Analyzing Spatial Difference And Modeling Influential Factors Of Trauma Based On Multi-source Big Data ——A Case Study Of Qingpu District And Minhang District In Shanghai

The development of various sectors such as industries,housing,and transportation facilities and infrastructures in many countries has been coupled with serious accidents causing trauma for individuals and society.Such accidents cause not only great harm to patients but also huge burdens of expenses to the families and public health sectors for treatment.Numerous incidences causing trauma have also been occurring in China,and the existing research shows that this country has higher rates of deaths and people with disability due to trauma compared with others.Despite their valuable findings,such research is limited to providing concrete information on the overall mechanism of trauma incidences and patients in society and their relationships with different environments due to various restrictions,including the lack of relevant and sufficient data.With the current information knowledge and communication technology,this data problem could be greatly addressed using multi-source big data that are available online.Such data include the points of interest(POI)data that provide detailed environmental and infrastructural information.Thus,based on the above,this research used descriptive analysis,exploratory analysis,negative binomial regression,and random forest as the algorithm framework to analyze spatio-temporal distributions of trauma incidences,relationships between those incidences and socio-economic status of their occurring environment and the internal and external factors influencing trauma incidences in society.All these analyses used the POI data from Qingpu and Minhang districts in Shanghai,China.A flow used to achieve this study's objectives started by selecting the research data and establishing a database of trauma cases with different impact factors.Next,a division of scale for the time and space and exploration of the relationships for the established impact factors of trauma accidents were followed to take a multi-factor and multi-dimensional research for the studied areas.Specifically,this research developed the following while implementing its methodology:1)It established a new database containing environmental characteristics of trauma incidences at three scales(i.e.,individual,street,and community)using multisource big data.Such data included the point of interest POI,house price data,resident population grid data,road data,land use data,per capita disposable income data.These data were studied using the network kernel density algorithm to determine the density of different types of trauma accidents.2)It conducted a comprehensive study on trauma's temporal and spatial distributions on multiple scales.A study focused on identifying e.g.,the high-incidence periods and road sections for different types of trauma using descriptive statistics and kernel density estimation algorithms.This information was processed to provide a reference for further exploratory analysis research.In addition,it was processed to help the relevant departments understand the hot spot areas of trauma incidences for providing more safety measures.3)It analyzed internal factors influencing trauma accidents based on their relationships with the socio-economic status in two ways.The first way used the data on the housing prices of the trauma patients to identify their relationships with socialeconomic status for individuals.The second used the per capita disposable income for the entire Qingpu district to determine its relationships with socio-economic status at the regional level.In both cases,correlation coefficients were calculated to understand the relationships between the two compared parameters.4)It analyzed external factors influencing trauma accidents using Minhang district as the study area.At this point,the paper designed the models of trauma incidences using the built environment factors(POI,land use data,etc.)at the macro and micro scales.At the macro-scale,the negative binomial regression algorithm was used to analyze the relationships between trauma incidences and the built environment factors at the regional level.At the micro-scale,the random forest models were constructed based on the road segment scale,road type,and traumatic accident type,and the fitness of those models were evaluated.In the end,a SHAP method(SHapley Addictive Translation)based on the additive features was used to explain the factors causing different types of trauma.The main results developed by this study are as follows:1)Through assessment of the socio-economic status(SES)of individuals and streets and the relationship between trauma incidences and their internal influencing factors,this study has determined a statistically significant negative correlation between two scales of the SES indicators and trauma incidences.At the individuals level,the higher the socio-economic status(housing price in the place of patient),the higher the incidences of trauma,and its reverse at the street levels showed similar patterns.Generally,many incidences of trauma in Qingpu are concentrated in the lowerend-middle areas with many residents whose per capital income is relatively low compared with others in different areas.This result provides a theoretical reference for the relevant institutions to formulate medical service policies and trauma cases prevention plans based on their daily activities.2)The results of analysing external factors showed that residential,green,and mixed land use areas have more occurrences of trauma outcomes at TAZ scales.And there are high fitness of the Random Forest models in different road sections.The SHAP values of these models have further demonstrated that more trauma incidences occur in areas which are more concentrated with e.g.,stores and restaurants.Thus,when taken all together,this paper contributes the following:First,taking trauma outcomes caused by traffic accident and falls into account,analyze comprehensively the external influence factors from macro and micro aspects.We found that there are differences in action mechanism of different factors between the two types of trauma at different research scales,and even for the same kind of influencing factor,the degree of its impact also varies under different time.In addition,housing prices and per capita disposable income were innovatively used as representative socioeconomic status indicators to study the relationship between socioeconomic status and the incidence of traumatic accidents at both individual and regional levels.Housing prices were found to be negatively correlated with the incidence of trauma at the individual level.And per capita disposable income was also found to be negatively correlated with the incidence of trauma at the regional level.The methods and findings of this work provide potential baseline information for other research and appropriate safety and health measures against trauma incidences in society.