A total of 8378 article titles and abstract were screened from all databases, of which 8348 were excluded based on the eligibility criteria after consensus were reached by the reviewers (AC, WT). 88 articles were sought for retrieval for full text to be read and reviewed independently by both reviewers, after which a further 58 articles were excluded. A total of 30 articles were included for data extraction for this review.
[Figure 1. PRISMA flow diagram]
Description of included studies on patient-physician risk communication
Studies identified were heterogenous in scope and focus. Of the 30 of articles included, 22 articles are related to CVDs (n=22/30, 73.3%), comprising 13 qualitative (n=13/22, 59.1%), 9 quantitative studies (n=9/22, 40.9%), and 1 mixed methods study (n=1/22, 4.5%). Out of 13 qualitative studies, 10 used a semi-structured interview approach to data collection (n=10/13, 76.9%), with 1 study using a qualitative descriptive approach (n=1/13, 7.7%), 1 that conducted focus group discussions (n=1/13, 7.7%), and 1 using both participant observation and interviews (n=1/13, 7.7%). Out of 9 quantitative studies related to CVDs, 7 were randomised controlled or controlled trials (n=7/9, 77.8%), 1 an interrupted time-series study (n=1/9, 11.1%), and 1 a cross-sectional study (n=1/9, 11.1%). 7 out of 30 articles included in this review are related to diabetes (n=7/30, 26.7%), comprising a total of 2 qualitative (n=2/7, 28.6%) and 5 quantitative studies (n=5/7, 71.4%). Out of the 2 qualitative studies, 1 used a semi-structured interview (n=1/2, 50.0%), and 1 a focus group discussion approach (n=1/2, 50.0%). For 5 quantitative studies related to diabetes, 3 studies were randomised controlled trials (n=3/5, 60.0%) and 2 studies were cross-sectional (n=2/5, 40.0%).
Studies selected were conducted in or referred to primary or secondary care settings. Of 22 articles on CVD, 17 studies referred to primary care settings (n=17/22, 77.3%), while 3 occurred in secondary care settings (n=3/22, 13.6%), and 3 in both primary and secondary care settings (n=3/22%). Primary care settings were mainly GPs (n=11/17, 64.7%), family practices (n=3/17, 17.6%), and community health centres (n=1/17, %). 2 studies did not state specifically the actual primary care type (n=2/17, 11.8%). Secondary care settings were mainly atrial fibrillation or transient ischemic attack clinics (n=3/17, 17.6%). 3 studies that included participants from both primary and secondary care settings did not specifically state the exact venue of study; only 1 mentioned primary care setting as GP sites. Out of 7 articles on diabetes, 3 studies referred to primary care settings (n=3/7, 42.9%), and 4 to secondary care settings (n=4/7, 57.1%). Primary care settings were GPs, community health service centres, and 1 which was not stated. Secondary care settings were a teaching hospital, university clinic, a surgery practice, and a diabetes centre. A summary of key characteristics of included studies can be found in Table 1.
Table 1 Key Characteristics of Included Studies
No.
|
Author (Year) / Country
|
Study Design / Methodological Orientation*
|
Data Collection
Method
|
Primary / Secondary Setting
|
Specific Study Setting
|
Study Participants (Patients/Physicians /HCPs/Public)**
|
Sample Size***
|
Specific CVD/Diabetes condition (if stated)
|
CVD – Quantitative Articles
|
1.
|
Adarkwah et. al. (2016) / Germany
|
Randomized controlled trial
|
Surveys
|
Primary
|
GPs
|
Patients
|
Intervention Group:
157 patients
Control Group:
147 patients
|
N.A.
|
2.
|
Adarkwah et. al. (2019) / Germany
|
Randomized controlled trial
|
Surveys
|
Primary
|
GPs
|
Patients
|
Intervention Group:
148 patients (Analysed)
Control Group:
146 patients (Analysed)
|
N.A.
|
3.
|
Casebeer et. al. (2009) / USA
|
Controlled trial
|
Surveys / Patient Electronic Records
|
Primary and Secondary
|
Not stated
|
Patients
|
Intervention Group:
355 patients
Control Group:
196 patients
|
N.A.
|
4.
|
Goodyear-Smith et. al. (2011) / New Zealand
|
Cross-sectional survey
|
Surveys
|
Primary
|
Family practice
|
Patients
|
934 patients
|
N.A.
|
5.
|
Krones et. al. (2008) / Germany
|
Randomized controlled trial
|
Surveys
|
Primary
|
Family practice
|
Patients
|
Intervention Group:
550 patients
(at index consultation)
460 patients
(at follow up)
Control Group:
582 patients
(at index consultation)
466 patients
(at follow up)
|
N.A.
|
6.
|
Roach et. al. (2010) / USA
|
Randomized controlled trial
|
Surveys
|
Primary
|
Not stated
|
Patients
|
Intervention Group:
51 patients
Control Group:
47 patients
|
N.A.
|
7.
|
Tawfik et. al. (2016) / Egypt
|
Randomized controlled trial
|
Surveys
|
Primary
|
family medicine outpatient clinic
|
Patients
|
Intervention Group:
127 patients
(at allocation)
107 patients
(at follow up)
Control Group:
128 patients
(at allocation)
|
N.A.
|
8.
|
Welchen et. al. (2012) / Netherlands
|
Randomized controlled trial
|
Surveys
|
Primary
|
GPs
|
Patients
|
Intervention Group:
132 patients
Control Group:
130 patients
|
N.A.
|
9.
|
Williams et. al. (2016) / Indonesia
|
Interrupted Time Series
|
Surveys
|
Primary
|
Community health centres
|
Patients
|
675 patients
|
N.A.
|
CVD – Qualitative Articles
|
10.
|
Barfoed et. al. (2015) / Denmark
|
Systematic text condensation
|
Semi-structured interviews
|
Primary
|
GPs
|
Physicians
|
10 GPs
|
N.A.
|
11.
