Given the absence of computerized information systems, this study was constrained by what data could feasibly be collected by hand. In particular this limited the study of unnecessary care, which was extremely time-consuming. If inpatient records had covered more years and more hospital years had been employed in the regression analyses, the factors that were not significant might have been statistically significant, and the goodness of fit of the models might have improved. Moreover, only two tracer conditions were studied, and only inpatient care was assessed. These conditions allow for only a limited degree of overprovision, especially in terms of lab tests and drugs, and so are likely to have underestimated the amount of unnecessary care.
Productivity assessment, for both unidimensional ratio analysis and DEA, had two major shortcomings: changes in the quality of care (other than the component of unnecessary care) were ignored, and changes in case mix were not adjusted for. Over time it is likely that the quality of care improved, and that case mix changed; thus this study is likely to have exaggerated the degree to which productivity declined over time. However, these points do not necessarily affect the finding that there was a lack of a statistically significant relationship between bonus type and hospital productivity. It is possible that hospital expansion and increased numbers of staff per hospital may have obscured any increase in quantity stimulated by the bonus payment.
The effect of bonus payment is likely to depend on both the type and the amount of bonus. However, data on the amount of bonus was a sensitive question, and panel hospitals either refused to provide data or provided data that were not considered reliable. In a related hospital census survey , it was found that there was no statistically significant difference in the average amount of bonus per doctor across the types of bonus. However, the variation of bonus payment among hospital doctors increased with the progression of bonus types. These findings suggest that exclusion of bonus amount in the analysis may not have introduced much error, because the average amount did not vary much, and it was the way in which the bonus was distributed that mattered.
In the analyses of individual hospitals and of trends, it was found that for some hospitals, the changes in indicators happened in the year of switch and for others the following year. There are two possible reasons. First, when hospitals responded to the bonus switch would depend on whether the bonus switch happened early or late in the year. Second, the speed of effects would also depend on how hospitals responded. Some may have responded quickly by admitting more patients, providing more existing services and prescribing more and costlier drugs. Others may have had to wait for the purchase of equipment (e.g. CT scanner) or the training of personnel (e.g. for open-heart surgery).
It was clear from the analysis that the increase in hospital cost recovery was not a result of improvement in hospital productivity, since between 1978 and 1997, cost recovery increased from 71% to 96% and the DEA efficiency score decreased from 97% to 73%. Apart from the increase in the proportion of unnecessary expenditure and the bonus system changes, there are at least four additional factors that may have affected the increase in hospital cost recovery.
First, over the previous 20 years the Chinese government had been raising medical care prices. Although these on average were set at about 50% of total cost , prices were higher relative to their cost than 10 years previously.
Second, due to the liberalization of the pharmaceutical market, the prices of drugs had increased considerably, doubling from 1980 to 1990 . This benefited hospitals, because they were allowed to sell drugs at a 20% mark-up.
Third, the development of new technologies encouraged the introduction of new treatments and drugs that usually had much higher regulated prices than traditional treatments and drugs. For example, before 1980, imported drugs accounted for less than 1% of the Chinese drug market, while by 1993 the sale of imported drugs made up 30%–55% of the market in major cities such as Beijing and Shanghai . Before 1980 there was no CT or MRI in China, while by 1995, CT scanners had became very popular in county hospitals and MRI could be found in any city at and above the prefecture level. Because the prices of high technology services could be set above cost, and the prices for the new imported drugs were 5 to 10 times the prices of the traditional and domestic drugs, hospitals obtained much more profit from using these services and drugs.
Finally, government budget reform increased the financial accountability of public hospitals. Reducing waste and saving costs became a major management concern, helping to improve hospital cost recovery. However, this study suggests that hospital characteristics explained around 55% of the variation in hospital cost recovery, implying that hospital management capacity and ability to control cost varied a great deal.
