Project Topics
Get Complete Project Material File(s) Now! »
Placebo DiD regression
To test whether the estimates obtained in Section 5 are not a ected by pre-reform changes uncontrolled for in out models, we preform DiD placebo estimation. These models include as covariates the interaction terms between the DRG incentive measure Δtarif f and dummies for of all semesters in 2010-2013, except for the second semester of 2011 serving as the baseline pre-reform period. Estimation results are presented in Table 1.5.
As mentioned in Section 5, results on single birth mothers who delivered in public hospitals may not stand this test of placebo DiD. In particular, model (1) shows 2 out 3 statistically signi cant coe cients for pre-reform periods (for 1st semester of 2010 and 1st semester of 2011, with 5% and %1 level of signi cance, respectively). As far as other hospital types and mothers groups are concerned, the results indicate no sign that pre-refom periods were a ected through channels other uncontrolled for in our functional speci cations.
As for the estimated post-reform e ects, the obtained results con rm the conclusion that DRG incentives did not signi cantly a ect the probability of scheduled C-section in private non-pro t and for-pro t clinics. A few sporadically signi cant coe cients in models (3) and (6) do not exhibit a clear pattern. In public hospitals 2 out 4 coe cients corresponding to both of 2012 semesters are negative and statistically signi cant. However, in light of the discussed pre-reform estimates, this estimate should be interpreted with considerable caution.
Finally, it is worth noting that similarly to the main models presented in the previous section and even regardless of the results of placebo DiD on the signi cance of pre-reform coe cients, the magnitude of the post-reform estimated e ects remains low and overall unlikely to impact the clinical practise on a scale posing any signi cant concern.
Models with deterministic severity
In this subsection we are testing a di erent approach to measuring DRG tari incentives for scheduled C-sections that relies on a set of simplifying assumptions. In line with other studies on tari incentives in obstetrics (as in, for example, Allin et al. (2015), Gruber et al. (1999)), we drop the tree structure of obstetrician/midwife decisions about child delivery modes and consider that tari can be known with certainty in advance based on ante-partum patient characteristics.
Pay-for-performance (P4P) and Prospective Payment Systems (PPS): review of existing research
A considerable body of research has been produced on the e ects of Prospective Payment Systems (PPS) on a multitude of healthcare indicators. In the economics literature on PPS, it has been argued that such systems can generate a number of positive outcomes. In particular, PPS systems can encourage care providers to favor cost-e ective treatments by limiting, for example, hospital length of stay (thus, mitigating the problem of long wait times and addressing the lack of care accessibility, which has long been a major concern for many countries, including Canada) and prescribing medications and interventions with a proven clinical bene t to patients. This same consideration may also push providers acting under PPS to specialize in procedures in which they are the most cost-e ective and referring other patients to a more suitable care provider. As a result, patients may receive a well-coordinated high-quality treatment. In addition, since PPS payments are known to care providers in advance, as long as a given treatment is reimbursed at a level even marginally higher than expected patient costs, hospitals would have a nancial interest to admit the maximum number of such patients. The fact that most PPS payments are adjusted based on severity of reported diagnoses motivates personnel to implement more rigorous and systematic practices with regards to coding diagnoses, which may contribute to a higher degree of transparency and information accuracy (Busse et al., 2011).
However, all these organizational bene ts may be compromised by a number of perverse incentives inherent to PPS. In this respect, Ellis (1998) studied a particular market envi-ronment, wherein patients and care providers act under complete information with respect to patient severity and hospital practice patterns. In addition, providers form a duopolis-tic market for the procedure demanded by fully-insured patients, while a third-party payer (insurer) is myopic to patient individual severity and sets in advance a payment sched-ule as a combination of a global budget and reimbursement on a per-patient basis. Ellis (1998) showed that in this setting the maximization of hospital pro ts can be achieved through decreasing costs per hospital stay by \skimping » on care quality, that is leaving certain patients with a sub-optimal level of care. A relatively broad interpretation of the term \quality » makes it possible to consider any preventable event susceptible to compro-mise treatment quality as \skimping » (for example, an unjusti ably early patient discharge from hospital). Finally, arguably the most radical strategy may involve denying hospital care altogether to patients with a perceived higher-than-average cost burden, a situation referred to as \dumping « . Ellis (1998) also suggests that \skimping » and \dumping » can be expected to be practised together.
Empirical studies overall nd moderate to no e ects of PPS on care quality. For example, a descriptive study by Schwartz and Tatter (1998) on patients who underwent colorec-tal cancer surgery in Mount Sinai Hospital in New York from 1983 to 1987 pointed to, among other things, a signi cantly lower rate of blood loss (down to 387cc from 550cc), insigni cant changes in margins of surgical resection (down to 13cm from 18cm) and a 5 percentage point decrease in the rate of post-operative complications. A more recent study by Shin (2018) analyzing DRG reimbursement changes within Medicare comes to the conclusion that increasing payments under PPS did not translate into improvements in healthcare quality measured by 30-day in-hospital mortality and 30-day readmission rates.
Under PPS hospital pro ts can be boosted not only by reducing cost but also by increasing revenue. Thus, hospitals professionals may \up-code » their patients, that is purposefully in ate the severity of diagnoses and/or adding more diagnoses than otherwise would be medically justi ed. Empirical evidence for recourse to such a strategy by hospitals is strong. For example, Dafny (2005) and the earlier mentioned Shin (2018) show that hospitals operating under Medicare and experiencing hikes in DRG reimbursement rates (caused by elimination of age criterion in the rst study, and a reformulation of geographic areas in the latter) may act under the in uence of presumed perverse PPS incentives and eventually alter DRG coding practices and/or, as also suggested by Shin (2018), shift patients into a higher paying tari group. In Europe, evidence for DRG upcoding was recently provided by Januleviciute et al. (2016), who nd, in particular, that surgical DRGs tend to be more liable to upcoding than medical ones.