|
Bengtsson et. al. (2021) / Sweden
|
Qualitative content analysis
|
Semi-structured interviews
|
Primary
|
GPs
|
Physicians
|
15 GPs
|
N.A.
|
12.
|
Bonner et. al. (2014) / Australia
|
Framework analysis
|
Semi-structured interviews
|
Primary
|
GPs
|
Physicians
|
25 GPs
|
N.A.
|
13.
|
Borg Xuereb et. (2016) / UK
|
Interpretative phenomenology
|
Semi-structured interviews
|
Secondary
|
AF Clinics
|
Patients and physicians
|
11 patients
16 physicians
|
Atrial Fibrillation
|
14.
|
Bridgwood et. al. (2020) / UK
|
Constant comparative method
|
Non-participant observation / Semi-structured interviews
|
Secondary
|
TIA Clinics
|
Patients and physicians
|
15 consultations with 9 stroke specialists observed
15 patients, with
12 accompanying relatives
|
Transient Ischemic Attack
|
15.
|
Durack-Brown et. al. (2003) / France
|
Not stated
|
Semi-structured interviews
|
Primary and Secondary
|
GPs and specialist consultants
|
Patients and physicians
|
27 patients
21 physicians
|
N.A.
|
16.
|
Fisseni et. al. (2008) / Germany
|
Qualitative content analysis
|
Semi-structured interviews
|
Primary
|
GPs
|
Physicians, healthcare professionals, public
|
6 GPs
4 healthcare assistants
12 laypeople
|
N.A.
|
17.
|
Hawking et. al. (2019) / UK
|
Thematic analysis
|
Semi-structured interviews
|
Primary
|
GPs
|
Patients
|
18 patients
|
N.A.
|
18.
|
Honey et. al. (2015) / UK
|
Qualitative descriptive study
|
Qualitative Descriptive Interviews
|
Primary
|
GPs
|
Patients
|
37 patients
|
N.A.
|
19.
|
Polak and Green (2015) / UK
|
Constant comparative method
|
Semi-structured interviews
|
Primary and Secondary
|
Not stated
|
Patients
|
34 patients^
|
N.A.
|
20.
|
Rosal et. al. (2004) / USA
|
Constant comparative method
|
Focus group discussions
|
Primary
|
Not stated
|
Physicians
|
11 physicians
|
N.A.
|
21.
|
Salmasi et. al. (2018) / Canada
|
Qualitative descriptive study
|
Semi-structured interviews
|
Secondary
|
AF Clinics
|
Patients and physicians
|
11 clinicians
10 patients
|
Atrial Fibrillation
|
22.
|
van Steenkiste et. al. (2004) / Netherlands
|
Constant comparative method
|
Semi-structured in-depth interviews
|
Primary
|
GPs
|
Physicians
|
15 GPs
|
N.A.
|
CVD – Mixed Methods
|
23.
|
Kirby and Machen (2009) / UK
|
Cross-sectional survey / Thematic analysis
|
Phase 1:
Surveys
Phase 2:
Focus Group Discussions
|
Primary
|
GPs
|
Patients and physicians
|
Phase 1
(nationwide survey results reported only)
202 GPs recalled using Factfile (out of 600 GPs from the national sample)
181 GPs who accessed risk calculator
Phase 2:
17 GPs
4 practice nurses
1 nurse practitioner
|
N.A.
|
Diabetes – Quantitative
|
24.
|
Denig et. al. (2014) / Netherlands
|
Pragmatic randomised controlled trial
|
Surveys
|
Primary
|
GPs
|
Patients
|
Intervention Group
225 patients
(at allocation)
199
(analysed for primary outcomes)
Control Group (usual care)
119 patients
(at allocation)
107 patients (analysed for primary outcomes)
|
N.A.
|
25.
|
Huang et. al. (2016) / USA
|
Pilot randomised trial
|
Surveys
|
Secondary
|
Clinics at a university
|
Patients
|
Intervention Group
75 patients
(at allocation and analysed)
Control Group:
25 patients
(at allocation and analysed)
|
N.A.
|
26.
|
Lyles et. al. (2012) / USA
|
Cross-sectional survey
|
Surveys
|
Primary
|
Not stated
|
Patients
|
569 patients
|
N.A.
|
27.
|
Ritholz et. al. (2017) / USA
|
Cross-sectional survey
|
Surveys
|
Secondary
|
Diabetes centre
|
Patients
|
128 adults
|
N.A.
|
28.
|
Rouyard et. al. (2018) / UK
|
Pilot randomised controlled trial
|
Surveys
|
Secondary
|
Surgery practice
|
Patients
|
Intervention Group
20 patients
(at allocation)
18 patients
(Analysed at follow up)
Control Group
20 patients
(at allocation)
18 patients
(Analysed at follow up)
|
N.A.
|
Diabetes – Qualitative
|
29.
|
Ledford (2011) / USA
|
Grounded theory
|
Semi-structured interviews
|
Secondary
|
Teaching hospitals
|
Physicians
|
12 physicians
|
N.A.
|
30.
|
Yao et. al. (2022) / China
|
Thematic analysis
|
Focus group discussions
|
Primary
|
Community health service centres
|
Patients
|
22 patients
|
N.A.
|
*Given that study design is often symbiotic with methodological approach for qualitative studies than for quantitative studies, we report methodological orientation for qualitative studies in this column.
** For RCTs, patients rather than both patients and physicians or other healthcare professionals, are indicated as study participants in this column if the primary and secondary outcomes of the study are based on data collected from patients.
***For RCTs, sample sizes are stated for analysis and follow up groups where applicable.
^ For Polak and Green (2015), participants were recruited in the community setting who were offered statin as part of primary and secondary prevention. Therefore, we take the assumption that participants are most likely patients as well.
