key: cord-1003361-vv1j93bh
authors: Winograd, Darren M.; Fresquez, Cara L.; Egli, Madison; Peterson, Emily K.; Lombardi, Alyssa; Megale, Allison; Tineo, Yajaira A. Cabrera; Verile, Michael G.; Phillips, Alison; Breland, Jessica Y.; Santos, Susan; McAndrew, Lisa M.
title: Rapid Review of Virus Risk Communication Interventions: Directions for COVID-19
date: 2021-01-20
journal: Patient Educ Couns
DOI: 10.1016/j.pec.2021.01.024
sha: 7758e3d408096319d63a04ef728f4966816d3aa7
doc_id: 1003361
cord_uid: vv1j93bh

Objective In response to COVID-19, we conducted a rapid review of risk communication interventions to mitigate risk from viruses to determine if such interventions are efficacious. Methods We searched for risk communication interventions in four databases: Medline, PsycInfo, the ProQuest Coronavirus Research Database, and CENTRAL. The search produced 1572 articles. Thirty-one articles were included in the final review. Results Results showed risk communication interventions can produce cognitive and behavior changes around viruses. Results were more consistently positive for interventions focused on HIV/AIDS as compared to influenza. There was no consistent best intervention approach when comparing peer health, audio/visual, and intensive multi-media interventions. Tailoring risk communication toward a target population, in comparison to not tailoring, was related to better outcomes. Conclusion The results suggest that risk communication interventions can be efficacious at reducing risk from viruses. They also highlight the complexity of risk communication interventions. Additional research is needed to understand the mechanisms that lead risk communication to reduce risk from viruses. Practical value Results support risk communication interventions to reduce risk from viruses.

To meet the inclusion criteria, articles had to: (1) be empirical, (2) be about a viral illness that can be transmitted to humans, (3) focus on transmission among humans, (4) employ some form of risk/health communication as a predictor, (5) and have a cognitive or behavioral outcome. Articles were excluded from the current review if they: (1) were not in English, (2) not peer-reviewed (e.g., editorial, comment, letter, or newspaper article),

(3) used a sample with individuals younger than 18, (4) did not evaluate a clear risk communication/dissemination intervention, (5) conducted an intervention not designed to assist individuals reduce their risk of infection (e.g., designed exclusively for those living with a virus), or (6) did not include quantitative data.

A total of 1572 articles were initially gathered for review (865 from PsycInfo; 684 from Medline; 23 from the ProQuest Coronavirus Research Database). Thirty-five were duplicates and removed. Eight authors each evaluated 195 article titles/abstracts for inclusion, with decisions double-checked by a second author. Authors reached initial agreement on inclusion decisions in 95% of cases, and 100% after discussion of J o u r n a l P r e -p r o o f discrepancies. During the title and abstract check, a total of 74 articles were excluded based on the length of intervention and use of qualitative data. In addition to the full text examination on the articles that met the initial pass, these 74 articles were also re-reviewed, with double checks again conducted to ensure accuracy in decision making. The re-review of these 74 initially excluded articles was done to determine the appropriateness of excluding interventions that were long (i.e., 5 hours at a time or over 2 weeks in length) and qualitative. Ultimately, lengthy interventions were included and wholly qualitative studies were excluded. Additionally, after further reviewing inclusion criteria, we decided to remove interventions designed exclusively for those diagnosed with a virus. This resulted in the removal of two originally included articles. Initial agreement on inclusion decisions was again high, achieved in 84% of cases and in 100% of cases post discussion (see Figure 1 for decision flow chart). Thirty-one articles were included in the review.

Extracted data included authors, year of publication, study design, virus of focus, number of participants, and sample demographics (see Figure 1 ). Extracted data on interventions included communication strategy, messaging content, length of intervention, setting, provider, and mode of transmission. Extracted data on analyses included outcome variables, measures, timeframe between comparisons, and achieved results (e.g., primary results, effect sizes). Outcomes were coded as: (1) cognitive risk perception, or how one views the risk, severity, or certainty of infection, (2) The heterogeneity of interventions and outcomes precluded meta-analysis. When meta-analytic procedures are not possible, synthesis without meta-analysis (SWiM) guidelines for the synthesis of quantitative data, suggest tallying the number of studies with positive, negative, and no effect [28] . In this review, all articles that included cognitive risk perceptions as an outcome sought to increase participant's perceptions of risk, therefore increasing cognitions of risk was coded as positive. Similarly, all articles that included cognitions about behaviors, behavioral intentions, or behaviors as outcomes sought to protect individuals from viral infection by reducing risky behaviors and/or increasing protective behavior, therefore changing cognitions about behaviors, behavioral intentions or behaviors to protect from viral infection (reduce risky behavior and/or increase protective behavior) was coded as positive.

