Cardiovascular disease risk communication and prevention: a meta-analysis

Mina Bakhit; Samantha Fien; Eman Abukmail; Mark Jones; Justin Clark; Anna Mae Scott; Paul Glasziou; Magnolia Cardona

doi:10.1093/eurheartj/ehae002

. 2024 Jan 19;45(12):998–1013. doi: 10.1093/eurheartj/ehae002

Cardiovascular disease risk communication and prevention: a meta-analysis

Mina Bakhit ^1,^✉, Samantha Fien ², Eman Abukmail ³, Mark Jones ⁴, Justin Clark ⁵, Anna Mae Scott ⁶, Paul Glasziou ⁷, Magnolia Cardona ⁸

PMCID: PMC10972690 PMID: 38243824

Abstract

Background and Aims

Knowledge of quantifiable cardiovascular disease (CVD) risk may improve health outcomes and trigger behavioural change in patients or clinicians. This review aimed to investigate the impact of CVD risk communication on patient-perceived CVD risk and changes in CVD risk factors.

Methods

PubMed, Embase, and PsycINFO databases were searched from inception to 6 June 2023, supplemented by citation analysis. Randomized trials that compared any CVD risk communication strategy versus usual care were included. Paired reviewers independently screened the identified records and extracted the data; disagreements were resolved by a third author. The primary outcome was the accuracy of risk perception. Secondary outcomes were clinician-reported changes in CVD risk, psychological responses, intention to modify lifestyle, and self-reported changes in risk factors and clinician prescribing of preventive medicines.

Results

Sixty-two trials were included. Accuracy of risk perception was higher among intervention participants (odds ratio = 2.31, 95% confidence interval = 1.63 to 3.27). A statistically significant improvement in overall CVD risk scores was found at 6–12 months (mean difference = −0.27, 95% confidence interval = −0.45 to −0.09). For primary prevention, risk communication significantly increased self-reported dietary modification (odds ratio = 1.50, 95% confidence interval = 1.21 to 1.86) with no increase in intention or actual changes in smoking cessation or physical activity. A significant impact on patients’ intention to start preventive medication was found for primary and secondary prevention, with changes at follow-up for the primary prevention group.

Conclusions

In this systematic review and meta-analysis, communicating CVD risk information, regardless of the method, reduced the overall risk factors and enhanced patients’ self-perceived risk. Communication of CVD risk to patients should be considered in routine consultations.

Keywords: Cardiovascular risk, Risk scoring, Risk communication, Absolute risk, Systematic review

Structured Graphical Abstract

See the editorial comment for this article ‘The art of deciphering and communicating cardiovascular risk: getting it right', by S.U. Khan et al., https://doi.org/10.1093/eurheartj/ehae103.

Introduction

The increasing prevalence of cardiovascular disease (CVD) and associated morbidity is now considered a global emergency. Cardiovascular disease mortality is estimated to account for a third of all deaths worldwide (17.9 million per year according to the World Health Organization).¹ The corresponding Australian estimate is 25%,² with heart attacks and strokes responsible for 85% of the yearly deaths.² Sedentary lifestyles, smoking, unhealthy diet, and poor screening behaviour are largely responsible for the escalation of this problem.²

For decades, clinicians and researchers have implemented effective measures to reduce, diagnose, and treat CVD.^3–5 This includes implementing public education campaigns aimed at improving awareness of the potential preventability of CVDs and accessibility to management.^3–5 Cardiovascular disease risk calculators to estimate individualized risk have also played a role in either preventing or modifying risk factors.⁶ Different international guidelines usually recommend to communicate individual CVD risk, which is calculated by combining risk factors in an empirical equation, as a first step to establish behaviour modification to decrease CVD risk even for asymptomatic people. Risk communication is challenging as multiple factors appear to affect risk perception, for example, how best to visually present this information and which numerical format to use. Timeframe (lifetime, 10 year) of the risk can also influence the perception of risk severity and intention to initiate treatment even in educated populations.^6,7

Various medical specialties have examined the gap between communicating risk information and inducing behaviour change. However, the effectiveness of different strategies to reduce this gap has shown mixed results.^8,9

