Reimbursement of Health Care Providers
Effects of Introducing Prospective Payment on Length of Stay in Psychiatric Care - Evidence from a recent reform in Germany
Franziska Valder, KU Leuven (Belgien)
Previous literature has shown that hospital discharge decisions are not purely medically motivated but that financial incentives play an important role. Prospective payment schemes set a strong incentive for hospitals to keep length of stay low. We analyse how financial incentives influence length of stay in psychiatric care in Germany. More specifically, we analyse a recent reform in Germany where a prospective payment scheme (PEPP - Pauschalierendes Entgeltsystem Psychiatrie und Psychosomatik) replaced constant daily lump sum charges for psychiatric patients. The system was introduced voluntarily since 2013, and is obligatory since 2018.
In a first step, we analyze selection of hospitals into the PEPP scheme. As a second step we try to disentangle to what extend changes in length of stay are influenced by selection of hospitals into the new reimbursement scheme or are a direct effect of the reimbursement scheme. Finally, we use relative change in reimbursement, differences in occupancy rates and kink-points in the reimbursement schedule to quantify the extend of early discharge triggered by the new system. We use the official hospital episodes statistics for Germany with information on diagnoses as well as hospital and patients characteristics for the years 2012 until 2017. Our main outcome of interest is patient length of stay (LOS). In order to disentangle raw effects of the PEPP reimbursement scheme from selection into the scheme we use staggered differences in differences estimations for hospitals with similar switching dates.
Our results show that there is large selection of hospitals into the new reimbursement in terms of higher average length of stay and higher average occupancy rates. Accounting for these selection effects, reveals large negative effects of prospective payment on length of stay, for the majority of diagnoses. The new payment system leads on average to a reduction of length of stay of 4.5 days.
Using a large payment scheme reform in Germany we analyse the effect of prospective payment on length of stay in inpatient mental health care. We find that specific hospitals select into such a system, with higher average length of stay and high average occupancy rates. Once we take this into account, we find very large negative effects of prospective payment on length of stay.
Franziska Valder, KU Leuven
SImon Reif, ZEW Mannheim
Harald Tauchmann, FAU Nürnberg
Can we design better contracts with GPs and specialists to manage cost containment?
Wolfgang Gick, IFN Stockholm & Unibz
The paper studies how health care organizations can be improved in terms of cost containment. Cost containment in physician agency depends on the agency costs arising in the organization. Agency costs arise when agents deviate in their actions from what is in the best interest of an organization. To reach incentive alignment among players, a program designer needs to pay information rents that may add up along the tiers of the health care organization under consideration. A decision maker may so, given her power to decide on treatment because of his freedom of interpreting a result (e.g. from a diagnostic technology) use a variety of treatment options – a fact well researched in the literature on DRGs. To contain costs thus makes it necessarily to study how this freedom results in additional inefficiencies in health care delivery and to a loss in cost containment.
The paper uses a contract-theoretic multi-agent framework to deliver a theoretical foundation for the ongoing analysis on the efficiency and cost contaiment in health care delivery. I consider a simple insurer-physician hierarchy to study the agency costs that emerge in a setup with endogenous information structures. A risk-neutral principal P (public health care authority, program designer) delegates the task of carrying out a medical service to two agents: a general practitioner (GP) and a specialist (S). Both GP and S have productive tasks in that they deliver health care services (therapy, treatment according to a prescribed output target), as well as informational tasks (diagnosis, learning a signal on the patient’s state of health). They face different costs when carrying out possible diagnostic tests and therapy. I so characterize the costs differences and agency rents that emerge in health care delivery.
A first result of the agency-theoretic setup is that more and better diagnostic options will not necessarily increase patient beneﬁt when information structures are not considered “hard” as explained above in one benchmark. Better diagnostic abilities may generally improve efficiency at the level of the GP as the latter may provide an adequate correspondence between patient’s health status and the best available treatment option. A second result is that improved diagnostic technologies also create freedom for interpretation. This all may be the source of additional costs that should be limited to improve on cost containment.
This paper delivers some novel theoretical foundations to study cost containment in health care agency where players face different forms of discretion to use diagnostic technologies and to carry out treatment. By showing how information rents accrue along the time line of the contracting game, I illustrate how speciﬁc forms of control and standardization may reduce unnecessary discretion to permit better cost containment.
