Research On Hospitalinfection Risk Prediction Model And Direct Economic Loss Evaluation

Research background and purpose:Hospital infection is one of the most prominent public health issues affecting patient safety.The incidence,mortality,and economic loss of hospital infections in patients in the Intensive Care Unit(ICU)are significantly higher than in other clinical departments.Therefore,the research on the prevention and control of hospital infections in ICU patients is of great significance..In the prevention and control of hospital infections,targeted interventions can prevent at least 50%of hospital infections.Targeted interventions are based on key risk factors identified by multiple risk prediction models.However,at present,the predictive models of hospital infection risk factors for ICU patients are mostly single models,and there are few studies on the combined application of multiple risk prediction models.Research on the development of a human-computer interactive hospital risk calculator(a key risk factor identified based on multiple risk prediction models)was not found.Hospital infections not only increase the suffering of patients,but also cause huge economic losses to patients,families,hospitals and the country by prolonging hospital stays.The recursive system model is used to analyze the direct and indirect contributions of various factors affecting the direct economic loss of hospital infections,calculate the total contributions,and find out the key factors of cost control.It can be seen that the recursive system model is of great significance to control its excessive growth.However,the current research using recursive system model to explore the influencing factors of direct economic loss of hospital infection in ICU patients has not been found.Therefore,in this study,ICU patients from six hospitals in Guizhou Province were taken as the research object.First,multiple risk prediction models were used to determine the key risk factors for hospital infection in ICU patients.The human-computer interactive hospital infection risk calculator developed based on the key risk factors.It can meet the usage habits of different medical staff,so as to better provide a decision basis for the ICU to formulate targeted hospital infection prevention and control measures.At the same time,the recursive system model was used to analyze the direct and indirect contributions of direct economic loss of hospital infections in ICU patients in the case group,calculate the total contributions,and find out the key factors controlling their costs.Methods:(1)Based on the real-time hospital infection monitoring system(Xinglin),collect the relevant medical records of all ICU patients in all six hospitals in Guizhou who have used the system for 1year from January 1,2017 to December 31,2017,and apply Rstudio Software programming,statistical description of the demographic characteristics of ICU patients and hospital infection patients,and Univariate generalized linear model for Univariate analysis of hospital infection risk factors.(2)Divide all the above samples into a training set and a validation set according to the ratio of3:1,and construct a generalized linear model,a decision tree model(the four algorithms of ID3,CART,C4.5,and C50)and a random forest on the training set.Multiple risk prediction models,including the model,are used to identify the risk factors for hospital infections in ICU patients,and then the model is validated in the validation set through three indicators:differentiation,calibration,and clinical effectiveness.Based on the key risk factors identified by multiple risk prediction models,human-computer interactive hospital infection risk calculators were developed to meet the different usage habits of medical staff.(3)Collect the general conditions and hospitalization costs of all ICU patients from January 1,2018 to December 31,2018 in four hospitals using the system in Guizhou Province.Use SPSS25.0software to make a 1:1 match according to the hospital,gender,age group,admission department,admission severity,and medical insurance method to form a case group and a control group to calculate the direct economic loss of patients with hospital infection.A paired rank sum test was used to compare the differences in hospitalization costs and length of stay in ICU patients between the case group and the control group.The recursive system model constructed by AMOS25.0software was used to analyze the direct and indirect contributions of various factors affecting the direct economic loss of hospital infections in ICU patients and calculate the total contributions.And find out the key factors that control the direct economic loss of hospital infection.Results:(1)(1)A total of 1782 ICU patients who met the inclusion exclusion criteria in this study.The hospital infection rate was 23.01%(410/1782),which was higher than that reported in the literature(1.62%to 7.23%)for ICU patients in developed regions.Among them,1191 were males,accounting for 66.84%.The number of patients with antibiotics?6 days was at most 1376(77.22%).The number of patients with hospitalized?37 days minimum was 236(13.24%).40.68%of patients had a history of surgery.Among them,410 cases(23.01%)of hospital infections occurred.The group with the highest incidence of hospital infections was?60 years(25.56%),male(25.10%),days of continuous fever?6 days(50.70%),and days of hospitalization?37 days(58.90%),antibiotic use days?13 days(49.02%),beds?2000(32.82%),pathogen culture(30.49%)required,others also include surgery history,infection history,multiple organ failure,cancer history and diabetes history,there was a statistical difference between the above project groups(P<0.05).