However, we found that even among similar risk groups, defined by established risk factors, risk variation can fluctuate significantly depending on how that group is defined, pointing to the need for more global assessments of risk that consider

multiple dimensions of risk. Typically, baseline risk is used to identify learn more optimal target groups for intervention, but the variability in risk is not considered. We show that in addition to baseline risk, risk dispersion is also an important consideration that can influence the benefit revised from a prevention intervention. We found that prioritizing target populations using an empirically derived cut-off would result in greater population benefit compared to single risk factor targets, even when

a similar proportion of the population would be targeted. The empirical risk cut-point we derived corresponds to a ‘moderate risk’ category according to existing individual risk calculators (Canadian Task Force on Preventive Health Care, 2012); however, these risk classifications were not statistically derived based on maximizing treatment benefit. This underscores the importance of improving who we target and using tools to ensure Ibrutinib in vitro our prevention strategies are appropriate for both the level and dispersion of risk in the population. Increasingly, the use of multivariate risk others algorithms are being encouraged to improve identification of individuals at risk by examining multiple dimensions of risk, but also to provide a more efficient way of a staged or multi-step screening approach at the individual level (Buijsse et al., 2011, Canadian Task Force on Preventive Health Care, 2012 and Tabak et al., 2012). A particularly novel contribution of this study is that

it provides a mechanism by which these principles can be applied to the population level, beyond individual risk screening tools that have been recommended to guide clinical prevention strategies (Buijsse et al., 2011). These algorithms are difficult to apply at the population level because of their reliance on detailed clinical measures; data that rarely exist at the population level. In addition, these models were designed to be used for individual clinical decision-making and not for population risk assessment. To date, a population risk algorithm that can be applied to existing self-reported data has not yet been validated or used for individual risk assessment. A recent systematic review of all diabetes risk scores and models published in 2011 found that of over 90 existing diabetes risk tools, DPoRT was the only tool built to inform population intervention strategies for diabetes (Noble et al., 2011).