Studies were tallied once for each type of outcome. Studies which reported multiple measures for one type of outcome (e.g., multiple measures of cognitive risk perception) could be tallied as having mixed results (e.g., mixture of positive and no effect). Therefore, studies were coded as: (1) positive (pos), (2) negative (neg), (3) no effect (NE), (4) mixture of both positive and no effect results (mixed pos&NE), and (5) mixture of both negative and no effect results (mixed neg&NE). No studies had a mixture of positive and negative outcomes. For ease of interpretation, result tallies for total positive (i.e., combination of pos and mixed pos&NE) and total negative (i.e., combination of neg and mixed neg&NE) are reported.

Initial data synthesis examined the data across all studies and then around virus type and intervention approach, with data arranged to examine patterns in results. Patterns were interpreted if there were at least three studies for a given outcome within the group (e.g., a specific virus or intervention approach). After examining the data, a secondary post-hoc synthesis was conducted to examine the efficacy of intervention message tailoring in producing cognitive or behavior changes for a virus. Tailored messaging was defined in accordance with the WHO definition as any effort to customize risk messaging for a specific target audience to make the messaging more accessible to said audience [29] . An evaluation of target audience tailoring was done because the initial data syntheses showed there was significant heterogeneity and tailoring messaging to a target audience may be efficacious for virus risk communication interventions [29] .

[INSERT FIGURE 1]

The Grading of Recommendations Assessment, Development and Evaluation approach (GRADE), a methodology to examine the scientific rigor of articles, was employed to rate the quality of included articles [30] . The GRADE methodology is a widely adopted and supported method for evaluating studies included in reviews. Per the GRADE methodology, articles were provided an apriori score of high for randomized control trials and low for non-randomized control trials. Scores were then downgraded in cases of within-study or publication biases, poor directedness (i.e., examination studied intended population, treatment, or phenomenon), poor precision (e.g., wide confidence interval margins, poor scale reliability), and result inconsistencies amongst studies/trials. Scores were also upgraded if the effects achieved in an article were likely smaller than the J o u r n a l P r e -p r o o f true effect, if effects were so large confounds likely did little to obscure true findings, or if the effect appeared proportional to intervention exposure. Scores achieved by included articles ranged from very low (n = 23) to low (n = 8).

Among the 31 studies included in the analysis, there was significant variability in design, demographic characteristics, and sample size (see Table 1 ). Most studies (n = 17) utilized a between-group design, comparing an intervention to some form of control group(s). The remaining 14 studies relied on within-group designs via pre-test to post-test comparisons. Results were relatively similar across within and between-group designs (see Table 1 ).

[INSERT TABLE 1] There was evidence that risk communication interventions for viruses can improve cognitive and behavior outcomes. Across studies, risk communication was shown to positively impact cognitive risk perceptions (e.g., greater perception of viral risk), cognitions about behaviors (e.g., greater efficacy beliefs in protective behaviors), behavioral intentions (e.g., greater intention to engage in protective behaviors) and behaviors to reduce risk (see Table 2 ).

[INSERT TABLE 2] Risk communication interventions focused on HIV/AIDS (n=19) showed consistent positive findings for changing all outcomes. In contrast, risk communication interventions focused on influenza (n=7) showed consistent positive findings for improving cognitions about behaviors, but little evidence for improving the other three outcome categories. Moreover, the single negative outcome within this review was for an communication interventions, as compared to non-tailored interventions, were consistently related to positive changes in cognitive risk perceptions and behavioral intentions (see Table 4 ).

