Age Estimation From Blood Test Results Using a Random Forest Model.

Abstract

Background and objectives: From a preventive medicine perspective, this study aims to clarify the role of screening data in aging and health problems by estimating age from screening data and verifying the number of data items required in widely used screening tests.

Materials and methods: A random forest model was applied to 11554 men and women (3043 and 8511, respectively) aged 0-95 years who underwent screening tests (60 blood tests, 8 urine tests and 2 saliva tests) between February 2020 and August 2023. All analyses were conducted in Python 3.10.12.

Results: Using all 71 items including gender, a high accuracy of R² = 0.7010 was achieved with 9243 training datasets (80% of total). R² decreased slightly to 0.6937 when data items were reduced to 15 by removing less important variables. When datasets numbered fewer than 800 or data items fewer than 7, R² fell below 0.6. Notably, postmenopausal women tended to have higher estimated ages compared to premenopausal women.

Conclusions: Age estimation from blood data using the random forest model (blood age) is sufficiently precise for assessing physical aging state. Blood age, as well as other biological ages estimated from various omics estimators, was shown to be a very promising method for exploring the problems of aging such as metabolic syndrome and frail syndrome.