From the perspective of the COH framework, combining hospital characteristic factors, price factors and patients characteristic factors, and using UEBMI data of Chengdu City from 2011 to 2015, our study empirically analyzed the effect of COH on medical expenditures by multivariate regression modeling. This study had a number of important findings.
Firstly, the hospital level had a significant influence on medical expenditure. The higher the hospital level, the lower the medical expenditures. This could be explained based on the following reasons: (a) The government has more strict supervision over the high-level hospitals than the low-level hospitals, and the medical price of high-level hospitals was relatively reasonable, (i.e. more rreflective of the true cost). This translates into a lower likelihood of excessive medical treatment and less unreasonable medical expenditure (overuse drugs and diagnostic testing.); (b) High- level hospitals were more standardized than low-level hospitals in terms of management systems and treatment processes; (c) High-level hospitals received more government subsidies than low-level hospitals. In order to generate profit, low-level hospitals prescribed more drugs and provided more high-technology tests. There may also have been over-prescribing behavior of physicians at low-level hospitals; (d) In China, there are more medical experts and more advanced medical equipment in high-level hospitals, so it is generally believed that healthcare service quality of the high-level hospitals were better. The provision of medical services was more efficient in high-level hospitals, saving on medical expenditures. The higher surgery expenditures may be related to the rigid price system for different COH in China.
Secondly, the reimbursement ratio of UEBMI had a significantly positive effect on various types of medical expenditures, indicating that the higher the reimbursement ratio was, the higher the medical expenditure would be. Therefore, the government should set the reimbursement ratio of medical insurance within a rational range which will enhance financial risk protection and improve population health, This in turn would not only control the growth of unreasonable medical expenditure, but also incentivize hospitals in getting more profit from medical insurance. The effect of the deductible on various types of medical expenditures was significantly negative, indicating that the higher the deductible was, the lower the medical expenditure. One possible explanation was that the higher deductible, the higher the share of out of pocket (OOP) expenditure. This may generate negative incentives for patients when seeking necessary medical services.
Thirdly, the effect of gender and age on medical expenditures was inconsistent. On the one hand, the older the age was, the higher the medical expenditure of drug, diagnostic testing and medical consumables. Conversely the older the age, the lower the nursing care, bed, surgery and blood expenditure. This may be related to the large proportion of elderly in the sample in this paper. Some studies have demonstrated that the time-to-death is a significant factor in medical expenditures [40, 41, 46]. The length of stay had a significantly positive effect on medical expenditure. The more serious the disease type, the higher the medical expenditures of, drug, diagnostic testing, nursing care and bed expenditure. Interestingly, the more serious the disease type, the lower the medical consumables, surgery and blood expenditures. This may be related to the large proportion of elderly in the sample used. Influenced by traditional Chinese culture, the elderly with more serious diseases (e.g. terminal cancer) may forego treatment.
Finally, after distinguishing for COH, hospitals of different levels had different influences on medical expenditures. In general, the higher the level of hospitals, the lower the total medical expenditure. After the addition of interaction items,including hospital levels and disease type and; hospital levels and the reimbursement ratio, the effect of hospital level, disease type and reimbursement ratio on medical expenditures was inconsistent. The higher the hospital level along with disease severity, and the higher the medical expenditure of, drugs and diagnostic testing. The higher the hospital level was, along with a high the reimbursement ratio, the lower the medical expenditures of drug, diagnostic testing, medical consumables and nursing care. This may be due to the relatively standardized hospital management system and treatment processes, which would save on medical expenditures. Thus China should strengthen hospital management (strict supervision and information technology) and clarify the standards of different classfications of hospitals to ensure that paitents can get effective treatment.
