Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Value in health in the era of vertical integration
(USC Thesis Other)
Value in health in the era of vertical integration
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
VALUE IN HEALTH IN THE ERA OF VERTICAL INTEGRATION by Jing Gu A Dissertation Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree DOCTOR OF PHILOSOPHY (HEALTH ECONOMICS) August 2020 Copyright 2020 Jing Gu ii Acknowledgements I would like to express my deepest appreciation to my advisor Dr. Neeraj Sood, who is wise, caring, and gave tremendous support and guidance to my research and life throughout my PhD study. I would also like to thank Drs John Romley, Seth Seabury, Peter Huckfeldt, Abby Alpert, and Rebecca Myerson, for their insightful guidance and feedback to my research. Special thanks to my parents and my husband, Jianshu Wu, for their love, support, trust and understanding along the way. iii Table of Contents Acknowledgements ........................................................................................................................ ii List of Tables ................................................................................................................................... v List of Figures ................................................................................................................................. vi Abstract ......................................................................................................................................... vii Chapter 1. Productivity growth of skilled nursing facilities in the treatment of post-acute-care- intensive conditions ......................................................................................................................... 1 Abstract ........................................................................................................................................ 1 Introduction .................................................................................................................................. 2 Methods ....................................................................................................................................... 5 Population ................................................................................................................................ 5 Production function framework ............................................................................................... 5 Measures .................................................................................................................................. 6 Analysis .................................................................................................................................... 8 Results .......................................................................................................................................... 9 Discussion .................................................................................................................................. 11 Appendices ................................................................................................................................. 23 References .................................................................................................................................. 27 Chapter 2. The effects of Accountable Care Organizations (ACOs) forming preferred Skilled Nursing Facility (SNF) network on patient volume, composition and outcomes ......................... 33 Abstract ...................................................................................................................................... 33 Introduction ................................................................................................................................ 34 Methods ..................................................................................................................................... 36 Study population .................................................................................................................... 36 Data sources ........................................................................................................................... 37 Study measures ....................................................................................................................... 38 Empirical approach ................................................................................................................ 39 Results ........................................................................................................................................ 41 Patient volume ........................................................................................................................ 42 Patient characteristics ............................................................................................................. 42 Patient outcomes .................................................................................................................... 43 Discussion .................................................................................................................................. 44 iv Appendices ................................................................................................................................. 53 Chapter 3. Do larger vertically integrated Pharmacy Benefit Managers (PBMs) benefit health plans and consumers? .................................................................................................................... 