(2)Hospital infections occur more frequently in winter(43.9%).The composition ratio of infection sites from high to low is 60.00%of the respiratory system,11.50%of the urinary system,and9.80%of the blood system.The most common pathogen is Acinetobacter baumannii 43.15%;The most common multi-drug resistant bacteria is Carbapenem-resistant Acinetobacter baumannii23.35%.The underlying diseases in which hospital infections occur are defined as diabetes,hypertension,chronic obstructive pulmonary disease,and cancer.The incidence of hospital infections in patients without underlying diseases was 17.53%,and the rates of hospital infections in patients with one,two,and three basic diseases were 26.92%,40.94%,and 60.53%,respectively.High to low are 62.28%of urinary catheters(thousand-day infection rate is 0.96‰),ventilator34.88%(thousand-day infection rate is 10.03‰),and central venous catheters are 32.48%(thousand-day infection rate is 1.58‰).The antimicrobial use days were?6 days,7-12 days,and?13 days,and the corresponding hospital infection rates were 6.50%,19.41%,and 49.02%.The differences between the three groups were statistically significant(P<0.05).The univariate generalized linear model found that risk factors that may cause hospital infections include:patient age,admission to clinical departments,surgery history,current need for pathogen culture,diabetes history,cancer history,hospitalization days,continuous fever days,antibiotic use days,hospital beds number,infection history,etc.(2)(1)Based on multiple risk prediction models,relatively comprehensive risk factors for hospital infections in ICU patients were found:the length of hospital stay during the last hospital stay,the number of days of antibiotic use,and the history of diabetes.History of surgery,history of cancer,number of hospital beds and admissions clinical departments;(2)The first four relatively stable risk factors are:the length of hospital stay during the last hospitalization(determined by 6models),the number of days of antibiotic use,and the history of diabetes(determined by 5 models).The current hospitalization requires the cultivation of pathogenic bacteria(determined by 4 models);(3)The comparison results of multiple risk prediction models are as follows:generalized linear model,random forest model,and four decision tree models.The discriminant C statistics are 0.857,0.849,0.745?0.778,and the corresponding calibration evaluation P(S:_P)Statistics:0.877,0.114,0.202?0.666,respectively;Net profit statistics for decision curve analysis are:11%,10.5%,8.3%?9.5%.According to the criterion that the index is larger and better,the generalized linear model is relatively optimal.(4)Based on the key risk factors of multiple risk prediction models,a variety of human-computer interactive hospital infection risk calculators that can meet the different usage habits of medical staff were developed.The first type is an interactive risk calculator based on Rstudio software.Clinical medical personnel who like to use code can dynamically visualize the risk probability of individual patients by clicking on it.The second type is an interactive risk calculator based on a smartphone/network.Or the medical staff on the network can scan the QR code or login URL through mobile phone to dynamically visualize the risk probability of individual patients.The third type is the Excel risk calculator,by clicking on the different characteristics of individual patients,their corresponding risk probabilities and measures to be taken can be displayed interactively in real time..(3)(1)In 2018,the ICU of four general hospitals admitted a total of 1403 patients,349 cases of hospital infection occurred,335 pairs were successfully matched,and 14 pairs were abandoned.The matching success rate was 95.99%.The direct economic loss of hospital infection in 335 ICU patients was 84073.34 yuan,the median hospitalization cost in the case group was 134455.96 yuan,and the control group was 37749.86 yuan;the median hospital stay in the case group was 25 days,and the control group was 7 days.The hospitalization expenses of the case group and the control group were statistically significant by paired rank sum test(P<0.05);(2)Among the direct economic losses of hospital infection,the median cost of western medicine was 31981.06 yuan(44.38%),followed by the median cost of treatment of 13,570.86 yuan(18.77%).(3)Traditional multiple linear regression model was used to find the risk factors that affect the direct economic loss of hospital infections in ICU patients:hospitalization days,antibiotic use days,and cancer history.Further recursive system model found that in addition to the above-mentioned influencing factors,the days of catheter use had only an indirect contribution on the direct economic loss of hospital infection.And the key factor to find the direct economic loss of hospital infections in ICU patients is the number of days of antibiotic use(The total contribution is the largest,with both direct and indirect contributions).Conclusion:(1)Compared with the literature report on the hospital infection rate of ICU patients in developed regions(1.62%?7.23%),the hospital infection rate of ICU patients in six hospitals in Guizhou Province is higher;(2)Multiple human-machine interactive hospital infection risk calculators developed using key risk factors identified by multiple risk prediction models can meet the different usage habits of medical staff.Their development and application can organically combine the construction of hospital infection monitoring and risk prediction models with clinical hospital infection prevention and control,and fill the gaps between hospital infection monitoring,model construction,and practical application,which has important practical significance.(3)In the study of the influencing factors of direct economic loss of hospital infection in ICU patients,the recursive system model is superior to the traditional linear model,which can clearly identify the direct and indirect contributions of direct economic loss of hospital infection in ICU patients,and calculate the total contribution for effective control.The direct economic loss of hospital infections provides an important basis.