Browsing by Subject "Health care"

In this dissertation, I investigate the impact of administrative burdens on the US health care sector. Using observational data---particularly medical claims from Medicare----and policy variation in the administrative burdens to which health care providers are exposed, I use causal inference methods to understand the effects of various administrative burdens on economic, health, and fiscal outcomes in multiple contexts within the US health care system. I also use theoretical and structural modeling techniques to highlight and quantify the trade-offs faced by economic agents in the health care system and the impact of policy choices. Using turnover in the identity of private contractors that administer Traditional Medicare, I first find that exposing providers to the increased administrative burden imposed by higher-denial contractors does not reduce Medicare spending despite increasing claim denials. The increased administrative burden also leads providers to invest in billing effort, consolidate into larger firms, and earn lower profits. Next, I use similar variation across contractors to show that Medicare coverage restrictions slow the adoption of new medical procedures. Furthermore, I find that the diffusion patterns induced by these administrative burdens are consistent with social learning by providers about the value of the innovations, motivating a structural model of provider learning that indicates that coverage restrictions slow the learning process of the medical community. Finally, I use the staggered roll out of a prior authorization regulation along with criminal and civil lawsuits to identify the effects of ``pay-and-chase" litigation and an administrative burden regulation on the prevalence of health care fraud. I find that prior authorization was extremely effective at reducing health care spending without causing any adverse patient health outcomes, while litigation was much less effective. In conclusion, this dissertation finds that the administrative burdens that permeate the US health care sector have major impacts on market structure, innovation, and health care fraud, with the benefits and costs of these burdens being highlight context dependent.

In this dissertation, we present four novel contributions to the field of statistics with the shared goal of personalizing medicine to individual patients. These methods are developed to directly address problems in health care through two subfields of statistics: probabilistic machine learning and causal inference. These projects include improving predictions of adverse events after surgeries, or learning the effectiveness of treatments for specific subgroups and for individuals. We begin the dissertation in Chapter 1 with a discussion of personalized medicine, the use of electronic health record (EHR) data, and a brief discussion on learning heterogeneous treatment effects. In chapter 2, we present a novel algorithm, Predictive Hierarchical Clustering (PHC), for agglomerative hierarchical clustering of current procedural terminology (CPT) codes. Our predictive hierarchical clustering aims to cluster subgroups, not individual observations, found within our data, such that the clusters discovered result in optimal performance of a classification model, specifically for predicting surgical complications. In chapter 3, we develop a hierarchical infinite latent factor model (HIFM) to appropriately account for the covariance structure across subpopulations in data. We propose a novel Hierarchical Dirichlet Process shrinkage prior on the loadings matrix that flexibly captures the underlying structure of our data across subpopulations while sharing information to improve inference and prediction. We apply this work to the problem of predicting surgical complications using electronic health record data for geriatric patients at Duke University Health System (DUHS). The last chapters of the dissertation address personalized medicine from a causal perspective, where the goal is to understand how interventions affect individuals not full populations. In chapter 4, we address heterogeneous treatment effects across subgroups, where guidance for observational comparisons within subgroups is lacking as is a connection to classic design principles for propensity score (PS) analyses. We address these shortcomings by proposing a novel propensity score method for subgroup analysis (SGA) that seeks to balance existing strategies in an automatic and efficient way. With the use of overlap weights, we prove that an over-specified propensity model including interactions between subgroups and all covariates results in exact covariate balance within subgroups. This is paired with variable selection approaches to adjust for a possibly overspecified propensity score model. Finally, chapter 5 discusses our final contribution, a longitudinal matching algorithm aiming to predict individual treatment effects of a medication change for diabetes patients. This project aims to develop a novel and generalizable causal inference framework for learning heterogeneous treatment effects from Electronic Health Records (EHR) data. The key methodological innovation is to cast the sparse and irregularly-spaced EHR time series into functional data analysis in the design stage to adjust for confounding that changes over time. We conclude the dissertation and discuss future work in Section 6, outlining many directions for continued research on these topics.