65 Abstract ...................................................................................................................................... 65 Introduction ................................................................................................................................ 66 Methods ..................................................................................................................................... 68 Study sample .......................................................................................................................... 68 Data sources ........................................................................................................................... 69 Study measures ....................................................................................................................... 69 Empirical approach ................................................................................................................ 70 Results ........................................................................................................................................ 71 Discussion .................................................................................................................................. 74 Appendices ................................................................................................................................. 83 References .................................................................................................................................. 86 v List of Tables Table 1.1 Summary statistics of patient characteristics and outcomes .......................................... 16 Table 1.2 Estimated percent of productivity growth and interaction with vertical integration ..... 22 Table 2.1 Background information for ACOs and their preferred SNF networks in the sample .. 48 Table 2.2 Percent of patients sent from hospitals to preferred SNFs in pre and post period (%) . 49 Table 2.3 Change in patient composition after preferred SNF network was formed .................... 50 Table 2.4 Change in patient outcomes after preferred SNF network was formed ........................ 52 Table 3.1 Summary statistics for UnitedHealthcare plans and plans of Pennsylvania Public Employees vs. other plans before and after the expansion of OptumRx ....................................... 77 Table 3.2 Event study .................................................................................................................... 78 Table 3.3 Regression results comparing United plans before and after the expansion (linear trend) .............................................................................................................................................. 79 Table 3.4 Regression results comparing Penn plans before and after the expansion (linear trend) ....................................................................................................................................................... 80 Table 3.5 Regression results comparing United plans before and after the expansion (quadratic trend) .............................................................................................................................................. 81 Table 3.6 Regression results comparing Penn plans before and after the expansion (quadratic trend) .............................................................................................................................................. 82 vi List of Figures Figure 1.1 Unadjusted and adjusted annual rates of SNF productivity growth for three conditions, 2006–2014 ..................................................................................................................................... 17 Figure 1.2 Annual rates of SNF productivity growth for three conditions, based on alternative definitions of quality, 2006-2014 .................................................................................................. 18 Figure 1.3 Annual rates of SNF productivity growth for three conditions based on costs vs. those based on payments, 2006–2014 ..................................................................................................... 19 Figure 1.4 Annual rates of SNF productivity growth for three conditions, sensitivity analysis results, 2006–2014 ......................................................................................................................... 20 Figure 1.5 Cumulative productivity change for three conditions, 2006–2014 .............................. 21 vii Abstract There has been rapid growth in vertical mergers and acquisitions across health care sectors in recent years: hospitals have acquired physician practices, health insurers have merged with hospitals and other health care providers, and pharmacy benefit managers (PBMs) have merged with health insurers (Frakt, Pizer, & Feldman, 2013; Fulton, 2017; Schulman & Richman, 2018). This growth in vertical integration is driven in part by the system’s emphasis on care coordination. In response to Medicare payment reforms such as Accountable Care Organizations (ACOs) and bundled payment, hospitals, post-acute care providers, insurers and all other entities of the healthcare system are building a closer relationship with each other. Theoretically, vertical integration poses both potential benefits and harms. Greater integration of providers across health