基于决策树的广州地区献血人群抑郁快速筛查模型的构建研究

Construction of a rapid depression screening model for blood donors in Guangzhou based on decision tree

目的对全血献血者人群中抑郁流行病学情况进行调查并作相关因素分析,在此基础上构建快速有效的献血者人群抑郁筛查模型。方法2020年5~8月,在本中心各街头献血点捐献全血的献血者共计13015名,指导其完成包含社会人口学信息及抑郁筛查量表(PHQ-9)的匿名电子问卷调查,PHQ-9量表评分10分及以上判定为筛查抑郁。用SPSS 26.0做Logistic回归分析抑郁相关因素并建立2级决策树,父节点/子节点中的最小样本数为30/10,进行10折交叉验证,剪枝PHQ-9的题项,形成献血者抑郁的快速筛查模型。结果13015名街头全血献血者中,364(2.80%)人的PHQ-9得分≥10分。年龄18~29岁(P<0.05)、未婚(P<0.05)、家庭人均年收入低于5万元的献血者(P<0.05)更易出现抑郁。决策树分析发现,81.96%PHQ-9得分“<10分”及3.85%“≥10分”的献血者被包含在PHQ-9的第6、2和4题项形成的两个终端节点中。由此3个题项构建出一个快速筛查模型,经10倍交叉法验证,该模型的误判率仅为1.70%。结论广州地区全血献血者人群中基于PHQ-9筛查的抑郁率为2.80%(95%CI:2.52%~3.09%),献血次数与初筛抑郁不相关。基于PHQ-9的第6、2和4题项可建立1个快速准确的献血者抑郁快速筛查模型。

Objective To investigate the prevalence of depression in blood donors and analyze the related factors,so as to develop a rapid depression screening model for blood donors.Methods A total of 13015 street whole blood donors in Guangzhou Blood Center during May to August,2020 filled in an anonymous e-questionnaire,including social demography information and the Patient Health Questionnaire-9 before donation.The cut-off value for detecting depression was 10.Logistic regression by SPSS 26.0 was used to analyze depression related factors.2-level decision tree with 30/10 as the minimum number of cases in parent/child node,10-fold cross validation was used to cut items of PHQ-9 to form the depression screening model.Results 364 out of 13015(2.80%)street whole blood donors reported a score≥10.Donors with 18-29 years old(P<0.05),unmarried(P<0.05),less than 50000 RMB household income per year(P<0.05)were more prone to depression.81.96%donors in“<10 scores”group,while 3.85%donors in“≥10 scores”group were in two terminal nodes formed by Item-6,2 and 4 of PHQ-9.After verification by the 10 fold crossover method,the estimated misclassification risk of the model was 1.7%.Conclusion The screening prevalence of depression based on PHQ-9 in Guangzhou blood donors was 2.8%(95%CI:2.52%-3.09%).Donation frequency was not re-lated to depression.A rapid and efficient depression screening model for blood donors based on item-6,2 and 4 of PHQ-9 was developed.