Predict Residents' Death Constitute Model Establishment And Evaluation

Research Purpose And SignificanceMonitoring of causes of death is an important part of disease surveillance in a link in anykind of disease surveillance, death as a final outcome of the disease, can be used to analyze theseverity of the disease and whether treatment is effective, and the strategy for the development ofprevention and treatment of diseases and a special focus on the object and the risk factorsprovide a strong basis. Description and analysis of different regions, different groups of people(including different genders, different age groups), the level of death and the dynamic changes ofthe various causes of death, reflect the socio-economic, cultural, education, health services forthe health of residents. However, in many regions of the country and no cause of death reportingsystem, only hospitals reporting data. Hospital mortality data and population-wide mortality datawith a certain degree of differences and commonalities, differences between the two is mainlydue to the different proportion of the different causes of death of the cases died in hospital, thecontact between the two is reflected in, if we can get different diseases died in hospital theproportion or probability, be able to use the hospital data to deduce the cause of deathdistribution of the population. In this study, analysis by monitoring the crowd cause of death,through the establishment of multi-factor model, the use of mortality data, it is estimated that thedeath of the crowd distribution. Be applied on the basis of the model validation and evaluation,so as to estimate the different groups of the main health problems in particular can not do themonitored area of the whole population the cause of death to provide a reference ideas.Research Methods2007-2010 cause of death statistics of the monitoring measurement data has been used ofDiseases and Related Health Problems International Classification (ICD-10) coded fordescriptive statistical analysis, with Excel2003, SPSS13.0, SAS9.0 software. Constitute the sixthcensus of 2010 Population by Age standardized as the standard data. 2007-2010 deathsurveillance data, affect whether an individual died in the hospital, multi-factor logisticregression model. Areas with hospital death information, you can take advantage of thesehospital death cases and the social dimension of indicators into the above model, figure out thecause of death distribution of the local crowd, and then get the local cause of death constitutes. The ResultsSystematic errors in the total model and individual model statistics show that the systematicerrors were 10.43% and 11.01% in about 10%.Proven, cause of death of the major categories of the total model CSMFs Xinghualing 2010and the relative error statistics show that when the cause of death divided into broad categories,Logistic regression model and individual indicators model for the actual cause of death isestimated that in 2010 the average relative error when causes of death subclasses, Logisticregression model of total model and individual indicators for the actual cause of death in 2010the estimated average relative error of 11.53% and 12.53% to 9.72% and 11.85%; Individualindicators model relative error is higher than the total model, the cause of death subclassesrelative error greater than the cause of death categories. Model the overall error of about 10%,the model was well fitted.Xinghualing 2011 cause of death of major categories of monitoring data model estimatedCSMFs and the relative error results show that the total model and individual model, the relativeerrors were 9.46% and 10.58% of the individual model relative error is slightly higher than therelative error of the total model. The total model is the relative error in accidental injuries(31.19%), the relative error is endocrine diseases (1.02%).The cause of death categories of individual indicators model is the relative error in accidentalinjuries (34.10%), the relative error is endocrine diseases (2.54%).2011 cause of death of the Xinghualing subclasses monitoring data model to estimate theCSMFs and the relative error, the relative error of the total model and individual model were13.03% and 14.22%, individual model relative error is slightly higher than the relative error ofthe total model, showing the same trend and the total model, and the cause of death subclassesrelative error is slightly larger than the cause of death categories of relative error. A total modelof the relative error is other respiratory diseases (38.78%). Relative error is the smallest ofischemic heart disease (1.35%).The cause of death subclasses individual indicators model relative error is other respiratorydiseases (38.78%). Relative error is the smallest of ischemic heart disease (0.48%).Conclusion1. In this study, the coroner’s monitoring sites, 2007-2010 Xinghualing population mortalitydata through the establishment of multi-factor Logistic regression model, and use of the existingdeath surveillance data to estimate the cause of death of different parts of the distribution. On the basis of the model validation and evaluation, thus estimated in different parts of the population ofthe main health problems refer to ideas and methods.2. This study was unable to do the monitored area of the whole population the cause ofdeath, the ideas and methods of the estimated crowd of death causes, to a certain extent canprompt false positives and false negatives of the Coroner’s monitoring for the cause of deathmonitoring a new approach.3. with a total model of the regional indicators and contain only individual indicators of themodel compared to the error of only 1% -2%, the rate of 1% -2% in the total contribution to themodel description of regional indicators, you can use the model of individual indicators insteadof the total model.