The use of visual risk communication and its significance for risk understanding and health literacy in out-clinic settings – a literature review

Patients frequently experience difficulties understanding communicated risks. The aim of this study was through a literature review to analyze if the use of visual risk communication tools improve risk understanding among patients in outpatient settings or general practice, and if one tool appears more useful than others. The electronic databases PubMed and PsycINFO were systematically searched. Relevant references were used for chain search to make sure all relevant literature was included. The main search revealed 1,157 titles. There were 13 eligible studies concerning visual risk communication in outpatient clinical settings. The design, quality and main findings of the studies were heterogeneous. However, most of the analysed studies found a significant positive effect of graphical, interactive and dynamic visual aids on risk communication.

There can be many factors influencing how well people assess and understand risk, including age, level of education, culture and gender (1). Research has shown that limited understanding of risk, particularly numerical risk information, can jeopardise preventive efforts as well as the understanding of diagnosis and treatment options (2). This is supported by findings where higher levels of verbatim knowledge (the ability to correctly read numbers from graphs) and gist knowledge (the ability to identify the essential points of the information presented) are significantly associated with making medically superior treatment choices (3).
The understanding of risk factors and the awareness of how to change them are important for the patient's motivation and compliance (4). The level of understanding may have an impact on whether the patients wish to change behaviour or participate in treatments, as the intention to change is related to the perception of risks (4). Low health literacy (the ability to understand, obtain and apply healthcare information in order to make appropriate health decisions and follow instructions for treatment) is generally associated with poor health (5). If the patient fails to initiate treatment as agreed with the general practitioner (GP) or does not comply with the treatment plan, it may cause health deterioration or reduced quality of life for the individual and increased healthcare expenses for the society (6).
During recent years, the focus on personalised risk communication and shared decision-making has increased (7, 8). The development of online devices has enabled patients to independently access risk information and reflect on questions before consulting a doctor (9, 10). However, the quality and coherence of the webbased information is often variable (11). Combined with a lexile level above most recommended guidelines, this may cause a reduction in the level of understanding for many patients (12). The introduction of computers in the clinical work has made it more accessible to communicate personalised risk information in a graphical format based on the risk factors and para-clinical tests of the patient (13). This innovative trend has inspired researchers to study if visual graphs impact the patients understanding of health-related issues, risk perception and health literacy in order to respond to information in a health-promoting manner (14). Method Systematic reviews within the area of visual risk communication are characterised by heterogeneity (7, 15) and metanalysis could not be done (7). This might be due to diverse methodological quality, a broad definition of visual communication (e.g. pictures, videos, interactive graphs and 3D phantoms) or inclusion of both primary and secondary settings (15). For the present study we excluded hospital settings, and focused on outpatient settings and general practice, where more than 80% of all prescriptions of medication are initiated in the Danish setting. The outpatient setting and general practice are comparable in relation to time frame, contact and an established longer-term relationship between doctor and patient. Outpatients can be more self-reliant compared to hospital settings, and possibly more confident rejecting a suggested treatment plan, if they do not understand the risks presented (16).

Search strategy
The electronic databases PubMed and PsycINFO were searched August 21 st 2018.
The search matrix was designed according to the PICO (population, interventions, comparison, outcomes) approach (17). In this study, the research question was transformed into four blocks covering: Population and setting (General practice, outpatient clinic and synonyms),  Studies were eligible for inclusion if they: 1) were published in peer-reviewed journals written in English, 2) involved adult populations (over 18 years old) and 3) were conducted in conditions similar to the primary care sector, outpatient clinics or diseases managed in general practice.
The visual intervention (VI) had to be actively or intentionally chosen by the participants themselves e.g. by accessing a website or participate in a trial. The VI could also be chosen by the doctor or used in the interaction between a patient and healthcare professionals. Interventions with passive exposure, like commercials or videos in a public area were excluded. To reflect the varying patient groups in general practice, the selection criteria did not include variables like gender, disease, or social demography.