In some cases, hospitals may also be incentivised to over-provide well-reimbursed services, even despite them having little to no expected clinical bene t for the patient, leading to what is referred to in the economics literature as supplier-induced demand. Evidence from Japan on the use of neonatal intensive care units (NICU) provided by Shigeoka & Fushimi (2014) suggests that introduction of PPS reimbursements for a vast majority of procedures except for those related to NICU (which remained to be reimbursed on a doctor fee-for- service basis), and the ensuing discrepancy between reimbursement mechanisms within an episode of care resulted in a protracted NICU utilization likely achieved by manipulating infants’ reported birth weights. Due to the fact that both Ontario and Japan currently combine hospital PPS and doctor fee-for-service payment schedules, policy implications of this Chapter can be of particular signi cance in the Canadian context. Anecdotal and statistical evidence also suggests that, for example, extensive use of X-ray scanners and other types of medical imaging (Baker, 2010), as well as a more frequent recourse to C-sections as compared to normal deliveries (Johnson & Rehavi, 2016) may at times by accounted for by pro t-making considerations, at least in the US context.
Perhaps due to the aforementioned potential negative rami cations associated with PPS systems, it is very uncommon to see them implemented in their pure form. In most countries, they are amalgamated with a form of global budgets (e.g. Germany from 2003, France from 2004), P4P (e.g. QOF in UK from 2004, HQID in USA from 2004) or fee-for-service arrangements (e.g. Ontario from 2012, Japan from 2003).
In practice, PPS systems are usually implemented through a set of tari s unique to each Diagnosis-related group (DRG) oftentimes adjusted by patient severity. Although this approach may arguably render hospital funding more structured and manageable, Geissler et al. (2012) have expressed concerns that in Europe DRGs may not predict patient costs as precisely as a basic set of patient characteristics, such as indicators for age groups, Charlson index, patient transfer during hospital stay and occurrence of an adverse event. However, in the context of the Ontario reform, it is di cult to compare e ectiveness of HBAM funding formulas since they were not made publicly available (for more information on HBAM see Subsection 3.2)
Pay-for-performance (P4P)
In theory, healthcare providers’ behaviour has been traditionally considered through the lens of contract theory with the goal of describing environments where principals, agents and nal service consumers can have areas of both overlapping and con icting interest. Holmstrom and Milgrom (1991) initiated the debate on the multitasking problem and its implications. Their model suggests that in the presence of noise in a set of performance in-dicators determining agents’ pay-o s, in order to prevent agents from skimping on poorly measured activities and to align agents’ behaviours with social interests, a risk-neutral regulator can exert its in uence through a broad range of leverages, such as in uencing as-set ownership structure, imposing restrictions on performing tasks, in uencing incentives to perform outside non-contractual activities, as well as imposing specialization among providers. A partial adaptation of this general framework to the healthcare context was proposed by Eggleston (2005), who models a benevolent purchaser reimbursing a single provider with a mixed payment scheme calculated as a fraction of incurred cost and a performance premium. The model predicts that under a pay-for-performance scheme reallocation of e ort towards enhancing performance with respect to a relatively more incentivized quality measure can be attenuated by introducing mixed payment systems, such as partial capitation, insofar as it reduces providers’ nancial burden (i.e. supply-side cost-sharing) of ensuring unrewarded dimensions of care.
However, this rather intuitive conclusion may no longer hold true if other features speci c to healthcare markets are allowed for in a model. Notably, as suggested, for example, by Mullen (2010), if dimensions of care are related one to another, such that e orts in incentivized dimensions of care can compensate for a decrease in unrewarded ones, then an overall negative impact of multitasking on unrewarded quality dimensions can be considerably dampened or even reversed. Including providers’ altruism motivation can have considerable theoretical implications on providers’ optimal behavior. Under perfect observability of all parameters and a set-up featuring providers with a heterogeneous level of altruism, Siciliani (2009) shows that an increase in prices for services may discourage production outputs by providers with a medium level of altruism, while increasing service volumes by doctors with low and high degrees of altruism. However, since these changes occur in opposite directions, the impact of price incentives on the overall production volume is ambiguous. Nevertheless, under a di erent set of assumptions, predictions can drastically change. For example, accord-ing to Markis and Siciliani (2013), if a limited nancial liability is ensured to partially altruistic providers who can select patients and whose cost functions are not observed by a purchaser, providers with low and high degrees of altruism will systematically skew their production levels compared to the rst-best solution, the direction of this output distortion depending on agents’ unobservable e ciency type.
Table of contents :
1 Impact of Tari Renement on the Choice Between Scheduled Cesarean Section and Normal Delivery: Recent Evidence from France
1 Introduction and literature review
2 Obstetric care in France: institutional context
3 Data
4 Empirical strategy
5 Main results
6 Robustness checks
7 Discussion and policy implications
8 Conclusion
9 Appendixes
2 Impact of Quality-based Procedures on orthopedic care quantity and quality in Ontario Hospitals
1 Introduction and literature review
2 Pay-for-performance (P4P) and Prospective Payment Systems (PPS): review of existing research
3 Canadian healthcare context
4 Research questions
5 Data
6 Models and main results
7 Robustness checks
8 Discussion and conclusion
9 Appendixes
3 Impact of Child Subsidies on Child Health, Well-being and Investment in Child Human Capital: Evidence from Russian Longitudinal Monitoring Survey 2011- 2017
1 Introduction and literature review
2 Maternity Capital institutional context
3 Data and descriptive statistics
4 Empirical strategy and main results
5 Robustness checks
6 Discussion and conclusion
7 Appendixes
General conclusion
Bibliography
