NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
McCormack L, Sheridan S, Lewis M, et al. Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence. Rockville (MD): Agency for Healthcare Research and Quality (US); 2013 Nov. (Evidence Reports/Technology Assessments, No. 213.)
This publication is provided for historical reference only and the information may be out of date.
Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence.
Show detailsIntroduction
This section presents the results for Key Question (KQ) 3: the effect of alternate ways of communicating uncertainty as it pertains evidence translation. The analytic framework for this question is presented as part of the Introduction and shows the effects of alternate ways of communicating uncertainty on both intermediate and distal outcomes. As we noted in our methods, the best studies to answer this question would be randomized controlled trials (RCTs). However, given the early state of research, we have also included studies with other experimental designs.
In this section, we present our results for four main types of uncertainty: directness of how evidence is presented; precision with which evidence is provided; various ways of depicting net benefit of the evidence, and overall strength of recommendations as reflected in the wording of various types of clinical recommendations. We found no eligible studies on overall strength of evidence, risk of bias, consistency, or applicability. Below, we subdivide our results by the specific alternate presentations of uncertainty presented and by outcomes studied to make strength of evidence (SOE) determinations. Tables below describe individual studies and their results and document our SOE grades. Detailed evidence tables for KQ 3 are in Appendix F.
Description of Included Studies
Figure 1 in the first results section (for KQ 1) depicts the flow of article exclusion and inclusion. After dual review at both the title/abstract and full-text article stage, we retained ten articles,123–132 meeting inclusion criteria and report on nine unique studies about alternative ways of communicating uncertainty. Of the nine included studies, we graded one as low risk of bias123 and eight as moderate risk of bias.124–132
Below we report on the nine studies with low or moderate risk of bias (i.e., good or fair quality). Of these studies, two were RCTs, four were factorial RCTs, one a non-controlled trial, and two quasi-experimental studies. Four studies reported on various presentations of precision;124,125,131 one tested alternative ways of communicating directness123; and four investigated different ways of communicating net benefit (with some studies making more than one comparison).123,127–130,132 One reported on the effects of alternate wordings of overall strength of recommendations.126 No studies reported on alternate presentations of overall strength of evidence, risk of bias, consistency, or applicability. Three studies reported the effects of alternate non-numeric presentations of uncertainty;123,126,128 three on alternate numeric presentations;124,125,131 one on numeric versus graphical presentations;124 one on alternate graphical presentations;124 and two on framing.129,132 Only one was directed to providers; all others were directed at patients.
Interventions were tested in study populations in the United States, Canada, and Switzerland. Sample sizes ranged from 120 participants to 2,944 participants. Outcomes studied included knowledge, perceived risk, accuracy of perceived risk, appropriate choices regarding care (e.g., selecting medications; obtaining screening), and decision satisfaction.
Key Question 3. Effects of Communicating Uncertain Health and Health Care Evidence to Patients and Clinicians
Key Points
- Communicating precision: Studies found mixed effects of presenting numeric risks as point estimates versus 95 percent confidence intervals (CIs), depending on the studied outcome, width of the confidence interval, and the presence or absence of comparative information about average population risk. Only a single small study examined the effects of changing the format of 95 percent CIs (numeric versus graphical) on perceived risk of colon cancer; this precludes definitive conclusions (one study, insufficient SOE). Further, only a single small study examined the effects of using clean versus blurry bar graphs to convey information about uncertainty (one study; insufficient SOE)
- Communicating directness: Choice of a cholesterol medication with direct evidence of benefit was better for patients receiving non-numeric advice or factual information encouraging consumers to choose the drug with direct evidence than for patients receiving usual care. However, medication choices did not differ by type of instruction (one study; low SOE).
- Communicating net benefit: Choice of a heartburn medication that was more likely to have net benefit was better for consumers receiving non-numeric advice or factual information encouraging consumers to choose the drug with greater net benefit than for patients receiving usual care, but medication choices did not differ by type of instruction (one study; low SOE). Receiving additional non-numeric information about benefits had little effect on refusals of cancer screening tests, but receiving more non-numeric information on harms significantly increased test refusals and significantly decreased decision satisfaction (one study; low SOE). Compared with usual care, giving men prostate cancer screening information alone or framed in the context of information about other, more beneficial screening services significantly increased prostate cancer knowledge (low SOE). However, giving prostate cancer screening information alone versus framed in the broader context of more beneficial services had differential effects on patient involvement and screening (two studies; insufficient SOE).