[INSERT Risk communication interventions to reduce risk from HIV/AIDS were most consistently related to improved cognitions and behaviors. In comparison, there was less evidence that interventions focused on reducing risk from influenza were efficacious. A possible explanation for this discrepancy is that influenza is acute and less severe while HIV/AIDS is chronic and more severe. It may be more difficult to change cognitions and behaviors for less severe conditions. The one negative finding, where the intervention resulted in lower perceived risk and fewer protective behaviors, were produced by a study on influenza. Another potential explanation for the distinct results around HIV/AIDS and influenza interventions may be how common influenza is. There was not strong evidence that one type of risk communication approach is more consistently effective at improving virus outcomes, in part because there were too few studies for most approaches to interpret outcomes. Audio/visual media, which was the least intensive and often included things like posters, was shown to change cognitive risk perception and cognitions about behaviors, but had mixed findings for other outcomes. It is promising that low resource interventions, like posters, can sometimes improve outcomes. However, this review suggests that they are not sufficient on their own. Intensive multi-media interventions, which included online risk communication interventions, were most likely to change behaviors, but there was not enough data to interpret other outcomes.

As behaviors are considered the most difficult to change, these interventions may be a promising avenue to change behaviors and should be studied more in the future.

Peer health communication also demonstrated initial positive findings for producing change in cognitive risk perceptions, cognitions about behaviors, and behaviors. This is consistent with previous research which has shown efficacy for peer communication interventions for other health threats [36] [37] [38] . Conceptually, peer health communication may provide social support for, or normalization of, protective health behaviors which in turn promotes their employment [39] . Peer health communication is also inherently tailored to the culture of the audience receiving the education.

The positive outcomes from peer health communication led us to hypothesize that tailoring of messaging to specific populations may be particularly beneficial. This is also consistent with the Common-Sense Model (CSM) which proposes that the public develops lay understanding of health threats and behaviors to reduce risk, and as a result, risk communication needs to be tailored to a target audience [10] . In response, we conducted a post-hoc analysis comparing tailored as compared to not tailored interventions. We found that interventions were tailored toward a target audience in multiple different ways including using focus groups, knowledge of the target population, providing narrative messages and using peer educators. Tailored interventions were consistently related to improvements in cognitive risk perception and behavioral intentions. This support for target audience tailoring is in line with the extant evidence of tailoring as particularly efficacious for producing cognitive and behavior changes for the mitigation of health threats [14] . However, less consistent results for the positive impact of tailoring on cognition about behaviors and behavior outcomes supports the need for further study on J o u r n a l P r e -p r o o f target audience tailoring on outcomes. Additional research is also needed on the best approaches to deliver tailored interventions, what aspects of communication need to be tailored, and if there is additional benefit to tailoring for the individual as compared to the target group. There were not enough studies to examine these questions in this review.

The results of this review were limited by the heterogeneity of the interventions which precluded meta-analytic procedures. There were also relatively few studies, on only a small number of viruses, which makes it difficult to conclude that any one type of approach is more efficacious than another. Moreover, the viruses included in this review are not perfect analogues for COVID-19. HIV/AIDS, which was the focus of the majority of interventions included in this review, is unique given its methods of transmission and the stigmatization of HIV/AIDS infection. Further, the politicalizing of COVID-19 is unique in comparison to other viruses and pandemic events. As such, the degree to which results apply to COVID-19 risk communication is unclear and in need of further study.

Overall, the quality of included studies was generally low. Further, there may be publication bias towards efficacious interventions, limiting the review. However, while conclusions must be viewed as tentative due to these limitations, the results may help guide COVID-19 risk communication.

The results suggest risk communication interventions may be an effective method for improving cognitions and behaviors to mitigate the risk of COVID-19 infection. We 

The results highlight the complexities inherent to risk communication about viral transmission. This review largely supports risk communication as efficacious in producing positive changes in individuals for the mitigation of viral risk. Results were more consistently positive for interventions focused on HIV/AIDS as compared to influenza.

