Objective: To optimize the warning threshold values of common communicable diseases in Gansu province, and improve the early warning effect. Method: An early warning model was set up for influenza, scarlet fever, other infectious diarrheal diseases, dysentery, typhoid and paratyphoid, viral hepatitis type E and hand foot and mouth disease (HFMD) respectively in Gansu by using the moving percentile method and cumulative sum method. By calculating the sensitivity, specificity, predictive value of positive test, predictive value of negative test, Youden' index and receiver-operating characteristic curve, the optimum early warning threshold values for communicable diseases in Gansu were selected. Results: The optimum early warning boundary values of influenza, scarlet fever, other infectious diarrheal diseases, dysentery, typhoid and paratyphoid, and viral hepatitis type E were P(90), P(80), P(95), P(90), P(80) and P(90) respectively. The optimum early warning parameters of HFMD were k=1.2, H=5σ. Under the optimum early warning boundary values/parameters, the early warning sensitivities of influenza, scarlet fever, other infectious diarrheal diseases, dysentery, typhoid and paratyphoid, viral hepatitis type E and HFMD were 86.67%, 100.00%, 91.67%, 100.00%, 100.00%, 100.00% and 100.00%, the specificities were 86.49%, 62.22%, 75.00%, 100.00%, 97.92%, 89.13% and 74.47%. The predictive values of positive test were 72.22%, 29.17%, 52.38%, 100.00%, 80.00%, 54.55% and 29.41%, and the predictive values of negative test were 94.12%, 100.00%, 96.77%, 100.00%, 100.00%, 100.00% and 100.00%, and the Youden' indexes were 0.73, 0.62, 0.67, 1.00, 0.98,0.89 and 0.74. Receiver-operating characteristic curve showed that the values/parameters of this warning boundary were the points closest to the upper left of the coordinate diagram. Conclusion: The early warning thresholds of influenza, other infectious diarrheal diseases, dysentery and hepatitis E in Gansu may be raised appropriately and the early warning parameters of HFMD need to be adjusted to improve the effectiveness of early warning.
目的： 优选甘肃省常见急性传染病的预警阈值，提高预警功效。 方法： 甘肃省流行性感冒（流感）、猩红热、其他感染性腹泻、细菌性和阿米巴性痢疾（痢疾）、伤寒/副伤寒、戊型肝炎（戊肝）采用移动百分位数法，手足口病采用累积和控制图法，分别建立预警模型，通过计算灵敏度、特异度、阳性和阴性预测值、约登指数及绘制受试者工作特征曲线（receiver operating characteristic curve，ROC），选择出适合甘肃省传染病的最优预警阈值。 结果： 流感、猩红热、其他感染性腹泻、痢疾、伤寒/副伤寒、戊肝的最优预警界值分别为P(90)、P(80)、P(95)、P(90)、P(80)、P(90)，手足口病的最优预警参数为k=1.2，H=5σ。在最优预警界值/参数下，流感、猩红热、其他感染性腹泻、痢疾、伤寒/副伤寒、戊肝、手足口病的灵敏度分别为86.67%、100.00%、91.67%、100.00%、100.00%、100.00%、100.00%，特异度分别为86.49%、62.22%、75.00%、100.00%、97.92%、89.13%、74.47%，阳性预测值分别为72.22%、29.17%、52.38%、100.00%、80.00%、54.55%、29.41%，阴性预测值分别为94.12%、100.00%、96.77%、100.00%、100.00%、100.00%、100.00%，约登指数分别为0.73、0.62、0.67、1.00、0.98、0.89、0.74，ROC显示此预警界值/参数均为最靠近坐标图左上方的点。 结论： 甘肃省流感、其他感染性腹泻、痢疾、戊肝的预警阈值可适当调高，手足口病预警参数需调整，从而提高预警功效。.
Keywords: Communicable disease; Cumulative sum; Early-warning threshold; Moving percentile method.