care settings may lead to improved information sharing and coordination of care through interoperable health information technology or more consistent management, potentially leading to less health care use or better patient outcomes (Cuellar & Gertler, 2006). On the other hand, vertical integration may lead to overuse of downstream vertically integrated providers and reduced patient choice of providers. Research investigating hospitals and physician groups finds strong evidence that vertical integration increases commercial prices and induces utilization, with limited evidence for improvements in care coordination, implying that the harms of vertical integration outweigh the benefits (Baker, Bundorf, & Kessler, 2014; Cuellar & Gertler, 2006; Neprash, Chernew, Hicks, & et al., 2015; Robinson & Miller, 2014) Contrarily, literature studying the impact of vertical integration between hospitals and Skilled Nursing Facilities (SNFs) find reductions in costs and improvement in patient outcomes (David, Rawley, & Polsky, 2013; Rahman, Norton, & viii Grabowski, 2016). This dissertation aims at empirically estimating the effects of vertical integration on outcomes, costs and value from different perspectives. Chapter 1 evaluated the trends in productivity of SNFs in the era of vertical integration and examined whether the trends in productivity are different for vertically integrated SNFs. We constructed an analytic file with the records of Medicare beneficiaries that were discharged from acute-care hospitals to SNFs with stroke, hip fracture, or lower extremity joint replacement (LEJR) between 2006 and 2014. We used ordinary least square regression to estimate growth in SNF productivity, measured by the ratio of “high-quality SNF stays” to total treatment costs. In order to study whether vertical integration has an effect on productivity growth, we added an interaction term between vertical integration rate and the trend variable in the regression. We find that SNFs improved their productivity in the treatment of patients with LEJR, stroke, and hip fracture by 1.1%, 2.2%, and 2.0% per year, respectively. Productivity first decreased and then increased, with a lowest point in 2011. Over the study period, quality continued to rise, but dominated by higher costs at first. Costs then started to decrease, driving productivity to grow. In conclusion, there has been substantial productivity growth in recent years among SNFs in the U.S. in the treatment of post-acute-care-intensive conditions. We also find that vertically integrated SNFs are more productive than non-integrated SNFs. If there were no decline in the percentage of vertically integrated SNFs in recent years, we might have observed even greater productivity growth among SNFs. Chapter 2 examined the effects of preferred SNF networks formed by ACOs on patient volumes, composition and outcomes. Preferred SNF networks are a form of informal vertical integration between hospitals and SNFs, as hospitals selectively form strong ties with SNFs but remain under separate ownership. First, we investigated whether hospitals were able to ix effectively "steer" patients to preferred SNFs by examining patient volume and composition after network formation. Next, we investigated whether preferred network formation improved outcomes and lowered costs through care coordination. We found that preferred SNF network formation was associated with modest changes in patient composition at preferred SNFs, with evidence suggesting that hospitals sent more complex patients to their preferred SNFs. This result indicates that ACOs developed preferred network of SNFs and targeted its care coordination and management efforts toward patients with more complex needs. We also found that preferred SNF network formation was not associated with higher patient volume or better outcomes for patients admitted to preferred SNFs. This situation might be ameliorated by giving ACOs greater control over patient choice of SNFs. While the first two chapters focused on the post-acute care sector, Chapter 3 studied vertical integration in the pharmaceutical market. We analyzed whether larger PBMs benefit consumers and health plans and whether these effects are different for health plans vertically integrated with the PBM. We used the significant expansion in the size of OptumRx in 2012 – the in-house PBM of UnitedHealthcare – as a natural experiment to analyze these effects. In particular, we used data from Medicare Part D market and compared monthly premiums and costs for top 100 drugs for United plans before and after the expansion relative to other plans in the same market. We also examined the experience of Pennsylvania Public School Employees plans, who used OptumRx as its PBM both before and after the expansion, relative to other plans in the same market. We find that United plans and consumers benefited from the increased size of OptumRx in the form of decreased premiums and drug costs. In contrast, we find no evidence that the larger size of OptumRx benefited Pennsylvania Public School Employee plans. Our results indicate that it is possible that an in-house PBM tend to pass cost savings to their x vertically integrated insurer but not other clients. Under such circumstances, it is possible that insurers would seek more opportunities to merge with PBMs and the insurer and PBM market would become even more concentrated. 