Selection and appraisal
Prior to the selection of the articles, the research group defined inclusion and exclusion criteria. The first author made the preliminary selection based on the agreed inclusion and exclusion criteria. If there were any concerns whether an article should be included or not, it was discussed at a meeting between all authors before the final decision to include or exclude.
The database search is shown in Figure 1. The database search provided 1258 hits in total, nine duplicates were removed, and 1249 articles were screened by title.
135 abstracts which matched the inclusion and exclusion criteria were reviewed. Of these, 45 papers were assessed in full-text for eligibility. The final analyses included 13 studies. The reason for excluding 90 out of 135 records after reading the abstract was that they focused on e.g. tele healthcare, text reminders on mobile phones, interventions in paediatric care, electronic health records, or were protocols or pilot studies, evaluated imaginary software for doctors, had less than ten participants, or did not have a full-text written in English. Backward citation search by the reference lists of the studies initially included and forward citation search by Web of Science revealed 31 other relevant studies.
After full text retrieval, 61 records out of the 76 were excluded. Among reasons to exclude the papers were that the main outcome was measured according to shared decision-making, patient satisfaction or patient anxiety. Other reasons were that the studies explored patient education without explicit evaluating understanding of risk message, risk perception or patient compliance, or that the intervention was transmitted on a screen in a waiting room without active recipients.

Data analysis
The remaining 15 studies were systematically reviewed focusing on study design, sample size, applicability, category of VI, risk of bias, confounders, statistical analysis, and significant results (23). The aim was to reveal strengths and limitations of the papers to determine the importance of the papers and weight the evidence of their findings. We used the hierarchy of evidence pyramid where highquality studies were preferred and weighted higher than those lower in the evidence hierarchy. The number of participants, the transparency of the statistical methods and the study types were important elements. If there was doubt about the statistical methods in the studies, a statistician was consulted. Two studies were excluded, the first one since only 27% of the participants were diagnosed with the disease that the intervention was targeting. The second study was excluded due to a discrepancy between the included population and the outcome measured. Finally, 13 studies met the criteria for the final analysis (Table 1).

Results
The 13 included studies were reviewed as described in the method section. Design, quality and main findings of the studies are generally heterogeneous as summarised in Table 1, although some parallel lines can be drawn.