This dissertation explores the industrial organization of two U.S. health care markets, outpatient dialysis and long term acute care hospitals, and examines how health care provision responds to market structure, ownership structure, and economic incentives. The first chapter introduces the research agenda presented in this document.

The second chapter looks at whether dialysis quality in the U.S. is an outcome of market structure and competition. This analysis uses a rich panel from the United States Renal Data System (USRDS) that includes data on virtually the universe of dialysis patients and providers in the U.S. from 1998 to 2012. I find that providers are able to exercise market power by reducing the clinical quality of dialysis and still capturing market share. This is possible because patients have horizontal preferences (travel costs) and often face congestion at dialysis facilities. Both of these sources should be considered when developing policies aimed at improving quality or access in this industry. I develop and estimate an entry game where providers choose capacity and compete on quality. I use the model to simulate policy counterfactuals that explore how to cost-effectively promote quality and access in dialysis. My simulation results reveal that offering providers more money for dialysis produces small improvements in patient welfare because providers are able to use local market power to capture the majority of the surplus such policies. However, policies targeting the sources of market power, such as subsidizing travel, provide more cost-effective improvements.

The third chapter explores the transference of corporate strategies to dialysis facilities that are acquired by for-profit national chains. I find evidence showing how acquired facilities change their internal practices to resemble facilities previously owned by the chain. These changes increase the revenue productivity of the acquired facility but yield weakly worse patient outcomes along many measures.

The final chapter examines the impact of Medicare's prospective payment system for long-term acute-care hospitals (LTCHs) on the timing of patient discharge. This payment policy provides modest per-diem reimbursements at the beginning of each patient's stay before jumping discontinuously to a large lump-sum payment after a pre-specified number of days. I find that LTCHs respond to the financial incentives associated with this system by disproportionately discharging patients shortly after they qualify for the lump-sum payment. I find that this occurs more often at for-profit hospitals, chain hospitals, and hospitals co-located with general hospitals. I then estimate a dynamic structural model to evaluate counterfactual payment policies that would provide substantial savings for Medicare.

The way in which health care providers make treatment decisions and the incentives which drive these choices are the subject of much policy and research discussion. Financial incentives have been used to steer provider treatments to more cost-effective options, such as in Medicare's recent Accountable Care Organization model, while acquisition of providers by a firm can provide incentives for providers to treat patients differently than prior to acquisition. In this dissertation, I use a variety of administrative data sources to study the effects of these financial incentives on both physicians and dialysis clinics.

In Chapter 1, I study the effects of integration between referring physicians and specialists in cardiology. To address concerns of endogeneity of integration, I exploit a change in Medicare payment rates which increased the financial benefit to vertically integrating for cardiologists. Instrumental variables estimates show that cardiologists who work in the same practice as cardiac surgeons are 7.7% more likely to refer patients for surgery rather than more conservative options. Patients diagnosed by integrated cardiologists in turn have worse mortality and readmission outcomes, with 18.7% higher mortality risk and 13.4% higher risk of readmission for AMI within 180 days. This is in spite of the fact that patients diagnosed by integrated cardiologists have 7.8% higher medical spending in the 180 days following diagnosis. I provide evidence that these effects are not driven by inherent risks of invasive surgery or selection on patient observables, but worse outcomes for patients receiving the most conservative treatment option.

