Barer D H, Mitchell J R
Department of Medicine, University Hospital, Queens Medical Centre, Nottingham.
Q J Med. 1989 Jan;70(261):27-39.
Data collected by a single observer on 362 patients taking part in an acute intervention trial were used to derive simple methods of predicting outcome in conscious stroke victims. The effectiveness of these univariate methods was then compared with that of multivariate models based on discriminant function analysis. The multivariate models were somewhat better at predicting death within the first month in newly-admitted patients, and in predicting lack of functional improvement between one and six months in those still in hospital at one month. Even so, these predictions could not be made with certainty and the improvements in probability offered by multivariate over univariate analysis are unlikely to be of practical benefit to the clinician. Single variables such as power in the affected arm and continence were as good as the discriminant functions at predicting discharge within the first month and first six months, respectively. The predictive power of these single simple clinical variables was validated in a separate group of 277 stroke patients. The various proposed methods of prognostication in stroke need to be compared in prospective studies, but until this is done, we recommend a simple approach using the best available clinical information.
由一名观察者收集的、参与急性干预试验的362例患者的数据，被用于推导预测意识清醒的中风患者预后的简单方法。然后将这些单变量方法的有效性与基于判别函数分析的多变量模型的有效性进行比较。多变量模型在预测新入院患者第一个月内的死亡情况，以及预测那些在一个月时仍住院的患者在1至6个月内功能未改善方面，表现稍好。即便如此，这些预测也无法确定，而且多变量分析相对于单变量分析在概率上的提高，对临床医生而言不太可能具有实际益处。诸如患侧手臂力量和大小便失禁等单个变量，在预测第一个月内和前六个月内出院方面，分别与判别函数表现相当。这些简单的单个临床变量的预测能力，在另一组277例中风患者中得到了验证。中风预后的各种建议方法需要在前瞻性研究中进行比较，但在此之前，我们推荐使用现有最佳临床信息的简单方法。