Study characteristics
Four of the 13 included studies were randomised controlled trials (RCT), seven were randomised trials without controls, one was controlled but not randomized and one  Table 1). The VI targeting the GP was associated with a higher rate of patients receiving an appropriate screening for CVD by measurement of their risk factors (p = 0.02) (26). There was a significantly higher number of new prescriptions or increase in numbers of medicines for the high risk cohort; antiplatelet (p <0.001), lipid-lowering (p < 0.001), and antihypertensive therapy (p = 0.02) (26). In the intervention group there was a higher proportion reaching the target set by the national BP guideline compared to the control group (p = 0.05) (26). Ruiz et al. (29) also evaluated a computer-based tutorial 'Your Cardiovascular Risk Score' in a small-scale RCT with 120 patients. The design had several limitations as it was performed in a laboratory setting, and risk perception was measured by self-report without measuring behavioural changes (Table 1). Differing from the study by Peiris et al., Ruiz et al. found that icon arrays may impair shortterm recall of CVD risk and therefore not necessarily results in a better recall of medical risk in all patients (29). Risk presented in icon arrays and frequencies resulted in an inferior accuracy of risk perception 20 minutes after the presentation compared to percentages or frequencies only (p = 0.001) (29). There were no differences when evaluating immediate risk understanding (p = 0.31) or recall at two weeks (p = 0.10), and participants with high graphical literacy performed significantly better than those with low graphical literacy at all times (p < 0.02) (29).
A study comparing diagnostic inferences (28) ( Table 1) between doctors and patients, found that additional presentation of a visual display of the numeric information in shaded blocs improved the diagnostic understanding, measured in accuracy of percentage and natural frequencies, for both doctors and patients (28).
The patients estimated the information as less useful when it was provided only numerically, as compared with the same information provided both numerically and visually (p = 0.023) (28). In contrast, the doctors found the information highly useful, with no statistical difference between the numerical or visual display (p = 0.322) (28). Overall, doctors had higher numerical skills than their patients (p = 0.001) (28). This correlates with results by Goodyear-Smith et al. who examined which presentation of hypothetical risks and benefits that would encourage statin users with a pre-existing heart disease, to take daily medicine and which one they preferred (33). They found that 57 % of the patients preferred information presented graphically (p < 0.001) (33).
The VI targeting both the doctor and the patient resulted in a significantly higher proportion of risk factors being measured in the patients, along with a significant increase in prescriptions of medicine for patients in high risk of disease (26). In summary, visual risk communication improved understanding of risks among patients as well as doctors, and the understanding and usefulness was affected by the patients' numeracy and graph literacy skills (3, 28), giving the patients with high level literacy the greatest benefits.
The best graphical format to increase risk understanding Numerical information can be presented by many different graphic designs (e.g. pie  Table 1). None of the animations improved any outcome, and most showed significant performance degradations (e.g. scatter with auto shuffle (p < 0.02)).
Static pictographs that grouped icons at the bottom of the array resulted consistently in better treatment choices and a higher gist knowledge of side effects than the animated icons (32).
McCaffery et al. (31) examined if the size of the numerator had an influence on graphical risk interpretation. The numerator size was categorized into small (<100), medium (100-499) and large (500-999), with the denominator fixed at 1000. The findings suggested that the optimal graph type for communicating risk information depended on the numerator size displayed. For adults with low education and literacy, pictographs were likely to be the best format to use when displaying small numerators as <100/1000 and bar charts for larger numerators as >100/1000 (Table   1) (31).
The optimal graphical format for increasing understanding of risk depends on the message to be conveyed. Static pictographs were highly useful for patients with lower numeracy and overall the best format for enhancing risk understanding across patient categories and educational levels.  Table 1). The BP goal (< 140/90 mmHg) was achieved more often in the intervention group with the colour-coded booklet (p = 0.037) (27). BP measured by the GP showed a significant decrease after six months in both groups compared with baseline measurements, with no difference between the groups. The antihypertensive therapy was changed overall in 63 % of the patients with no difference between the two groups (p = 0.367) (27).
Fraccaro et al. tested 20 patients' ability to determine the need for medical attention, after viewing a graphical presentation of hypothetical laboratory tests (24) ( Table 1). The results demonstrated the patients' difficulties in interpreting laboratory test results with 65 % of the participants underestimating the need for action across all presentations at least once, and with 70 % of the participants overestimating the need for action at least once even when abnormal values were highlighted using colours and graphical cues (24). The results indicate that care must be taken, when communicating visual health related risks through patient portals, without consulting a doctor.
To summarise the above, visual tools for use at home, must be ensured with an action plan that is easy to understand for the patient, to promote health literacy and support the patient in responding appropriately to the information given.