Between 1978 and 1997, the visits/admission ratio and admissions/operation ratio went down. The analysis showed that a bonus switch from one with a weaker economic incentive to one with a stronger incentive was one of the factors explaining the decreases in the visits/admission ratio and in the admissions/operation ratio, suggesting there must be other factors affecting these ratios. The number of visits to county hospitals was increasing until around the middle of the 1980s, and then started falling. Beginning at the end of the 1970s, rural economic reform brought about a rapid increase in peasants' income, which could have encouraged an increased demand for health care . This may have been felt particularly at higher-level health institutions, such as county hospitals, because of the collapse of the rural Cooperative Medical System and the decrease in the number of rural doctors after the rural economic reform . However, the increase in the number of rural doctors after the mid-1980s, when rural private practice was permitted, may have pulled patients back from county hospitals. In addition, the increase at that time in the number of county level health institutions, such as stations of maternal and child health and hospitals of Chinese traditional medicine, and the increase in medical prices, would have decreased the demand for county hospital care.
It is difficult to explain fully the steady increase in the number of admissions and the number of operations. Since inpatients can be admitted only through the outpatient department and only inpatients can be operated on, it is obvious that the hospitals admitted more and more from among the outpatients and performed more and more operations for the inpatients. There are several possible reasons for this. First, the case mix may have changed so that more patients needed to be admitted and operated on. Second, the development of technology and changes in medical criteria for admissions and operations may have led to more patients' being admitted and operated on. Third, the increase in the numbers of beds and doctors may have permitted needed admissions and operations that had not been possible before due to lack of inputs. Finally, related to the major hypothesis of this research, the changes in doctor payment system may have motivated staff to provide more unnecessary admissions and surgical operations. The study has shown that a bonus switch appeared to bring about an increase in the visits/admission ratio, but case notes did not permit a judgement as to whether an admission was necessary or unnecessary.
Although the DEA efficiency score decreased when panel hospitals switched from non-bonus to flat bonus, this does not mean that the bonus switch helped to decrease hospital productivity. This is, first, because the DEA efficiency score was generally decreasing over time; and second, because in two of the three hospitals (Liangshan and Changyi), the bonus switch appears to have helped to slow down the rate of decrease in the DEA efficiency score.
Conclusions and policy implications
Based on these analyses, we can draw several conclusions. First, there was a steady increase in hospital revenue, and bonus type was a significant factor explaining its variation across hospitals and years.
Second, a considerable proportion of unnecessary expenditure out of total expenditure was identified, and there was a relationship between the bonus system and unnecessary care. Analyses showed that the bonus system was positively correlated with the unnecessary care indicator, implying that the higher the expected incentive of the bonus system, the higher the proportion of unnecessary expenditure.
Third, although hospital productivity decreased over time, a bonus switch from flat bonus to revenue-related bonus appeared to increase hospital productivity in the year of the switch. A bonus switch from non-bonus to flat bonus was not similarly able to reverse the trend of hospital productivity, but it seemed that the rate of decrease in hospital productivity was slowed down by a bonus switch.
Fourth, hospital cost recovery increased over time. The study suggests that the bonus switch brought about an increase in cost recovery and that the bonus system was positively correlated with hospital cost recovery.
Last and in general, the research suggests that the bonus change over time contributed significantly to the increase in hospital service revenue and hospital cost recovery. The increase in unnecessary care and increase in the number of admissions out of the existing number of outpatients, with the bonus system switching from one with a weaker incentive to one with a stronger incentive, suggests that the improvement in hospital cost recovery was achieved at least in part through the provision of more unnecessary care and drugs and through admitting more patients.
There are two policy implications from this study. First, there is little evidence that the performance-related pay system as designed by Chinese public hospitals is socially desirable. It could improve hospital financial sustainability, but did not necessarily lead to improvements in efficiency from a social perspective. The key barrier to achieving the social objectives of performance-related pay was the inappropriate link (whether direct or indirect) between bonus payment and hospital revenue. Hospital bonus distribution should be based on doctor performance measured by indicators that are in line with the desired overall performance of the health care system.
Second, reforms in various countries are characterized by increased exposure of public hospitals to financial risk, in order to increase financial accountability, efficiency and productivity. However there is a risk of encouraging revenue maximization and rent-seeking. Chinese experiences show that when increasing public hospital autonomy, hospitals should be monitored closely by the government, and regulations applied to limit opportunistic behaviour. Otherwise, the containment of government financing to public facilities may result in an increase in the provision of unnecessary care, an increase in health costs to society and a waste in social resources.