Several systematic reviews have investigated the effect of risk information on clinical outcomes either by looking at the effect of personalized risk information as general descriptors¹⁰ or providing CVD risk scores (irrespective of the model used e.g. Framingham) or comparing the effect of the different calculators on accuracy of risk perception.^11,12 The systematic reviews reported mixed results and the impact of risk communication interventions on changes in the accuracy of risk perception, patient intentions and actual behaviour change, and clinician’s medications prescribing (in response to the knowledge of their patients’ CVD risk) remains unclear. Hence, an updated synthesis with the inclusion of results from newly published articles was warranted to enhance understanding of the total effects of risk communication on patients and doctors for primary and secondary prevention.

This review investigated the following research questions: (i) What is the impact of CVD risk communication on patient-perceived CVD risk, actual change in CVD risk factors, psychological responses, and self-reported behaviours? (ii) What is the impact of CVD risk knowledge on clinician’s prescribing behaviour?

Methods

This systematic review is reported following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement.¹³ The protocol was developed prospectively and registered on the Open Science Framework (https://doi.org/10.17605/OSF.IO/SNAKV).

Eligibility criteria

Participants

We included studies of adults aged 30 years and above with (secondary prevention) or without (primary prevention) established CVD. Studies of adults with genetic predisposition of CVD or with a sample size less than 40 were excluded for the validity of normal approximation of estimates.¹⁴

Interventions and comparators

We included trials that compared any type of CVD risk communication strategy covering online, paper-based, or verbal administration of any scoring system or tool presenting global CVD risk score or the risk of any specific clinical CVD events (e.g. heart attack, stroke, or atrial fibrillation). This was then compared with usual care with or without any CVD risk communication strategy including comparisons against different risk communication formats (e.g. visual, verbal, and numeric) or communication strategies (e.g. threat and efficacy).

Setting

Eligible interventions were those implemented in any health setting and delivered by any healthcare provider (e.g. cardiologist, general practitioner, nurse, or pharmacist).

Types of outcome measures

Primary outcome

Accuracy of risk perception (patient-reported) as defined and reported eligible by study authors, which could be reported using a scale, numerical, or categorial and then assessed for its accuracy (where the numerator is the self-perception of risk and the denominator is the objectively calculated risk, or where categories of low–medium–high are assigned to % risk levels).

Secondary outcomes

Participant-reported outcomes

Psychological responses to CVD risk information (e.g. decisional conflict and depression).
Behaviour changes either intention to change (e.g. smoking cessation, exercise, and start medications) or actual reported and recorded changes (e.g. quit smoking, lost weight, and dietary changes).

Clinician-reported outcomes

Change in predicted global CVD risk or event rates.
Changes in blood pressure, lipids, and glucose levels.
Prescribing of new/additional medications (e.g. lipid-lowering and/or antihypertensive medications) or lifestyle changes (e.g. smoking cessation).

Study design

We included randomized controlled trials (RCTs) of any design (e.g. parallel, cluster, and crossover) if at least one of the primary or secondary outcomes of interest was reported. Excluded were as follows: observational studies, pre–post or qualitative designs, or studies using hypothetical case scenarios of risk perception. Reviews of primary studies (e.g. systematic reviews and literature reviews) were excluded, but the reference lists of any relevant reviews were checked for any additional, relevant primary studies.

Timeframe

We included eligible outcomes reported immediately after the intervention and other follow-up timeframes ranging from 2 weeks to over 12 months.

Search strategies

Database search

PubMed (MEDLINE), Embase (Elsevier), and PsycINFO (OVID) were searched from inception to 6 June 2023. One of the review authors (J.C.), designed the search string in PubMed, refined and pilot tested with two other authors (M.B. and M.C.), and then translated it for use in the other databases using the Polyglot Search Translator.¹⁵ The complete search strategy for all databases is provided in Supplementary data online, Box S1. No restrictions by language or publication date were imposed, but only publications that were available in full were includable. Conference abstracts were excluded unless they had a clinical trial registry record, or other public report, with the additional information required for inclusion. We supplemented our search with forward and backward citation searches of included studies and by contacting authors of studies not yet published, which we identified in conference abstracts or randomized trial protocols.