1 Chapter 1 Productivity growth of skilled nursing facilities in the treatment of post-acute-care- intensive conditions* Abstract Health care is believed to be suffered from a “cost disease,” in which a heavy reliance on labor limits opportunities for efficiencies stemming from technological improvement. Although recent evidence shows that U.S. hospitals have experienced a positive trend of productivity growth, skilled nursing facilities (SNFs) are relatively “low-tech” compared to hospitals, leading some to worry that productivity at skilled nursing facilities will lag behind the rest of the economy. The objective of this study is to assess productivity growth among skilled nursing facilities in the treatment of conditions which frequently involve substantial post-acute care after hospital discharge. We constructed an analytic file with the records of Medicare beneficiaries that were discharged from acute-care hospitals to SNFs with stroke, hip fracture, or lower extremity joint replacement (LEJR) between 2006 and 2014. We used ordinary least square regression to estimate growth in SNF productivity, measured by the ratio of “high-quality SNF stays” to total treatment costs. We find that SNFs improved their productivity in the treatment of patients with LEJR, stroke, and hip fracture by 1.1%, 2.2%, and 2.0% per year, respectively. Productivity first decreased and then increased, with a lowest point in 2011. Over the study period, quality continued to rise, but dominated by higher costs at first. Costs then started to decrease, driving productivity to grow. In conclusion, there has been substantial productivity growth in recent years among SNFs in the U.S. in the treatment of post-acute-care-intensive conditions. * Gu, J., Sood, N., Dunn, A., & Romley, J. (2019). Productivity growth of skilled nursing facilities in the treatment of post- acute-care-intensive conditions. PloS one, 14(4). 2 Introduction Health care is believed to suffer from a “cost disease,” which means that the costs rise at a rate significantly greater than the rate of inflation, because the quantity of labor required to produce health care services is difficult to reduce (Baumol, 2012). Although new technologies that aim to increase efficiency have been introduced into health care, they have done little to lower costs (Cutler & McClellan, 2001). As a result, there is wide interest in increasing the productivity of the U.S. health care system. According to the Institute of Medicine, “the only sensible way to restrain costs is to enhance the value of the health care system, thus extracting more benefit from the dollars spent” (Kibria et al., 2013). The Bureau of Labor Statistics defines multifactor productivity as “output per unit of labor, capital, and other measurable inputs,” which reflects “intangible influences…such as improvements in efficiency and technology.”(Harper, Khandrika, Kinoshita, & Rosenthal, 2010) There is recent evidence of a positive trend in hospitals’ multifactor productivity. A study from the Centers for Medicare and Medicaid Services (CMS) showed that the most recent 10-year moving-average growth rate of hospitals’ multifactor productivity (MFP), ending in 2013, was 0.1–0.5% (Spitalnic, Heffler, Dickensheets, & Knight, 2016). Using Medicare data from 2002 to 2011, Romley et al. found that the rate of annual quality-adjusted MFP growth among hospitals was 0.8%, 0.6%, and 1.9% in the treatment of patients with heart attack, heart failure, and pneumonia, respectively (J. A. Romley, Goldman, & Sood, 2015). By contrast, the Bureau of Labor Statistics reported that MFP for hospitals together with nursing and residential care facilities decreased by 0.4% annually from 2006 to 2014 (Bureau of Labor Statistics, 2017), which may indicate that nursing homes are not sharing the same 3 productivity growth that hospitals have shown. In addition, nursing homes are relatively “low- tech” compared with hospitals, making them theoretically more vulnerable to cost disease (Zinn & Mor, 1998). Skilled nursing facilities (SNFs) are a particular type of nursing home that provide short- term, skilled nursing care and rehabilitation services, such as physical and occupational therapy and speech-language pathology services, to patients following a stay in an acute-care hospital (Medicare Payment Advisory Commission, 2017). SNF services are covered by Medicare, so detailed administrative data are available for the study of SNF performance. According to the National Health Expenditure Accounts, Medicare is the second largest payer for nursing home care, paying 23% of the total in 2016 (Centers for Medicare and Medicaid Services, 2016c). There is wide concern about Medicare’s increasing expenditures on SNFs and its payment system. Medicare payments for post-acute care have grown faster than most other categories of spending, and SNFs account for over 50% of the total Medicare expenditures on post-acute care (Mechanic 2014). There is concern that the current prospective payment system to pay SNFs for each day of service might induce SNFs to keep patients longer than necessary and/or furnish therapy services that are unrelated to a given patient’s condition, which threatens the quality of care and the productivity of the SNFs (Einav, Finkelstein, & Mahoney, 2017; Medicare Payment Advisory Commission, 2017). The Medicare payment policy also results in highly fragmented health care delivery (Medicare Payment Advisory Commission, 2017). Under the payment policy, acute-care hospitals and SNFs each receive a separate payment for providing acute and post-acute care, and the policy does not reimburse any entity for