MACHINE LEARNING-BASED ROUTINE LABORATORY VARIABLES SCREENING FOR ONE-YEAR COGNITIVE AND FUNCTIONAL DECLINE IN INDIVIDUALS AGED 75+ YEARS: THE PIETÀ STUDY

Abstract

Background: Cognitive and functional decline are important health problems in older adults. Objectives: To investigate Machine Learning (ML) algorithms of routine laboratory variables in predicting cognitive and functional decline in a population-based sample aged 75+ years within one-year. Methods: 132 individuals were selected from a population-based project. Functional and cognitive performances were evaluated at baseline and one year after. Results of routine laboratory tests obtained at baseline were used to generate three ML algorithms. Results: Random forest (RF), including triglycerides, glucose, hematocrit, RDW, albumin, hemoglobin, globulin, HDL, TSH, creatinine, lymphocyte, erythrocyte, platelet/leucocyte (PLR) and neutrophil/leucocyte (NLR) ratios, ALT, leukocyte, LDL, cortisol, GGT and eosinophil, showed the best performance to predict the cognitive decline (accuracy = 0.80). For functional decline (accuracy =0.92), the most important RF variables were platelet, PLR and NLR, hemoglobin, globulin, cortisol, RDW, glucose, basophil, B12 vitamin, creatinine, GGT, ALT, AST, eosinophil, hematocrit, erythrocyte, triglycerides, HDL and monocyte. Conclusions: Our results suggest that ML presents a good accuracy to predict cognitive and functional decline in oldest-old subjects using routine laboratory variables.