Working Diagnosis.
Cardiovascular

PREDICT-CVD Risk

Five-year cardiovascular risk for primary prevention, calibrated for New Zealand. Comparison with Framingham, QRISK3, SCORE2, and ACC/AHA PCE below.

Consultation support tool. For primary prevention in patients without established CVD, CKD, or familial hypercholesterolaemia - these warrant separate assessment pathways. This tool approximates PREDICT-CVD using the published algorithm structure; confirm treatment decisions with predict.org.nz or the HeartsScore app.
Patient variables
Additional risk factors
Enter patient variables and
press Calculate
NZ Guidelines

Treatment thresholds - what the numbers mean

NZ Heart Foundation and BPAC use 5-year PREDICT risk as the primary decision threshold for statin initiation and blood pressure management, replacing the Framingham 10-year approach used until 2018.

<5%
Low
Lifestyle advice only. Reassess in 5 years (or 2-3 years if borderline).
5-10%
Moderate
Lifestyle intervention. Discuss statin benefit with patient. Shared decision-making.
10-15%
High
Medication usually recommended alongside lifestyle. BP and lipid treatment targets apply.
≥15%
Very High
Medication strongly recommended. Consider early specialist input. Treat all modifiable risk factors.
Lower thresholds for Māori, Pacific, and Indian patients: NZ guidelines recommend considering treatment at the 10-15% band (rather than waiting for ≥15%) for Māori, Pacific, and Indian patients aged under 55, given the higher burden of premature CVD in these groups and residual under-estimation of risk even after ethnicity adjustment.
International Context

How the tools compare - and why it matters

The same patient will receive meaningfully different risk estimates depending on which tool you use. Understanding what each was built on helps you explain the differences to patients and colleagues - and know when to trust the number in front of you.

Tool Origin / derivation Timeframe Endpoint Key variables NZ use
PREDICT-CVD
New Zealand NZ guideline
 ~500,000 NZ primary care patients; derived and validated 2012-2018. Includes Māori, Pacific, Indian, NZ European cohorts. Updated equations published BMJ 2018 (Pylypchuk et al.). 5 years Composite fatal + non-fatal CVD (MI, stroke, HF hospitalisation, PVD, CVD death) Age, sex, ethnicity, SBP (± treatment), TC:HDL, smoking, diabetes (T1/T2 separate), family Hx, deprivation (NZDI), AF, eGFR, prior CVD Recommended - BPAC, NZ Heart Foundation, MoH
QRISK3
United Kingdom UK guideline
 UK CPRD database, ~10 million patients, 2017 derivation (Hippisley-Cox et al., BMJ 2017). Includes English deprivation (Townsend score). Annual updates. 10 years Fatal + non-fatal MI or stroke Similar to PREDICT plus: SLE, RA, severe mental illness, erectile dysfunction, migraine, systemic corticosteroids, TIA, BMI, height, deprivation Not used - UK-population basis, Townsend deprivation not applicable
Framingham Risk Score
USA (Wilson 1998 / D'Agostino 2008)
 Framingham Heart Study cohort, predominantly white American participants. D'Agostino 2008 update (Circulation) extended to general CVD events. Widely used globally for decades. 10 years General CVD (MI, angina, HF, stroke, PVD, claudication) Age, sex, total cholesterol, HDL, SBP (± treatment), smoking, diabetes Superseded - used before PREDICT; tends to under-estimate risk in Māori and Pacific; no deprivation or ethnicity term
ACC/AHA PCE
USA US guideline
 Pooled Cohort Equations (Goff 2014, Circulation). Derived from multiple US cohort studies. Separate equations for White and African-American men and women. 10 years First ASCVD event (MI, stroke, coronary death) Age, sex, race (White/Black), TC, HDL, SBP (± treatment), smoking, diabetes Not used - "White" and "Black" race categories not applicable; known to over-estimate in some cohorts; no deprivation
SCORE2
Europe ESC guideline
 ESC 2021 - pooled European cohort data, ~677,000 individuals. Country-specific calibration by region (low, moderate, high, very-high risk). Replaces SCORE (fatal CVD only) with SCORE2 (fatal + non-fatal). 10 years Fatal + non-fatal CVD (MI, stroke) Age, sex, smoking, SBP, non-HDL cholesterol. Simplified - no diabetes, no family history Not used - no NZ calibration available; no ethnicity or deprivation term; notably simpler variable set
SCORE2-OP
Europe - older persons
 ESC 2021 extension for patients ≥70 years, where standard SCORE2 tends to over-estimate. Same derivation cohort, age-specific calibration. 5 or 10 years Fatal + non-fatal CVD As SCORE2 Not used
Understanding the differences

Why PREDICT gives a different number

Three questions GPs commonly encounter when comparing PREDICT to overseas tools.

PREDICT was derived from NZ primary care records with a median follow-up of around 4-5 years. A 10-year endpoint would have required extrapolation beyond the observed data, which introduces uncertainty. The 5-year horizon also fits more naturally into GP review cycles and makes the benefit of treatment more tangible to patients - "reducing your risk over the next 5 years from 18% to 10%" is easier to grasp than "over 10 years."

