A Study On Application And Comparison Of Time Series Analysis Technology In Drug Cost Of Outpatients With Three Chronic Disease

ObjectivesTo analyze the drug data of three chronic diseases(diabetes,hypertension,coronary heart disease)in Chongqing from 2015 to 2019,to master the basic situation of the drug cost of the outpatient chronic diseases in the third class A hospitals in Chongqing.To establish a single prediction model and a combined prediction model for the average drug costs of diabetes,hypertension,and coronary heart disease clinics,and to select the optimal model.So as to provide some methodological support for the prediction of the average drug cost of outpatients with chronic diseases and necessary data support for the effective regulation and control of costs about chronic diseases in Chongqing.Simultaneously,in order to provide decision-making reference for the government or relevant departments to deepen the reform of medical and health system by evaluating the implementation effect of Zero Mark-up Drug Policy(ZMDP)in Chongqing top three hospital with the interruption time series model.MethodsThe cost data of outpatients with three chronic diseases(diabetes,hypertension and coronary heart disease)in the third class A hospitals in Chongqing was collected,we analyzed the basic characteristics of these diseases and built single prediction models of BPNN,WNN,ENN,LSTM and ES based on the average drug cost of outpatients,as well as combined prediction models of ES-WNN,ES-BPNN,and ES-ENN.The best prediction model is selected by evaluated indicators of RMSE,MAE and MAPE.At the same time,we explored the trend and long-term changes of relevant variables and reasonably evaluated the implementation effect of ZMDP by the segmented regression model of interruption time series based on the cost data before and after the full implementation of ZMDP in Chongqing.Results(1)From January 2015 to February 2019,the total number of outpatients for the three chronic diseases(diabetes,hypertension and coronary heart disease)in a top three hospital in Chongqing was 350848,including 176340 male patients(50.26%)and 174508 female patients(49.74%).From January 2015 to February 2019,the outpatient volume of the three chronic diseases(diabetes,hypertension,coronary heart disease)in the institution showed an overall downward trend.The average age ofpatients with diabetes,hypertension and coronary heart disease were 63.74 years old,64.78 years old,67.53 years old.The outpatient visits were197227,130993 and 22628 respectively.The average cost of drugs were736.56,601.00 and 778.36 yuan,respectively.(2)The predictive performance of the five single models for the average drug cost of diabetes was LSTM>WNN>ES>BPNN>ENN.The predictive performance of hypertension was LSTM>WNN>ES>BPNN>ENN.And the predictive performance of coronary heart disease was LSTM > WNN > BPNN > ES > ENN.From the comprehensive evaluation indicators(RMSE,MAE,MAPE),the LSTM model had the highest prediction accuracy,and its prediction performance was the best,followed by the WNN,BPNN,ES model,and the ENN model was the worst.(3)The three combined models of the average drug cost,both in terms of prediction precision and accuracy,have improved compared with the single model.Similarly,as for the evaluation index synthesis,the prediction performance of the ES-WNN combination model was the best,followed by the ES-BPNN and ES-ENN combination model.(4)The results of interruption time series showed that the full implementation of ZMDP had significant instantaneous and long-term effects in various degrees.The instantaneous level of total average drug cost of three chronic diseases dropped by 49.92 yuan(P=0.008).Meanwhile,although there was no instantaneous change in the total outpatient volume of the three chronic diseases(P>0.05),the long-term trend decreased significantly(P<0.001)and decreased by about 191 people per month.The average drug cost of diabetes decreased by 45.51 yuan(P=0.029)and showed a long-term trend(P=0.029)after the implementation of the policy;the average drug cost of hypertension patients did not change(P=0.276),but its long-term decline was significant(P<0.001);the average drug cost of coronary heart disease decreased significantly(P <0.05)and continued to decrease by 10.23 yuan per month.Conclusions(1)Although the medical expenses of outpatients with chronic diseases have been relieved to a certain extent,it still remains high,and the follow-up still needs to continuously adjust the medical price health service system and effectively adjust the drug cost.(2)The combination prediction model achieves higher precision and accuracy for drug cost prediction,which is better than the single prediction model.Among them,the prediction performance of ES-WNN is optimal,it should be more suitable for the prediction of the cost of outpatient drugs for chronic diseases,which provides a new method for the cost prediction,and can also provide some ideas for the subsequent study of the prediction model of the cost of other diseases.(3)The analysis of interruption time series shows the characteristic changes of ZMDP before and after implementation.It effectively evaluates the actual effectiveness of the full implementation of ZMDP in Chongqing public hospitals.It is confirmed that the drug cost of patients after the implementation of ZMDP is significantly reduced,and the economic burden of patients is reduced.