[Table 1 Key characteristics of included studies]
Most studies on patient-physician risk communication were conducted and concentrated in several high-income countries (HICs) in Europe, North America, and Oceania, such as UK (n=7/30, 23.3%), Germany (n=4/30, 13.3%), Netherlands (n=3/30, 10.0%), Denmark (n=1/30, 3.3%), Sweden (n=1/30, 3.3%), France (n=1/30, 3.3%) in Europe; USA (n=7/30, 23.3%), Canada (n=1/30, 3.3%) in North America, and 1 each in Australia and New Zealand respectively. Only 1 (1/30, 3.3%) (Yao et. al., 2022) study was conducted in an upper middle-income country (UMICs) (China), and only 2 (2/30, 6.6%) (Tawfik et. al., 2016, Williams et. al., 2016) conducted in 2 lower middle-income countries (LMICs) (Egypt and Indonesia), suggesting a paucity of research from non-HIC countries. There were no studies from countries in Africa, South America or other parts of Asia other than China. A summary of countries where studies are conducted can be found in Table 2.
Table 2 Summary of countries where studies were conducted
HIC
|
Continent
|
Country
|
Number
|
Percentage (%)
|
Europe
|
UK
|
7
|
23.3%
|
Germany
|
4
|
13.3%
|
Netherlands
|
3
|
10.0%
|
Denmark
|
1
|
3.3%
|
Sweden
|
1
|
3.3%
|
France
|
1
|
3.3%
|
North America
|
USA
|
7
|
23.3%
|
Canada
|
1
|
3.3%
|
Oceania
|
Australia
|
1
|
3.3%
|
New Zealand
|
1
|
3.3%
|
UMIC
|
Asia
|
China
|
1
|
3.3%
|
LMIC
|
Africa
|
Egypt
|
1
|
3.3%
|
Asia
|
Indonesia
|
1
|
3.3%
|
Total
|
30
|
100%
|
Understanding risk in the context of patient-physician communication
CVD related risk information
15 articles, consisting of 8 qualitative (Bonner et. al, 2014, Bridgwood et. al., 2020, Durack-Brown et. al., 2003, Fisseni et. al., 2008, Hawking et. al. 2019, Honey et. al., 2015, Polak and Green et. al., 2015, Salmasi et. al., 2018) 6 quantitative studies (Adarkwah et. al., 2016, Adarkwah et. al., 2019, Goodyear-smith et. al., 2011, Roach et. al., 2010, Tawfik et. al., 2016, Welschen et. al., 2012) and 1 mixed methods study (Kirby et. al., 2009) focused on different aspects of patient-physician risk communication. We summarise and cluster the articles into 3 thematic areas emergent from articles on CVDs: (1) understanding and recalling risk information in the context of patient-physician communication, (2) risk formats and its effects on the risk communication process, and (3) perception of risk information over time.
Understanding and recalling risk information
Many patients tend to perceive risk in binary terms, such as whether they were ‘at risk’ or ‘not’ (Hawking et. al., 2019), or understood future risks in generic, non-numeric terms, even if numeric values were often used to discuss weight, blood pressure and medication dosage (Polak and Green et. al., 2015). A qualitative study of patients with high 10-year CVD risk in the UK found that most patients do not remember receiving explanations about their CVD risk score or what their scores mean (Honey et. al., 2015). As such, most healthcare professionals tend to explain risk narratively rather than describe risk in percentage terms to patients (Honey et. al., 2015). In a study conducted in Vancouver, Canada, physicians describe how atrial fibrillation (AF) patients tend to overestimate their bleeding risk regarding anti-coagulants and have difficulty weighing risk against benefits (Salmasi et. al., 2018). In response to a lack of interest or inability of patients to understand their own risk, physicians communicate an individual’s risk of stroke to a patient less often unless there is a need to do so, such as when patients show resistance towards medications, if there is an unjustified fear of bleeding, or where there is poor understanding towards how medications can reduce risk (Salmasi et. al., 2018).
For patients with asymptomatic conditions, a common problem is recognizing and acknowledging risks that may not be apparent due to a lack of symptoms. A qualitative study to understand the experience of transient ischaemic attack (TIA) patients during consultations sessions find that prior knowledge and health beliefs influences actions taken by patients, and that a lack of symptoms leads toless recognition of risks (Bridgwood et. al., 2020). Many patients, such as those with high cholesterol who do not have manifest symptoms, find their risks unpredictable, unstable and abstract. These patients also have a poor understanding of CVD risk factors and do not perceive hypercholesterolemia to be a risk factor for CVD (Durack-Brown et. al., 2003). Physicians of AF patients find it worrying that patients often associate symptom severity with risk of stroke, who correspondingly believe that having a lack of symptoms implies not being at risk. (Salmasi et. al., 2018)
One intervention study using probabilistic scenarios conducted with GPs, healthcare assistants, and laypeople to test the level of minimum absolute risk required for participants to justify prescribing a hypothetical tablet able to prevent heart attack over 5 years, find that most participants think it makes no difference if a the drug is consumed over 10 years instead or 5, even if the benefit was greater over a longer time period, suggesting challenges in risk estimation even for healthcare professionals (Fisseni et. al., 2008). An intervention study that aims to facilitate the communication of CVD risks between patients and physiciansby providing patients with a tablet computer containing a series of educational modules that patients have to watch prior to consultation sessions, find sthat such an educational intervention makes it easier for patients to speak to their physicians and have a better understanding of why controlling CVD risk factors is important (Roach et. al., 2010).
Different risk formats and its effect on the risk communication process
Multiple studies focused specifically on risk formats and how it shapes the risk communication process. 1 qualitative study that interviewed GPs in New South Wales, Australia, suggests how pragmatic considerations can affect how physicians choose to convey risk. The study finds that physicians prefer using qualitative formats for communicating risk to patients who have lower numeric literacy and who are of lower risk, given how the discussion of numbers with patients may take up a substantial amount of time,. Absolute, relative risk, and risk displayed in a frequency format, were preferred formats used by GPs to convey information to patients who are at high risk or who had a poorly managed CVD condition (Bonner et. al., 2014). For some patients, there was the perception that providing both absolute and relative risk calculations may actually be unnecessary and confusing, since limited information about ways to reduce risk were already received by patients in the first place. Some patients had strong objections to the word ‘absolute’, which was seen as ambiguous and that seem to convey a risk score that was ‘conclusive’ or ‘definitive’ (Kirby et. al., 2009).
