Applied econometrics 2

Einleitung / Introduction

Baumol’s (1967) model of ‘unbalanced growth’ yields a supply-side explanation for the ‘cost explosion’ in health care. Applying a testing strategy suggested by Hartwig (2008), a sprawling literature affirms that the ‘Baumol effect’ has both a statistically and economically significant impact on health-care expenditure growth. Skeptics maintain, however, that the proliferation of hi-tech medicine in acute care is clearly at odds with the assumption underlying Baumol’s model that productivity-enhancing machinery and equipment is only installed in the ‘progressive’ (i.e. manufacturing) sector of the economy. They argue that Baumol’s cost disease may affect long-term care, but not acute care.

Methode / Method

To examine the sensitivity of the individual variables on per-capita HCE growth, we apply (variants of) EBA, as suggested by Leamer (1985) and Levine and Renelt (1992). This approach, which has been widely used in the economic growth literature, has become a popular tool for economists who want to test the robustness of the results of their empirical work. In addition, the EBA provides an opportunity to test whether a particular determinant is robustly related to the dependent variable. Our testing strategy also consist in combining Extreme Bounds Analysis (EBA) with an outlier-robust MM estimator.

Ergebnisse / Results

Using panel data for 23 OECD countries, our results provide robust and statistically significant evidence that expenditures on both acute care and long-term care are driven by Baumol's cost disease, even though the effect on long-term care expenditures is more pronounced.

Zusammenfassung / Conclusion

With regard to the EBA estimates, the Baumol variable is found to be greater for LTCE than for ACE. The use of the two-sample t-test reveals that this difference in Baumol coefficients between ACE and LTCE is in both OLS and MM significant at the 1% level. The results reveal also that next to these two variables quite different variables should be considered as further determinants for LTCE and ACE indicating varying drivers.

Authors

Kaan Celebi, Technische Universität Chemnitz

Jochen Hartwig

Anna Pauliina Sandqvist

Einleitung / Introduction

There is complex interplay between nature and nurture. Estimating this interaction between genes and environmental factors is the focus of a new strand of literature. It aims to answer the central question of whether environmental factors can influence the relationship between genes and an outcome of interest? This interaction can go two ways: Environment may be able to 1) reinforce this link or 2) offset genetic differences. The recent availability of genetic information in surveys has made it possible to use genetic information directly. A central problem is that environmental factors are often endogenous. The usual approach is to instrument environment using a natural experiment or reform and estimate the gene-environment-interactions using standard two-stage least-squares regression. This solves the endogeneity problem convincingly. However, individuals may also self-select into environments according to their unobserved potential gains (…associated to their genetic endowment). In this case, traditional two-stage least-squares estimators conflate two different things into a single estimate of the interaction effect: 1) potentially different effects of education by genetic endowment (which we are interested in) and 2) potentially different complier groups by genetic endowment. To obtain effects for well-defined subgroups, we use the marginal treatment effect framework. This allows us to control for differences in observed and unobserved characteristics between complier groups with different genetic endowments.

Methode / Method

Using a simulation example, we show that under selection into gains the two-stage least-squares estimate can be different from the true interaction effect and how estimating marginal treatment effects can solve the problem. Subsequently, show this using data. We investigate the question whether education influence the link between genes and cognitive ability at old age using data from the English Longitudinal Study of Ageing (ELSA).

Ergebnisse / Results

Our results are two-fold: 1) The interpretation of gene-environment interactions is not retained under selection into gains when estimating standard two-stage least-squares models. This selection can generate a bias of unknown size and direction. Marginal treatment effects can recover the true gene-environment interaction effects. 2) For our application, we find a positive gene-environment interaction. Education likely reinforces the effect of genes on old-age cognitive abilities.

Authors

Johannes Hollenbach, RWI – Leibniz-Institut für Wirtschaftsforschung

Hendrik Schmitz, Universität Paderborn

Matthias Westphal, FernUniversität in Hagen

Einleitung / Introduction

Non-adherence to medication regimens immediately after hospital discharge is a widespread problem worldwide. Even short interruptions in pharmaceutical therapy directly after hospital discharge are associated with an increased risk of a discontinuation of medication, readmission due to comparable indications or death, particularly in the context of cardiovascular diseases. On October 01, 2017, an expansion of the hospital discharge management came into effect in Germany. Hospitals now have the option of prescribing drugs. The ability to fill drug prescriptions directly after hospital discharge, without having to visit an outpatient physician first, potentially reduces the supply gap between hospital discharge and prescription filling and might hence improves clinical outcomes such as readmission or death.

Methode / Method

We employ survival analysis methods (Kaplan-Meier, Cox Proportional Hazard) in order to investigate the relation between hospital prescriptions and pharmaceutical persistence (time to first interruption) and event-free survival. The treatment group consists of patients hospitalized for cardiovascular diseases (ICD I.20 - I.25), who got a hospital prescription for associated pharmaceuticals (ATC: B01AC, C07, C08, C09, C10A) during this stay. The control group consists of patients without hospital prescription. We adjust for risk using Entropy Balancing (Elixhauser, age, gender). The analysis is based on anonymized claims data of about 5 million insured from 19 health insurance funds in Germany between 2015 and 2021.

Ergebnisse / Results

Preliminary results indicate, that patients, who receive a prescription during hospital discharge are much more likely to fill the first post-discharge prescription within 14 days after discharge (HR 5.44, p<0.001). Simultaneously, these patients are less likely to interrupt or abort the medical therapy within the first three years after hospitalization (HR 0.92, p<0.001). While we do not observe differences between both groups regarding all-cause mortality (HR 1.01, p>0.9), the likelihood of a second hospitalisation due to a similar indication is greater for those patients with hospital prescriptions (HR 1.39, p<0.001).

Zusammenfassung / Conclusion

The observed positive effect on persistence is in line with existing literature. However, contrary to the subsequent expectation, we did not observe an improvement regarding mortality. One possible explanation may be found in the severity of the hospital cases in question. We observe that severe cases in particular received a discharge prescription during discharge. Particularly severe cases have a higher probability of re-admission or a second medical incident. However, risk adjustment has yet to be finalized.

Authors

Fabian Grünwald, Hamburg Center for Health Economics, University of Hamburg

Christian Kümpel, GWQ Serviceplus AG

Tom Stargardt, Hamburg Center for Health Economics, University of Hamburg

Einleitung / Introduction

Financial incentives matter for physician behavior. We study the financial incentives of physicians when it comes to prescribing generic drugs versus brand-name drugs in Switzerland.

Methode / Method

Based on claims data at the physician level, we use variation in the dispensing regime across cantons to identify the causal effect on prescribing brand-name drugs rather than generic drugs if such are available. We implement our empirical strategy using Entropy Balancing. In particular, we compute the counterfactual drug costs per patient for each physician had she prescribed the cheapest generic drug available instead of the brand-name drug. In other words, we estimate the potential savings for each physician.

Ergebnisse / Results

Preliminary results, show that physicians in cantons allowing self-dispensation have lower potential savings. This could be explained by self-dispensing physicians benefiting from higher markups for brand-name drugs.

Zusammenfassung / Conclusion

This has implications for regulation and health policy regarding self-dispensing regimes and reimbursement, respectively.

Authors

Andreas Kohler, ZHAW

Manuel Langhart, tarifsuisse ag

Louie Saracho, Moser Health Care Solutions AG