Research On The Development Of Risk Prediction Model And Evidence-based Nursing Prevention Program For Central Venous Access Device-related Thrombosis In Hospitalized Children

Objectives:(1)Under the guidance of operational research thinking procedures,the environment and constraints of central venous access device-related thrombosis(CRT)in hospitalized children were discussed from four aspects:children-treatment-central venous access device(CVAD)-nursing prevention based on evidence-based method,and knowledge,attitude/belief,practice(KAP)theory.It was expected to provide practical and operable basis for the construction of risk prediction model and evidence-based nursing prevention program for CRT in hospitalized children.(2)The risk prediction model of CRT in hospitalized children were constructed based on machine learning(ML),in order to provide technical guidance for the construction of evidence-based nursing prevention program for CRT in hospitalized children.(3)The evidence-based nursing prevention program and path plot for CRT in hospitalized children were constructed.The purpose of this graded and whole process nursing prevention program is to provide theoretical support and practical basis for standardizing the scientificity of nursing prevention and control management of pediatric CRT in China.Methods:This study included the following four stages:(1)Systematic evaluation and Meta-analysis of risk factors for CRT in hospitalized children:Seven databases at home and abroad were searched to collect the research on the risk factors of CRT in hospitalized children from the establishment of the database to March 22,2020.Research literature was managed through endnote X7 software.Newcastle Ottawa scale was used to evaluate the quality of studies.The general reverse variance random effect model was used for meta-analysis integration.The combined effect value was expressed by odds ratio and its 95%confidence interval.Q test and I~2 were used to evaluate and quantitatively analyze the heterogeneity between studies.The source of heterogeneity was explored by sensitivity analysis and subgroup analysis.Egger’s test was used to evaluate the potential publication bias of the studies.(2)KAP status and influencing factors of pediatric nurses in preventing CRT in hospitalized children:Guided by the KAP model,on the basis of literature research and previous meta-analysis results,we developed the KAP questionnaire for pediatric nurses to prevent CRT in hospitalized children.The formal questionnaire was formed by Delphi expert consultation and pre-survey.Further,a multi-stage sampling method was used to cross-sectional survey on nurses engaged in pediatric clinical nursing in 14 tertiary grade A general hospitals and 7 tertiary grade A specialized pediatric hospitals in China.Descriptive statistics,independent sample t test or approximate t test,one-way analysis of variance,nonparametric Kruskal-Wallis H test and multiple linear regression analysis were used to study the current situation and influencing factors,and Pearson correlation analysis was used to explore the relationship between each dimension of the questionnaire.(3)Construction and validation of CRT risk prediction model for hospitalized children:A retrospective study was conducted to collect 503consecutive cases that met the inclusion and exclusion criteria in a tertiary grade A general hospital in Changsha,Hunan,and randomly divided them into the modeling group:the internal validation group by 7:3 ratio.At the same time,a retrospective study was conducted to collect 85 consecutive cases that met the inclusion and exclusion criteria in two tertiary grade A general hospitals in Urumqi,Xinjiang Uygur Autonomous Region for external validation.Based on ML,LR,RF,and ANN were selected to construct CRT risk prediction models for hospitalized children,and the prediction efficiency between each prediction model was compared.(4)Construction of evidence-based nursing prevention program for CRT in hospitalized children:Through the process of searching and screening the existing evidence-based guidelines,systematic evaluation,expert consensus,evidence synthesis,best practices and recommended practices-literature quality evaluation-best evidence summary-evidence classification and strength recommendation evaluation,the first draft of evidence-based nursing prevention program and path plot for CRT in hospitalized children was formed.Further,we demonstrated and constructed the final draft of evidence-based nursing prevention program and path plot for CRT in hospitalized children through expert meeting.Results:According to the four stages,the research results were summarized as follows:(1)The results of meta-analysis showed that there were statistically significant differences among hospitalized children with CRT in the following factors(P<0.05):(1)for children-related factors,there were history of thrombosis,gastrointestinal/liver disease,hematologic disease and cancer;(2)for treatment-related factors,there were parenteral nutrition(PN)/total PN(TPN),hemodialysis,extracorporeal membrane oxygenation,cardiac catheterization and antibiotic therapy.(3)In terms of CVAD-related factors,there were placement type(PICC vs tunneled lines),placement type(PICC vs totally implantable venous access port),placement vein(subclavian vein vs femoral vein),number of catheters(multiple vs single),number of catheter lumen(multiple vs single),catheter-related bloodstream infection and catheter malfunction.