In Chapters 2 and 3, which are joint with Paul Eliason, Ryan McDevitt, and James Roberts, we use a rich panel of Medicare claims data for nearly one million dialysis patients to advance the literature on the effects of mergers and acquisitions by studying the precise ways in which providers change their behavior following an acquisition and the effects of bundled payment reforms. We base our empirical analysis on more than 1,200 acquisitions of independent dialysis facilities by large chains over a twelve-year period and find that chains transfer several prominent strategies to the facilities they acquire. Most notably, acquired facilities converge to the behavior of their new parent companies by increasing patients' doses of highly reimbursed drugs, replacing high-skill nurses with less-skilled technicians, and waitlisting fewer patients for kidney transplants. We then show that patients fare worse as a result of these changes: outcomes such as hospitalizations and mortality deteriorate, with our long panel allowing us to identify these effects from within-facility or within-patient variation around the acquisitions. Because overall Medicare spending increases at acquired facilities, mostly as a result of higher drug reimbursements, this decline in quality corresponds to a decline in value for payers. We conclude the paper by considering the channels through which acquisitions produce such large changes in provider behavior and outcomes, finding that increased market power cannot explain the decline in quality. Rather, the adoption of the acquiring firm's strategies and practices drives our main results, with greater economies of scale for drug purchasing responsible for more than half of the change in profits following an acquisition.

Chapter 3 studies the effect of a bundled payment reform in 2011 for dialysis providers. Using an instrumental variables strategy, exploiting a biological interaction between a patient's elevation of residence and their health outcomes, we show that bundled payment reform yielded better hospitalization outcomes for patients, but worse transfusion outcomes. This is consistent with the decreased use of drugs to prevent blood transfusions observed after the reform. In addition, we find significant patient and firm heterogeneity in responses.

This dissertation explores the effects of private equity (PE) investment on product quality among healthcare providers. In the first essay, I study the determinants of PE manager behavior, focusing on the role of product market competition. Using the nursing home setting as a backdrop, I consider the broader question of whether and how product market competition shapes the impact of PE acquisitions on consumers. By studying acquisitions of skilled nursing facilities by PE firms, I find that PE-owned providers exhibit greater competitive sensitivity—in that they compete more aggressively when competitive incentives are comparatively strong, and exploit market power more aggressively when competitive incentives are comparatively weak.

To investigate whether PE managers respond differently than non-PE managers to competition, I consider two sources of variation in competitive incentives facing nursing homes. First, I exploit the fact that nursing homes compete with one another in geographically segmented markets to contrast facilities according to the levels of local competition they face. I find significant heterogeneity in the effect of PE ownership according to levels of local market concentration. In highly competitive markets, PE owners increase staffing by $101,783 worth of care annually (enough to increase registered nurse (RN) hours by 20.8% of the mean), while actually reducing staffing in less competitive markets. Second, I show that PE-owned nursing homes respond more strongly to policies intended to spur competition. I study the introduction of the Five-Star Quality Rating System, a policy that increased the salience of staffing for consumers. Following its introduction, PE-owned facilities increased their staffing by an average of $39,118 worth of care more than their non-PE counterparts. Moreover, PE managers more aggressively shift their staffing composition towards RNs in response to the rating system's specific emphasis on RN staffing (RN expenditure increasing by 14.7% of the mean, with licensed practical nurse (LPN) expenditure decreasing by 4.9% of the mean): in total, the share of RN staffing increased by 1.9 percentage points (17.3% of the mean) more than non-PE facilities.

In the second essay, I assess how PE acquisitions influenced the readiness and outcomes of nursing facilities during the onset of the COVID-19 pandemic. With over 40% of U.S. COVID-19 deaths occurring in nursing homes, long-term care is a critical setting in which we must better understand the impact of PE ownership during the coronavirus pandemic. I find PE ownership to be associated with a mean decrease in the probability of confirmed COVID-19 cases among residents by 7.1 percentage points and confirmed staff cases by 5.4 percentage points. PE was also associated with a decreased probability of PPE shortages—including N95 masks, surgical masks, eyewear, gowns, gloves, and hand sanitizer. However, facilities previously (but not presently) owned by PE firms did not fare similarly well. I observe that prior PE ownership may result in increased PPE shortages and a potentially greater likelihood of resident outbreaks. This suggests that the contribution of PE ownership to improved COVID-19 outcomes is a result of active management during the pandemic, rather than the legacy of interventions undertaken beforehand.