Development and validation of cardiovascular risk prediction equations in 76 000 people with known cardiovascular disease.

AIMS: Multiple health administrative databases can be individually linked in Aotearoa New Zealand, using encrypted identifiers. These databases were used to develop cardiovascular risk prediction equations for patients with known cardiovascular disease (CVD). METHODS AND RESULTS: Administrative health databases were linked to identify all people aged 18-84 years with known CVD, living in Auckland and Northland, Aotearoa New Zealand, on 1 January 2014. The cohort was followed until study outcome, death, or 5 years. The study outcome was death or hospitalization due to ischaemic heart disease, stroke, heart failure, or peripheral vascular disease. Sex-specific 5-year CVD risk prediction equations were developed using multivariable Fine and Gray models. A total of 43 862 men {median age: 67 years [interquartile range (IQR): 59-75]} and 32 724 women [median age: 70 years (IQR: 60-77)] had 14 252 and 9551 cardiovascular events, respectively. Equations were well calibrated with good discrimination. Increasing age and deprivation, recent cardiovascular hospitalization, Mori ethnicity, smoking history, heart failure, diabetes, chronic renal disease, atrial fibrillation, use of blood pressure lowering and anti-thrombotic drugs, haemoglobin A1c, total cholesterol/HDL cholesterol, and creatinine were statistically significant independent predictors of the study outcome. Fourteen per cent of men and 23% of women had predicted 5-year cardiovascular risk <15%, while 28 and 24% had ≥40% risk. CONCLUSION: Robust cardiovascular risk prediction equations were developed from linked routine health databases, a currently underutilized resource worldwide. The marked heterogeneity demonstrated in predicted risk suggests that preventive therapy in people with known CVD would be better informed by risk stratification beyond a one-size-fits-all high-risk categorization.

Using regionwide New Zealand health databases, methods of predicting hospitalization risk in patients with existing heart disease were developed. Using only data from health databases, it was possible to predict the risk accurately.Among patients with existing heart disease, the predicted risk varied markedly which could help improve preventive strategies.

Predicting cardiovascular risk from national administrative databases using a combined survival analysis and deep learning approach.

BACKGROUND: Machine learning-based risk prediction models may outperform traditional statistical models in large datasets with many variables, by identifying both novel predictors and the complex interactions between them. This study compared deep learning extensions of survival analysis models with Cox proportional hazards models for predicting cardiovascular disease (CVD) risk in national health administrative datasets. METHODS: Using individual person linkage of administrative datasets, we constructed a cohort of all New Zealanders aged 30-74 who interacted with public health services during 2012. After excluding people with prior CVD, we developed sex-specific deep learning and Cox proportional hazards models to estimate the risk of CVD events within 5 years. Models were compared based on the proportion of explained variance, model calibration and discrimination, and hazard ratios for predictor variables. RESULTS: First CVD events occurred in 61 927 of 2 164 872 people. Within the reference group, the largest hazard ratios estimated by the deep learning models were for tobacco use in women (2.04, 95% CI: 1.99, 2.10) and chronic obstructive pulmonary disease with acute lower respiratory infection in men (1.56, 95% CI: 1.50, 1.62). Other identified predictors (e.g. hypertension, chest pain, diabetes) aligned with current knowledge about CVD risk factors. Deep learning outperformed Cox proportional hazards models on the basis of proportion of explained variance (R2: 0.468 vs 0.425 in women and 0.383 vs 0.348 in men), calibration and discrimination (all P <0.0001). CONCLUSIONS: Deep learning extensions of survival analysis models can be applied to large health administrative datasets to derive interpretable CVD risk prediction equations that are more accurate than traditional Cox proportional hazards models.

Cardiovascular risk prediction in type 2 diabetes before and after widespread screening: a derivation and validation study.