(2)(1)Questionnaire development and evaluation:After two rounds of expert consultation and statistical entry screening after pre-survey,the KAP questionnaire for pediatric nurses to prevent CRT in hospitalized children were finally formed,with a total of 54 entries.In the exploratory factor analysis,5,3,and 3 common factors were extracted for each dimension and explained 60.552%,89.829%and 84.258%of the total variance respectively.The content validity of the questionnaire was 0.968,1.000,and 1.000 respectively,and the content validity of the items was0.833～1.000.The Cronbach’sαcoefficients for the total questionnaire and each dimension were 0.958,0.926,0.973,and 0.970,respectively,and the retest reliability for the total questionnaire and each dimension were 0.898,0.688,0.797,and 0.850,respectively.(2)The results of the current situation survey:pediatric nurses’knowledge,beliefs and practices scores of CRT nursing prevention were(17.92±6.62),(44.91±6.73),and(72.59±17.18)respectively,and the level of knowledge dimension was the lowest with a standard score of(13.19±4.97).The highest education level of pediatric nurses,whether they had received training related to CRT prevention and control,whether their existing knowledge related to CRT was sufficient to help cope with their clinical work,and the attention of CRT in children in their hospitals/units were the main influencing factors on the current state of knowledge,beliefs and practice of pediatric nurses in preventing CRT.Pearson correlation analysis revealed a positive correlation between knowledge,beliefs and practice(r=0.242～0.767,P<0.01).(3)(1)In terms of incidence,the incidence of CRT in hospitalized children in the two regions was 29.8%and 38.8%,respectively,and63.3%and 48.5%of the children had asymptomatic CRT.(2)The risk prediction model was constructed by applying LR and its predictive efficacy was assessed by discrimination,calibration and clinical fitness.The discrimination results showed that the area under the receiver operating characteristic curve(AUC)were 0.737,0.710 and 0.761 for the modeling group,the internal group and the external validation group,respectively,with an accuracy of 78.7%,71.5%and 70.6%,which had good discrimination.The calibration results showed that the calibration curves fit well for the data of each group.The results of the clinical fitness of each group’s analysis showed that the model had a good net patient benefit.A CRT score system for hospitalized children was further developed with a score range of 0-15,and cluster analysis was performed to stratify the risk into low risk(0～2 points),intermediate risk(3～5points),and high risk(≥6 points).Meanwhile,a Nomogram diagram was drawn to predict the risk of CRT in hospitalized children.(3)Based on the model constructed by RF,the internal verification results showed that the AUC value was 0.536 and the prediction accuracy was 61.5%.The results of external validation of this model showed that the AUC value was 0.515and the prediction accuracy of the model was 87.1%,which was a poor prediction performance.(4)Based on the model constructed by ANN,the internal verification results showed that the AUC value was 0.865 and the prediction accuracy was 82.9%,and the results of external validation of this model showed an AUC value of 0.833 and a prediction accuracy of81.2%,which was a high prediction performance.(4)In this phase,we constructed an evidence-based nursing prevention program for CRT in hospitalized children,including 15 topics:personnel qualification,hospitalized children selection,risk assessment,CVAD selection and use,catheter tip position,catheter failure risk reduction,basic prophylaxis,pharmacologic prophylaxis,imaging,health education,nursing documentation,follow-up,embolism screening,extubation,and other management.There were 77 best evidences in total,of which 62 were strong recommendations and 15 were weak recommendations.Also,a path plot of graded evidence-based nursing prevention program was constructed,including in-hospital risk assessment,targeted nursing measures based on risk stratification and discharge follow-up,so as to promote the feasibility and applicability of the best evidence in clinical practice.Conclusions:(1)Based on the systematic evaluation and analysis of the environment and constraints of CRT in hospitalized children from four aspects of children-treatment-CVAD-nursing prevention,it is found that the risk of CRT is influenced by the disease characteristics of the hospitalized children,the treatment,and CVAD related factors.The current status of CRT nursing prevention knowledge,beliefs and behavior of pediatric nurses in our country need to be improved,leaving hidden dangers for children’s tube safety to some extent.It suggests an urgent need to construct the risk prediction model and evidence-based nursing prevention program for CRT in hospitalized children.(2)The incidence of CRT in hospitalized children is high,and asymptomatic CRT is predominant.The risk prediction model of CRT in hospitalized children are constructed based on ML,and the prediction efficiency of the three models are compared.The ANN model has the best prediction performance,and the RF has the worst performance.The LR model has been further matched with the corresponding scoring system,and risk stratification(low,medium and high risk)has been provided.A Nomogram has also been drawn for visualization.LR model can better predict the risk of CRT in hospitalized children,which may be more convenient for clinical operation.(3)Based on evidence and expert conference validation,the evidence-based nursing prevention program and path plot of CRT for hospitalized children were constructed.The provision of graded evidence-based nursing prevention path throughout the whole process of CVAD catheterization and extubation is of great significance to effectively promote the scientization and standardization of CRT protection for hospitalized children,provide theoretical basis and technical standard for clinical pediatric nurses to strengthen management of the risk of hospitalized children’s CRT,and ensure the tube safety of children.