Effects of fee-for-service, diagnosis-related-group, and mixed payment systems on physicians’ medical service behavior: experimental evidence

Abstract

Background Healthcare reforms in many countries have shown a movement from pure payment systems to mixed payment systems. However, there remains an insufficient understanding of how to design better mixed payment systems and how such systems, especially Diagnosis-Related-Group (DRG)-based systems, benefit patients. We therefore designed a controlled laboratory experiment to investigate the effects of fee-for-service (FFS), DRG, and mixed payment systems on physicians’ service provision. Methods A total of 210 medical students were recruited from Capital Medical University as subjects. They, in the role of physicians, were randomly divided into seven groups and chose the quantity of medical services for different patient types under pure FFS, pure DRG, or mixed payment schemes that included two FFS-based mixed payment schemes and three DRG-based mixed payment schemes. There were five rounds of each group of experiments, and each subject made 18 decisions per round. The quantity of medical services provided by subjects were collected. And relevant statistics were computed and analyzed by nonparametric tests and random effects model. Results The results showed that the physicians’ overprovision (underprovision) of services under FFS (DRG) schemes decreased under mixed payment schemes, resulting in higher benefit to patients under mixed payment schemes. Patients’ health conditions also affected physicians’ behavior but in different directions. Higher disease severity was associated with higher deviation of physicians’ quantity choices from the optimal quantity under DRG and DRG-based mixed payment schemes, while the opposite was found for FFS and FFS-based mixed payment schemes. Conclusions Mixed payment systems are a better way to balance physicians’ profit and patients’ benefit. The design of mixed payment systems should be adjusted according to the patient’s health conditions. When patients are in lower disease severity and resource consumption is relatively small, prospective payments or mixed systems based on prospective payments are more suitable. While for patients in higher disease severity, retrospective payments or mixed systems based predominantly on retrospective payments are better.

Physician altruism under the change from pure payment system to mixed payment schemes: experimental evidence

“…Specifically, from the pure DRG (FFS) to mixed-DRG (mixed-FFS), π ( q ) increased from 8.28 (7.98) to 9.62 (9.50); and r increased from 0.26 (0.40) to 0.53 (0.50) ( p < 0.001, matched-pairs WSR test), respectively. In comparisons of the increase of r of different l ( x , y , z ) between the pure payment system and mixed payment schemes, we found similar results to the improvement of quantity in our published paper [ 39 ]; namely, for x , y, and z , the r increased from 0.26 to 0.64 (Mix-DRG-4, p < 0.001, matched-pairs WSR test), 0.23 to 0.79 (Mix-DRG-6, p < 0.001, matched-pairs WSR test), and 0.61 to 0.78 (Mix-FFS-8, p = 0.001, matched-pairs WSR test), respectively. They are detailed in Table 1 .…”

“…Based on the OLS regressions, we further implemented Tobit regressions to conduct robustness tests and yielded similar results (see additional file 3 ). In addition, the random effects model in our previous study also yielded similar results [ 39 ]. Therefore, we believe that causality is relatively reliable even if selection bias of subjects exists based on these analyses.…”

“…Of note, α had a significant negative correlation with π ( q ) (Spearman’s ρ = − 0.777, p < 0.01) and a significant positive correlation with B ( q ) (Spearman’s ρ = 0.913, p < 0.01). Our previous study also showed an increase in B ( q ) when switching from the pure payment system to mixed payment scheme [ 39 ]. We further explored the relationship between α and B ( q ) in different payment systems.…”

“…In order to compare the impact of the change in payment system on physicians’ altruism, we used a within-subject design, which means subjects had to participate in two parts of the experiment: Part I was incentivized by non-blended FFS or non-blended DRG, and Part II by mixed payment schemes, which is a mixture of DRG and FFS in different proportions. The design of FFS, DRG, and a mixture of the two, as well as the parametric of R ( q ), π ( q ), c ( q ) and B ( q ) are detailed in our recently published article [ 39 ]. Although the stylized decision setting abstracts from the complexity of real-world physicians’ decision-making, it can isolate the trade-off between the physicians’ profit and the patients’ health benefit and still inherit the incentives occurring in a real-world setting [ 26 ].…”

“…The experiment involved seven groups (see [ 39 ]), but this study only focused on the first five groups. They are three groups of DRG and mixed-DRG payment schemes (Mix-DRG-2, Mix-DRG-4 and Mix-DRG-6) and two groups of FFS and mixed-FFS payment schemes (Mix-FFS-6 and Mix-FFS-8).…”