Fang H, Lu B, Wang X, Zheng L, Sun K, Cai W
Shenzhen Hospital, Southern Medical University, Shenzhen, China.
The Third Affiliated Hospital of Sun Yat-Sen University, Guangzhou, China.
Braz J Med Biol Res. 2017 Aug 17;50(10):e6638. doi: 10.1590/1414-431X20176638.
This study proposed a decision tree model to screen upper urinary tract damage (UUTD) for patients with neurogenic bladder (NGB). Thirty-four NGB patients with UUTD were recruited in the case group, while 78 without UUTD were included in the control group. A decision tree method, classification and regression tree (CART), was then applied to develop the model in which UUTD was used as a dependent variable and history of urinary tract infections, bladder management, conservative treatment, and urodynamic findings were used as independent variables. The urethra function factor was found to be the primary screening information of patients and treated as the root node of the tree; Pabd max (maximum abdominal pressure, >14 cmH2O), Pves max (maximum intravesical pressure, ≤89 cmH2O), and gender (female) were also variables associated with UUTD. The accuracy of the proposed model was 84.8%, and the area under curve was 0.901 (95%CI=0.844-0.958), suggesting that the decision tree model might provide a new and convenient way to screen UUTD for NGB patients in both undeveloped and developing areas.
本研究提出了一种决策树模型，用于筛查神经源性膀胱（NGB）患者的上尿路损伤（UUTD）。病例组招募了34例患有UUTD的NGB患者，对照组纳入了78例无UUTD的患者。然后应用一种决策树方法，即分类与回归树（CART）来构建模型，其中将UUTD用作因变量，将尿路感染史、膀胱管理、保守治疗和尿动力学检查结果用作自变量。发现尿道功能因素是患者的主要筛查信息，并作为树的根节点；最大腹压（Pabd max，>14 cmH2O）、最大膀胱内压（Pves max，≤89 cmH2O）和性别（女性）也是与UUTD相关的变量。所提出模型的准确率为84.8%，曲线下面积为0.901（95%CI = 0.844 - 0.958），这表明该决策树模型可能为欠发达和发展中地区的NGB患者筛查UUTD提供一种新的便捷方法。
