Think A=positive and B=test positive then plug into Bayesian formula P(A|B)=P(B|A)P(A)/P(B) covered by AP statistics.
A MD needs to know these basics including the concepts of prevalence, sensitivity and specificity, because you need to rank your solutions by importance, likelihood and economy.
it is critical to apply statistics to medical problems
you need to evaluate the probabilities of competing diseases before you take a solution. A women may have 10+% chance to suffer from breast cancer in her life time. It varies from person to person depending on family history and other risk factors. Do you want to remove yours irregardless your risk factors? Your baby may suffer from hundereds of genetic diseases each with a tiny possibility. Do you want to test all these possibilities before you give birth to a baby?
In fact, treating a medical problem has no difference than other problems in life, you maximize your expected return within your acceptable risk.
可我的直觉不错, 思维 OK 的.
键哥出题
假设某癌患病率是千分之五，也就是平均一千人里有5人患癌，再假设我们有一种上佳的癌症测试方法，确诊率高达99.5，也就是平均1000个真真癌病者中有995人能被检测出阳性，而误诊率只有1%，也就是平均100个健康人用此法测试有一人会误呈阳性。现在某人做了测试，结果是阳性，请问其有多大可能真真不幸患上癌症? 大约 99%，大约 75%，大约 50%，大约 33%?
铃兰答不上来
这题目有个 bug, 如果确诊率为 99.5%, 那么误诊率就是 0.5%, 而不是 1%.
我从直觉出发，基于信息，得出 99% 患癌的可能性.
但是, 切切记住, 凭任何单一或 / 和单次的检测, 是不能准确诊断疾病的. 人是一个复杂的生物, 不是机器, 不能完全靠数学推导出诊断和治疗方案.
奇了怪了, 铃兰上网明明轻松来着, 宣称不学习的, 偏偏自投罗网, 钻进这种烧脑, 内涵, 硬朗的圈套,还点赞了呢, 真是自虐得莫名其妙呀
可以理解為大數量無病人口中産生的誤诊數把小數量病人中的確診數稀释了。
高达99.5%确诊率是怎么得来的？可以告医疗公司做假广告吗？
如果测试结果表明患病, 实际上病人也患有该疾病，称之为真阳性.
如果检测结果表明患病, 但实际上病人没有罹患该疾病, 称之为假阳性.
如果检测结果表明没有患病, 实际上病人也没有罹患该病, 称之为真阴性.
如果检测结果表明没有患病, 但病人实际上患有该疾病, 称为之假阴性.
确诊率高达99.5—— 字面理解是，随机抽取1000人，有995人被确诊患病。而不应该说成是1000本来患病的人被正确诊断出疾病。
这个99.5， 是不是指该设备的准确率？
則愚矣。
思考有帮助.
阳性预测值（positive predictive value, PPV）指检查结果阳性并且确实罹患该疾病的比率
如果 10 个阳性检查结果中 9 个为正确的（真阳性), PPV 为 90％. 因为所有阳性检查结果包含其真阳性及假阳性数, 因此 PPV 可以用于描述某个特定人群阳性检查结果中真阳性的可能性。
阴性预测值（negative predictive value, NPV）指检查结果阴性并且确实没有罹患该疾病的比率
如果 10 个阴性测试结果中有 8 个是正确的（真阴性), NPV 为 80％. 因为不是所有阴性检查结果都为真阴性, 一些检查结果阴性的患者事实上患有该疾病. NPV 可以用于描述某个特定人群阴性检查结果中真阴性的可能性。
p = 0.99 * 0.005 / (0.99*0.005 + 0.01*0.995) = 0.3322
好像没有什麽難度啊。
看來很可能我學過但忘了。
还好，像小谢兄这样的眼光打上去，就能算出来的人，属于极少数，而且他也不会给你讲题。God bless。
概率及检查的特征, 计算验后概率.
将检验数据 / 结果, 正确整合到临床环境中 (接近真相), 利于作出临床决策.
Think A=positive and B=test positive then plug into Bayesian formula P(A|B)=P(B|A)P(A)/P(B) covered by AP statistics.
A MD needs to know these basics including the concepts of prevalence, sensitivity and specificity, because you need to rank your solutions by importance, likelihood and economy.
you need to evaluate the probabilities of competing diseases before you take a solution. A women may have 10+% chance to suffer from breast cancer in her life time. It varies from person to person depending on family history and other risk factors. Do you want to remove yours irregardless your risk factors? Your baby may suffer from hundereds of genetic diseases each with a tiny possibility. Do you want to test all these possibilities before you give birth to a baby?
In fact, treating a medical problem has no difference than other problems in life, you maximize your expected return within your acceptable risk.