蒋鸿琳, 童懿昕, 徐宁, 周艺彪, 姜庆五. 传染病在低感染率状态下大规模人群筛检的假阳性模拟分析[J]. 上海预防医学, 2022, 34(4): 314-317. DOI: 10.19428/j.cnki.sjpm.2022.22018
引用本文: 蒋鸿琳, 童懿昕, 徐宁, 周艺彪, 姜庆五. 传染病在低感染率状态下大规模人群筛检的假阳性模拟分析[J]. 上海预防医学, 2022, 34(4): 314-317. DOI: 10.19428/j.cnki.sjpm.2022.22018
JIANG Honglin, TONG Yixin, XU Ning, ZHOU Yibiao, JIANG Qingwu. False positives of screening in large-scale population with low infection rate of an infectious diseasea: a modeling analysis[J]. Shanghai Journal of Preventive Medicine, 2022, 34(4): 314-317. DOI: 10.19428/j.cnki.sjpm.2022.22018
Citation: JIANG Honglin, TONG Yixin, XU Ning, ZHOU Yibiao, JIANG Qingwu. False positives of screening in large-scale population with low infection rate of an infectious diseasea: a modeling analysis[J]. Shanghai Journal of Preventive Medicine, 2022, 34(4): 314-317. DOI: 10.19428/j.cnki.sjpm.2022.22018

传染病在低感染率状态下大规模人群筛检的假阳性模拟分析

False positives of screening in large-scale population with low infection rate of an infectious diseasea: a modeling analysis

  • 摘要:
    方法 通过数据模拟的方式,假定某人群的总人口为2 000万,人群疾病感染率为0.1%~5.0%,模拟不同灵敏度(99.0%、99.5%、100.0%)和特异度(97.0%、97.5%、98.0%、98.5%、99.0%、99.5%、99.9%)的组合情况,计算各组合的PPV、真阳性数和假阳性数。
    结果 在低感染率(≤5.0%)状态下,与灵敏度相比,特异度对PPV的影响更大,且随着感染率的下降,特异度提高时PPV的增幅加大。当感染率>1.0%时,特异度越趋近99.9%,PPV越接近100.0%;而当感染率<1.0%时,PPV的最大值也仅约为90.0%。当人群感染率为0.1%、灵敏度≥99.0%时,在2 000万的人群中筛检试验可以发现的真阳性人数约为2万人;当特异度为97.0%时,假阳性人数约为 59.9万人,PPV为3.2%;当特异度达到99.9%时,假阳性人数降到约2万人,PPV升高到50.0%。当人群感染率为1.0%时,灵敏度≥99.0%和特异度≥97.0%的筛检试验在2 000万的人群中可发现的真阳性数为19.8万~20.0万人;当特异度由97.0%升高至99.9%时,假阳性人数由59.4万人降至2万人。当人群感染率为5.0%时,灵敏度≥99.0%和特异度≥97.0%的筛检试验在2 000万的人群中可发现的真阳性数为99.0万~100.0万人;当特异度由97.0%升高至99.9%时,假阳性人数由57.0万人降至1.9万人。在总人口为 2 000万的人群中,当感染率≤5.0%时,即使灵敏度和特异度分别达到最大值100.0%和99.9%,仍存在约2万例的假阳性。
    结论 当人口数量较多而感染率较低时,运用筛检试验进行大规模人群筛查除了要提高特异度外,大量假阳性所产生的问题不可忽视。

     

    Abstract:
    Objective To explore the positive predictive value (PPV) and false positive (FP) number of screening test in mass testing when the prevalence of infection is low.
    Methods Assuming a population of 20 million with the prevalence of disease infection ranging from 0.1% to 5.0%, PPV, true positive (TP) and FP numbers were calculated under different scenarios of combination of sensitivity (99.0%, 99.5%, and 100.0%) with specificity (97.0%, 97.5%, 98.0%, 98.5%, 99.0%, 99.5%, and 99.9%).
    Results For low infection prevalence (≤5.0%), specificity has a greater impact on PPV than sensitivity; with the decrease of infection prevalence, the increase in PPV elevates when the specificity increases. When the infection prevalence is >1.0%, the closer the specificity is to 99.9%, the closer the PPV is to 100.0%. However, when the infection prevalence is <1.0%, the maximum PPV is only about 90.0%. When the infection rate is 0.1%, a screening test with more than 99.0% sensitivity could detect about 20 thousand TP cases in a population of 20 million. Additionally, the FP and PPV are estimated to be 599 thousand and 3.2% if the specificity is 97.0%, and 20 thousand and 50.0% if the specificity is 99.9%. When the infection rate is 1.0%, a screening test with ≥99.0% sensitivity and ≥97.0% specificity could detect about 0.198‒0.200 million TP cases; and the number of FP decreases from 594 thousand to 20 thousand when the specificity increases from 97.0% to 99.9%. When the infection rate is 5.0%, a screening test with ≥99.0% sensitivity and ≥97.0% specificity could detect about 0.99‒1.00 million TP cases; and the number of FP decreases from 570 thousand to 19 thousand when the specificity increases from 97.0% to 99.9%. When the infection prevalence is ≤5.0% in a total population of 20 million, there are about 20,000 FP cases even if the sensitivity and specificity reach the maximum values of 100.0% and 99.9%, respectively.
    Conclusion When the population is large and the infection prevalence is low, in addition to improving the specificity of the screening test in mass testing, the problem of a large number of false positives cannot be ignored.

     

/

返回文章
返回