脑出血患者家属重症监护室后综合征风险预测模型构建

Construction of risk prediction model for post intensive care syndrome in family members of patients with intracerebral hemorrhage

目的构建脑出血患者家属重症监护后综合征(PICS-F)风险预测模型,并对模型预测效能进行评估。方法采用便利抽样法,选取2020-05-01-2022-03-31厦门大学附属东南医院重症监护室(ICU)治疗的脑出血患者233名家属作为研究对象,根据是否发生PICS-F将研究对象分为发生组和未发生组,通过二元logistic回归分析确认影响因素后构建风险预测模型,并通过受试者工作特征(ROC)曲线验证预测模型的有效性。结果233名研究对象中,172名(73.82%)发生了PICS-F,61名(26.18%)未发生PICS-F。单因素分析结果显示,PICS-F发生组年龄(t=2.252,P=0.025)、女性占比(χ^(2)=4.165,P=0.041)、ICU入住时长(t=2.307,P=0.022)和急性生理学及慢性健康状况Ⅱ(APACHEⅡ)评分(t=2.180,P=0.030)均大于未发生组,格拉斯哥昏迷指数(GCS)评分(t=3.343,P=0.001)小于未发生组,差异均有统计学意义;二元logistic回归分析结果显示,年龄(OR=1.188,95%CI为1.114~3.266,P<0.001)、性别(OR=2.981,95%CI为1.073~8.288,P=0.036)、APACHEⅡ评分(OR=2.838,95%CI为1.670~5.043,P=0.030)、ICU入住时长(OR=3.067,95%CI为1.033~9.108,P=0.044)和格拉斯哥昏迷指数(GCS)评分(OR=0.642,95%CI为0.549~0.951,P=0.028)均为脑出血患者PICS-F的独立影响因素。构建脑出血患者PICS-F风险预测模型为P=1/{1+exp〔-(13.647+0.172×年龄+1.092×性别+0.609×APACHEⅡ+1.121×ICU入住时长-0.443×GCS评分)〕}。模型预测曲线下面积(AUC)为0.821(95%CI为0.656~0.903),最佳预报截断值为0.534时,模型预测灵敏度为68.60%,特异度为88.52%。结论成功构建了脑出血患者PICS-F风险预测模型,经验证模型具有较好预测效能。

Objective To establish a risk prediction model for post intensive care syndrome-family(PICS-F)members of pa-tients with intracerebral hemorrhage,and the to evaluate the prediction efficiency of the model.Methods Totally 233 family members of intracerebral hemorrhage patients in the intensive care unit(ICU)of Dongnan Hospital of Xiamen Uni-versity from May 1,2020to March 31,2022were selected by the convenient sampling method as the research objects.All of them were divided into the occurrence group and the non-occurrence group according to whether PICS-F occurred.The risk prediction model was constructed after the influencing factors were confirmed by binary logistic regression analy-sis,and the the efficiency of the prediction model was verified by the receiver operating characteristic curve(ROC).Re-sults Of 233subjects,172(73.82%)had PICS-F and 61(26.18%)had non PICS-F.Univariate analysis showed that age(t=2.252,P=0.025),female proportion(χ^(2)=4.165,P=0.041),length of stay in ICU(t=2.307,P=0.022),and acute physiology and chronic health evaluationⅡ(APACHEⅡ)scoring system(t=2.180,P=0.030)in PICS-F group were higher than those in the non-PICS-F group,while Glasgow coma scale(GCS)score(t=3.343,P=0.001)was lower than that in the non-PICS-F group,the differences were statistically significant.Binary logistic regression showed that age(OR=1.188,95%CI:1.114-3.266,P<0.001),gender(OR=2.981,95%CI:1.073-8.288,P=0.036),APACHEⅡ(OR=2.838,95%CI:1.670-5.043,P=0.030),length of stay in ICU(OR=3.067,95%CI:1.033-9.108,P=0.044)and GCS score(OR=0.642,95%CI:0.549-0.951,P=0.028)were independent influencing fac-tors of PICS-F in intracerebral hemorrhage patients.The PICS-F risk prediction model for intracerebral hemorrhage patients:P=1/{1+exp[-(13.647+0.172×Age+1.092×Gender+0.609×APACHEⅡ+1.121×length of stay in ICU-0.443×GCS score)]}.The area under curve(AUC),optimal value,sensitivity and specificity of the prediction model were 0.821(95%CI:0.656-0.903),0.534,68.60%and 88.52%,respectively.Conclusion The PICS-F risk prediction model for intracerebral hemorrhage patients is successfully constructed,which has high predictive value for clinical application.