Absolute cardiovascular risk assessment by Australian early-career general practitioners: a cross-sectional study

Cardiovascular disease burden

Cardiovascular disease (CVD) is the leading global cause of disability-adjusted life-years and mortality, accounting for approximately one-third of deaths globally.1 In Australia, CVD carries a significant burden of disease.2 3 Internationally, CVD-related morbidity and mortality rates have improved significantly in recent decades. Between 2010 and 2019, there was an 11.1% decrease in age-standardised rates of death due to CVD (14.7% for cerebrovascular disease and 9.7% for ischaemic heart disease).4 5 Despite these improvements, the global decline in CVD incidence and mortality has faltered, and even reversed in younger people.6 7 A further aspect of the Australian context is that Aboriginal and/or Torres Strait Islander peoples experience CVD at higher rates than non-Indigenous Australians but are undertreated for risk factors and receive poorer ongoing management,8 9 outcomes reflected in First Nations populations in Canada10 and the USA.11

Absolute cardiovascular risk assessment

Modifiable risk factors for CVD are well established (including high blood pressure, high blood cholesterol, smoking and diabetes). Better management of these risk factors has been a major reason for declines in rates of CVD in advanced countries in recent years. Nine potentially modifiable risk factors have been deemed to account for 90% of the Population Attributable Risk of myocardial infarction in men and 94% in women in a major international study.12

The additive effects of individual risk factors in CVD and a need to target preventive treatment to those patients at greatest risk of CVD has led to the concept of a management tool based on absolute cardiovascular risk (ACVR) rather than independent management of individual risk factors.13 Calculation of ACVR involves the use of multivariable risk assessment tools (CVD risk scores or calculators) to estimate an individual’s CVD risk. These scores or calculators incorporate relevant CVD risk factors and arrive at a summary estimate of the individual’s personal absolute risk of experiencing a major cardiovascular event (typically, in the following 5 or 10 years).14 The intensity of management (including antihypertensive and lipid-lowering treatment) should then be considered accordingly.15 As long ago as 1993, recommendations have been made for antihypertensive and lipid-lowering treatment decisions to be based on ACVR calculation rather than relative risk.16 17

ACVR algorithms include those for estimating risk of fatal or non-fatal coronary heart disease developed from Framingham, USA data,18 for estimating risk of any CVD mortality from European data,19 and for estimating risk of incident CVD (coronary heart disease, stroke and transient ischaemic attack) from the UK data.20 In Australia, use of the Australian cardiovascular risk calculator, based on the Framingham Risk Equation and recalibrated for the Australian population, is recommended by the National Heart Foundation.14

Implementation of ACVRa in general practice

General practice is the primary mechanism for CVD risk identification and management in Australia.21 All GPs and GP registrars (vocational specialist trainees in family medicine/general practice) will have access to ACVR risk calculators in their practice software, plus easily accessible online availability. Hardcopy ACVR risk calculation algorithms (using colour-coded charts) are also widely available.22 Australian GP practice software programs (and online ACVR calculators) produce a percentage estimate of ACVR (eg, % risk of incident CVD in the next 5 years) based on Framingham data. Printed materials (and some online sites) produce an ACVR risk stratification classification, for example, ‘low risk’ (less than 10% risk of CVD within the next 5 years), ‘moderate risk’ (10%–15%) and high risk’ (greater than 15%).14

Routine use of ACVRa has been shown to promote better health outcomes through improved cardiovascular management.23–25 However, implementation in Australian general practice has proved problematic: a 2020 survey found that 78% of Australian GPs use ACVRa, with just 45% reporting high assessment rates.26 In baseline audits of general practice in a recent Australian RCT, only 48% of patients aged 45–69 had ACVRa recorded in their notes.27 Thus, there appears to be only modest uptake of ACVRa in Australian general practice. Available literature demonstrates similarly poor utilisation of CVD risk assessment tools among primary care practitioners internationally, with reported rates of frequent/regular use as low as 17% in the USA28 and 32.4% in Ireland.29 In Belgium, 53% of GPs reported never using a global (absolute) CVD risk assessment tool.30 While a 2020 survey in the UK revealed high self-reported utilisation of CVD risk assessment among GPs across a number of CVR risk tools, how frequently these tools are used remains unclear.31

GPs continue to focus on individual risk factors (blood pressure and cholesterol) rather than absolute CVD risk.32 33 This represents a significant evidence-practice gap in CVD management.34 35 It has been estimated that only 7% of Australians aged 45–74 years attending general practice have had an ACVR score calculated.36 Qualitative Australian research suggests that implementing ACVR-based management in general practice is complex and may require multifaceted approaches.37 38 There remains significant scope for increasing the proportion of eligible patients receiving ACVRa.