Study selection and screening

Two pairs of review authors (M.B. and E.A. or M.C. and S.F.) independently screened the title and abstract of retrieved records against the inclusion criteria. Screening was conducted using the Screenatron feature of the Systematic Review Accelerator.¹⁶ Disputes were identified using the Disputatron feature of the Systematic Review Accelerator and were resolved by discussion or by consulting a third author (M.C. or P.G.).^15,16 See Figure 1 for the PRISMA flow diagram outlining the selection process and Supplementary data online, Table S1 for the complete list of excluded full-text articles with reasons for exclusions.

Preferred Reporting Items for Systematic Reviews and Meta-Analyses flow diagram. *At least one outcome in the article was included in the meta-analysis. Adapted from: Page MJ, McKenzie JE, Bossuyt PM, Boutron I, Hoffmann TC, Mulrow CD, *et al*. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ 2021; 372:n71. doi: 10.1136/bmj.n71

Data extraction

A data extraction form was piloted on two of the included studies and modified based on feedback from within the team as required. Two pairs of review authors (M.B. and E.A. or M.C. and S.F.) independently extracted the following data:

Study characteristics: author, publication year, country, setting, and design.
Participants’ characteristics: target group setting, risk factors, sample size, age, sex, and cardiovascular risk score.
Intervention and comparator description: type of risk communication strategy, who delivered the risk information, how, where, and when the information was delivered (as per the TIDieR framework¹⁷).
Primary and secondary outcomes: as reported above.

Assessment of risk of bias in included studies

Paired authors (M.B. and E.A. or M.C. and S.F.) independently assessed the risk of bias for each included RCT using the Cochrane Risk of Bias 1.0 tool.¹⁸ Risk of Bias Tool 1.0 was used in preference to the Risk of Bias Tool 2.0 as the former allows the assessment of biases from conflict of interest and funding (under the domain: other sources of bias), whilst the latter does not. The following domains were assessed: random sequence generation, allocation concealment, blinding (participants and personnel), blinding (outcome assessment), incomplete outcome data, selective reporting, and other sources of bias (e.g. funding and reported conflicts of interests). Each potential source of bias was graded as low, high, or unclear, and each judgement was supported by a quote from the relevant study. Any disagreements were resolved by discussion between screeners or by referring to a senior author (M.C. or P.G.).

Measurement of effect and data synthesis

Where feasible, dichotomous data were expressed as odds ratios (ORs) or risk ratios with 95% confidence intervals (CIs). Continuous data were expressed as mean differences (MDs) or standardized MDs with 95% CIs. Meta-analysis was only undertaken when meaningful (i.e. when ≥2 studies or comparisons reported the same outcome) using the individual as the unit of analysis, where possible, and due to the anticipated heterogeneity among included studies, a random effects model was used (DerSimonian and Laird random effects method). We used the I² statistic to measure heterogeneity among the included trials. Publication bias was intended to be assessed using a funnel plot, provided there were greater than 10 trials included in the analysis. When meta-analysis was not feasible, studies were plotted to facilitate interpretation and narrative explanation was provided.¹⁹

For psychological response data, the only meta-analysable outcome under this domain was decisional conflict, a concept encompassing ambiguity on the next steps to take on lifestyle modifications. Decisional conflict scales generally covered subcomponents of uncertainty, lack of clarity, information deficits, perception of support levels, and quality of the decision. The outcome was measured as a difference between follow-up and baseline score points.

We also intended to analyse the effects by the following a priori-defined subgroups, when feasible: type of comparator (e.g. usual care or another risk communication strategy), setting, and time to follow-up. We intended to analyse the data according to the TiDIER intervention components and to conduct a sensitivity analysis by including vs. excluding studies with three or more Cochrane risk of bias domains rated at high risk of bias; however, it was not possible due to the paucity of data. We reported the intervention components in Supplementary data online, Table S2.