There was no consistent best intervention approach when comparing peer health, audio/visual, and intensive multi-media interventions, with results suggesting that a variety of modalities can be efficacious. There was evidence that interventions tailored to a population can be efficacious when compared to non-tailored interventions. Article Screening and Selection Summary. Demographics of included articles (n = 31) Table 2 Primary results of included articles (n = 31) Table 3 J o u r n a l P r e -p r o o f Summarized results of included articles (n = 31) Table 4 Summarized results of included articles around tailoring of messaging (n = 31) Notes: HBV = Hepatitis B; IPC = interpersonal counseling; aOR= adjusted odds ratio; OR = odds ratio; SC = score change; PR = prevalence rate; RR = risk ratio; aPR = adjusted prevalence rate; aRR = adjusted risk ratio B = beta coefficient; N.S. = not significant; S. = significant J o u r n a l P r e -p r o o f Table 3 Summarized results of included articles (n = 31) *Pos=positive effect of risk communication intervention changing cognition/behavior in the intended way (i.e., increasing perceptions of risk, increasing risk mitigation behaviors, reducing risk behaviors), neg=negative effect of risk communication intervention changing cognition/behavior in unplanned direction (i.e., reducing perceptions of risk, decreasing risk mitigation behaviors, increasing risk behaviors, no effect/NE=no effect of risk perception intervention; mixed (pos&NE)=mixture of both positive and no effect results; mixed (neg&NE)=mixture of both negative and no effect results. Active control group Agreement with the following statements: (1a) "Common side effects of the flu shot are a sore arm where the shot is given and body aches" (1b) "You can give the flu to others even before you have symptoms" (1c) "The only way to catch the flu is to come in contact with someone who has the flu" (1d) "You get the flu from others who cough and sneeze while they are ill, or by touching something that has the flu virus on it" (1e) "I think the flu shot causes the flu" (1f) "The flu shot is not a cure for the flu and will not help you if you are already ick with the flu"

Intervention exposure related to significantly higher agreement with (1c) "you can give the flu to others even before you have symptoms," and decreased agreement with statement (1f) "I think the flu shot causes the flu" *Pos=positive effect of risk communication intervention changing cognition/behavior in the intended way (i.e., increasing perceptions of risk, increasing risk mitigation behaviors, reducing risk behaviors), neg=negative effect of risk communication intervention changing cognition/behavior in unplanned direction (i.e., reducing perceptions of risk, decreasing risk mitigation behaviors, increasing risk behaviors, no effect/NE=no effect of risk perception intervention; mixed (pos&NE)=mixture of both positive and no effect results; mixed (neg&NE)=mixture of both negative and no effect results.

The differential psychological distress of populations affected by the COVID-19 pandemic

A nationwide survey of psychological distress among Chinese people in the COVID-19 epidemic: implications and policy recommendations

Nationwide survey of psychological distress among Italian people during the COVID-19 pandemic: immediate psychological responses and associated factors

Protests in Hong Kong during the Covid-19 pandemic

The socio-economic implications of the coronavirus pandemic (COVID-19): a review

General information on risk communication

The effectiveness of mass media in changing HIV/AIDSrelated behaviour among young people in developing countries

The impact of communicating genetic risks of disease on risk-reducing health behavior: systematic review with meta-analysis

Message framing in vaccine communication: a systematic review of published literature

Risk perception and communication

Combining risk communication strategies to simultaneously convey the risks of four diseases associated with physical inactivity to socio-demographically diverse populations

Flood risk assessment, future trend modeling, and risk communication: a review of ongoing research

Perception and communication of flood risks: a systematic review of empirical research

Health-related disaster communication and social media: mixed-method systematic review

HIV risk behavior among college students in the United States

Effects of empathic and positive communication in healthcare consultations: a systematic review and meta-analysis

Strategies for addressing vaccine hesitancy-a systematic review

A scoping review of rapid review methods

Synthesis without meta-analysis (SWiM) in systematic reviews: reporting guideline

Note: One article (de witt et al, 2008) compared types of tailored messaging (narrative vs statistical messaging); within the article this comparison accounted for a mixed (pos&NE) tally in cognitive risk perception and a no effect tally in behavioral intentions

increasing perceptions of risk, increasing risk mitigation behaviors, reducing risk behaviors), neg=negative effect of risk communication intervention changing cognition/behavior in unplanned direction (i.e., reducing perceptions of risk, decreasing risk mitigation behaviors, increasing risk behaviors, no effect/NE=no effect of risk perception intervention; mixed (pos&NE)=mixture of both positive and no effect results

We acknowledge the experts and colleagues at the University at Albany, SUNY who were consulted and assisted with the development of the search syntax, most notably Justin Kimber and Deborah LaFond.

tiab((eHealth OR mHealth OR communicat* OR messag*))) AND (mesh((Health Behavior)) OR tiab((adopt OR behavior* OR behavior* OR change OR cognition* OR belief* OR perception OR "risk assessment"))) NOT (pt((editorial OR comment OR letter OR newspaper article))) NOT (mesh((animals)))) AND la.exact("English") AND PEER(yes) Appendix B