Performance of risk prediction scores for cardiovascular mortality in older persons: External validation of the SCORE OP and appraisal

• Published in: PloS one 2020-04, Vol.15 (4), p.e0231097-e0231097
• Main Authors: Piccininni, Marco, Rohmann, Jessica L, Huscher, Dörte, Mielke, Nina, Ebert, Natalie, Logroscino, Giancarlo, Schäffner, Elke, Kurth, Tobias
• Format: Article
• Language: English
• Subjects: Biology and Life Sciences; Calibration; Cardiovascular disease; Cardiovascular diseases; Cholesterol; Database industry; Diabetes; Health risks; Heart attacks; Insurance; Insurance companies; Insurance industry; Mathematical analysis; Medicine and Health Sciences; Middle age; Mortality; Myocardial infarction; Neurosciences; Older people; People and Places; Population; Population studies; Public health; Public health movements; Risk assessment
• ISSN: 1932-6203; 1932-6203
• DOI: 10.1371/journal.pone.0231097

European guidelines recommend the use of the Systematic COronary Risk Evaluation (SCORE) to assess 10-year risk of fatal cardiovascular events in people aged 40 to 65. The SCORE Older Persons (SCORE OP, 5-year and 10-year versions) was recently developed for people aged 65 or older. We assessed the performance of these risk scores in predicting fatal cardiovascular events in older persons in Berlin. Data from the Berlin Initiative Study (BIS), a prospective, population-based study of older persons recruited from a German public health insurance company database were used. 1,657 participants aged 70 or older without reported previous myocardial infarction were included. We assessed calibration by comparing predicted risks to observed (for 5-year versions, 5y) or projected (for 10-year versions) probabilities. During follow-up (median: 4.8 years), 118 cardiovascular deaths occurred. The calibration assessment of the SCORE OP-H 5y and SCORE OP-L 5y equations revealed 2.1- and 1.5-fold overestimation. Comparing 10-year versions, the SCORE OP showed better discrimination ability compared to the SCORE (C-indices of around 0.80 compared to 0.72) and the SCORE for high-risk regions showed the best calibration (chi-square = 29.68). The SCORE OP overestimated the true risk; 519 and 677 events were predicted using the low-risk and high-risk region SCORE OP equations compared to 397 to 399 events projected based on BIS follow-up data (predicted/actual ratios of 1.3 and 1.7). Given the low transportability of the SCORE OP observed in our population, we caution against its use in routine clinical practice until further information is available to avoid possible overtreatment among older persons in Berlin.