Results

Included studies

Our search identified 1782 records, of which 1093 remained after duplicates were removed. After full-text screening of 136 records, 62 studies reported in 68 articles met our inclusion criteria.^20–86 A total of 55 studies were included in the meta-analysis (reporting at least one outcome), and seven studies did not contribute to the meta-analysis^{22,31,32,41,49,69,73} (Figure 1).

Characteristics of included studies

Most studies were individually randomized trials of face-to-face single-component interventions conducted in primary care in Europe and the USA (Table 1 and Supplementary data online, Table S3).

Table 1.

Summary characteristics of included studies

	Number of studies (references)
Study design
Randomized controlled trials (RCTs)	44^{21–23,25–27,29–33,35,37–42,47–52,55,58–60,64,66,67, 70–72,74,75,80–86}
Cluster RCTs	18^{20,24,28,34,43,45,46,54,56,57,62,63,65,73,76–79}
Study location
Europe^a	26^{20,21,23,24,28–31,38,42,43,45,46,54–58,66,70–74,76, 78,86}
North America	22^{22,32,33,37,39–41,47,50–52,59,60,64,65,69,75,79–82}
Oceania	9^{25,26,35,49,62,67,77,83,84}
South America	4^27,34,48,63
Africa	1⁸⁵
Number of intervention arms
Two arm trials	47^{20,21,24,26–30,32–35,37–41,43,45–47,50–52,57–60, 62–67,69,71,73,75–82,85,86}
Three or more arms	15^{22,23,25,31,42,48,49,54–56,70,72,74,83,84}
Type of cardiovascular prevention
Primary prevention	42^{20–25,27–33,38,40,41,43,45,46,49,50,55,57,60,63–65, 69–75,77–79,82–86}
Secondary prevention	16^{26,34,35,37,39,42,47,48,51,52,54,58,59,66,67,81}
Mixed	4^56,62,76,80
Setting
Primary care	33^{20,21,24,27,31,34,37,42,43,45,46,50,51,54–57,60, 62–67,70,73–78,81,86}
Community	14^{22,23,25,33,38,48,49,52,71,72,80,83–85}
Secondary care	7^{29,30,32,58,69,79,82}
Tertiary care	5^{26,35,39–41}
Mixed settings	3^28,47,59
Intervention description
Mode of delivery
Face-to-face	41^{20–22,24,26–35,37–41,43,45–52,54–58,62,63,66,73–76,78}
Remote^b	15^{23,25,42,59,64,65,67,69–72,82–85}
Mixed	6^{60,77,79–81,86}
Type of intervention provider^c
General practitioners^d	15^{21,29–31,35,42,43,46,54,57,59,62,76,78,86}
Specialists	13^{24,28,37,39–41,47,49–51,63,75,79}
Researchers	15^{23,25,48,55,65–67,69–72,82–85}
Allied health	11^{20,22,26,27,32,45,52,58,73,74,81}
Mixed	4^34,56,77,80
Other	3^33,60,64
Target participants
Patients	32^{22,23,25,31–33,37,38,42,49,52,55,58–60,64–67, 69–72,74,77,80–86}
Clinicians	13^{28,41,43,46–48,50,51,54,56,57,75,76}
Mixed^e	17^{20,21,24,26,27,29,30,34,35,39,40,45,62,63,73,78,79}
Structure of interventions
Single component	40^{21–23,25–27,31,32,37–41,43,46–48,50–52,54–56,58, 63,65,66,69–76,79,82–84,86}
Multi-faceted	22^{20,24,28–30,33–35,42,45,49,57,59,60,62,64, 67,77,78,80,81,85}

Open in a new tab

^aIncluding one study with multi-EU countries.

^bIncluding online/phone/mail.

^cOne study did not report the type of provider.

^dIncluding family doctors.

^eTwo studies targeted system/organisational changes.