In general, the larger the scale of the hospital, the more resources they had at their disposal, including advanced medical devices, and management systems. At the same time, these types of hospitals received more government subsidies and medical insurance funding; as well as benefitted from more formalized and structured operational systems. Lower-level hospitals had less government subsidies, operated fewer hospital beds and had less advanced technology. In order to generate profit, lower-level hospitals overly prescribed drugs and high-technology tests. Therefore, the medical expenditures of the high-level hospital was not always higher than that of the low-level hospital. Thus per the severity of the disease, different diseases should be treated in different levels of hospitals.
Our findings are consistent with other studies. Newhouse [4] studied the factors driving increased medical expenditure. He found that factors affecting medical expenditures can be roughly divided into three levels. The first level mainly referred to price factors, such as the medical insurance system [5,6,7,8,9,10,11,12,13], government subsidies [14,15,16,17], essential medicines programmed (EMP) [18,19,20] and separation of hospital revenue from drug sales [21,22,23]. These policies influence medical expenditures by affecting the relative or absolute prices of medical services and product [24]. The second level mainly referred to hospital characteristic factors, such as the diagnosis and treatment level [25,26,27], hospital management [28,29,30,31,32,33] and use of advanced medical devices [34,35,36,37,38,39]. The third level mainly referred to the patients characteristics, such as age [40], gender [41], education status [42], and health status [43,44,45]. In summary, our study found price, hospital characteristic and patients factors affected medical expenditures, consistent with Newhouse.
Our study has some interesting findings which can potentially be used for policy recommendations. As it relates to hospital management in China, the hospital level has become an important factor influencing the effective allocation of health resource. In order to improve the quality of medical services and control the increase of medical expenditure, the classification of hospital should become an important focus of health-care reform. Secondly, China should focus on improvements to the hospital management system and medical insurance system, and strengthen the government oversight of hospitals of different levels. In addition, the medical insurance system should be designed to encourage behaviors of physicians and hospitals in order to improve the quality and efficiency of medical services. Further, China examines the marginal effect of factors influencing medical expenditures in order to adopt reform of the healthcare system. Lastly China should implement differentiated reimbursement ratios for different disease types treated in different hospitals levels in order to improve medical insurance reimbursement policies.
Our study has several important contributions. Firstly, the use of the UEBMI data of Chengdu, is the first time a study on the effect of COH on medical expenditure has been undertaken. Secondly, establishing a framework of COH, which can be divided into hospital characteristic factors, price factors and patients characteristic factors was unique. Finally, by adopting multivariate statistical analysis and introducing interaction items into the model, we could better describe the mechanism of the effect of hospital grade, reimbursement ratio, and disease type on medical expenditures. This can be of use for the implementation and improvement of a tiered delivery system. However, the effect of Grade II Level A hospitals, Grade II Level B hospitals, Grade II Level C hospitals and below on different types of medical expenditures were different, needs to be further analyzed.
Our study also has some limitations. First, due to the data limitations, we only studied Chengdu City and the patients were mainly elderly, which may overestimate the effect of COH on HE. Second, the current HE don’t include ER visit and medications. Third, also due to the data limitations, we didn’t address Grade I and III hospitals, only Grade II Level A, B and C hospitals. This in turn may underestimate the effect of COH on HE. Although we only used data with Grade II Level A, B and C hospitals, we could still have similar findings with Grade I and III hospitals. On the one hand, different grade hospitals had the same classification standards and management system in China, such as hospital’s scale, service provision, medical technology and equipment, medical research and so on. The differences among different grade hospitals were very similar. On the other hand, there were many different grade hospitals with a large number of patients in China. According to China's Health Statistics Bulletin in 2019 [47], we found that the number of Grade I, II and III hospitals were 11,264, 9,687 and 2,749 and the number of patients were 12,090,000, 81,770,000 and 92,920,000, respectively. The number of Grade II hospitals were large enough to stand for Grade I and III hospitals to estimate the effect of COH on HE. While there are differences, the effect of COH on HE is an important topic in China, and any evaluation provides important insights. Finally, our data was a compilation of non-severe to severe DRGs (within the same DRG) and associated costs by severity could not be examined.