With the disease burden of CVD projected to increase significantly in this decade,39 ‘increased efforts are needed to tackle these major risk factors’.40 GP registrars (specialist vocational trainees in general practice/family practice) are a group of particular interest given that registrars (during their 18 months of general practice-based training) comprise approximately 13% of current GP workforce by head count41 and are at a career stage when they are establishing clinical approaches and practices that may be long-lasting. Exploring the use of ACVRa by GP registrars will be valuable in determining uptake and understanding associations of ACVRa. An understanding of these and other contextual aspects of ACVR use will inform the multifaceted approaches required to increase the use of ACVRa in primary care, including during the vocational training of GPs.

In this study, we aimed to establish the prevalence and associations of GP registrars performing in-consultation ACVRa.

Registrar Clinical Encounters in Training

ReCEnT is an ongoing inception cohort study of GP registrars and the content of their consultations. It is a multisite study, conducted since 2010. The data for this analysis were collected from 2017 to 2018.

For this analysis, participants were GP registrars training with three of Australia’s nine then-current regional training organisations (RTOs) across three Australian states (New South Wales, Victoria and Tasmania) and the Australian Capital Territory. RTOs were government-funded, not-for-profit, geographically defined educational and training organisations. They delivered regular education sessions for their registrars and co-ordinated and managed registrars’ placement in individual training practices where training is conducted within an apprenticeship-like model with supervision by experienced GPs. Registrars spend 18 months to 2 years of their training in the practice environment. Though they have recourse to supervisor advice and assistance, they practice with considerable clinical autonomy. The participating RTOs were responsible for the training of 44% of Australian GP registrars during the period of data collection.42

The ReCEnT methodology has been reported elsewhere.43 Briefly, registrars collect data on 60 consecutive office-based consultations, at approximately the midpoint of each of three 6-month (full-time equivalent) general practice-based training terms. Data collection is via dedicated Case Report Forms (CRFs). The CRFs are composed of an extensive set of items which are ongoing. There is also capacity to add items exploring new topics for a limited number of rounds of data collection. In the case of ACVR, this item was included for three rounds of data collection in 2017 and 2018. ReCEnT is a routine component of registrars’ educational programme, with individual written feedback provided to prompt individual registrars’ reflection on their clinical and educational experiences.44 Registrars may consent for their data to be also used for research purposes.

In addition to the in-consultation data, registrar demographic data and training practice characteristics are collected (via questionnaire) 6 monthly.

Study population

The study population included all registrars from participating RTOs who were in one of their first three (6-month full-time equivalent) GP training terms.

Outcome

The outcome of interest was whether an ACVRa was performed.

Registrars were also asked to record whether the ACVRa outcome was categorised as low risk (<10% 5-year risk), moderate risk (10%–15%), high risk (>15%) or whether the risk was deemed to be ‘automatically clinically categorised as high risk’ (eg, if the patient had a pre-existing condition that categorised them as high risk).

This data were obtained via items included in the ReCEnT CRF for a limited number of rounds of data collection.

Independent variables

Independent variables related to registrar, practice, patient, consultation and consultation actions.

Registrar factors were age, gender, part-time or full-time status, training term, number of years since graduation, whether they worked at the practice previously, and if they qualified as a doctor in Australia. Practice factors included size of practice (number of full-time equivalent GPs), billing practice (does the practice bulk bill, ie, routinely do not charge the patient a consultation fee at the point of care), rurality (based on the Australian Standard Geographical Classification Remoteness Area classification),45 practice location Socio-Economic Index for Area (SEIFA-IRSD) relative index of disadvantage decile46 and RTO region. Patient factors were age, gender, identification as Aboriginal and/or Torres Strait Islander, non-English speaking background and patient/practice status (whether the patient was an existing patient of the practice, new to the registrar or new to the practice). Consultation factors were duration of consultation, number of problems/diagnoses addressed in each consultation, if the problem/diagnosis was new, and if the registrar sought information or assistance for diagnosis and/or management of the problem (if they consulted their supervisor and/or other sources of information). The consultation action factors were if medication was prescribed, pathology or imaging ordered, referrals made, follow-up organised and if any learning goals were generated by the registrar.

