EHR-AGE: measuring biological age based on medical history data

Abstract Aging clocks have emerged as the primary tools for measuring biological age, providing useful tools for personalized medicine and aging interventions. However, in clinical practice, data for most of the current aging clocks (e.g., omics-based clocks) are not always available as they are too expensive or time-consuming. In contrast, patient medical histories are often readily available through electronic health record (EHR) systems. Here, we used a dataset that contained the diagnosis codes (based on the International Classification of Diseases, ICD-10) of the whole Hungarian population (approx. 9.5 million people) between 2010 and 2021. We trained machine learning algorithms to predict patients’ chronological age based on their ICD-10 diagnosis history of the previous 3 years. The final model (EHR-AGE) predicted the age of independent test samples with high performance (Pearson r = 0.89, MAE = 7.78 years). The age acceleration of the EHR-AGE model was associated with all-cause mortality, suggesting that EHR medical history-based age prediction can be a complementary measurement of biological age, and in the future, it can be a useful tool in clinical practice.