There was consensus among patients that risk was generally difficult to understand. In one mixed methods study conducted in the UK to evaluate the use of the JBS2 risk calculator and chart within GP settings, patients do not recall seeing a risk assessment tool used, although they agree that the use of tools can increase confidence in risk assessment and aid patients in understanding risks (Kirby et. al., 2009). Patients prefer a risk calculator that indicates risk in the form of a thermometer rather than paper charts, highlighting how a visual thermometer is more appealing and easier to understand and can even be motivational, although this may cause anxiety to those whose risk is very high (Kirby et. al., 2009). A qualitative study conducted in the UK to understand the experiences of participants presented with a personalised risk report that includes heart age and QRISK2 risk score that indicates a person’s risk of having a stroke or heart attack within the next 10 years, finds that patients tend to recall heart age easier rather than a probabilistic score (Hawking et. al., 2019).
In terms of format preferences, a cross-sectional study examining the preferences of patients attending GP practices in Auckland, New Zealand found that relative risk (n=603/934, 64.5%) was the highest ranked mode of risk presentation preferred, followed by absolute risk (n=131/934, 14.0%), then natural frequencies (n=91/934, 9.7%), when it comes to the format that would help a patient to decide. In this study, relative risk was ranked first by participants who were more numerate (OR = 1.2; 95% CI, 1.0-1.4), those who were more concerned about a heart attack (OR = 1.1; 95% CI, 1.01-1.2), and less by Pacific Islanders (OR = 0.4) or Asian (OR = 0.4) participants (ethnicity overall, 95% CI, 0.7-0.8). Pictures were preferred over numbers by those who had less schooling (OR = 1.2; CI, 1.1-1.3) and by those who were less numerate (OR = 1.1; CI, 1.01-1.2) (Goodyear-smith et. al., 2011).
In a multi-component RCT conducted in Egypt to investigate the accuracy of CVD risk perception among patients with diabetes, patients were provided a combination of absolute and relative risk scores conveyed in percentage and frequency formats, and given advice framed positively by physicians on how to change their risk based on WHO/ISH guidelines. Agreement between perceived and objective CVD scores increased substantially for the intervention group (n=107) (pre-/post- intervention, kappa = 0.271+/-5.2%, p = 0.0 to 0.837+/-4.4%, p = 0.0), compared to the control group (kappa = 0.088/+/-4.5%, p = 0.052 to 0.105+/-4.6%, p = 0.022), which increased only marginally and remained low (Tawfik et. al., 2016).
Perception of risk information over time
Two intervention studies suggests that patient’s perception of risk, although mediated by formatting and visualisation elements that aims to improve understanding, tends to taper over a longer period of time. A non-inferiority RCT conducted in Germany to test a time to event (TTE) format versus emoticons in representing a patient’s 10-year absolute risk of CVD finds TTE to have a stronger effect on risk perception than emoticons (Adarkwah et. al., 2016), although the effect on perception waned somewhat after 3 months (Adarkwah et. al., 2019). Another RCT intervention conducted with Type 2 diabetes patients newly referred to a diabetes care system in the Netherlands, that used a 6-step CVD risk communication method, found that patients in the intervention group were able to estimate their risk of developing CVD more accurately than those in the control group in the short term (appropriateness of risk perception, intervention 0.33 vs. control -0.1, difference = 0.48, CI 0.02 to 0.95 (p = 0.04)), but the effect of risk perception diminished after 12 weeks. The intervention used a combination of tools that include conveying to patients their absolute risk scores calculated using the UK prospective diabetes study (UKPDS) risk engine, with a risk card containing a population diagram, and having positive framed messages of risk conveyed to patients (Welchen et. al., 2012).
Diabetes related risk information
There were a only a small number of studies on patient-physician risk communication focused on diabetes. Five articles, consisting of 2 qualitative (Ledford, 2011, Yao et. al., 2022) and 3 quantitative studies (Huang et. al., 2016, Ritholz et. al., 2017, Rouyard et. al., 2018), focused on how patients may not be provided sufficient information about diabetes related complications and risks by physicians..
Understanding and recalling risk information
One descriptive cross-sectional study describes how only about 23% (32/138) and 14% (19/138) of patients diagnosed with diabetes respectively recall their providers providing them with factual information and warning them about the implications of complications (Ritholz et. al., 2017). The low proportion for this study conducted at a diabetes centre seem to imply a relatively limited number of patients who are conveyed actual risk of complications by physicians.. A focus group discussion conducted with Type 2 diabetes patients from community health service centres in Guangzhou, China, describe how patients understand normal blood glucose and HbA1c levels as reflective of a stable condition, and that higher numbers or fluctuating numbers are a source of worry. Patient participants were not aware that diabetes was a risk factor for CVDs, though they were concerned about diabetes complications. Patients found that consultations with physicians to be too brief and wanted more information about how diabetes can progress and develop further into complications (Yao et. al., 2022). Physicians who treat patients with diabetes, describe engaging more with communicatively active patients (CAP) who are able to recognize and respond to new or evolving medication risk information (Ledford, 2011).
