Miettinen Olli S, Flegel Kenneth M, Steurer Johann
Department of Epidemiology, Biostatistics, and Occupational Health, Faculty of Medicine, McGill University, Montreal, QC, Canada.
J Eval Clin Pract. 2008 Apr;14(2):343-50. doi: 10.1111/j.1365-2753.2007.00873.x.
Clinical diagnosis of pneumonia is a concern when a patient presents with recent cough--new or worsened--together with fever as the chief complaint. Given this presentation, the doctor would benefit from having access to software that specifies, first, what diagnostic indicators experts typically use in that diagnosis; then, upon entry of those facts, what experts' typical probability of pneumonia is in such a case; and finally, how much this probability might change upon adding the facts from chest radiography.
We specified a set of 36 hypothetical presentations of this type by patients 20-70 years of age, involving a comprehensive set of clinical-diagnostic indicators. Members of three separate expert panels independently set the probability of pneumonia in each of these cases, and also the range of possible post-radiography probabilities. A logistic function of the diagnostic indicators was fitted to the medians of the probabilities.
The median probability of pneumonia was a joint function of the patient's age and current rate of cigarette smoking; history as to the cough's duration, the fever's maximum, dyspnea (including whether on effort only) and rigors; and physical examination as to temperature, signs of upper respiratory infection, prolongation of expiration, dullness on percussion and some auscultation findings. Non-contributory were history of wheezing, pain on inspiration, type of sputum and signs of cold or influenza. This probability function, and the post-radiography functions based on the same indicators, are accessible at http://www.evimed.ch/pneumonia.
The expert inputs to clinical diagnosis that were derived and made readily accessible provide for expertly clinical diagnosis of pneumonia, relevant for decisions about radiography and treatment without it.
当患者以近期咳嗽（新出现或加重）伴发热为主诉就诊时，肺炎的临床诊断成为一个关注点。鉴于这种临床表现，医生若能使用一款软件将有所助益，该软件首先能明确专家在该诊断中通常使用的诊断指标；其次，在输入这些事实后，能得出专家在此类病例中诊断为肺炎的典型概率；最后，能显示在加入胸部X光检查结果后该概率可能会有多大变化。
我们设定了一组由20至70岁患者出现的36种此类假设性临床表现，涵盖了一整套临床诊断指标。三个独立专家小组的成员分别独立设定每种情况下肺炎的概率，以及X光检查后可能的概率范围。将诊断指标的逻辑函数拟合到概率中位数上。
肺炎的中位数概率是患者年龄、当前吸烟率、咳嗽持续时间、发热最高温度、呼吸困难（包括是否仅在用力时出现）和寒战病史，以及体温、上呼吸道感染体征、呼气延长、叩诊浊音和一些听诊结果等体格检查情况的联合函数。喘息病史、吸气时疼痛、痰液类型以及感冒或流感体征并无助于诊断。此概率函数以及基于相同指标的X光检查后函数可在http://www.evimed.ch/pneumonia获取。
所推导并易于获取的专家对临床诊断的输入信息有助于对肺炎进行专业的临床诊断，这对于在不进行X光检查的情况下做出关于X光检查和治疗的决策具有重要意义。