coordinating patient transitions across providers (Medicare Payment Advisory Commission, 2012). The lack of coordination and fragmentation of care potentially lower productivity. In response to those 4 concerns, the Affordable Care Act included several Medicare reforms such as penalties for readmissions, “Accountable Care Organizations,” and “Bundled Payments” to improve care coordination, particularly after hospital discharge (Sood & Higgins, 2012). Since then, there has been an increase in “virtual” integration between hospitals and SNFs, meaning that some hospitals started to form a “preferred” SNF network and to discharge most of their patients to SNFs within their network (Evans, 2015). Studies have shown that such integration potentially lowers costs and improves quality of care, resulting in higher productivity (Afendulis & Kessler, 2011; David, Rawley, & Polsky, 2013; Rahman, Foster, Grabowski, Zinn, & Mor, 2013). It is unclear, however, how productivity growth among SNFs has changed over the years and whether recent Medicare payment reforms emphasizing care coordination have made a difference. Few studies have examined MFP at SNFs. A study of residential care facilities in Canada found that labor productivity (which is influenced by MFP) was virtually unchanged from 1984 to 2009 (Gu & Li, 2015). However, an important limitation of that study is that it did not account for differences in the quality of care received by patients during SNF stays. Productivity might appear to be decreasing over time if SNFs are treating more severely ill patients, providing better quality of care, or both (J. A. Romley et al., 2015). In quantifying trends in SNF productivity, it is important to address all such potentially confounding factors. To our knowledge, no study has directly reported on the quality-adjusted productivity of SNFs in the U.S. This study is intended to fill the research gap by analyzing quality-adjusted productivity for SNF stays involving the most common conditions in post-acute-care settings. 5 Methods Population We analyzed data from Medicare fee-for-service beneficiaries discharged to SNFs from short-term acute-care hospitals from 2006 to 2014. We restricted our analysis to SNF stays that started within 90 days of hospital discharge and were the first post-acute care (including care received in Home Health Agencies, or HHA) that the beneficiary received after being discharged from the hospital. We identified patients with lower extremity joint replacement (LEJR), hip fracture, or stroke according to the principal diagnosis code used during hospitalization (Huckfeldt, Sood, Escarce, Grabowski, & Newhouse, 2014). Those three conditions are common in the elderly population of the U.S. and result in high usage of post-acute care (Sood, Buntin, & Escarce, 2008). We used Medicare Provider Analysis and Review (MedPAR) files as our primary data source (Centers for Medicare and Medicaid Services, 2016b). The University of Southern California Institutional Review Boards has approved this study. The form of consent was not obtained because the data were analyzed anonymously. Production function framework MFP is defined as a measure of economic performance that compares the amount of goods and services produced (output) to the amount of combined resources (inputs) used to produce those goods and services (Ashby, Guterman, & Greene, 2000; Caves, Christensen, & Diewert, 1982). The inputs may include labor, capital, energy, materials, and purchased services (Bureau of Labor Statistics, 2017). We characterized the output as the number of “high-quality” stays produced by a SNF for patients with a condition in a year. We summarized the inputs as the total treatment costs of a SNF in treating a condition in a year. It is important to consider control 6 variables such as the severity of illness because the outcomes and the amount of resources needed to achieve them are likely to depend on the initial condition of the patient. Building on our prior work on hospitals, we applied a productivity framework to SNFs that uses the logarithm of the ratio of the output to the input as the dependent variable (J. A. Romley et al., 2015). Measures Output We characterized the output in the production function as the number of high-quality SNF stays. A high-quality stay was one in which the patient was alive and had returned to the community (with or without HHA care) as of 90 days after SNF admission. We obtained information about mortality from Master Beneficiary Summary files. We determined whether patients were residing in the community (potentially receiving HHA services at home) using combined information from MedPAR, the Minimum Data Set (MDS), and Home Health Standard Analytic Files (Buntin, Colla, Deb, Sood, & Escarce, 2010; Centers for Medicare and Medicaid Services, 2012; Sood et al., 2008). The MDS contains assessments of health status for all residents of nursing homes certified by Medicare or Medicaid. We applied an algorithm to group the assessments into stays linked with our sample (Appendix 3), which enabled us to ascertain readmissions to facilities and residence in facilities as of day 90 after SNF admission, including custodial nursing homes financed by Medicaid. We also considered two alternative definitions for high-quality stays. The first alternative defined high-quality stays as stays where patients were discharged from hospitals directly to home or HHA, as indicated by the discharge destination codes from Medicare claims. The 7 second alternative