Читать книгу Diabetic Retinopathy and Cardiovascular Disease - Группа авторов - Страница 21

External validation of risk scores is essential because prediction models generally perform better within the cohort in which they were created than when applied to another population. A systematic review of cardiovascular risk prediction models for patients with type 2 diabetes found that just under a third of published risk scores had been validated in external study cohorts [31]. Where there are substantial differences between populations, the predictive model may need to be recalibrated. Large studies of pooled international observational cohorts have shown that hazard ratios for particular risk factors are similar in most Western and Asian populations (there are limited data for African or Latin American cohorts) [33]. However, rates of cardiovascular disease vary considerably in different populations and geographical regions for numerous reasons. Disease rates also vary significantly over time within 1 location. Therefore, average absolute risk of cardiovascular events varies and this variability needs to be factored into risk scores. Rather than continually re-creating risk scores in each population and periodically over time, models can be recalibrated as long as population-specific data is available to determine contemporary mean risk factor levels and rates of cardiovascular disease.

Statistical assessment of risk scores involves 2 key factors: discrimination and calibration. Discrimination is the ability of the tool to identify those who will develop the disease and those who will not. This is commonly measured using the area under the curve (AUC) on a receiver operating characteristic curve, which incorporates both sensitivity and specificity. A similar measure is the concordance statistic or “c-statistic” [32]. Values range from 0.5, indicating no discrimination, to 1.0, indicating perfect discrimination. Calibration describes the correlation between risk predicted by the tool and the observed event rate in the population. There are a few methods for assessing calibration, including the Hosmer-Lemeshow test, which compares mean predicted risk to observed outcome rates across deciles of the distribution of expected risks [32].