Troëng T, Janzon L, Bergqvist D
Department of Surgery, Central Hospital, Karlskrona, Sweden.
Eur J Vasc Surg. 1992 Nov;6(6):628-35. doi: 10.1016/s0950-821x(05)80840-0.
to compare two different statistical methods in predicting the outcome of surgery for chronic leg ischaemia.
the present study from the Swedvasc registry is based on an inception cohort of 1635 patients with chronic leg ischaemia (intermittent claudication in 609 and critical ischaemia in 1026 patients), who have been followed until 1 year after surgery. Outcome was classified as improved vs. not improved, amputated or dead in claudication and as the intact leg vs. amputation or death in critical ischaemia.
logistic regression analysis was compared to the inductive expert system program, Assistant Professional, in the prediction of outcome. Seventy per cent of cases in the data base were used to create a risk factor model including 17 variables registered in Swedvasc. These variables included an assessment of patients overall health status, severity of disease, the surgeon's experience and surgical procedures. This model was then evaluated using the remaining 30% of the patients in the data base.
a risk score indicating a probability of an adverse outcome exceeding 0.5 was, in patients with intermittent claudication, associated with a sensitivity of 38% using logistic regression and 26% using Assistant Professional. The percentages of correctly predicted adverse outcomes were 29 and 50%, respectively. In patients with critical ischaemia, the sensitivities with the two methods were 68 and 38% and the predictive values 42 and 57%, respectively.
the risk scores created with the two methods gave low sensitivities. It is concluded that risk functions could not be used to predict an adverse outcome in patients operated on for chronic leg ischaemia with the data set used.
比较两种不同的统计方法在预测慢性下肢缺血手术结果方面的差异。
本研究来自瑞典血管登记处，基于一个起始队列，共1635例慢性下肢缺血患者（609例间歇性跛行患者和1026例严重缺血患者），对其进行随访直至术后1年。结果分为改善与未改善、间歇性跛行患者截肢或死亡，以及严重缺血患者保肢与截肢或死亡。
将逻辑回归分析与归纳专家系统程序“专业助手”用于预测结果进行比较。数据库中70%的病例用于创建一个风险因素模型，该模型包含瑞典血管登记处记录的17个变量。这些变量包括对患者总体健康状况、疾病严重程度、外科医生经验和手术操作的评估。然后使用数据库中其余30%的患者对该模型进行评估。
在间歇性跛行患者中，风险评分表明不良结局概率超过0.5时，逻辑回归分析的敏感度为38%，“专业助手”的敏感度为26%。正确预测不良结局的百分比分别为29%和50%。在严重缺血患者中，两种方法的敏感度分别为68%和38%，预测值分别为42%和57%。
两种方法创建的风险评分敏感度较低。得出的结论是，使用所采用的数据集，风险函数无法用于预测接受慢性下肢缺血手术患者的不良结局。
