既往数据较少时预测患病率的不同方法比较

发布时间：2018-04-23 01:22

本文选题：高血压 + 患病率　；参考：《中国卫生统计》2014年03期

【摘要】：目的比较既往数据较少时不同方法在疾病预测中的应用。方法利用我国建国以来四次大规模高血压抽样调查的数据,分别采用几何级数法、logistic函数模型和GM(1,1)灰色模型对患病率进行估计,并通过对比各方法的拟合效果、理论依据和临床意义等,评价其可靠性和准确性。结果所有方法用于高血压患病率的短期预测(2~3年),结果都比较相似,2005年患病率的估计值均处在(19.3±1.0)%之间。但随着预测时间的延长,不同方法给出的疾病趋势则有了较大幅度的差异,截至2050年,GM(1,1)灰色模型(R2=0.9863)的预测结果已达到86.44%,几何级数法给出的估计值也超过75%,而logistic函数模型(R2=0.6933)认为患病率仍只有33.53%。其中logistic模型估计较保守,并假设患病率始终不会超过上限40%;而几何级数法和GM(1,1)模型给出的预测值将快速增长,于2060年之前突破100%。结论几何级数法结合了数据的数学意义和临床意义,适用于中短期的患病率估计,而logistic回归模型给出的患病率趋势在理论上符合人们对未来疾病发展的预估,更适合远期推断。GM(1,1)灰色模型与原始数据的拟合程度较高,且不要求提供大量相关信息,因此也较为适合数据点较少的短期预测。
[Abstract]:Objective to compare the application of different methods in disease prediction with less previous data. Methods based on the data of four large-scale sampling surveys of hypertension since the founding of the people's Republic of China, the geometric series logistic function model and the GM-1 1) grey model were used to estimate the prevalence rate, and the fitting results of the methods were compared. The theoretical basis and clinical significance were used to evaluate its reliability and accuracy. Results all the methods were used to predict the prevalence of hypertension for 2 ~ 3 years, and the results were similar. The estimated prevalence rate in 2005 was between 19.3 卤1.0%. However, with the prolongation of the prediction time, the disease trends given by different methods vary considerably. By 2050, the predicted results of the grey model R2 + 0.9863) had reached 86.44 and the estimated value given by the geometric series method was more than 750.6933, while the logistic function model R20.6933) considered that the prevalence rate was still only 33.53. The logistic model is conservative and assumes that the prevalence rate will never exceed the upper limit of 40. Conclusion the geometric series method combines the mathematical and clinical significance of the data, and is suitable for estimating the prevalence rate in the short and medium term. However, the prevalence trend given by the logistic regression model is theoretically in line with the prediction of the future disease development. The grey model is more suitable for the short-term prediction with fewer data points because of its high fitting degree with the original data, and does not require a large amount of relevant information.
【作者单位】：北京协和医学院中国医学科学院国家心血管病中心阜外心血管病医院心血管疾病国家重点实验室;
【分类号】：R311

【参考文献】