BACKGROUND: Until recently, most patients with diabetes worldwide have been diagnosed when symptomatic and have high cardiovascular risk, meaning most should be prescribed cardiovascular preventive medications. However, in New Zealand, a world-first national programme led to approximately 90% of eligible adults being screened for diabetes by 2016, up from 50% in 2012, identifying many asymptomatic patients with recent-onset diabetes. We hypothesised that cardiovascular risk prediction equations derived before widespread screening would now significantly overestimate risk in screen-detected patients. METHODS: New Zealanders aged 30-74 years with type 2 diabetes and without known cardiovascular disease, heart failure, or substantial renal impairment were identified from the 400 000-person PREDICT primary care cohort study between Oct 27, 2004, and Dec 30, 2016, covering the period before and after widespread screening. Sex-specific equations estimating 5-year risk of cardiovascular disease were developed using Cox regression models, with 18 prespecified predictors, including diabetes-related and renal function measures. Equation performance was compared with an equivalent equation derived in the New Zealand Diabetes Cohort Study (NZDCS), which recruited between 2000 and 2006, before widespread screening. FINDINGS: 46 652 participants were included in the PREDICT-1° Diabetes subcohort, of whom 4114 experienced first cardiovascular events during follow-up (median 5·2 years, IQR 3·3-7·4). 14 829 (31·8%) were not taking oral hypoglycaemic medications or insulin at baseline. Median 5-year cardiovascular risk estimated by the new equations was 4·0% (IQR 2·3-6·8) in women and 7·1% (4·5-11·2) in men. The older NZDCS equation overestimated median cardiovascular risk by three times in women (median 14·2% [9·7-20·0]) and two times in men (17·1% [4·5-20·0]). Model and discrimination performance measures for PREDICT-1° Diabetse equations were also significantly better than for the NZDCS equation (eg, for women: R2=32% [95% CI 29-34], Harrell's C=0·73 [0·72-0·74], Royston's D=1·410 [1·330-1·490] vs R2=24% [21-26], C=0·69 [0·67-0·70], and D=1·147 [1·107-1·187]). INTERPRETATION: International treatment guidelines still consider most people with diabetes to be at high cardiovascular risk; however, we show that recent widespread diabetes screening has radically changed the cardiovascular risk profile of people with diabetes in New Zealand. Many of these patients have normal renal function, are not dispensed glucose-lowering medications, and have low cardiovascular risk. These findings have clear international implications as increased diabetes screening is inevitable due to increasing obesity, simpler screening tests, and the introduction of new-generation glucose-lowering medications that prevent cardiovascular events. Cardiovascular risk prediction equations derived from contemporary diabetes populations, with multiple diabetes-related and renal function predictors, will be required to better differentiate between low-risk and high-risk patients in this increasingly heterogeneous population and to inform appropriate non-pharmacological management and cost-effective targeting of expensive new medications. FUNDING: Health Research Council of New Zealand, Heart Foundation of New Zealand, and Healthier Lives National Science Challenge.

Prediction of first cardiovascular disease event in 2.9 million individuals using Danish administrative healthcare data: a nationwide, registry-based derivation and validation study.

Aims: The aim of this study was to derive and validate a risk prediction model with nationwide coverage to predict the individual and population-level risk of cardiovascular disease (CVD). Methods and results: All 2.98 million Danish residents aged 30-85 years free of CVD were included on 1 January 2014 and followed through 31 December 2018 using nationwide administrative healthcare registries. Model predictors and outcome were pre-specified. Predictors were age, sex, education, use of antithrombotic, blood pressure-lowering, glucose-lowering, or lipid-lowering drugs, and a smoking proxy of smoking-cessation drug use or chronic obstructive pulmonary disease. Outcome was 5-year risk of first CVD event, a combination of ischaemic heart disease, heart failure, peripheral artery disease, stroke, or cardiovascular death. Predictions were computed using cause-specific Cox regression models. The final model fitted in the full data was internally-externally validated in each Danish Region. The model was well-calibrated in all regions. Area under the receiver operating characteristic curve (AUC) and Brier scores ranged from 76.3% to 79.6% and 3.3 to 4.4. The model was superior to an age-sex benchmark model with differences in AUC and Brier scores ranging from 1.2% to 1.5% and -0.02 to -0.03. Average predicted risks in each Danish municipality ranged from 2.8% to 5.9%. Predicted risks for a 66-year old ranged from 2.6% to 25.3%. Personalized predicted risks across ages 30-85 were presented in an online calculator (https://hjerteforeningen.shinyapps.io/cvd-risk-manuscript/). Conclusion: A CVD risk prediction model based solely on nationwide administrative registry data provided accurate prediction of personal and population-level 5-year first CVD event risk in the Danish population. This may inform clinical and public health primary prevention efforts.

Ethnic differences in cardiovascular risk profiles among 475,241 adults in primary care in Aotearoa, New Zealand.

AIM: In Aotearoa, New Zealand, cardiovascular disease (CVD) burden is greatest among Indigenous Maori, Pacific and Indian people. The aim of this study was to describe CVD risk profiles by ethnicity. METHODS: We conducted a cross-sectional analysis of a cohort of people aged 35-74 years who had a CVD risk assessment in primary care between 2004 and 2016. Primary care data were supplemented with linked data from regional/national databases. Comparisons between ethnic groups were made using age-adjusted summaries of continuous or categorical data. RESULTS: 475,241 people (43% women) were included. Fourteen percent were Maori, 13% Pacific, 8% Indian, 10% Other Asian and 55% European. Maori and Pacific people had a much higher prevalence of smoking, obesity, heart failure, atrial fibrillation and prior CVD compared with other ethnic groups. Pacific and Indian peoples, and to a lesser extent Maori and Other Asian people, had markedly elevated diabetes prevalence compared with Europeans. Indian men had the highest prevalence of prior coronary heart disease. CONCLUSIONS: Maori and Pacific people experience the most significant inequities in exposure to CVD risk factors compared with other ethnic groups. Indians have a high prevalence of diabetes and coronary heart disease. Strong political commitment and cross-sectoral action to implement effective interventions are urgently needed.