definition was the strictest one. It defined high-quality stays as stays where the patient continuously resided in the community for 30 days without any readmission during a 90- day window. We assessed the outcomes using a 90-day window after SNF admission. To test whether the length of the window would influence our results, we performed three sensitivity analyses using a 30-day window, a 60-day window and a 120-day window, respectively. Inputs We defined the input as the total costs incurred during SNF stays. We converted Medicare payments to costs using cost-to-revenue ratios that SNFs submit as part of their cost accounting reports to the CMS (Buntin et al., 2010; Sood et al., 2008). This input measure accounts for the opportunity cost to society of employing scarce resources for treatment (J. A. Romley et al., 2015). We also performed a supplementary analysis from the payer perspective using Medicare payments. We adjusted all costs and payments for inflation to 2014 dollars using market basket indices from the CMS (Centers for Medicare and Medicaid Services, 2018a). Controls We obtained information about patient characteristics such as age, sex, race/ethnicity, and Medicare enrollment from the Master Beneficiary Summary files. Elixhauser comorbidities were based on hospitalization claims (Elixhauser, Steiner, Harris, & Coffey, 1998) and constructed from the first 10 diagnosis codes of the index hospitalization. In sensitivity analysis, we added the length of hospital stay and the number of days between hospital discharge and SNF admission as control variables, as they may indicate the 8 severity of illness. In a separate sensitivity analysis, we followed the calculation method for inpatient quality indicators published by the Agency for Healthcare Research and Quality (AHRQ) and included the risk-adjusted mortality rate for each condition in order to separate the quality produced by the hospitals from that produced by the SNFs (Agency for Healthcare Research and Quality). To determine whether vertical integration between hospitals and SNFs has an impact on productivity growth, we constructed an indicator of integration for each SNF using the percentage of admitted patients who were discharged from the SNF’s parent hospital. We used the variable “related provider number” from Medicare Provider of Services (POS) files to identify each SNF’s parent hospital (Centers for Medicare and Medicaid Services, 2018b). Analysis In order to study productivity growth and how adjustment for quality affects the trend, we first looked at unadjusted productivity growth by defining output as the number of patient stays. Productivity was measured as the ratio of output to inputs. For each condition at each SNF in each year, we regressed the logarithm of unadjusted productivity on the trend variable (year) using the Ordinary Least Square (OLS) model, in order to get the compound growth rate from 2006 through 2014, as shown in equation (1). Next, we performed the same regressions but included patient characteristics and severity of illness as control variables, as shown in equation (2), where X is a vector for all control variables. Finally, we incorporated quality of care by defining the output as high-quality stays in which the patient was alive and had returned to the community within 90 days after SNF admission, as shown in equation (3) below. All regressions were weighted by the number of patients at each SNF. 9 log(𝑢𝑛𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦) = 𝛽 8 𝑦𝑒𝑎𝑟 (1) log(𝑢𝑛𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦) = 𝛽 8 𝑦𝑒𝑎𝑟+𝑋𝛽 < (2) log(𝑞𝑢𝑎𝑙𝑖𝑡𝑦 𝑎𝑑𝑗𝑢𝑠𝑡𝑒𝑑 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦) = 𝛽 8 𝑦𝑒𝑎𝑟+𝑋𝛽 < (3) We also regressed costs, payments, and the rate of return of patients to the community on the trend variable separately (Appendix 4). In addition, we analyzed productivity growth in each year by replacing the single trend variable with binary variables for each year. As a supplemental analysis, we added an interaction term between the vertical integration rate and the trend variable to study whether vertical integration has an effect on productivity growth. Results Our sample included 1,076,066 patient stays with a diagnosis of LEJR at 14,394 SNFs, 315,546 patient stays with a diagnosis of stroke at 14,154 SNFs, and 739,608 patient stays with a diagnosis of hip fracture at 14,588 SNFs. Table 1.1 shows a summary of the patient characteristics and outcomes. The average age of the patients in our sample ranged from 79 years for LEJR to 84 for hip fracture. Each patient had 2–3 comorbid conditions on average. The average length of hospital stay was 5–6 days, with less than 1 day between hospital discharge and SNF admission. The inpatient risk-adjusted mortality rate was 0.1% for hip replacement, 3.1% for hip fracture and 8.0% for stroke. The rate of survival and return to community was 77%, 56%, and 57% for LEJR, stroke, and hip fracture, respectively. The amount of Medicare payments for SNF care ranged from $11,441 to $15,679, and the corresponding costs ranged from $9,497 to $12,993. 