One pilot RCT study with diabetes patients to assess the feasibility of adopting a new risk communication intervention tool based on behavioural and psychological concepts in primary care, focused on diabetes as a risk factor for CVD, find that recall for effective heart age is significantly better than other formats such as 10-year CVD risk, both immediately and 12 weeks after intervention (Rouyard et. al., 2018). Another pilot intervention conducted at 2 clinics in the University of Chicago, USA, find that using a web-based decision support tool can improve risk understanding. The intervention consists of an education support module, a model for calculating life expectancy and risk of developing CVDs, a treatment preference questionnaire, and geriatric screening component consolidated in the form of a personalised report to be delivered by patients prior to a patient’s visit with a physician. The study showed that decisional conflict (DC) are reduced in the intervention group more than the control group (Overall DC score pre-/post- intervention 52.7 +/- 33.0 to 24.5 +/- 26.7, pre-/post- control 51.2 +/- 35.5 to 36.6 +/- 33.8, p=0.07). Although the results for the overall DC scale was not significant, the informed DC subscale which asks about knowledge and risk understanding related to A1c goals is significant (Informed DC subscale score pre-/post- intervention 54.0 +/- 40.1 to 18.3 +/- 33.7, pre-/post- control 56.0 +/- 38.4 to 31.0 +/- 41.0, p=<0.001) (Huang et. al., 2016)
Strategies and approaches used by physicians in patient-physician risk communication
Eight articles (5 CVD, 3 diabetes), consisting of 7 qualitative (Barfoed et. al., 2015, Bonner et. al., 2014, Borg Xuereb et. al., 2016, Honey et. al., 2015, Ledford, 2011, Rosal et. al., 2004, Yao et. al., 2022) and 1 quantitative study (Ritholz et. al., 2017), describe strategies and approaches used by physicians to communicate risk information to patients. Strategies used to convey CVD risk include the use of fear or scare tactics (Bonner et. al., 2014, Borg Xuereb et. al., 2016, Honey et. al., 2015), use of strong language to evoke fear (Rosal et. al., 2004), use of positive language (Bonner et. al., 2014), downplaying risk (Honey et. al., 2015), or use of metaphors and analogies such as associating heart function with an electrical system to simplify risk information and improve patient’s understanding (Borg Xuereb et. al., 2016). Other approaches used include presenting different CVD scenarios to those who are less adherent to medication (Barfoed et. al., 2015), speaking indirectly to patients (Bonner et. al., 2014), intervening strategically during ‘teaching’ moments (Rosal et. al., 2004), emphasising gradual and continuous change (Rosal et. al., 2004), and prioritising discussion points (Rosal et. al., 2004).
For patients with diabetes, strategies used by physicians include setting goals (Yao et. al., 2022), using specific words such as ‘common’ or ‘rare’ to describe risks (Ledford, 2011), avoiding statistics (Ledford, 2011), varying presentation style to different types of patients (Ledford, 2011), and withholding information such as low-level risks that may affect a patient’s medication intake (Ledford, 2011). Additional approaches include using dramatic images such as illustrations of amputations to persuade patients (Yao et. al., 2022). Physicians who provided prompt feedback to patients using clear language and had positive body language were viewed positively by patients (Yao et. al., 2022). diabetes patients, like patients treated for CVD conditions, similarly mentioned how physicians used fear as a motivator to warn about complications (Ritholz et. al., 2017).
The use of fear and scare tactics was a recurring theme to direct patients towards desired behaviour. If physicians perceived patients to be of higher risk but are generally unmotivated about their own health status (Bonner et. al., 2014), the consequences of risky behaviour and complications that may occur in the future (e.g., such as being bed bound) were used by physicians to persuade patients to change (Borg Xuereb et. al., 2016, Rosal et. al., 2004, Ritholz et. al., 2017). If a patient was not ready and receptive to risk information, then physicians would avoid conveying risks directly to avoid alarming or affecting patients negatively (Bonner et. al., 2014). On the other hand, for patients who were motivated or self-activated, a positive strategy is taken to reassure and encourage patients to focus on achievable change (Bonner et. al., 2014), since a communicatively active patient is cognitively ready and can be provided more details about risks (Ledford, 2011).
A summary of articles and strategies and approaches used by physicians for CVDs and diabetes related conditions are described in table 3.
Table 3 Strategies and approaches used by physicians in patient-physician risk communication
No.
|
Author (Year) / Country
|
Strategies and approaches used in patient-physician risk communication
|
Description of strategies and approaches used
|
CVDs
|
1.
|
Barfoed et. al. (2015) / Denmark
|
- Presenting different CVD scenarios
|
- 2 different ways of managing compliance by GPs were identified, ‘resigned’ and ‘confrontational’.
- For the resigned group, compliance is seen as a patient’s choice, which the GP should respect, or is not seen as a problem at all.
- For the confrontational group, there were dual approaches. Patients were presented with different CVD scenarios that may occur if drug treatment were not adhered to, or there were repeated consultations to reveal barriers in following agreed treatment plans.
|
2.
|
Bonner et. al. (2014) / Australia
|
- Positive
- Scare Tactics
- Indirect
|
- ‘Positive’ strategies were used where GP’s perception of patient’s CVD risk was low and patients were determined in managing their own condition. Approach involves reassuring and motivating patients, focusing on achievable changes.
- ‘Scare tactic’ strategies were used when a GP’s perception of patient’s CVD risk was high, where patients are unmotivated, with a focus on how future risk may be even higher possibly leading to a CVD event.
- ‘Indirect’ strategies were used by GPs where patients, who may be of low CVD risk and have lifestyle factors that needs to change, tend to be more resistant, where the conveyance of risk may not be helpful and may elicit a negative response in patients.
|
3.
|
Borg Xuereb et. (2016) / UK
|
- Use of metaphors and analogies
- Use of fear
|
- Physicians used diagrams and metaphors to explain about AF and to convey the risks of stroke and bleeding.
- Metaphors focused on bodily physiology and analogies to describe bodily function (e.g. Heart function as electrical wiring).
- A few clinicians mentioned using fear of stroke to persuade patients to make the right decision.
|
4.
|
Honey et. al. (2015) / UK
|
- Downplaying risk
- Use of fear
|
- Patients described healthcare professionals as downplaying their high-risk scores, which affect the level of significance a patient will attribute their own risk.
