• Health policy
• effectiveness
• developing countries
• social epidemiology
• preventive medicine
• public health
• heart disease
• eastern Europe

In this issue, Manuel et al1 suggest that multivariate risk algorithms, widely used by clinicians, have great potential for the assessment of population risk and optimisation of population-based health planning. We agree with this proposition, but argue that this algorithm-based approach needs to consider the local context, particularly when transferring the risk instruments derived for Western populations to non-Western settings.

Traditionally, the examination of population risk has focused on the outcome component of the exposure–outcome relationship (often using historical estimates of disease incidence to predict future trends)—an approach which is, essentially, retrospective. However, true risk prediction should be prospective and provide a reliable estimate of future outcome probability, which is based on the currently observed exposure to multiple risk factors. The additional benefits of assessing population risk with the algorithms based on quantitative estimates of the exposure–outcome association, which can be used when reliable incidence data are unavailable, when the baseline risk varies considerably …