- Communicating strength of recommendations: Only a single small study examined the effects of different ways of wording recommendations to convey strong or weak recommendations for care; this precludes definitive conclusions (one study; insufficient SOE).
Tables 32–35 document findings and SOE grades.
Table 32
Strength of evidence: Communicating precision.
Table 33
Strength of evidence: Communicating directness.
Table 34
Strength of evidence: Communicating net benefit.
Table 35
Strength of evidence: Communicating strength of recommendation.
Detailed Synthesis
Communicating Precision
Three studies reported on the effect of various ways to communicate the preciseness with which evidence is presented (Table 36). Three reported on the effects of using alternate numeric presentations (95% CIs or point estimates),124,125,131 one on the effect of presenting 95 percent CIs in numeric versus in a visual format,124 and one on presenting 95 percent CIs in alternative visual formats.124
Table 36
Studies of communicating precision.
Intervention: Alternate Numeric Presentations
One factorial randomized trial124 and two quasi-experiment studies125,131 reported the effects of using alternate numeric presentations of precision (CIs versus point estimates) on various outcomes including perceived risk and accuracy of perceived risk. We graded the strength of evidence by study outcome.
Two studies assessed the effects of alternate presentations of precision on perceived risk. In the factorial randomized trial, investigators randomly assigned a convenience sample of 240 members of a Web survey panel to receive presentations of either point estimates or 95 percent CIs describing their hypothetical risk of developing colon cancer.124 The sample was subsequently randomized to either text or horizontal bar graph presentations. All presentations were done with and without comparative information on the risk of colon cancer in the general population. In analyses with no comparative risk information, the perceived risk of colon cancer did not differ significantly when risk was presented with 95 percent CI (5 to 13%), versus as a point estimate (9%) in either text (+0.3 on 0 to 5 scale, p>0.05) or graphical format (−0.5 on 0 to 5 scale, p>0.05). However, when risk was presented in the context of comparative risk information, 95 percent CIs had small, statistically significant effects on perceived risk. This approach reduced perceived risk compared with presentations of point estimates in text format (−0.58 on a 0–5 scale, p=0.03) and increased perceived risk compared with presentation of point estimates in the bar graph format (+0.2 on a 0–5 scale, p=0.03).
In the quasi-experimental study,131 investigators gave participants 3 randomly ordered presentations of their hypothetical risk of developing temporary skin discoloration from an acne medication: a point estimate (20 out of 100), a 95 percent CIs with a small range (16–24 out of 100), and a 95 percent CIs with a large range (8–32 out of 100). The source of this information (either doctor or pharmaceutical company) was randomly varied for each participant. In combined analysis (regardless of source), this study supported conclusions of no difference in the effect of a 95 percent CIs with a small range and a point estimate (+0.13 on a 0–5 scale; 95% CI −0.04 to 0.30). However, a 95 percent CIs with a large range resulted in significantly higher perceived risk then either 95 percent CIs with a small range (+0.23 on a 0–5 scale; 95% CI 0.06 to 0.4) or a point estimate (+0.36; 95% CI 0.19 to 0.53). Based on these two studies focusing on the outcome of perceived risk, we graded the strength of evidence as insufficient. This grade was based on the fact that studies appeared to reach different conclusions based on the range of the confidence interval and accompanying information about comparative risk.
Two quasi-experimental studies assessed the effects of alternate presentations of precision on accuracy of risk perception and came to different conclusions. The quasi-experimental study already mentioned above,131 showed that 95 percent confidence intervals with either small or large range produced significantly less accurate risk perception than point estimates (small range: −60%, p<0.001; large range: −59.3%, p<0.001) based on proportion of individuals able to correctly answer 3 questions about the risk. The quasi-experimental study by Brewer,125 on the other hand, showed that, among a convenience sample of 143 breast cancer patients, a graphical presentation of the point estimate of breast cancer recurrence accompanied by text about the 95 percent CI, did not affect the accuracy of risk perception compared with a text or graphical presentation of the point estimate alone. However, in this study, accuracy was based only on ability to state the point estimate.