- Other risk communication approaches include appealing to fear by emphasizing the worst possible consequences that may occur, and the use of humour to communicate risk.
|
5.
|
Rosal et. al. (2004) / USA
|
- Use of strong language to evoke fear
- Intervening at strategic timepoints
- Emphasising gradual and continuous change
|
- Strategies used by physicians to motivate lifestyle change include using strong language to evoke fear in patients about risky behaviour or intervening during “teaching moments”. Other strategies include emphasising continuous and gradual change by asking patients to adopt one change at a time.
|
Diabetes
|
6.
|
Ledford (2011) / USA
|
- Use words like ‘common’ or ‘rare’ to present risks
- Avoid statistics
- Varying present style to different types of patients
|
- Physicians use various strategies in presenting risk information to patients, such as using words like ‘common’ or ‘rare’, avoiding statistics, personalising presentation to patients, and varying presentation style by medication.
|
7.
|
Yao et. al. (2022) / China
|
- Setting goals
- Using illustrations such as pictures of amputations
- Prompt provision of feedback
- Use of clear and frank words
- Positive body language
|
- There were instances where GPs set goals for patients, although patients may adhere less required than advised.
- Some patients mentioned that their GPs attempted to use dramatic illustrations such as pictures of amputations to persuade them.
- Good communication experiences include prompt provision of feedback, use of clear and frank words, and positive body language.
|
[Table 3 Strategies and approaches used by physicians in patient-physician risk communication]
Communication tools that facilitate patient-physician risk communication
8 articles (7 CVD, 1 diabetes), consisting of 6 qualitative (Barfoed et. al., 2015, Bengtsson et. al., 2021, Bonner et. al., 2014, Borg Xuereb et. al., 2016, Hawking et. al., 2019, van Steenkiste et. al., 2004), 1 quantitative (Lyles et. al., 2012) and 1 mixed methods study (Kirby et. al., 2009), identify tools used to facilitate the communication of CVD risks between patients and physicians. Studies describe the use of personalised risk reports to aid the understanding of risks (Hawking et. al., 2019), that contain pictorial information such as colour codes to indicate CVD risk status with visual depictions of arterial plaques (Bengtsson et. al., 2021). Other tools used include risk charts (Bonner et. al., 2014, Kirby et. al., 2009), risk tables (van Steenkiste et. al., 2004), drawn diagrams (Borg Xuereb et. al., 2016), CVD guidelines (Barfoed et. al., 2015), risk assessment tools (Barfoed et. al., 2015), results from risk calculators such as the JBS2 calculator in the UK (Bonner et. al., 2014, Kirby et. al., 2009), and images such as cholesterol spikes or how the brain looks like during a stroke (Bonner et. al., 2014).For communication tools used by diabetes patients, 1 study found that although most risk factor discussion with physicians still occurs in-person, those who use phone and secure messages to communicate risk factors tend to also have higher in-person provider visits, insulin use and poorly controlled Hba1c (Lyles et. al., 2012).
Interventions to improve the communication of risk information
Interventions related to CVD conditions
Eight interventions were related to CVDs, of which 6 were RCTs (Adarkwah et. al. 2016, Adarkwah et. al. 2019, Krones et. al., 2008, Roach et. al., 2010, Tawfik et. al., 2016, Welschen et. al., 2012), 1 a controlled trial without randomisation (Casebeer et. al., 2009), and 1 an interrupted time series design (Williams et. al., 2016). Most interventions used a combination of training and interventional materials for physicians or healthcare providers, along with reinforcing tools provided to patients that emphasise risk management. Interventions usually involve the use of relative and absolute CVD risk scores generated by risk calculators or algorithms (Krones et. al., 2008, Tawfik et. al., 2016, Welschen et. al., 2012, Williams et. al., 2016), or reports that present CVD risk stratification using colour codes (Casebeer et. al., 2009). A pocket guideline from the National Health Cholesterol Education Program and WHO/ISH guidelines were used by Casebeer et. al. (2009) and Tawfik et. al. (2016) respectively. Additional intervention tools include the use of script like decision aid (Krones et. al., 2008), use of smiley faces (Krones et. al., 2008), emoticons or time to event graphics to visualise risk (Adarkwah et. al. 2016, Adarkwah et. al. 2019), population diagrams (Welschen et. al., 2012), positive framing (Welschen et. al., 2012), risk cards (Welschen et. al., 2012), educational worksheets (Williams et. al., 2016), checklists (Williams et. al., 2016), patient contract or pledges (Casebeer et. al., 2009), and chart stickers (Casebeer et. al., 2009). 1 study used a tablet with multiple educational modules, linked to a patient’s electronic health records (EHR), that requires a patient to access and watch before attending a consultation session with a physician, spread over a period of 5 visits (Roach et. al., 2010).
Five interventions (Casebeer et. al., 2009, Krones et. al., 2008, Roach et. al., 2010, Welschen et. al., 2012, Williams et. al., 2016) required patients to actively respond to the risk information and educational materials provided to them by the physician or health professional. This includes pledging to commit to the medical regimen (statin therapy) of the intervention (Casebeer et. al., 2009), participating in a shared decision-making process with the physician in modifying one’s risk (Krones et. al., 2008), talking to the physician shortly after watching educational modules on a tablet (Roach et. al., 2010), thinking aloud to check if one can self-explain one’s own risk (Welschen et. al., 2012), or completing a checklist to count the number of tick marks one has and the risk category one falls into (Williams et. al., 2016). 1 intervention study, the Heart Health Counts program, sent 5 print mailings over a 4 month period to patients who are new to statin therapy with information focusing on various aspects of CVD risk (Casebeer et. al., 2009).
Before the start of each intervention, training or meeting sessions are sometimes initiated between physicians and other healthcare professionals to familiarise about study aims and protocol (Williams et. al., 2016), discuss the epidemiology of CVD risk calculations (Krones et. al., 2008), explain the meaning of absolute and relative risks (Tawfik et. al., 2016), and discuss practical strategies of how to communicate risk information to patients (Krones et. al., 2008, Tawfik et. al., 2016, Welchen et. al., 2012, Williams et. al., 2016). Sessions are also held to train physicians and healthcare professionals on how to use risk prediction tools such as the adapted Framingham algorithm (Krones et. al., 2008), WHO/ISH CV risk prediction chart (Tawfik et. al., 2016), or the UKPDS risk engine (Welchen et. al., 2012); to discuss the causes and consequences of CVD risk (Welchen et. al., 2012), and discuss the ethics of shared decision making (Krones et. al., 2008). For 1 intervention, training include explaining behavioural theories such as the Theory of Planned Behaviour and Self-Regulation Theory to physicians (Tawfik et. al., 2016), although engagement with theoretical concepts was not common among most interventions included in this review. Role playing was included in some of the intervention training sessions to reinforce learning (Krones et. al., 2008, Williams et. al., 2016). Other than a general outline of training procedures, interventions described in the articles included do not provide more details about the process of training sessions.