Personalized Prediction of Cardiovascular Benefits and Bleeding Harms From Aspirin for Primary Prevention: A Benefit-Harm Analysis.

Background: Whether the benefits of aspirin for the primary prevention of cardiovascular disease (CVD) outweigh its bleeding harms in some patients is unclear. Objective: To identify persons without CVD for whom aspirin would probably result in a net benefit. Design: Individualized benefit-harm analysis based on sex-specific risk scores and estimates of the proportional effect of aspirin on CVD and major bleeding from a 2019 meta-analysis. Setting: New Zealand primary care. Participants: 245 028 persons (43.6% women) aged 30 to 79 years without established CVD who had their CVD risk assessed between 2012 and 2016. Measurements: The net effect of aspirin was calculated for each participant by subtracting the number of CVD events likely to be prevented (CVD risk score × proportional effect of aspirin on CVD risk) from the number of major bleeds likely to be caused (major bleed risk score × proportional effect of aspirin on major bleeding risk) over 5 years. Results: 2.5% of women and 12.1% of men were likely to have a net benefit from aspirin treatment for 5 years if 1 CVD event was assumed to be equivalent in severity to 1 major bleed, increasing to 21.4% of women and 40.7% of men if 1 CVD event was assumed to be equivalent to 2 major bleeds. Net benefit subgroups had higher baseline CVD risk, higher levels of most established CVD risk factors, and lower levels of bleeding-specific risk factors than net harm subgroups. Limitations: Risk scores and effect estimates were uncertain. Effects of aspirin on cancer outcomes were not considered. Applicability to non-New Zealand populations was not assessed. Conclusion: For some persons without CVD, aspirin is likely to result in net benefit. Primary Funding Source: Health Research Council of New Zealand.

Predicting Bleeding Risk to Guide Aspirin Use for the Primary Prevention of Cardiovascular Disease: A Cohort Study.

Background: Many prognostic models for cardiovascular risk can be used to estimate aspirin's absolute benefits, but few bleeding risk models are available to estimate its likely harms. Objective: To develop prognostic bleeding risk models among persons in whom aspirin might be considered for the primary prevention of cardiovascular disease (CVD). Design: Prospective cohort study. Setting: New Zealand primary care. Participants: The study cohort comprised 385 191 persons aged 30 to 79 years whose CVD risk was assessed between 2007 and 2016. Those with indications for or contraindications to aspirin and those who were already receiving antiplatelet or anticoagulant therapy were excluded. Measurements: For each sex, Cox proportional hazards models were developed to predict major bleeding risk; participants were censored at the earliest of the date on which they first met an exclusion criterion, date of death, or study end date (30 June 2017). The main models included the following predictors: demographic characteristics (age, ethnicity, and socioeconomic deprivation), clinical measurements (systolic blood pressure and ratio of total-high-density lipoprotein cholesterol), family history of premature CVD, medical history (smoking, diabetes, bleeding, peptic ulcer disease, cancer, chronic liver disease, chronic pancreatitis, or alcohol-related conditions), and medication use (nonsteroidal anti-inflammatory agents, corticosteroids, and selective serotonin reuptake inhibitors). Results: During 1 619 846 person-years of follow-up, 4442 persons had major bleeding events (of which 313 [7%] were fatal). The main models predicted a median 5-year bleeding risk of 1.0% (interquartile range, 0.8% to 1.5%) in women and 1.1% (interquartile range, 0.7% to 1.6%) in men. Plots of predicted-against-observed event rates showed good calibration throughout the risk range. Limitation: Hemoglobin level, platelet count, and body mass index were excluded from the main models because of high numbers of missing values, and the models were not externally validated in non-New Zealand populations. Conclusion: Prognostic bleeding risk models were developed that can be used to estimate the absolute bleeding harms of aspirin among persons in whom aspirin is being considered for the primary prevention of CVD. Primary Funding Source: The Health Research Council of New Zealand.

Performance of a Framingham cardiovascular risk model among Indians and Europeans in New Zealand and the role of body mass index and social deprivation.