Wolfgang Gick, IFN Stockholm & Unibz
It probably worked: A Bayesian approach to evaluating Israel’s 2013 hospital reform
Martin Siegel, TU Berlin
Israel changed its reimbursement scheme in hospital care from per-diem to procedure-specific lump sum payments for a set of elective procedures in 2013. The aim was to decrease waiting times by incentivizing hospitals to treat more patients in less time. This study evaluates the effect of this reform on annual case numbers and on patients’ length of stay (LoS) in hospitals.
We use administrative data from Israel’s Ministry of Health on 14 medical procedures in three medical fields for the years 2005-2016. We employ a duration analysis framework to estimate the probability that the reform decreased patients’ LoS. We then use the absolute changes in the numbers of patients treated per year to assess whether annual changes after the reform differ from those before the reform. We opt for a Bayesian estimation approach, which allows us to add a different perspective to the interpretation of policy impact analyses that goes beyond a merely dichotomous decision over hypotheses.
A total of n=83,318 cases were included in our analysis. The point estimates derived from the posterior distributions of the time ratios suggest that LoS decreased after the reform for 12 out of 14 procedures. The estimated decreases vary considerably between different procedures, ranging from 2.5% to 20%. Adding the Bayesian perspective, our analysis yields a probability of over 95% to observe a shorter LoS after the reform in 1 out of 5 surgical procedures, in 6 out of 7 urological procedures and in 1 out of 2 gynecological procedures. Somewhat surprisingly, only two procedures yield a probability of over 95% to observe an increase in absolute changes in case numbers. We interpret these high probabilities analogously to p-values such that a probability of over 95% to have a shorter LoS after the reform roughly corresponds to a significant effect with a p-value below 5% (when testing against a one-sided null hypothesis).
LoS decreased with a high probability in 8 out of 14 procedures after this policy change, but the extent varied considerably between procedures and across medical fields. Finding a clear effect on the numbers of patients treated in only two procedures is somewhat surprising in the light of the decreased LoS. A potential explanation may be that hospitals increase the lengths of stay after procedures which may still be paid per-diem, or shift their efforts to procedures with more lucrative lump-sum reimbursements. We speculate that balancing economic and medical considerations depends on the particularities of the medical procedures under consideration and may lead to different responses in different medical fields.
Martin Siegel, TU Berlin
Ruth Waitzberg, The Smokler Center for Health Policy Research, Ben-Gurion University of the Negev
Wilm Quentin, TU Berlin
Reinhard Busse, TU Berlin
Dan Greenberg, Ben-Gurion University of the Negev
Rewards for information provision in patient referrals: a theoretical model and an experimental test
Mathias Kifmann, Universität Hamburg, Hamburg Center for Health Economics
When patients are referred from one physician to another, the provision of information by the referring physician is important for medical decision making. However, empirical studies find that many referrals do not include a transfer of information and, if they do, the provided information is often insufficient. We therefore study whether bonus payments for information provision can improve the information flow.
We present a theoretical model based on altruistic primary care physicians (PCPs) who can transfer no, low- or high quality-information. Either the patient or the receiving physician can benefit from information provision. PCPs’ preferences may exhibit loss aversion. We develop hypotheses relating the information transfer to the benefit generated and the bonus payment for information provision. We predict how information transfers change if the payment exceeds thresholds relating to the cost of information provision. PCPs can be expected to provide more low- and high-quality information as the bonus payment increases. We test our predictions in a controlled laboratory experiment. Subjects in the role of PCPs decide on passing on information to subjects in the role of specialists while referring a patient. The monetary value of patient benefit is transferred to the German branch of Doctors of the World. Experimental conditions vary with regard to who benefits from information provision (patient vs. specialist), who has higher earnings (PCP vs. specialist/patient), and, whether the bonus payment is introduced cost neutrally.
PCPs in the experiment pass on more low- and high-quality information as the bonus payment increases. If the bonus payment is at least as high as the costs for the provision of high-quality information, PCPs provide less low-quality information and more high-quality information than in decision tasks with lower bonus payments. This is in line with our model considering loss aversion in addition to altruism. Moreover, we observe that PCPs’ reactions to increases in the bonus payment are similar regardless of whether the bonus payment is introduced cost neutrally or not. If specialists benefit from information provision, PCPs mainly focus on their own profit and provide less high-quality information than if patients benefit. However, this effect depends on the relative earnings of PCPs and specialists.
Our theoretical model and experimental observations reveal that a bonus payment for information provision can improve the information flow between physicians in patient referrals. However, low- as well as high-quality information are induced by such a payment. It is therefore not clear whether a bonus payment is welfare increasing. The case for the introduction of a payment is strongest if implemented cost neutrally.