Interventions related to diabetes
3 interventions were related to diabetes, consisting of 3 RCTs (Denig et. al., 2014, Huang et. al., 2016, Rouyard et. al., 2018). All the interventions used a personalised decision aid or support tool to present risks to patients in the form of a report (Denig et. al., 2014, Huang et. al., 2016, Rouyard et. al., 2018). 1 pilot intervention was specifically guided by ideas from behavioural economics and psychology to shape the design of intervention, factoring in concepts such as optimistic bias, affect or representative heuristic, risk aversion, present biasness and limited attention span to determine the structure of risk format and description of risk information (Rouyard et. al., 2018). The intervention study, however, did not go into detail about how concepts are translated into risk information. Personalised reports usually include a patient’s screening test results (Denig et. al., 2014, Huang et. al., 2016), including treatment options or recommended actions that can be taken for specific risk factors (Denig et. al., 2014, Rouyard et. al., 2018). Reports also conventionally includes a calculator that estimates life expectancy or risk of developing complications such as developing a heart attack or risk of amputation or blindness (Huang et. al., 2016), or that estimates heart age (Rouyard et. al., 2018), with an education module (Huang et. al., 2016). Risks are described using natural frequencies to convey outcome probabilities relevant to each patient (Denig et. al., 2014, Huang et. al., 2016, Rouyard et. al., 2018). 1 intervention mentioned setting achievable goals for patients to aim towards to (Denig et. al., 2014), while 2 interventions required patients to discuss their treatment options with a physician after a personalised report was received by the patient (Denig et. al., 2014, Huang et. al., 2016).
A summary of articles and associated intervention components described for included articles is described in table 4.
Table 4 Intervention components of studies to improve patient-physician risk communication
No.
|
Author (Year) / Country
|
Study Design
|
Intervention Components
|
Description of Intervention
|
CVDs
|
1.
|
Adarkwah et. al. (2016) / Germany
|
Randomized controlled trial
|
- Risk report
- Risk scores, represented by either emoticons or time to event (TTE) illustrations
|
- Patients were randomized to the emoticons or the TTE illustration group during consultation with GPs who entered a study ID into the decision support software.
- GPs learned about each patient’s allocation by the illustration displayed by the software, then started a discussion with their patients based on the allocated display (i.e. emoticons or TTE)
- After consultation, patients fill in a questionnaire covering the immediate outcome assessments. GPs recorded the decision made, such as specific medications, dose adjustments, behavioural measures or no change at all.
- 3 months later, patients were contacted by telephone to assess their adherence.
|
2.
|
Adarkwah et. al. (2019) / Germany
|
Randomized controlled trial
|
Same as Adarkwah et. al., 2016
|
See above, Adarkwah et. al., 2016
- A score was developed by the study group to measure adherence as no validated instrument was available for use.
- For each item, a distinction was made according to whether a patient had been completely adherent, partially adherent, or nonadherent. We labelled the grade of adherence as 2= fully adherent, 1=partially adherent, 0=non-adherent.
|
3.
|
Casebeer et. al. (2009) / USA
|
Controlled trial
|
- Risk report, including colour codes to indicate CVD risk
- Counselling kit for physicians
- Patient contract/pledges
- National Cholesterol Program Pocket guidelines
- Chart stickers
- Multiple mailings of educational materials to patients
|
- The goal of the Heart Health Counts (HHC) program is to facilitate statin adherence and provide tools for physicians to increase clarity in CVD risk dialogue.
- Materials address cardiovascular risk and provide context for patients who might be presented with lab values but may not have a context for severity of risk.
- The 1-minute health manager and other HHC materials used colour coding with green, yellow, and red to represent cardiovascular risk stratification, and to serve as a call to action for patients who are in the yellow and red zones.
- Physicians receive a counselling kit including:
(1) a set of 1-minute health manager patient education tools used to describe cholesterol risks, (2) patient contracts or pledges designed to confirm a patient's commitment to the prescribed medical regimen, (3) a copy of the National Cholesterol Education Program pocket guidelines and, (4) a set of chart stickers.
- Following office visit, patients who are new to statin therapy receive 5 HHC mailings over a 4 month period, with the four-color print mailings focused on various aspects of cardiovascular health and cardiovascular risk.
|
4.
|
Krones et. al. (2008) / Germany
|
Randomized controlled trial
|
- Training sessions for physicians, including role-playing and feedback
- Decision aid with script for physicians
- Risk report
- Absolute and relative risk scores, represented by emoticons (smiley faces)
|
For physicians:
- Attend 2 Continuing Medical Education (CME) sessions lasting 2 hours each, to discuss the epidemiological background of global CVD risk calculation and ethics of shared decision making, including practical communication strategies.
- The script-like decision aid was practiced through role playing with participants receiving feedback from peers. After completion of the trial, physicians in the control arm were offered meetings on ARRIBA-Herz and materials.
For patients:
- The patient’s perspective on prevention of CVD was addressed, and patients were invited to a shared decision-making process.
- Physicians calculated each patient’s absolute risk for stroke and myocardial infarction based on an adapted Framingham algorithm with the decision aid.
- Individual prognosis was compared with age- and sex-adjusted population risk.
- For patients in secondary prevention, we assumed about 50% absolute risk for stroke or myocardial infarction in the next 10 years.
- Individual prognosis was displayed through marked smiley faces (smileys).
- The possible effects of single or multiple interventions were calculated by applying the specific relative risk reduction on the calculated and demonstrated absolute risk, which was visually supported by smileys being crossed out, i.e., events prevented.