10 The growth rate of unadjusted productivity (without controlling for patient characteristics and severity of illness) was negative for all three conditions (Fig 1.1). The negative growth estimates mainly resulted from increasing treatment costs from 2006 to 2014. After we controlled for patient characteristics and the severity of illness, productivity growth was still negative, and the annual rate did not change substantially compared to the unadjusted rate. However, productivity for LEJR and stroke decreased somewhat more with severity adjustment. When we controlled for patient characteristics and comorbidities and defined the output as stays where the patient survived and was residing in the community 90 days after SNF admission, the annual rate of productivity growth became positive for all three conditions. The productivity growth was due to improvement over time in the quality of care as measured by survival and return to the community. We had two alternative definitions of a high-quality SNF stay. As shown in Fig 1.2, there was growth in productivity over the study period for all three conditions no matter which definition of high-quality stay we used. Fig 1.3 shows a comparison of the productivity growth rate for each condition calculated on the basis of costs versus that calculated on the basis of Medicare payments. The productivity growth based on costs was higher than the productivity growth based on Medicare payments for all three conditions. For LEJR, the productivity growth was negative when the input was measured using Medicare payments. When we changed the 90-day window to 60 days or 120 days, the productivity growth rate was still positive for all three conditions (Appendix 1). For sensitivity analyses, we performed regressions using several covariates. When we added the length of hospital stay and the number of days between hospital discharge and SNF 11 admission as two additional control variables, the productivity growth rates decreased by roughly half but were still significantly positive. When we included the inpatient risk-adjusted mortality rate, the productivity growth was similar to our primary specification (Fig 1.4). We also regressed the logarithm of the productivity on year dummies and all other covariates to show the change in productivity in each year from 2007 to 2014 compared with the base year 2006. The productivity mostly decreased until 2011 and then increased through 2014 (Fig 1.5). That trend was mainly driven by changes in costs. Over the study period, the rate of high-quality SNF stays continued to rise, but the productivity was dominated by increasing costs at first. Later on, the costs started to decrease, resulting in productivity growth overall. SNFs with higher degree of vertical integration had faster productivity growth in the treatment of LEJR; however, vertical integration had no significant effect on the rate of productivity growth in the treatment of the other two conditions (Table 1.2). For all three conditions, productivity was positively correlated with the degree of integration. The rate of vertical integration decreased annually from 2006 to 2014 (Appendix 2), which indicates that SNFs with higher productivity were crowded out of the market. Since 2013, the integration rate increased slightly, except for the stroke condition. Discussion We assessed productivity growth from 2006 to 2014 among SNFs treating Medicare beneficiaries who had been admitted to hospitals with common conditions that frequently involve post-acute care. For all three conditions, the unadjusted productivity growth was negative. However, after adjusting for disease severity and quality of care, we found substantial 12 productivity growth ranging from 1.1% to 2.2% per year. The results were robust across several sensitivity analyses. An understanding of productivity trends is important, as the Affordable Care Act now requires that the market basket percentage under the Medicare Prospective Payment System be reduced annually by a productivity adjustment (Spitalnic et al., 2016). A rationale for the adjustment is that Medicare should benefit from productivity gains in the economy at large. However, if productivity grows relatively rapidly in the broader economy, productivity gains among providers might not be sufficient to offset slower growth in reimbursement and resources under the Affordable Care Act, and providers might struggle to maintain quality of care (J. A. Romley et al., 2015). Our prior results suggest that such concerns might be overstated, at least in the inpatient setting (J. A. Romley et al., 2015). However, this finding would not necessarily have generalized to the post-acute care setting. Studies have shown that post-acute care is the largest driver of the geographic variation in Medicare spending, and there is much uncertainty about the clinical appropriateness of post- acute care sites for particular patients (Institute of Medicine & Newhouse, 2013; Medicare Payment Advisory Commission, 2017; Newhouse & Garber, 2013a, 2013b). The Bureau of Labor Statistics reported a negative rate of productivity growth for hospitals plus nursing and residential care facilities (Bureau of Labor Statistics, 2017). That report did not account for variation in disease severity and quality of care, however. We addressed those factors by using “high-quality stays” as the output of SNFs, and also tried to separate the quality of care produced by acute-care hospitals from that produced by SNFs. The definition of “high-quality stays” – survival and return to community as of 90 days after SNF admission – was motivated by CMS policies with respect to public reporting and reimbursement. For example, CMS and Hospital 13 Quality Alliance (HQA) publicly report 30-day mortality measures for certain conditions since 2007 as an important indicator for quality of care (Centers for Medicare and Medicaid Services, 2017). In terms of skilled nursing facilities, CMS