Mining biomarkers from routine laboratory tests in clinical records associated with air pollution health risk assessment

Clinical laboratory in hospital can produce amounts of health data every day. The purpose of this study was to mine biomarkers from clinical laboratory big data associated with the air pollution health risk assessment using clinical records. 13, 045, 629 clinical records of all 27 routine laboratory tests in Changsha Central Hospital, including ALB, TBIL, ALT, DBIL, AST, TP, UREA, UA, CREA, GLU, CK, CKMB, LDL-C, TG, TC, HDL-C, CRP, WBC, Na, K, Ca, Cl, APTT, PT, FIB, TT, RBC and those daily air pollutants concentration monitoring data of Changsha, including PM2.5, PM10, SO2, NO2, CO, and O3 from 2014 to 2016, were retrieved. The moving average method was used to the biological reference interval was established. The tests results were converted into daily abnormal rate. After data cleaning, GAM statistical model construction and data analysis, a concentration-response relationship between air pollutants and daily abnormal rate of routine laboratory tests was observed. Our study found that PM2.5 had a stable association with TP (lag07), ALB (lag07), ALT (lag07), AST (lag07), TBIL (lag07), DBIL (lag07), UREA (lag07), CREA (lag07), UA (lag07), CK (lag 06), GLU (lag07), WBC (lag07), Cl (lag07) and Ca (lag07), (P < 0.05); O3 had a stable association with AST (lag01), CKMB (lag06), TG (lag07), TC (lag05), HDL-C (lag07), K (lag05) and RBC (lag07) (P < 0.05); CO had a stable association with UREA (lag07), Na (lag7) and PT (lag07) (P < 0.05); SO2 had a stable association with TP (lag07) and LDL-C (lag0) (P < 0.05); NO2 had a stable association with APTT (lag7) (P < 0.05). These results showed that different air pollutants affected different routine laboratory tests and presented different pedigrees. Therefore, biomarkers mined from routine laboratory tests may potentially be used to low-cost assess the health risks associated with air pollutants.