [DOI] [PMC free article] [PubMed] [Google Scholar]
O'Keefe 2007
- O'Keefe DJ, Jensen JD. The relative persuasiveness of gain‐framed loss‐framed messages for encouraging disease prevention behaviors: a meta‐analytic review. Journal of Health Communication 2007;12:623‐44. [DOI] [PubMed] [Google Scholar]
O'Keefe 2009
- O'Keefe DJ, Jensen JD. The relative persuasiveness of gain‐framed and loss‐framed messages for encouraging disease detection behaviors: a meta‐analytic review. Journal of Communication 2009;59:296‐316. [DOI] [PubMed] [Google Scholar]
Van Den Noortgate 2003
- Noortgate W, Onghena P. Estimating the mean effect size in meta‐analysis: bias, precision, and mean squared error of different weighting methods. Behavior Research Methods, Instruments, & Computers 2003;35(4):504‐11. [DOI] [PubMed] [Google Scholar]
Wilson 1988
- Wilson DK, Purdon SE, Wallston KA. Compliance to health recommendations: a theoretical overview of message framing. Health Education Research: Theory and Practice 1988;3:161‐71. [Google Scholar]