Video as a visual tool
The database search found two randomised studies with a video intervention (34, 35). The studies had several limitations (Table 1) though abnormal values were highlighted using colours and graphical cues. These findings emphasize the importance of the graphical design regarding type, colour, cues, complexity, scale and animations in order to inform rather than confuse the patients. Tailoring the graph format to the type of information needed for a particular medical decision would likely produce the most informed patient (3) and thereby hopefully the best decisions. Clear evidence for an association between level of understanding and level of decision-making is still to be investigated.
It is essential that the effects of an intervention can be measured by an outcome.
The examination of a colour-coded BP diary showed no significant difference in values of BP, change in antihypertensive treatment or adherence to the diary between the groups after six months (27). Many of the patients (66 %) had already used home BP measurement before the study, which may have reduced the effect that could be measured by introducing the book. The authors highlight that BP control (< 140/90 mmHg) was achieved more often in the intervention group (p = 0.044). These results must be interpreted with caution since the study design did not include any guideline or action plan according to the BP values measured at home and the proportion of patients with BP control at baseline had not been measured.
These findings support the importance of the study design, in order to develop a VI that can facilitate a more informed discussion, contribute to shared decision making, and increase health-efficacy. Thus, assisting the patients in the lower sociodemographic groups in making the most beneficial health choices.
Communicating risk through visual tools appears beneficial for the patients' understanding. The optimal type of visual tool for communicating risk depends on the message to be conveyed (gist or verbatim knowledge, the size of the risk etc.), health literacy level, and socioeconomic status of the patient. The results support introducing a personalised approach to risk communication based on graphical/visual risk presentation together with numerical information, like tables, in order to enhance risk understanding.

Other aspects of visual communication
There are many aspects that need to be considered when evaluating the usefulness and benefits of a visual communication tool. It is difficult to measure benefits versus costs such as resources needed for implementation or education, licences etc. vs. benefits such as a more informed patient together with lower health costs if co-morbidity and the need for hospitalisation or other healthcare services is reduced. Peiris et al. (26) found that a screen pop-up for the GP improved the frequency in which the patients' risk factors were screened (p = 0.02). It is known that screening may result in overtreatment (36), but the results showed no significant differences in prescription rates for AHT, statins or antiplatelets for those at low risk of CVD (26) (25). Thus, indicating that VI did not generate unnecessary medication prescriptions for people with low risk of CVD. There were significant escalations of new prescriptions or an increased number of medicines prescribed in the high-risk cohort, but not a significantly higher proportion of patients receiving medication as prescribed by guidelines. It would have been relevant to explore if the significant increase in screening of patients was associated with reduced incidence of CVD in patients who had not yet been diagnosed or classified as high risk according to a cost-efficacy perspective.
The graphical presentation was preferred by 57 % of the patients (p < 0.001) (33) and the most complex graphics were the least preferred by the participants (32).
This correlates with Garcia-Retamero et al. who report that patients find information less useful when provided only numerically, in contrast to the doctors who perceived the information as highly useful, with no statistical difference between the numerical or visual display (28). Consequently, the doctors may not experience the same benefit from the VI as the patients; which is an important observation as decision aids are most often introduced by the health care specialist. When the level of numeracy was statistically controlled for, the type of participant no longer had a significant impact on the understanding. This suggests that the preference for VI's is not related to profession but to numeracy. The examination of an online visual decision aid used in primary care showed no significant extension of consultation time, and the variation of the duration of the consultations was significantly lower in the intervention group. The authors suggest that the VI to some extent could result in a more systematic and reproducible discussion between the patient and the doctor (24). With these findings in mind, it would be reasonable to evaluate if the use of a validated VI in primary care could result in a more focused dialogue with a well-prepared patient, a standardisation of the information given, and a more informed health choice without requiring additional resources from the GP. The potential of "video" in risk communication Shukla et al. (34) found that an educational DVD equalled the understanding of a second-grade reading brochure, and at the same time outperformed the understanding obtained by brochures of higher reading levels. Hence, it is relevant to study if the video format has the potential to compensate for impaired reading skills along with reduced numeracy or graph literacy and enhance risk understanding. Valázquez-López et al.'s findings suggest that adding a multimedia tool to conventional nutritional therapy is associated with an improvement in health outcomes (35). The two studies have only used the DVD at the clinic, but the video has the potential to be used as infinite repetition of information at home and a way to involve family members by sharing the information given. If the VI was watched as a preparation to an appointment at the doctor, it may also have the potential to facilitate a more informed discussion as proposed in the paragraph above (4.3).
Since the video only has to be recorded once, it does not require resources consecutively. Based on the limited evidence, it appears that video as a supplemental risk communication tool could be a way of improving health as well as health literacy significantly. Future studies should investigate if the video format has potential to enhance risk understanding, if it will be more cost effective and/or whether it has a potential to be used at home for enhanced understanding and involvement of patients as well as relatives.

