不同死亡率标准下区域长寿死亡占比指标一致性分析

Consistency Analysis of Regional Proportion of Longevity-related Deaths According to Different Mortality Criteria

目的 分析死亡漏报对区域长寿相关死亡占比水平的影响,以期为利用人口死亡登记数据开展区域长寿水平评价提供依据。方法 基于2019年全国人口死亡登记数据计算各个区县的粗死亡率。按照不同的粗死亡率标准(粗死亡率≥4.5‰,≥5.0‰,≥6.0‰,≥7.0‰)分别计算相应纳入标准情况下百岁老人、长寿老人、高龄老人死亡占比等9个指标的全国水平,采用标准差和变异系数评价各个指标的稳定性。在省级和地市水平分别计算纳入全部监测区县和只纳入粗死亡率≥6.0‰区县情况下的9个死亡占比指标,采用基于百分位数法的Bland-Altman散点图法评价两种情况下各个指标的一致性。结果 在全国水平,100岁及以上死亡占比指标的标准差最小(0.01~0.02),其次是90岁及以上死亡占比指标(0.13~0.18),80岁及以上死亡占比指标的标准差最大(0.20~0.33);变异系数最小的指标是80+/60+和100+/90+(0.43~0.44),变异系数最大的指标是100+/all和90+/all(1.69~1.80)。在省级水平和地市水平,9个死亡占比指标的95%LoA均超出了相应的专业可接受范围,意味着两种情况下的指标测量结果不可互相替换。结论 死亡漏报对全国水平长寿死亡占比指标影响较小,但是对省级水平和地市水平影响较大。在依托人口死亡登记数据报告区域长寿死亡占比指标时须遴选数据完整性较高(粗死亡率≥6.0‰)的区县进行测算,以保证结果的准确性。

Objective To analyze the effect of death under-reporting on the regional longevity-related deaths proportion and provide some evidence for its application in the regional longevity level evaluation.Methods National deaths registration data in 2019 were obtained from the National mortality surveillance system and the crude mortality of each county were calculated.According to different crude mortality criteria(≥4.5‰,≥5.0‰,≥6.0‰,≥7.0‰,respectively),a total of 9 indicators of death proportion of the national level were computed based on the corresponding number of counties included.Standard deviation and coefficient of variation were adopted to evaluate the stability of each indicator.At the provincial level and prefecture level,Bland-Altman plot was adopted to evaluated the consistence of the 9 indicators of death proportion in two different conditions,namely all the monitored counties included and only those counties with the crude mortality greater than 6.0‰included,respectively.Results At the national level,the proportions of centenary death had the smallest standard deviation(0.01~0.02)followed by the proportions of longevity death(0.13~0.18),while the proportions of deaths aged 80 years and older had the largest standard deviation(0.20~0.33).Those indicators named as 80+/60+and 100+/90+had the smallest coefficient of variation(0.43~0.44),while 100+/all and 90+/all had the largest(1.69~1.80).At the provincial and prefecture scale,the 95%limits of agreement(LoA)of all the 9 proportions were beyond the corresponding professional acceptable variation,which means that those indicators in the two conditions cannot be replaced with each other.Conclusion Death under-reporting has little effect on the proportion of longevity-related deaths at the national level but has a great impact on provincial level or prefecture level.Counties or districts with high death data completeness(crude mortality greater than 6.0‰)can only be included to obtain the longevity-related deaths proportion at provincial or prefecture level to ensure their accuracy.