Based on these two studies focusing on the accuracy of perceived risk, we graded the strength of evidence as insufficient. Studies drew different conclusions based on differences in both presentation format and their measure of the accuracy of risk perception.
Intervention: Numeric Versus Visual Presentations of Precision
The factorial randomized trial (N=240) mentioned above also examined the effect of numeric versus visual (graphical) presentations of 95 percent CIs.124 Perceived risk of colon cancer did not differ significantly when risk was presented in graphical format versus text format (−0.4 on a 0–5 scale, p not significant). Based on this single small study, we graded the overall strength of evidence insufficient.
Intervention: Alternate Visual Presentations of Precision
One small randomized trial (N=135) examined the effect of alternate horizontal bar graph presentations of 95 percent CIs: a solid bar graph versus a blurred bar graph that was intended give a better indication of the uncertainty.124 In this trial, perceived risk of colon cancer did not differ by format (effect size not reported). Based on this single small study, we graded the overall strength of evidence insufficient.
Communicating Directness
One large (N=2,934) high-quality (low risk of bias) factorial randomized trial conducted in a nationally representative sample examined the effect of giving non-numeric advice versus factual statements to consumers to encourage use of cholesterol-lowering drugs that have proven efficacy in lowering heart attack (the distal and direct outcome) rather than just cholesterol levels (the proximate and indirect outcome) (Table 37).123 Those in the advice group were told that “Surrogates do not always translate into patient outcomes. Ask for a drug to reduce heart attacks”; those in the factual statement group were told only that “Surrogates do not always translate to patient outcomes.” Compared with usual care, both advice and factual statements improved appropriate choice of medications (directive: +12 percentage points, 95% CI, 7 to 18; nondirective: +12 percentage points, 95 percent CI, 7 to 18), but medication choices did not differ by type of instruction. Based on this single study with low risk of bias, direct evidence and precise findings, we graded the overall strength of evidence low.
Table 37
Study on directness.
Communicating Net Benefit
Four studies reported on the effects of alternative ways of communicating net benefit (i.e., taking both benefits and harms into account).123,127–130,132 One focused on alternate non-numeric ways to promote health care with maximum net benefit, one on what happens to choices and decision satisfaction by varying the amount of non-numeric information about benefits and harms, and two on the effect of presenting net benefit information framed in the context of other, more beneficial health services (Table 38).
Table 38
Studies about providing information about net benefit (balance of benefits and harms).
Intervention: Alternate Non-Numeric Presentations of Net Benefit
One large, good-quality trial mentioned above also examined the effect of giving advice or factual statements to consumers to encourage their choice of a heartburn drug more likely to have a net benefit because of its longer evidence of safety.123 Those in the advice group were told that “It takes time to establish the safety of drugs. Ask for a drug with a longer track record,” whereas those in the factual statement group were told only that “It takes time to establish the safety of drugs.”
Compared with usual care, both the advice and factual statements improved consumers’ appropriate choice of medications (directive: +19 percentage points, 95% CI, 13 to 24; nondirective: +19 percentage points, 95% CI, 13 to 24). The choice of medications did not differ, however, by type of instruction. Based on this single large study with low risk of bias, we graded the overall strength of evidence low.
Intervention: Varying the Amounts of Non-Numeric Benefit and Harm Information
A large (N=2,333) factorial randomized trial conducted on the general population in one Swiss canton examined the effects of providing varying amounts of non-numeric benefit and harm information on test refusals and decision satisfaction.127,128 Investigators first randomized participants to receive one of three levels of benefit information about a screening test for an unnamed cancer: no information, information on survival benefit, or information on survival benefit plus information on the relief individuals experience from screening. Participants were subsequently randomized to receive one of three levels of harms information about the same test: no information, information about false-positive results, and information about false-positive and false-negative results.
Although additional benefit information had little effect on test refusals (odds ratio [OR] for full information versus no information: 1.0; 95% CI, 0.7 to 1.3), increasing information on harms significantly increased test refusals (OR for partial vs. no harm information: 2.5; 95% CI, 1.8 to 3.4; OR for full vs. no harm information: 3.0, 95% CI, 2.2 to 4.2). Furthermore, harms (although not benefits) significantly decreased decision satisfaction (−5.1, 95% CI, −6.6 to −3.6 on a scale of 0 to 100). Based on this single large study with moderate risk of bias, we graded the overall strength of evidence low.