Perspectives for future research
This review has revealed a lack of RCT studies in the field of visual risk communication. The majority of studies published has been made in small-scale or with hypothetical scenarios. Studies have shown that tests of hypothetical decisions differ from behavioural change (37). Consequently, it would be beneficial to measure outcomes that relate to factual behaviour or biochemical parameters instead of risk understanding. This, to make sure that the VI has an impact on the actual health decisions of the patients and not only affects the more theoretical and not so quantifiable risk understanding. Based on the results in Table 1

Ethics approval and consent to participate
Not applicable as this is a literature review

Consent to publish
Not applicable as this is a literature review

Availability of data and materials
This is a literature review, and all data presented and analyzed are available in the referenced papers.

Competing interests
The authors declare that they have no competing interests  Daily home BP measurement (HBPM) noted in either a schematic standard non-coloured BP booklet (control group) or a colour-coded booklet (intervention). The scheme in the coloured book was divided into three zones, according to the BP value: green, yellow and red. The duration of the study was six months. Clinical parameters and medication changes were recorded at 0, 3 and 6 months. The outcome measurements: Adherence to HBPM measurements, BP values at follow up at the general practitioner and prescription of antihypertensive medication. The target was to test optimal graphic risk communication formats when presenting small probabilities using graphics with a denominator of 1000. The experimental computer-based manipulation compared three types of graphics; bar charts and pictographs with blocks or dots across horizontal or vertical orientation. The numerator size was divided into three groups: small < 100, medium 100-499 and large 500-999.
Participants were asked two questions concerning the treatment of the medical condition "X". One focussing on gist knowledge and one on verbatim knowledge. Three trainings were completed to ensure that the participants understood the tasks, and how to record their responses before the trial. The study evaluated eight different animated risk graphics presented by icons arrays (blocs). They were viewed on a PC screen that incorporated different combinations of three basic animations: 1) building risk one unit at a time, 2) settling scattered risk into a grouping and 3) shuffling scattered risk to reinforce randomness. Participants received all risk information in 1 out of 10 possible pictograph formats. Outcome: To test if animated icon array pictographs, displaying risks of side effects, could improve participants' ability to select the treatment with the lowest risk profile, as compared with seeing static images of the same risks. Outcome measurements: The ability of the participants to choose the less risky treatment (choice accuracy), gist knowledge of side effects (knowledge accuracy), and graph evaluation ratings, controlled for subjective numeracy, and need for cognition. Imaginary scenario in general practice with a choice between two different types of medication to avoid a bypass surgery. One treatment was designed as superior according to its risk profile and beneficial effects. Numerical risk information was given in one out of the following six graph formats; bar graph, pictograph, modified pictograph (sparkplug), pie chart, modified pie graph (clock graph) or in a table.
The aim was to evaluate what impact these six graphical formats had on answers about treatment risks and benefits. The outcome measurements: Verbatim knowledge (the ability to correctly read numbers from graphs) and gist knowledge (the ability to identify the essential points of the information presented). Patients with a pre-existing heart disease and users of statin.
Patients were interviewed about their preference for methods expressing the preventive benefit of a hypothetical medication. Benefits were expressed in numerical formats (relative risk, absolute risk, number needed to treat, odds ratio and natural frequency) and one graphical (bar chart). The outcome measurements: Could information presented in a different way encourage the patient to take the medication daily, which method was preferred to express the benefit of the medication and if the patient preferred positively or negatively framed information.
No matter ho of the patien would be enc medication.
(68 %) prefe benefits ove patients cou presentation (57 %) prefe graphically ( preferred op Most patient framing (des treatment) a (description treated). A g benefits was the most.  Figure 1 Search history for literature review.