基于常规体检血液检测指标和衰弱指数的社区老年人群死亡风险预测模型构建及评价

Construction and evaluation of mortality risk predictive model for the community elderly population based on routine physical examination blood test indicators and frailty index

摘要:

目的 利用东莞市大朗镇社区老年衰弱随访队列，构建基于常规体检血液检测指标和衰弱指数的社区老年 人群死亡风险预测模型。方法 以基线衰弱评估结果和血液实验室检测结果为预测变量、以随访期内死亡为结局变量， 采用Cox回归分析构建风险预测模型，Kaplan-Meier生存曲线、ROC曲线下面积（AUC）进行模型评价并生成在线风险 评分计算工具。结果 多因素Cox分析结果显示存活组和死亡组的白细胞、血红蛋白、血小板分布宽度、中性粒细胞百分 比、红细胞分布宽度标准差、淋巴细胞百分比、总胆红素、谷丙转氨酶、葡萄糖、肌酐、甘油三酯、谷草转氨酶 12 项基线血 液检测结果差异均有统计学意义（P<0.05）。在测试集中，以此 12 项血液检测结果构建的FI-lab 6 年平均AUC为 0.826， 基于衰弱指数量表自我报告的FI-self-report 6 年平均AUC为 0.809，联合两者的FI-combined 6 a平均AUC为 0.834。结论 基于常规体检血液检测结果和衰弱指数构建的风险预测模型能够有效预测老年人死亡发生风险，在线风险评分计算工 具便于临床应用。

Abstract:

Objective To constructe a mortality risk predictive model for community elderly population based on routine physical examination blood test indicators and frailty index using a community-based geriatric frailty follow-up cohort in Dalang Town, Dongguan City. Methods A frailty risk score model was constructed using Cox regression analysis with baseline frailty assessment results and blood laboratory test results as predictor variables and survival status during the followup period as an outcome variable. Kaplan-Meier survival curves, and the area under the receiver operating characteristic curve (AUC) were used to evaluate the model, and generate an online risk score calculation tool. Results The results of multifactorial Cox analysis showed significant intergroup differences (P<0.05) between the surviving and dying groups for the 12 baseline blood test results for leucocyte, haemoglobin, platelet distribution width, neutrophil percentage, standard deviation of erythrocyte distribution width, lymphocyte percentage, total bilirubin, ghrelin, glucose, creatinine, triglyceride, and azelaic acid aminotransferase levels. In the test set, the 6-year mean AUC of the FI-lab constructed from these 12 blood test results was 0.826, the 6-year mean AUC of the FI-self-report based on self-report of the Frailty Index scale was 0.809, and the 6-year mean AUC of the FI-combined combining the two was 0.834. Conclusion The frailty risk prediction model constructed on the routine physical examination blood test results and frailty index can effectively predict the risk of mortality in the elderly, and an online risk score calculation tool facilitates clinical implementation.