Statistical analysis

Analysis was at the level of the individual problem/diagnosis and performed on data from three rounds of data collection 2017–2018.

Analysis was confined to Aboriginal and Torres Strait Islander patients aged 35 years and older, and to non-Indigenous patients aged 45 years and older, reflecting National Vascular Disease Prevention Alliance (NVDPA) guidelines for ACVRa.14

Proportions of problems/diagnoses that involved an ACVRa were calculated with 95% CIs, adjusted for clustering of observations within registrars.

The main analyses used logistic regression, within the generalised estimating equations framework to account for repeated measures within registrars. An exchangeable working correlation structure was assumed.

Four models were constructed, all were with the outcome ACVRa performed/ACVRa not performed.

1. The first two models included ‘all problems/diagnoses’.

   The first model included patient-related, practice-related and registrar-related covariates. The second model included these same covariates plus consultation-related and consultation-related covariates. The rationale for the sequential models is that whether the registrar performs an ACVRa will plausibly be influenced by patient, registrar and practice factors, but evaluation of these influences may be compromised by inclusion in the multivariable model of factors operating once the consultation is progressing.

2. The third and fourth models included only specific problems/diagnoses for which an ACVRa could plausibly be indicated. These problems/diagnoses were determined by a panel of GP study investigators to be atrial fibrillation, chronic kidney disease, diabetes mellitus type I, diabetes mellitus type II, hyperlipidaemia, hypertension, obesity, smoking and ‘check-up’.

Univariable analyses were conducted on each covariate, with the outcome. Covariates with a univariable p<0.20 were considered for inclusion in the multiple regression model.

Once the model with all significant covariates was fitted, model reduction was assessed. Covariates which were no longer significant (at p<0.2) in the multivariable model were tested for removal from the model. If the covariate’s removal did not substantively change the resulting model, the covariate was removed from the final model. A substantive change to the model was defined as any covariate in the model having a change in the effect size (OR) of greater than 10%.

Diagnostic tests were conducted to assess goodness of fit, using the Hosmer-Lemeshow test for logistic models.

The regressions modelled the log-odds that ACVRa was performed during the consultation. Effects were expressed as ORs with 95% CI. Significance was declared at the conventional 0.05 level, with the magnitude and precision of effect estimates also used to interpret results.

Analyses were programmed by using STATA V.16.0 and SAS V.9.4.

Patient and public involvement

There was no patient involved. The participants in the ReCEnT study are registrars. Registrars were involved in the design of the study (in 2009) and participating registrars regularly provide feedback on the ongoing study. Results are disseminated to study participants in monthly GP training newsletters.

Associations of an ACVRa being performed

The characteristics associated with an ACVRa being performed for ‘all problems/diagnoses’ in the included age groups are presented in online supplemental table 1. The associations with an ACVRa being performed (logistic regression models) are presented in table 2.

Significant multivariable associations of an ACVRa being carried out included patient age 55–64 years (OR 1.19 (95% CI 1.00 to 1.41)), and age 65–74 years (OR 1.18 (95% CI 1.00 to 1.39). Patient age 75+ was associated with a considerably lesser odds of an ACVRa being carried out (OR 0.66 (95% CI 0.52 to 0.83)) (all compared with age <55 years). Female gender was associated with reduced odds of ACVRa being carried out (OR 0.61 (95% CI 0.54 to 0.68)). There was some evidence (p=0.10) that Aboriginal and/or Torres Strait Islander people in the general population may be more likely to receive ACVRa (OR 1.40 (95% CI 0.94 to 2.08)). There were also associations with variables related to continuity of care with reduced odds of ACVRa if the patient was new to the registrar (OR 0.65 (95% CI 0.57 to 0.75)) new to the practice (OR 0.24 (95% CI 0.15 to 0.38)); or the problem was new (OR 0.68 (95% CI 0.59 to 0.78)); and increased odds if personal follow-up was organised (OR 1.43 (95% CI 1.24 to 1.66)). Consultations in which an ACVRa was conducted were statistically significantly longer, but the effect size was small.

Regression diagnostics for the final model showed no violations of the assumptions of heteroscedasticity or normality. Goodness-of-fit tests indicated good model fit and there were no influential observations.