- Physicians were taught to calculate and show the effect of several preventive measures simultaneously.
|
5.
|
Roach et. al. (2010) / USA
|
Randomized controlled trial
|
- Multimedia educational tool (tablet) with personalised health information and educational videos for patients
|
- The tool uses multimedia technology to educate patients with Type 2 diabetes about their personal CVD risk, explain therapeutic options, and most importantly, motivate them to discuss therapy options with their provider.
- Program software incorporates personalized health information from the EHR to create personalized risk messages.
- Effective interaction with the tool does not require computer skills.
- Patients interact with the computer-based program just before visiting their primary care provider.
|
6.
|
Tawfik et. al. (2016) / Egypt
|
Randomized controlled trial
|
- Training for physicians
- Risk report
- Absolute and relative risk scores, in frequency and percentage formats
- CVD guidelines
|
- Treating physicians receive introductory training on the importance of providing information to patients on CVR including the definition of risk, an explanation of absolute and relative risks, methods of risk communication, self-regulation theory (SRT), theory of planned behaviour (TPB), and a demonstration on how to use the WHO/ISH CVR prediction chart and on how to discuss CVR with patients.
- Discussion of CVR was based on discussion of having a risk of heart attack and include: (a) providing a clear and simple message on CVD and what actions can be taken to prevent it, (b) explaining in frequency and percentage formats, including absolute and relative risk score for a person with Diabetes as similar age and sex as the patient, (c) explaining to patients how to change their CVR based on WHO/ISH CVD guidelines.
|
7.
|
Welchen et. al. (2012) / Netherlands
|
Randomized controlled trial
|
- Training for nurses and dieticians
- Absolute risk scores, represented by natural frequencies
- Risk card, with population diagram
- Feedback from patients after risk information is conveyed
|
- Diabetes nurses and dieticians receive training for two half-days in learning to perform the intervention.
- Intervention consists of six steps:
1) Introducing risk communication and explanation about health risks related to Type 2 Diabetes, including the cause and consequences of CVD risk.
2) Communication of absolute risk of developing CVD in the next 10 years using the UK Prospective Diabetes Study (UKPDS) risk engine, explained through natural frequencies format.
3) Visual communication by means of a risk card, with the help of a population diagram.
4) Positive framing explanation that lifestyle changes by the patient can help to reduce the risk.
5) Communication with the patient for a reaction: after finishing the explanation of the risk, the diabetes nurse asked the patient to give a reaction to the information that was given using open questions.
6) Thinking aloud; patient have to explain the risk to him/herself and was encouraged to think aloud.
|
8.
|
Williams et. al. (2016) / Indonesia
|
Interrupted Time Series
|
- Training for physicians and nurses, including role-playing
- Checklist of modifiable risk factors
- Risk charts (report)
- Educational worksheet with patient’s input
|
- Training was conducted with 20 providers on the study aims, intervention protocol, roles, and intervention materials.
- Trainers than demonstrated the intervention behaviours to the physicians and nurses, after which they role-played and received feedback.
- During the intervention, nurses assessed patients for the study inclusion criteria and recorded this on the physician intervention material, containing a checklist of modifiable stroke risk factors (i.e., blood pressure, smoking, diet, body weight, physical activity). The checklist in placed in the patient’s charts for the physician to discuss with the patient.
- Patients receive a one-page educational worksheet covering the same list of stroke risk factors to complete in the waiting room prior to their consultation with the physician.
- The worksheet asked patients to think about their status on each of the stroke risk factors, then asked them to count the number of check marks and provided an interpretation of “high risk for stroke” / “caution” / “low risk for stroke” in accordance with the count.
|
Diabetes
|
9.
|
Denig et. al. (2014) / Netherlands
|
Pragmatic randomised controlled trial
|
- Decision aid for patients
- Personalised risk report
- Treatment goals and options for patients
- Risk scores using graph and text, represented by natural frequencies
|
- A decision aid was developed for patients with Diabetes, with individually tailored information on risks and treatment options for multiple risk factors.
- Key features include a personal status report including test results and current drug treatment; the presentation of tailored information on achievable treatment goals and possible treatment options for specific risk factors; a combination of graphs and text using natural frequencies for outcome probabilities; the presentation of pros and cons of all treatment options; and asking patients to think about treatment options.
- Should be used by patients before a regular quarterly check-up and discussed together with healthcare provider during consultations to help prioritise treatment that will maximise relevant outcomes.
|
10.
|
Huang et. al. (2016) / USA
|
Pilot randomised trial
|
- Training for physicians
- Training for patients by research assistants
- Decision support tool for physicians
- Personalised risk report
- Absolute risk scores
- Treatment preferences for patients
|
- Intervention physicians underwent 1 hour of in-person training on the principles of geriatric diabetes and the use of the decision support tool.
- Patients met with a research assistant 1 hour prior to physician appointment and were given brief instructions on how to use the tool.
- Main components of the decision support tool include: 1) an interactive diabetes education module, 2) a simulation model that calculates life expectancy and risk of developing complications, 3) treatment preference elicitation, 4) geriatric condition screening, and 5) a personalized patient printout.
- Model calculates a range of values including life expectancy, lifetime risk of developing a heart attack (at A1C of 7%), and risk of amputation or blindness (at A1C of 7%, 8%, and 9%). In addition, personalized risks, patient preferences, and geriatric screening results were summarized in the printout.
- Patients received 2 copies with instructions to give 1 copy to their doctor.
|
11.
|
Rouyard et. al. (2018) / UK
|
Pilot randomised controlled trial
|
- Personalised Risk report
- Risk scores
|
- A risk communication intervention aiming to target 7 decision-making biases that is known to influence decision-making process of people with T2DM.
- The intervention aims to improve insight and recall of diabetes-related risks to nudge people with T2DM towards better self-management.
- Approaches include mitigating optimistic bias, increasing emotional impact and representativeness of risk, taking advantage of people’s loss aversion tendencies and mitigating present biasness, increase people’s limited attention, and personalising risk using a gained framed message.
|
[Table 4 Intervention components of studies to improve patient-physician risk communication]