Abstract: Objective To study the principal component analysis and prediction model of key test indicators in component transfusion for liver disease. Methods A retrospective study was conducted, collecting data from patients with liver disease and those without, who underwent component transfusion from January 2017 to December 2022. Patients were categorized based on the types of transfusion received: suspended red blood cell transfusion, virus-inactivated frozen plasma transfusion, and single-donor platelet transfusion. Genera patient information and pre-transfusion laboratory indicators were gathered, including hemoglobin (Hb), hematocrit (HCT), platelet count , coagulation function indicators, liver function indicators, and transfusion conditions. Differences in these indicators between liver disease and non-liver disease groups were analyzed using T-tests and variance analysis. The suitability of factor analysis was confirmed by the KMO test, Bartlett's sphericity test, and Scree Test. Principal component analysis (PCA) was utilized to observe the variance contribution of each indicator and evaluate their correlations. Receiver operating characteristic (ROC) curve analysis was performed to assess the predictive value of each test indicator for different component transfusions. Results A total of 96 liver disease patients and 216 non-liver disease patients were included in the study. Among the liver disease patients, 57.30% received plasma transfusion (55/96 cases), while 54.20% of the non-liver disease patients received red blood cell transfusion (117/216 cases). The average Hb levels were 70.61 g/L for liver disease patients and 82.82 g/L for non-liver disease patients HCT levels averaged 20.80% and 24.47%, alanine aminotransferase(ALT) were 45.94 U/L and 25.43 U/L, and total bilirubin(TBil) levels were 44.38 μmol/L and 19.31 μmol/L, respectively, for liver disease and non-liver disease groups. These four indicators showed significant differences between the groups (P<0.05). In the plasma transfusion group, the average Hb levels were 73.45 g/L for liver disease patients and 111.43 g/L for non-liver disease patients. HCT levels averaged 21.70% and 31.06%, ALT levels were 59.33 U/L and 28.33 U/L, aspartate aminotransferase(AAT) levels were 44.35 U/L and 22.52 U/L, and INR values were 1.43 and 1.07, respectively, for liver disease and non-liver disease patients. These indicators also showed significant differences (P<0.05). In the platelet transfusion group, the average platelet counts were 36.70×10⁹/L for liver disease patients and 50.76×10⁹/L for non-liver disease patients, ALT levels were 54.20 U/L and 31.19 U/L, PT was 15.95 s and 12.98 s, APTT was 54.42 s and 29.90 s, and INR values were 1.36 and 1.11, respectively, for liver dosease and non-liver disease patients. These five indicators showed significant differences (P<0.05). PCA revealed that the primary and secondary components of pre-transfusion indicators in liver disease patients were blood and liver function indicators, respectively, whereas in non-liver disease patients, the primary components were liver function and coagulation indicators. ROC curve analysis demonstrated that the area under the curve(AUC) for HCT in the red blood cell transfusion group was 0.912; in the plasma transfusion group, the AUCs for INR and PT were 0.964 and 0.953, respectively. In the platelet transfusion group, the AUC for INR was 0.938. Conclusion This study establishes a foundation for correlation analysis and predictive modeling of various pre-transfusion indicators, particularly aiding in the selection of appropriate component transfusions for liver disease patients.

Key words: Liver disease, Component transfusion, Detection indicators, Principal component analysis (PCA), Predictive model