Wang Yawen, Shen Zhongzhou, Jiang Yu
School of Public Health, Chinese Academy of Medical Sciences/Peking Union Medical College, Beijing, China.
Medicine (Baltimore). 2019 Feb;98(6):e14384. doi: 10.1097/MD.0000000000014384.
Maternal mortality rate (MMR) in China has reduced during a decade but still higher than many countries around the world. Rural China is the key region which affects over all maternal death. This study aims to develop a suitable model in forecasting rural MMR and offer some suggestions for rural MMR intervention. Data in this study were collected through the Health Statistical Yearbook (2017) which included the overall MMR in China and urban and rural mortality rate. A basic grey model (GM(1,1)), 3 metabolic grey models (MGM), and a hybrid GM(1,1)-Markov model were presented to estimate rural MMR tendency. Average relative error (ARE), the post-test ratio (C), and small error probability (P) were adopted to evaluate models' fitting performance while forecasting effectiveness was compared by relative error.The MMR in rural China reduced obviously from 63.0 per 100,000 live births in 2005 to 21.1 per 100,000 live births in 2017. One basic GM(1,1) model was built to fit the rural MMR and the expression was X^((1)) (k + 1) = 553.80e^0.0947k - 550.00 (C = 0.0456, P > .99). Three MGM models expressions were X^((1)) (k + 1) = 548.67e^0.0923k - 503.17 (C = 0.0540, P > .99), X^((1)) (k + 1) = 449.39e^0.0887k - 408.09 (C = 0.0560, P > .99), X^((1)) (k + 1) = 461.33e^0.0893k - 425.23(C = 0.0660, P > .99). Hybrid GM(1,1)-Markov model showed the best fitting performance (C = 0.0804, P > .99). The relative errors of basic GM(1,1) model and hybrid model in fitting part were 2.42% and 2.03%, respectively, while 5.35% and 2.08%, respectively, in forecasting part. The average relative errors of MGM were 2.07% in fitting part and 17.37% in forecasting part.Data update was crucial in maintain model's effectiveness. The hybrid GM(1,1)-Markov model was better than basic GM(1,1) model in rural MMR prediction. It could be considered as a decision-making tool in rural MMR intervention.
中国的孕产妇死亡率（MMR）在过去十年中有所下降，但仍高于世界上许多国家。中国农村地区是影响整体孕产妇死亡的关键区域。本研究旨在建立一个合适的模型来预测农村孕产妇死亡率，并为农村孕产妇死亡率干预提供一些建议。本研究的数据通过《卫生统计年鉴（2017）》收集，其中包括中国的整体孕产妇死亡率以及城乡死亡率。提出了一个基本灰色模型（GM(1,1)）、3个新陈代谢灰色模型（MGM）和一个GM(1,1)-马尔可夫混合模型来估计农村孕产妇死亡率趋势。采用平均相对误差（ARE）、后验差比（C）和小误差概率（P）来评估模型的拟合性能，同时通过相对误差比较预测效果。中国农村地区的孕产妇死亡率从2005年的每10万例活产63.0例明显下降到2017年的每10万例活产21.1例。建立了一个基本GM(1,1)模型来拟合农村孕产妇死亡率，表达式为X^((1)) (k + 1) = 553.80e^0.0947k - 550.00（C = 0.0456，P > 0.99）。三个MGM模型的表达式分别为X^((1)) (k + 1) = 548.67e^0.0923k - 503.17（C = 0.0540，P > 0.99）、X^((1)) (k + 1) = 449.39e^0.0887k - 408.09（C = 0.0560，P > 0.99）、X^((1)) (k + 1) = 461.33e^0.0893k - 425.23（C = 0.0660，P > 0.99）。GM(1,1)-马尔可夫混合模型显示出最佳的拟合性能（C = 0.0804，P > 0.99）。基本GM(1,1)模型和混合模型在拟合部分的相对误差分别为2.42%和2.03%，而在预测部分分别为5.35%和2.08%。MGM模型在拟合部分的平均相对误差为2.07%，在预测部分为17.37%。数据更新对于维持模型的有效性至关重要。GM(1,1)-马尔可夫混合模型在农村孕产妇死亡率预测方面优于基本GM(1,1)模型。它可被视为农村孕产妇死亡率干预的决策工具。
