Katz M H, Nicholson B W, Singer D E, Kelleher P A, Mulley A G, Thibault G E
Department of Medicine, Massachusetts General Hospital, Boston 02114.
J Gen Intern Med. 1988 Nov-Dec;3(6):533-9. doi: 10.1007/BF02596094.
The authors studied the clinical courses of 216 prospectively selected patients with cardiogenic pulmonary edema presenting to an emergency ward (EW) to identify which patients should be triaged to the intensive care unit (ICU). The first four hours were considered the EW or pre-triage phase of hospitalization. During the EW phase, 108 patients remained stable; 33 of them developed cardiopulmonary complications over the next two days. Logistic regression identified four significant independent features that distinguished these 33 patients from the remaining patients: four-hour diuresis less than 1L, history of prior pulmonary edema, T-wave abnormalities, and jugular venous distention. A model containing the four variables predicted hospital complications with a sensitivity of 81% and a specificity of 65%. In comparison, the sensitivity of physicians in admitting to the ICU patients who would go on to have complications was 70%, with a specificity of 63%. In a model containing a term for the physicians' actual triage decision, all four independent predictors remained significant, producing an overall sensitivity of 81% and an overall specificity of 69%.
作者研究了216例前瞻性选择的因心源性肺水肿而就诊于急诊科（EW）的患者的临床病程，以确定哪些患者应被分诊至重症监护病房（ICU）。最初的四个小时被视为住院的急诊科或预分诊阶段。在急诊科阶段，108例患者病情稳定；其中33例在接下来的两天内出现心肺并发症。逻辑回归确定了四个显著的独立特征，可将这33例患者与其余患者区分开来：四小时尿量少于1升、既往有肺水肿病史、T波异常和颈静脉扩张。包含这四个变量的模型预测医院并发症的敏感性为81%，特异性为65%。相比之下，医生将最终出现并发症的患者收入ICU的敏感性为70%，特异性为63%。在一个包含医生实际分诊决定项的模型中，所有四个独立预测因素仍然显著，总体敏感性为81%，总体特异性为69%。
