刘奕男, 李锐, 黄俊, 赵琬. 两种糖尿病筛查模型在社区应用的比较与评估[J]. 环境与职业医学, 2010, 27(3): 127-130.
引用本文: 刘奕男, 李锐, 黄俊, 赵琬. 两种糖尿病筛查模型在社区应用的比较与评估[J]. 环境与职业医学, 2010, 27(3): 127-130.
LIU Yi-nan , LI Rui , HUANG Jun , ZHAO Wan . Evaluation on Effect of Two Screening Models for Diabetes[J]. Journal of Environmental and Occupational Medicine, 2010, 27(3): 127-130.
Citation: LIU Yi-nan , LI Rui , HUANG Jun , ZHAO Wan . Evaluation on Effect of Two Screening Models for Diabetes[J]. Journal of Environmental and Occupational Medicine, 2010, 27(3): 127-130.

两种糖尿病筛查模型在社区应用的比较与评估

Evaluation on Effect of Two Screening Models for Diabetes

  • 摘要: 目的 探索适合社区居民糖尿病筛查经济且高效的方法。

    方法 在 2007年上海市闵行区社区 40岁以上无糖尿病史人群流行病学调查资料的基础上, 建立危险因素评分模型及 BP人工神经网络模型, 并比较两者的筛查效度和效益。训练组(1 970人)用于危险因素评分模型及 BP人工神经网络模型的建立; 测试组(1 977人)用于验证两种模型筛查的真实性和可靠性, 并对两种模型的效率和效益进行比较。

    结果 当以危险因素的累计分值 < 27作为判别阈值时, 危险因素评分模型的灵敏度为 63.64%, 特异度为 72.28%; 当以网络输出值为 ≥ 0.16作为判别阈值时, BP人工神经网络 2型糖尿病(T2DM)模型的灵敏度为 71.82%, 特异度为 63.60%。危险因素评分模型发现每例 T2DM或糖调节受损(IGR)患者的花费为 24.29元人民币; BP人工神经网络模型为 27.04元人民币。

    结论 两种筛查方法在筛查效度上差异不大; 在效益上, BP人工神经网络模型发现 1例 T2DM或 IGR患者的成本比危险因素评分模型增加了 11.32%; 且危险因素评分模型更易于操作, 适合在基层日常应用。

     

    Abstract: Objective To explore an economical and effective screening method of diabetes suitable for application in community.

    Methods Based on an epidemiological survey of diabetes mellitus in the community populalion aged above 40 who have no diabetes history in Minhang District of Shanghai in 2007, risk factors scoring model and BP artificial neural network model were established to evaluate the efficiency and economic benefits of the screening method.The original data was divided into training group(1 970 cases)and testing group(1 977 cases). The two models were developed based on training group and the validity and reliability of the models were validated and compared based on the testing group.

    Results Taking the risk factors score 27 as the threshold value, the sensitivity and specificity of risk factors scoring model were 63.64% and 72.28% respectively. Taking the network output value 0.16 as the threshold value, the sensitivity and specificity of BP artificial neural network model were 63.64% and 63.60% respectively. The direct medical cost per case of risk factors scoring model was 24.29 yuan, whereas the cost of BP artificial neural network model was 27.04 yuan.

    Conclusion The screening effect between risk factors scoring model and BP artificial neural network model was not significantly different. Compared with risk factors scoring model, the cost of BP artificial neural network model was 11.32% higher. Risk factors scoring model was easier to operate and might become a routine diabetes screening method for practice in community.

     

/

返回文章
返回