基于多变量GM(1,N)灰色模型的2009—2015年河南省恶性肿瘤死亡预测

目的 探索河南省环境因素与肿瘤死亡关系,并以此建立恶性肿瘤死亡灰色预测模型。方法 基于2009—2015年河南省居民肿瘤死亡率数据及2000—2015年环境污染数据,采用灰色关联分析方法,对环境因素与肿瘤死亡的相关性及环境污染致肿瘤死亡的潜伏期做定量分析。并建立灰色预测模型,与传统Poisson回归模型对比,择优预测恶性肿瘤死亡率。结果 环境污染指标与肿瘤死亡率关联强度排序为烟(粉)尘>废水>二氧化硫>化学需氧量>工业固体废物>氨氮排放,致居民肿瘤死亡的潜伏期依次分别为8、0、5、6、9、4年。Poisson回归模型、GM(1,4)模型,时滞改进GM(1,4)模型的平均绝对百分误差分别为15.11%、4.94%和4.29%。时滞改进GM(1,4)模型预测2016、2017年河南省恶性肿瘤死亡率分别为157.97/10万、156.41/10万。结论 烟(粉)尘等环境污染指标与肿瘤死亡密切相关,且环境污染具有滞后效应,可根据环境污染指标预测肿瘤死亡并制定肿瘤长期防治策略。

Prediction of cancer death in Henan province, 2009-2015 based on multivariate GM (1, N) grey model

Objective To explore the association between environmental factors and tumor death in Henan province so as to establish a grey prediction model for cancer mortality. Methods Based on the data about the cancer mortality of Henan residents in 2009-2015 and environmental pollution in 2000-2015, grey relational analysis was used to explore the correlation between environmental factors and cancer death, and the incubation period of tumor death caused by environmental pollution was quantitatively analyzed. The grey prediction models were established, and then compared with the traditional Poisson regression model. The optimal model was chosen to predict the cancer mortality. Results The descending order of correlation between environmental pollution indicators and tumor mortality was smoke (dust) > wastewater emission > sulfur dioxide > chemical oxygen demand > industrial solid waste > ammonia-nitrogenous wastewater, with 8 years, 0 year, 5 years, 6 years, 9 years and 4 years incubation period for tumor deaths, respectively. The mean absolute percentage error (MAPE) of the Poisson regression model, the GM (1,4) model and the time-delay GM (1,4) model were 15.11%, 4.94%, and 4.29%, respectively. The cancer mortality rates in Henan province in 2016 and 2017 predicted by the time-delay GM (1,4) model were 157.97/100,000 and 154.61/100,000 respectively. Conclusions Environmental pollution indicators such as smoke (dust) are closely related to cancer death with hysteresis effects. Cancer death can be predicted according to environmental pollution indicators; meanwhile, long-term prevention and treatment strategies for cancer can be formulated.