Associations with an ACVRa being performed for problems/diagnoses for which an ACVRa is indicated

The characteristics associated with an ACVRa being performed for problems/diagnoses for which an ACVRa is indicated are presented in online supplemental table 2. The associations with an ACVRa being performed (logistic regression models) are presented in table 3.

For the analysis of problems/diagnoses for which an ACVRa is indicated, findings were broadly similar to those for ‘all problems/diagnoses’, above. One difference was that the association with Aboriginal and/or Torres Strait Islander status was now significant at the p<0.05 level (OR 1.60 (95% CI 1.04 to 2.46)). There was also evidence of regional variability with two of the five regions having significantly greater odds of ACVRa performance (OR 1.69 (95% CI 1.16 to 2.48) and 1.52 (95% CI 1.10 to 2.10) compared with the referent region). The age groups 55–64 and 65–74 were no longer significantly different to the referent age group (<55). The effect size of the association with female gender was less (OR of 0.88 compared with 0.61) and no longer significant at the p<0.05 level (OR (95% CI 0.77 to 1.01), p=0.070).

Regression diagnostics for the final model showed no violations of the assumptions of heteroscedasticity or normality. Goodness-of-fit tests indicated good model fit and there were no influential observations.

Summary of main findings

Registrars conducted an ACVRa in patients 45 years or older (or 35 years and older in Aboriginal and/or Torres Strait Islander patients) for 2.5% of ‘all problems/diagnoses’ and, for problems/diagnoses for which an ACVRa is plausibly indicated, in 11.8% of instances.

We found that registrars are significantly less likely to conduct ACVRa in female patients, but this was attenuated when limited to the ‘at-risk problem/diagnosis’ population. In the ‘all problems/diagnoses’ population, there is some evidence that Aboriginal and/or Torres Strait Islander patients are more likely to have ACVRa carried out. The effect size increased and was statistically significant at the 0.05 level when confined to the ‘at-risk’ problems/diagnoses population. In the ‘all problems/diagnoses’ population, registrars are more likely to carry out ACVRa with increasing age of the patient (55–64 years and 65–74 years), but less likely after the age of 75 years. These findings are attenuated when confined to ‘at-risk’ problems, but those over 75 years of age were still significantly less likely to receive ACVRa. Follow-up being arranged to see the registrar (but not other GPs in the practice) is significantly associated with ACVRa. A patient being new to the practice, being new to the registrar, and the patient presenting with a new problem is associated with reduced odds of ACVRa. Association of ACVRa with increased consult duration is statistically significant but was of modest effect size.

Interpretation of findings and comparison with existing literature

The findings of frequency of ACVRa in our study are of limited import. Our data does not elucidate what proportion of patients for who an ACVRa is indicated have already received one at some time (and in a timely manner). Intuitively, 11.8% of appropriate problems/diagnoses prompting an ACVRa at individual presentations suggests registrars are proactively performing risk assessments, but the extent of potential for improvement is unclear. The findings regarding associations of performing an ACVRa, however, are relevant and potentially important to clinical GP and to GP vocational training:

Implications for education and training

These findings have implications for GP Training. While there is some evidence of ACVRa being employed effectively by registrars in the ‘at-risk’ problems/diagnoses population, in the wider population ACVRa of female patients is poorer—potentially contributing to mismanagement (either undermanagement or overmanagement). Also, ACVRa of Aboriginal and/or Torres Strait Islander patients could be improved to contribute to greater equity in CVD outcomes. A concerted effort to improve ACVRa performance in these groups within clinical practice is warranted. Improved education of registrars in provision of ACVRa must be prioritised to bridge evidence-practice gaps.35

Strengths/limitations

To our knowledge, this was the first study exploring ACVRa among GP registrars/trainees. The large number of consultations, high response rate for studies of GPs61 and geographical coverage (classifications of rurality from major city to remote) were strengths of this study. The findings have good generalisability to Australian GP vocational training and other apprenticeship-like practice-based GP training programmes internationally.62

A limitation is that, as the analyses are cross-sectional, association only (not causality) can be inferred.

A further limitation of our data is that it is a ‘snapshot’ of what occurred in the consultation, and we do not have patients’ medical histories or medicines regimens. The data being confined to single consultations, so not allowing calculation of proportion of patients having had an ACVRa within a specified time period, is thus a limitation of our study.

As the data are self-recorded, there is potential for social desirability bias, but this should be minimal given that the registrars record a large number of potential clinical actions in every consultation, so ACVRa would not be a particular area of focus.