For rough comparison with 10-year tools: multiply the 5-year PREDICT risk by approximately 1.75-2.0 (the relationship is non-linear and higher-risk patients have proportionally higher 10-year risk). So a PREDICT 5-year risk of 15% equates very roughly to a 10-year risk of 25-30%.

PREDICT was explicitly derived and validated in Māori, Pacific, Indian, and NZ European cohorts separately. The ethnicity terms in the equation reflect residual risk that isn't explained by the shared risk factors - meaning that at identical age, BP, cholesterol, and smoking status, Māori patients have meaningfully higher observed CVD event rates.

The adjusted hazard ratios published by Pylypchuk (2018) are approximately +50% for Māori and +30% for Pacific and Indian patients relative to NZ European, after controlling for all other factors. This is the main reason PREDICT gives a higher number than Framingham for the same patient - Framingham has no ethnicity term at all.

Importantly, even after these adjustments, there is evidence of residual under-estimation in younger Māori patients, which is why guidelines recommend lower treatment thresholds for this group.

Socioeconomic deprivation predicts CVD events independently of the standard clinical risk factors - probably through mechanisms including chronic stress, reduced health-seeking, poorer medication adherence, higher dietary sodium and ultra-processed food intake, and environmental exposures. The NZ Deprivation Index (NZDep) is a census-derived area-based measure scored 1-10 by meshblock.

In the PREDICT equations, higher deprivation contributes a modest but statistically significant additional risk. The practical implication: two patients with identical clinical variables, one in decile 2 and one in decile 9, will have modestly different predicted risks. This is clinically relevant when discussing how lifestyle and social circumstances interact with pharmacological risk reduction.

QRISK3 includes the Townsend deprivation score (England), making it the closest overseas equivalent in this regard. Framingham, SCORE2, and the ACC/AHA PCE include no deprivation variable.

PREDICT is a primary prevention tool. Do not use it when:

  • Established CVD - prior MI, CABG, PCI, stroke or TIA, peripheral arterial disease. These patients are at high risk by definition and go directly onto secondary prevention pathways.
  • Familial hypercholesterolaemia (FH) - lifetime lipid exposure makes standard 10-year calculators unreliable. Refer to FH-specific guidelines (Dutch Lipid Clinic Network score, Simon Broome criteria).
  • eGFR <30 (CKD stages 4-5) - risk is very high by definition; CKD-specific CVD management pathways apply.
  • Type 1 diabetes with end-organ disease - high-risk category treated outside the PREDICT threshold model.
  • Age <35 or >74 - outside PREDICT's validated age range. Clinical judgement applies; some guidelines extend to 75 with caution.

For age >74, consider that treatment decisions become more nuanced (benefit vs. polypharmacy burden, falls risk with antihypertensives, statin myopathy risk). Shared decision-making and geriatric frailty assessment become more relevant than the raw number.

QRISK3 is more complex, incorporating 20+ variables including SLE, RA, severe mental illness, erectile dysfunction, migraine, and corticosteroid use. Its C-statistic (discriminatory ability) is marginally better than QRISK2 in UK populations. But "more variables" doesn't automatically mean "better for your patient."

A risk model is only as good as its calibration in the population you're applying it to. QRISK3 was derived and validated exclusively in UK CPRD populations. Its deprivation term uses the English Townsend score (not NZDep). Its ethnicity categories map to UK census categories, which don't map cleanly onto NZ priority populations. There is no NZ calibration available.

The key advantage of PREDICT in New Zealand isn't its C-statistic - it's that it was built on NZ patients, including Māori and Pacific cohorts who are systematically under-represented in overseas derivation studies. A well-calibrated simpler model will always outperform a complex mis-calibrated one.

References
  1. Pylypchuk R, Wells S, Kerr A, et al. Cardiovascular disease risk prediction equations in 400 000 primary care patients in New Zealand: derivation and validation. BMJ. 2018;361:k1668. doi:10.1136/bmj.k1668
  2. BPAC NZ. Cardiovascular risk assessment: using PREDICT in primary care. Best Practice Journal. Updated 2023. Available at: bpac.org.nz
  3. NZ Heart Foundation. Cardiovascular Risk Assessment and Management for Primary Care. Updated 2021.
  4. D'Agostino RB Sr, Vasan RS, Pencina MJ, et al. General cardiovascular risk profile for use in primary care. Circulation. 2008;117(6):743-753.
  5. Goff DC Jr, Lloyd-Jones DM, Bennett G, et al. 2013 ACC/AHA guideline on the assessment of cardiovascular risk. Circulation. 2014;129(25 Suppl 2):S49-73.
  6. SCORE2 working group and ESC Cardiovascular risk collaboration. SCORE2 risk prediction algorithms. Eur Heart J. 2021;42(25):2439-2454.
  7. Hippisley-Cox J, Coupland C, Brindle P. Development and validation of QRISK3 risk prediction algorithms to estimate future risk of cardiovascular disease. BMJ. 2017;357:j2099.
  8. Wells S, Kerr A, Broad J, et al. PREDICT-CVD: validating the primary prevention algorithm in New Zealand primary care. Heart. 2012;98(16):1217-1222.