Analysis On The Influencing Factors Of Hospitalization Cost For The Insured Inpatients With Angina Pectoris In Tianjin

Objective: Through the analysis of the hospitalization expenses of patients with angina pectoris in Tianjin,explore the cost composition and find the internal and external factors affecting costs.Besides,data support and theoretical advice are provided for the control of hospitalization expenses for angina in the future.Through the in-depth mining of data,the rules for the excessive hospitalization expenses of angina pectoris patients are discussed.Three different prediction models are constructed to compare the pros and cons of prediction performance and select the optimal model to predict the excessive hospitalization expenses.Methods: Data was obtained from the Tianjin Urban Employees' Basic Medical Insurance Register System database during January 2003 to December 2013.The patients with angina pectoris were recruited according to the ICD-10 code(ICD-10 codes of angina pectoris were I20.0,I20.1,I20.8,I20.9).The cases meet with one of the following situations was excluded: less than one day of hospitalization or the hospitalization cost were out of the normal range.Descriptive analysis of the general demographic characteristics and the basic conditions of hospitalization expenses of inpatients with angina pectoris was completed through R software.New grey relational analysis was used to explore the correlation between each hospitalization cost and overall hospitalization cost,the value of structure variation analysis was used to measure the change of each hospitalization cost.Combining the previous two results,the internal factors that affect hospitalization costs were obtained.The LASSO regression model was used to analyze the external influences of hospitalization costs affecting patients with angina pectoris.Association rules data mining was used to explore the implicit rules of excessive hospitalization expenses.Support Vector Machines,Random Forests,and AdaBoost were used to predict the excessive hospitalization expenses respectively and the optimal prediction model was sought.Results: 22 958 cases were recruited finally,including 12 279 males and 10 679 females.The age of patients was ranged from 27 to 98 years old and the median age was 63.00(15.00)years old.In terms of personnel category,4 743 cases were in-service,18 095 were retired and 120 were old workers.4391 cases received treatment in primary hospital,8502 cases in secondary hospital and 10065 cases in tertiary hospital.5 405 cases received surgery and the other 17 553 patients did not.341 cases received blood transfusion and the other 22 617 patients did not.The days of hospitalization ranged from 1 to 90.The result of new grey relational analysis showed that the top four single item costs which was most relevant with total hospitalization expenses were medicine cost(0.9470),medical materials cost(0.9287),treatment cost(0.8132)and examination cost(0.8037).The DSV analysis showed that the top four individual expenses with the largest structural change contribution rate were the drug cost(32.99%),medical materials cost(31.84%),the treatment cost(13.68%)and the Chinese medicine fee(11.71%).The result of LASSO regression analysis showed that hospital level,gender,type of personnel,received surgery or not,received blood transfusion or not,age,actual conceit ratio,and length of hospital stay are factors influencing hospitalization costs for patients with angina pectoris.The hospitalization cost of patients with angina pectoris in tertiary hospitals is higher than that in patients with angina pectoris in secondary hospitals and primary hospitals.The hospitalization cost of patients with angina pectoris is lower than that of male patients with angina pectoris.The cost of hospitalization for older workers with angina is higher than that of active employees and retirement.The hospitalization costs for patients with angina underwent surgery are higher than those without surgery.The hospitalization expenses for patients with angina pectoris who received blood transfusion is higher than the patients who did not receive blood transfusion.The expenses of patients with hospitalization no less than seven days was higher than that patients with less than seven hospitalization days.The patients whose self-pay ratio was 0.05 resulted in less hospitalization expenses compared with those whose self-pay ratio was 0.15.The hospitalization costs for patients aged 60-74 years are higher than for other age groups.The association rule model has unearthed 9 meaningful rules.For inpatients received surgery and blood transfusion therapy,the probability of developing excessive hospitalization expenses was 6.59 times that of before.For male inpatients received treatment in tertiary hospital,accompanied by received blood transfusion,the probability of developing excessive hospitalization expenses was 6.39 times that of before.For inpatients received treatment in tertiary hospital,accompanied by received surgery and no less than 21 days of hospitalization,the probability was 6.27 times that of before.For inpatients received treatment in tertiary hospital,accompanied by received blood transfusion,the probability was 6.20 times that of before.For male inpatients who received blood transfusion,the probability was 6.15 times that of before.For inpatients aged 60-74 years,accompanied by received blood transfusion,the probability was 5.99 times that of before.For inpatients whose percentage of individual payment was 0.1,accompanied by received blood transfusion,the probability was 5.76 times that of before.For inpatients who retired,accompanied by received blood transfusion,the probability was 5.76 times that of before.For inpatients ? 21 days of hospitalization,accompanied by underwent surgery,the probability was 5.76 times that of before.The model prediction results showed that SVM is better than random forest and AdaBoost in specificity,AUC and G-mean.The difference is statistically significant.There is no significant difference in prediction performance between random forest and AdaBoost algorithm.Conclusions: It is necessary to strengthen the standardized use of drugs through incentive mechanisms,guide the rational pricing of drugs,promote the separation of medicines,and reduce the proportion of drug costs.For medical materials and inspection fees,the doctor and patient's awareness of the cost of medical materials should be increased and the medical staff's diagnosis and treatment behavior should be regulated.Besides,more attention should be paid to Chinese medicine fees.The implementation of grading diagnosis and treatment should be promoted from single disease to avoid waste of medical resources.At the same time,the clinical path management should be constructed to shorten the hospitalization days of patients,thereby reducing the cost of hospitalization for angina pectoris.Last but not least,Support vector machine can be used to predict the over-limitation of hospitalization expenses for patients with angina pectoris in Tianjin.