School of Population Health

PREDICT in Primary Care

PREDICT is a web-based clinical tool used during patient consultations. It is based on a series of templates (203.7 KB PDF) that are filled in and submitted by healthcare providers (usually a general practitioner or practice nurse). The PREDICT clinical tool then provides the following information to the health provider:

  1. A predicted 5-year risk of a new CVD event
  2. The patient's Heart Forecast and Heart Age
  3. A summary of recommendations
  4. A detailed set of recommendations from evidence-based guidelines
  5. Patient advice described in lay language that can be printed or emailed directly to the patient.
    This includes links to National Heart Foundation and Diabetes New Zealand patient resources.


All PREDICT data are stored both in the patient’s electronic medical record held by their GP and on secure web servers operated by Enigma Solutions Ltd. Enigma developed the PREDICT software platform in partnership with the University of Auckland PREDICT/VIEW research group. The data are unidentifiable and not available to any other groups, including VIEW researchers, without written permission from the health provider using PREDICT software and without national ethical approval. At the University, the VIEW research team link these data to national health datasets (mortality, hospitalisations, drug prescribing, laboratory test claims and Primary Health Organisation enrolment) and regional datasets (Auckland laboratory test results) using a third-party encryption process to ensure the anonymity of all data. The linked data make up the PREDICT cohort study. This large, and growing, cohort is being used to develop New Zealand-specific risk prediction tools and assess the quality of CVD and diabetes care for patients. Research findings are disseminated in medical journals, at conferences and hui and as reports to health provider organisations, funders and planners.

Risk prediction

The risk prediction equation used for the primary care PREDICT cohort uses the Framingham predictors:

  • Age
  • Gender
  • Smoking
  • Diabetes
  • Systolic blood pressure
  • Total cholesterol/HDL

As well as the following New Zealand-adjusted Framingham predictors:

  • Ethnicity (Maori, Pacific Island, Indo-Asian)*
  • Family history of premature CVD*
  • Total cholesterol > 8mmol/L OR total cholesterol/HDL > 8*
  • Blood pressure > 170/100 mmHg*
  • Type 2 diabetes > 10 years OR with HbA1c > 64mmol/mol OR with microalbuminuria*

* once-only 5% added to 5-year risk

The outcomes used for risk prediction are all vascular or athersclerotic CVD hospitalisations and deaths: ischaemic CHD events including CABG and PCI, ischaemic and haemorrhagic cerebrovascular events, including TIAs, excluding trauma, PVD including arterial aneurysms and prcoedures unless specified non-athersclerotic causes, and CHF and cardiomyopathies unless specified non-ischaemic causes.


Research objectives

  1. Investigate the CVD predictive power (independent of other multiple risk factors) of a range of individual factors including family history, extreme blood pressure and lipid values, blood glucose and HbA1c, microalbuminuria, creatinine and estimated glomerular filtration rate (GFR)
  2. Develop risk prediction algorithms for multiple population groups (e.g. with and without CVD, the elderly, people with diabetes, Maori, Pacific and other high-risk populations) and for multiple vascular outcomes, for use in routine clinical practice and for modelling treatment effects
  3. Investigate the impact of drug treatment during follow-up on the accuracy of risk prediction
  4. Investigate the impact of competing risk on the accuracy of risk prediction.
  5. Describe the extent and equity of CVD risk screening and appropriateness of drug treatment in relation to estimated CVD risk; and, identify the degree of under- and over-treatment and their determinants
  6. Describe patterns of risk factor monitoring including blood glucose, HbA1c, lipid profile and renal function during follow-up by socio-demographic factors and CVD risk profile at entry
  7. Determine whether health service system performance of screening, initiation of treatment and long-term maintenance to treatment are improving over time

The PREDICT Cohort

We are currently analysing data from May 2012 in which the PREDICT primary care cohort included 272,682 patients (covering all ages and both patients with and without a history of CVD). The cohort had increased to approximately 380,000 by May 2013. The mean follow-up time to July 2012 is approximately 3 years (median approximately 2.5 years). About 25% of the cohort have been followed for more than 4 years.

Baseline numbers to May 2012 by ethnicity and gender
(ages 30-84 years, excluded if previous history of CVD)

  Men Women
TOTAL 128,419 102,978
European 61.933 51,089
Maori 15,036 13,731
Pacific 16,817 13,959
Indian 10,110 7,361
Other Asian 9,030 7,844
Other 2,395 1,590

First CVD events to July 2012
(ages 30-74 years, excluded if previous history of CVD)

  Non Fatal Fatal
TOTAL 7,377 615
CHD 49% 61%
Stroke 14% 24%
CHF 19% 10%
PVD 7% 5%
TIA 6% 0%
Other CVD 6% 0%

VIEW:PREDICT Participant Information Sheet