Article Text

Download PDFPDF
How do cardiovascular risk prediction equations developed among 30–74 year olds perform in older age groups? A validation study in 125 000 people aged 75–89 years
  1. Suneela Mehta1,
  2. Rod Jackson1,
  3. Katrina Poppe1,
  4. Andrew J Kerr1,2,
  5. Romana Pylypchuk1,
  6. Sue Wells1
  1. 1Section of Epidemiology and Biostatistics, The University of Auckland, Auckland, New Zealand
  2. 2Cardiology Services, Middlemore Hospital, Auckland, New Zealand
  1. Correspondence to Dr Suneela Mehta, Section of Epidemiology and Biostatistics, Private Bag 92019, University of Auckland, Auckland 1142, New Zealand; suneela.mehta{at}


Background Cardiovascular disease (CVD) risk prediction equations are being used to guide risk management among increasingly older individuals. We examined the performance of recent equations, derived from a 2006 cohort including almost all New Zealanders aged 30–74 years, among older people.

Methods All New Zealanders aged 75–89 years in contact with state-funded health services in 2006 without prior CVD or heart failure and with complete predictor data were identified by anonymised individual-level linkage of eight national administrative health datasets. Baseline 5-year CVD risk was estimated using sex-specific New Zealand risk equations, and CVD hospitalisations or deaths occurring between 2007 and 2011 inclusive were ascertained. Performance was assessed with calibration plots and standard metrics.

Results Among 124 358 New Zealanders aged 75–89 years old, 30 152 CVD events were recorded during follow-up. Sex-specific equations derived from 30–74 year olds slightly underestimated CVD risk among women and slightly overestimated risk among men aged 75–89 years. Discrimination metrics were poor in both sexes and the risk equations explained only 9.4% of the variation in time to CVD event among women and 6.0% for men. In the 5-year age bands, progressively worsening underprediction in women, overprediction in men and poorer performance metrics were observed with increasing age.

Conclusion Entire-population CVD risk equations developed among 30–74 year olds do not perform well among older people. Existing risk algorithms developed from primarily middle-aged or early-retirement cohorts should be used with caution in those aged ≥75 years until carefully validated in narrow age bands to avoid masking poorer performance in older age groups.

  • cardiovascular disease
  • elderly
  • prevention
  • record linkage

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.


  • Contributors SM, RJ and SW planned the study. SM (and Grant Hanham) constructed the study dataset. SM analysed the data and SM, RJ, KP, AK, RP and SW were involved in interpreting the results. SM drafted the manuscript. SM, RJ, KP, AK, RP and SW critically revised the manuscript and approved the final version. SM is the guarantor for the study.

  • Funding This work was supported by the Health Research Council of New Zealand (grant numbers 11/800, 14/010 to SM) and the Stephenson Foundation (SW). The study funders/sponsors had no role in the study design, collection, analysis or interpretation of data.

  • Competing interests SW has received a grant from Roche Diagnostics for research outside the submitted work. SM, RJ, KP, AK and RP declare that they have no conflict of interest.

  • Patient consent for publication Not required.

  • Ethics approval This study was performed in accordance with the 1964 Declaration of Helsinki and is part of the VIEW research programme that was approved by the Northern Region Ethics Committee Y in 2003 (AKY/03/12/314), with subsequent annual reapproval by the national Multi-Region Ethics Committee since 2007 (MEC/01/19/EXP).

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data availability statement Data are available on reasonable request, subject to the required approvals. All data were obtained from the New Zealand Ministry of Health. Access to the statistical code and the dataset can be requested from the corresponding author at Requests will be granted after consideration by the VIEW research programme Governance Group, agreement by the New Zealand Ministry of Health and ethical approval by the New Zealand Multi-Region Ethics Committee.