| Background:Accurate measurement of mortality level in each county(district)in China is fundamental for achieving the goals set out in the Healthy China 2030 Action Plan.The National Cause-of-Death Reporting System(CDRS)covers 605 counties out of 2844 countries in China,which needs in-depth qualitative research to improve the quality of cause of death data reporting and obtain accurate death information in each county.Meanwhile,majority of counties in China do not have causes of death surveillance system data and miss basic mortality data for the past decades.Therefore,it is necessary to develop county-level mortality estimation models with existing mortality data in some counties to provide reference for other counties.Method:1)From October to November,in 2020,twenty-nine staff,participating or participated in CDRS work at different level in province A and province B,were recruited for semi-structured interviews.The thematic framework analysis was used for analyzing the interview contents.Meanwhile,the quantitative analysis was used to evaluate cause of death surveillance data quality in 7 counties from the above two provinces.2)Based on the causes of death surveillance system data and population data of 14 counties(districts)along the Huaihe River Basin,with the relative economic data from the above counties’statistical yearbooks,a Poisson Regression(PR)or Negative Binomial Regression MixedEffect model(NBR)was constructed to estimate mortality for people aged 5 years and above at county level.3)Referring to the "co-variate model",four typical counties(districts)-Lingbi,Yongqiao,Xuyi and Jinhu were chosen to explore the "borrow strength"by borrowing auxiliary variable or geography,respectively.Four PR/NBR models were structured,respectively,by getting rid of one of the above four counties(districts)one by one.In addition,the difference between the true mortality rate and the estimated mortality rates by using two borrowing approaches was compared by its 95%confidence interval(CI),respectively.Result:1)The procedure of reporting death information in the two surveyed provinces mainly met with the national CDRS requirements.The popularization of modern communication tools improved the timeliness of data reporting.In the process of reporting,the selection manner of the underlying cause of death is the key problem affecting the data accuracy;There are some factors to affect the completeness of mortality data,including using the average national mortality rate of 6‰ as an assessment index,nonstandard reporting procedure of under-reported information and difficult data exchange.In addition,there are some problems in the grass-roots staff,such as high mobility and insufficient capacity.2)By using the mortality data from 14 counties(districts)along the Huaihe River Basin,the county-level mortality estimation model has been established as below.ln(Yjtas)=ln(Pjtas)-6.3745+0.6247gender+3.48 55age+0.000002449GDPjt-0.00209year2+0.01307gender·year-0.3463gender·age-0.0142age ·year-0.01732gender·year · age+μj3)The four models have been constructed successfully to estimate mortality for Lingbi,Yongqiao,Xuyi and Jinhu counties,separately.The difference between the predicted and true values of mortality using the auxiliary variable borrowing and geographic borrowing approaches were-28.041 per 100000 people(95%CI:-59.430,3.347)and-8.181 per 100000 people(95%CI:-40.208,23.846)for Lingbi County,respectively.For Yongqiao District,the above two differences were 18.428 per 100000 people(95%CI:-32.617,69.474)and-33.865 per 100000 people(95%CI:-83.887,16.157),respectively.The results showed that two borrowing approaches could get good estimations for both Lingbi and Yongqiao counties because of the above 95%CI values all included 0.However,there are no good estimations by using two borrowing approaches for both Xuyi and Jinhu counties.Xuyi County’s mortality were generally higher than the true mortality when estimated with the auxiliary variable borrowing approach,with the difference of 25.735 per 100000 people(95%CI:4.455,47.016),and lower than the true mortality when estimated with geographical borrowing approach,with the difference of79.407 per 100000 people(95%CI:-101.213,-57.601).In Jinhu County,the results were all overestimated,and the average difference values for the auxiliary variables borrowing and geographical borrowing methods were 154.445 per 100000 people(95%CI:103.216,205.673)and 140.130 per 100000 people(95%CI:89.811,190.448),respectively.Conclusion:At present,there is a lack of mortality data in all counties in China.For counties(districts)with CDRS,some problems existed in the procedure of reporting,suggesting that our country should pay more attention to them,promote multi-sectional data exchange,strictly control the reporting process,and strengthen the quality of training for grassroots staff,etc.For counties(districts)without CDRS,the idea of small area estimation(SAE)could be borrowed to construct a model to estimate mortality rates;In the application of the model,the supporting information of the estimating and borrowing counties,geographical location and other factors related to the mortality level should be taken into consideration in order to find a suitable "borrowing strength".It is also recommended that national departments should strengthen the collection and disclosure of basic data at the county(district)level to facilitate the construction of estimation models. |
PDF Full Text Request