Objectives: To evaluate a Framingham 5-year cardiovascular disease (CVD) risk score in Indians and Europeans in New Zealand, and determine whether body mass index (BMI) and socioeconomic deprivation were independent predictors of CVD risk. Methods: We included Indians and Europeans, aged 30-74 years without prior CVD undergoing risk assessment in New Zealand primary care during 2002-2015 (n=256 446). Risk profiles included standard Framingham predictors (age, sex, systolic blood pressure, total cholesterol/high-density lipoprotein ratio, smoking and diabetes) and were linked with national CVD hospitalisations and mortality datasets. Discrimination was measured by the area under the receiver operating characteristics curve (AUC) and calibration examined graphically. We used Cox regression to study the impact of BMI and deprivation on the risk of CVD with and without adjustment for the Framingham score. Results: During follow-up, 8105 and 1156 CVD events occurred in Europeans and Indians, respectively. Higher AUCs of 0.76 were found in Indian men (95% CI 0.74 to 0.78) and women (95% CI 0.73 to 0.78) compared with 0.74 (95% CI 0.73 to 0.74) in European men and 0.72 (95% CI 0.71 to 0.73) in European women. Framingham was best calibrated in Indian men, and overestimated risk in Indian women and in Europeans. BMI and deprivation were positively associated with CVD, also after adjustment for the Framingham risk score, although the BMI association was attenuated. Conclusions: The Framingham risk model performed reasonably well in Indian men, but overestimated risk in Indian women and in Europeans. BMI and socioeconomic deprivation could be useful predictors in addition to a Framingham score.

Is general practice identification of prior cardiovascular disease at the time of CVD risk assessment accurate and does it matter?

AIMS: To determine the accuracy of general practice recording of prior cardiovascular disease (CVD) at the time of CVD risk assessment and whether recording impacts on CVD management. METHODS: Prior CVD status entered at the time of a first CVD risk assessment from 2002-2015 was compared to prior ischaemic CVD hospitalisations from national datasets using anonymous linkage with an encrypted National Health Index identifier. Clinical factors associated with inaccurate recording of prior events were identified using multivariable logistic regression. The impact of recording accuracy was assessed by the dispensing of CVD preventive medications in the six months after first CVD risk assessment. RESULTS: Among 454,369 people aged 35-74 years who had CVD risk assessments, 30,924 (6.8%) had previously been admitted with ischaemic CVD. Of these people, only 61% were recorded as having prior CVD during risk assessment, with better recording for coronary and stroke events than for peripheral vascular procedures. Inaccurate primary care recording was more likely for younger people (<55 years), women, Maori, Pacific, Indian and Asian ethnic groups whereas smokers and people with diabetes were more likely to have prior CVD correctly identified. Over more than a decade, the odds of inaccurate recording during risk assessment increased [OR 1.09 (95% CIs 1.08-1.10)]. If prior CVD was entered at the time of risk assessment then dispensing of blood pressure-lowering, lipid-lowering, antiplatelet/anticoagulant medications, separately or together, was higher (86%, 85%, 83% and 69%, respectively) than if not recorded (70%, 60%, 60% and 43%). CONCLUSIONS: Overall, 39% of people with prior CVD hospitalisations were not recorded as having prior CVD when their CVD risk was first assessed in general practice. This was associated with inequities in evidence-based risk management. System-based measures are required for robust data sharing at the time of clinical decision making.

Can traditional risk factors explain the higher risk of cardiovascular disease in South Asians compared to Europeans in Norway and New Zealand? Two cohort studies.

OBJECTIVES: The objective was to prospectively examine potential differences in the risk of first cardiovascular disease (CVD) events between South Asians and Europeans living in Norway and New Zealand, and to investigate whether traditional risk factors could explain any differences. METHODS: We included participants (30-74 years) without prior CVD in a Norwegian (n=16 606) and a New Zealand (n=129 449) cohort. Ethnicity and cardiovascular risk factor information was linked with hospital registry data and cause of death registries to identify subsequent CVD events. We used Cox proportional hazards regression to investigate the relationship between risk factors and subsequent CVD for South Asians and Europeans, and to calculate age-adjusted HRs for CVD in South Asians versus Europeans in the two cohorts separately. We sequentially added the major CVD risk factors (blood pressure, lipids, diabetes and smoking) to study their explanatory role in observed ethnic CVD risk differences. RESULTS: South Asians had higher total cholesterol (TC)/high-density lipoprotein (HDL) ratio and more diabetes at baseline than Europeans, but lower blood pressure and smoking levels. South Asians had increased age-adjusted risk of CVD compared with Europeans (87%-92% higher in the Norwegian cohort and 42%-75% higher in the New Zealand cohort) and remained with significantly increased risk after adjusting for all major CVD risk factors. Adjusted HRs for South Asians versus Europeans in the Norwegian cohort were 1.57 (95% CI 1.19 to 2.07) in men and 1.76 (95% CI 1.09 to 2.82) in women. Corresponding figures for the New Zealand cohort were 1.64 (95% CI 1.43 to 1.88) in men and 1.39 (95% CI 1.11 to 1.73) in women. CONCLUSION: Differences in TC/HDL ratio and diabetes appear to explain some of the excess risk of CVD in South Asians compared with Europeans. Preventing dyslipidaemia and diabetes in South Asians may therefore help reduce their excess risk of CVD.