Intervention descriptions

Of note, 22 interventions in this review were multi-component (Table 1 and Supplementary data online, Table S2) and elements went beyond the communication of the risk, being supplemented with other strategies such as educational materials for patients to keep, motivational counselling, attendance to physical activity sessions, knowledge re-testing, interdisciplinary referrals, prescriptions, and follow-up appointments to check on progress. Fifteen trials used some form of decision aid or shared decision-making approach.^{21,29,30,39,40,43,45–47,51,63,69,79,82,85}

Risk of bias

The studies generally had low risk of selection bias with appropriate random sampling, low attrition rates, and complete reporting of intended outcomes. Some concerns were apparent on allocation concealment in almost half of the studies. High risk of bias was found in 41 studies that did not blind participants and/or personnel, with some concerns or unclear reporting for the remaining studies. Risk of bias for blinding of outcome assessment was rated as high or unclear for 20 and 22 studies, respectively. A total of 13 articles did not report on conflicts of interest or funding sources or both (Figure 2 and Supplementary data online, Figure S1).

We were able to produce funnel plots to assess the potential for reporting bias for five outcomes, which are shown in Supplementary data online, Box S2 (accuracy of risk perception, changes in CVD risk score, actual changes in medication, blood pressure, and cholesterol changes). Visual inspection of the plots indicated the potential for publication bias related to studies reporting these outcomes.

Primary outcomes

Accuracy of risk perception

Meta-analysable studies

The accuracy of risk perception assessed immediately after the risk communication in 10/55 studies was higher among intervention participants with or without established disease (10 studies, pooled estimate OR = 2.31, 95% CI = 1.63, 3.27) than among the control participants, although heterogeneity was high (I² = 70%). This was apparent regardless of whether the control strategy was usual care or active control (Figure 3). A test for subgroup difference showed no evidence that the effect of intervention differed across primary versus secondary prevention subgroups (P = .79). The high heterogeneity for primary prevention vs. usual care was related to settings, personnel administering the intervention, and intervention components.

Accuracy of risk perception (n = 10 studies). CI, confidence interval; SE, standard error

Non-meta-analysable studies

Among the eight non-meta-analysable studies for this outcome (Table 2 and Supplementary data online, Table S4), only three primary prevention interventions led to significantly more accurate risk perception: two trials using decision aids with outcomes at 3- and 6-month follow-up, respectively^30,63 and a multi-component study where the Heart Age message achieved higher accuracy at 4 weeks than intervention without the CVD risk communication.⁷²

Table 2.

A summary of findings for results of non-meta-analysable studies for primary and secondary outcomes

	Immediate	2 w	1 M	2–3 M	6–9 M	12 M
Primary outcomes
Accurate risk perception (n = 8)	? ^30,51,71,72		⁷²	?? ^22,51,63,70	? ^29,30,51
Secondary outcomes
Change in CVD risk (n = 10)			? ³²	^43,66,74,81	? ⁴⁹	? ^{20,49,50,52,56}
Psychological response (n = 9)	^25,71	^72,78	^25,42	^66,70	^42,58,78	³¹
Smoking
Intention to change (n = 3)	? ^30,69	²⁵			³⁰
Actual change (n = 15)				^43,70,74,81	? ^58,59,78	? ^{20,27,34,43,45,48,49,52,67}
Physical activity
Intention to change (n = 2)		²⁵			? ⁶⁹
Actual change (n = 3)			³²		⁶⁰	⁴⁸
Dietary
Intention to change (n = 3)	? ⁶⁹	²⁵	⁷²
Actual change (n = 3)			? ^32,72	? ³²	⁵⁹
Medication adherence
Intention to change (n = 1)					? ⁶⁹
Actual change (n = 1)						⁷⁶
CVD risk factors
Blood pressure (n = 5)				⁵⁴	? ^49,57	? ^20,49,52
Lipids (n = 4)			⁷³		^49,65	^c ⁵²
Glucose^b (n = 0)
Clinicians’ prescribing (n = 3)					^a ^57,59	²⁸

Open in a new tab

Inline graphic Favours risk communication P < .05.

Inline graphic Favours risk communication non-significant.

Inline graphic Against risk communication non-significant.

Inline graphic Against risk communication P < .05.

? Not clearly reported.

^aSignificant for antihypertensive medications and non-significant for lipid-lowering medications.

^bAll studies were meta-analysed.

^cOnly for HDL.

CVD, cardiovascular disease; M, month; W, week.