Intervention: Presenting Net Benefit Framed in the Context of Services With Different Net Benefit
Two studies examined the effect of presenting net benefit in the context of services with different net benefit. One nonrandomized trial in three communities in North Carolina examined the effect of giving male participants decision aids with information only about prostate cancer screening or about prostate screening framed in the context of other, more beneficial screening tests for men. Outcomes included knowledge, involvement in decisionmaking, and PSA screening at 12 months.129,130 Compared with usual care, both prostate cancer screening information alone and framed in the context of other more beneficial screening services increased prostate cancer knowledge (prostate alone vs. usual care: +0.9 on a 0–10 scale, p<0.05; prostate information in context versus usual care: +1.5 on a 0–10 scale, p<0.001). Both decision aids also slightly increased the proportion of individuals reporting active involvement in decisionmaking (prostate information alone versus usual care: +4%, p=0.064; prostate information in context versus usual care: +3%, p=0.045). However, prostate cancer screening information alone increased screening (+7%, p NR), whereas prostate cancer screening information framed in the context of the other screening services decreased screening (−3%, p NR).
The second study was also focused on the effects of presenting prostate cancer information either alone or in the context of other more beneficial services.132 This study included two similarly conducted practice based RCTs that compared the effects of a highway safety video with the effects of a prostate cancer screening decision aid and coaching tool (framed, in one of the trials, in the context of other more beneficial services). Finding no difference in trial outcomes by prostate information frame, this study combined trial results using a random effects model and showed that its prostate cancer screening decision aid increased knowledge (+34%, 95% CI, 19% to 50%) and reduced 9-month screening rates (−22%, 95% CI, −38 to −7%), but had no effect on patient involvement in decisionmaking.
Based on these two studies of framing net benefit information, we graded the overall strength of evidence as low for knowledge given that both the intervention in both trials increased knowledge regardless of frame. We graded strength of evidence for other outcomes as insufficient. Effects on screening and involvement varied by trial, perhaps due to setting (community versus practice).
Communicating Overall Strength of Recommendation
One factorial randomized trial examined the effect of alternative non-numeric ways of communicating strength of recommendations on guideline-concordant choices for congestive heart failure or inflammatory bowel disease care (Table 39).126 Strength of recommendations was reflected in the ways that various groups word their recommendations about health care. These mimicked current wording used by the American College of Chest Physicians (e.g., “we recommend,” “we suggest”), the National Institute for Health and Clinical Excellence (e.g., “clinicians should,” “clinicians might”), or the Grading of Recommendations Assessment, Development, and Evaluation group (e.g., “we recommend,” “we conditionally recommend”). In this trial, investigators randomly assigned 341 medical residents to one of three groups with different wording for health care recommendations and subsequently randomized them to receive one strong (i.e., “we recommend”) and one weak (i.e., “we suggest”) recommendation either “for” or “against” some specific health care option.
Table 39
Studies of communicating strength of recommendation.
The effect of various strong recommendations (e.g., “we recommend” vs. “clinicians should” vs. “we recommend”) either for care or against care did not differ much. Relatively few residents who received the strong recommendation either for or against care indicated that they would prescribe care appropriate with the intended recommendation of the guideline (i.e., guideline-concordant care; “for,” 8 percent on average across three groups; “against,” 47 percent on average across three groups).
- Communication and Dissemination Strategies to Facilitate the Use of Health-Relat...Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence
- Results—Key Question 3: Communicating Uncertainty - Communication and Disseminat...Results—Key Question 3: Communicating Uncertainty - Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence
- Results—Key Question 2: Dissemination Strategies to Clinicians and Patients - Co...Results—Key Question 2: Dissemination Strategies to Clinicians and Patients - Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence
- Results—Key Question 1: Communication Strategies - Communication and Disseminati...Results—Key Question 1: Communication Strategies - Communication and Dissemination Strategies to Facilitate the Use of Health-Related Evidence
- A Framework To Facilitate the Use of Systematic Reviews and Meta-Analyses in the...A Framework To Facilitate the Use of Systematic Reviews and Meta-Analyses in the Design of Primary Research Studies
Your browsing activity is empty.
Activity recording is turned off.
See more...