has the SNF Quality Reporting Program (QRP) that publicly reports SNF provider performance on the quality measures, and one of the five measures is the rate of successful return to home or community from an SNF (Centers for Medicare and Medicaid Services, 2018c). The outcomes analyzed here are widely studied in health services and health policy (Birkmeyer et al., 2002; Jena, Sun, & Romley, 2013; John A Romley, Jena, & Goldman, 2011; John A Romley, Jena, O’Leary, & Goldman, 2013). Romley et al. focused on hospital stays, and therefore considered mortality and readmissions (J. A. Romley et al., 2015). This study also considers return to community, because it is an important outcome in post-acute care (Buntin et al., 2010; DeJong et al., 2009; Deutsch et al., 2005; Kramer, Holthaus, Goodrish, & Epstein, 2006). We also considered two alternative definitions for high-quality stays. The first alternative defined high-quality stays as stays where patients were discharged from hospitals directly to home or HHA, as indicated by the discharge destination codes from Medicare claims. That was the approach used in the Improving Medicare Post-Acute Care Transformation (IMPACT) Act of 2014, which required post-acute care providers to report such information as a quality metric (Centers for Medicare and Medicaid Services, 2016a). That approach fails, however, to capture readmissions to facilities and residence. The second alternative definition was the strictest one. It defined high-quality stays as stays where the patient continuously resided in the community for 30 days without any readmission during a 90-day window. We used 90 days as the primary window because of several considerations. Firstly, 90 days is a common length of episode used in many published studies about post-acute care (Li, Morrow-Howell, & Proctor, 2004; Morley, 14 Bogasky, Gage, Flood, & Ingber, 2014; Sood et al., 2008). Some studies used even longer window, such as 120 days (Buntin et al., 2010). Secondly, SNF patients typically need a longer time to recover and the length of stays are usually longer than hospital patients. Thus we used a longer window to look at SNF patients’ outcomes than 30 days, which is commonly used in studies on hospitals. We do recognize that there’s no perfect standard about the length of episodes when studying outcomes among SNF patients. Therefore, we did sensitivity analysis using different length of windows to define outcomes, including 30-day, 60-day, 90-day and 120-day, in order to show robustness. Our results do not support the hypothesis that SNFs suffer from a cost disease. That suggests that similar to hospitals, SNFs have had positive quality-adjusted productivity growth in recent years. Thus, the concerns about a decline of quality of care when reimbursement does not keep up with health care cost inflation may be overstated in post-acute care setting, too. Conventional measures of health care prices tend to ignore quality of care and overstate true inflation (Lucarelli & Nicholson, 2009; J. A. Romley et al., 2015). Moreover, innovative payment and delivery approaches under the ACA and among private payers may increase the incentives for SNFs to achieve productivity gains in the future. In addition, we found that vertically integrated SNFs are more productive than non-integrated SNFs. If there were no decline in the percentage of vertically integrated SNFs in recent years, we might have observed even greater productivity growth among the SNFs. Our results indicate that there is a need for some reassessment of the performance of the U.S. health care system and for payment reforms emphasizing integration and coordination between hospitals and post-acute care facilities. Our study has several limitations. Quality is complex and difficult to measure. MDS files lack discharge dates for some nursing home stays, which causes some measurement error. To 15 address that issue, we defined quality in three different ways. All three definitions of quality led to similar results. We also performed analyses using different follow-up periods and found similar productivity growth for three different time windows. As in our prior work, we focused on the site of care, instead of the full episode, which gives a more comprehensive measure of health care output. Productivity improvement might be better or worse if it is measured in terms of episodes rather than sites of care. That is an important direction for future research. In conclusion, our results suggest that similar to hospitals, SNFs have not suffered from what has been called a cost disease, in which technological change does not generate efficiencies to offset increasingly costly labor. Much work remains to be done on this important topic. 16 Tables and Figures Table 1.1 Summary statistics of patient characteristics and outcomes Lower Extremity Joint Replacement Stroke Hip Fracture N=80,480 SNF- years N=52,790 SNF- years N=76,570 SNF- years Age 79.4 (4.4) 82.6 (4.9) 84.2 (4.0) Male (%) 26.8 (24.3) 33.9 (28.9) 24.5 (21.8) Race/ethnicity White (%) 90.1 (21.0) 85.0 (26.2) 92.3 (17.8) Black (%) 6.0 (16.7) 10.6 (22.6) 3.9 (13.1) Other (%) 3.9 (13.9) 4.4 (14.6) 3.7 (12.4) Elixhauser comorbidity counts 2.4 (0.7) 2.4 (0.8) 2.6 (0.7) Hospital length of stay 5.0 (2.0) 6.4 (3.1) 5.9 (2.0) Days between hospital discharge and SNF admission 0.3 (1.9) 0.7 (3.1) 0.5 (2.4) Inpatient risk adjusted mortality rate (%) 0.1 (2.1) 8.0 (5.6) 3.1 (4.2) Alive and residing in community in 90 days post SNF admission (%) 78.1 (21.4) 56.9 (26.6) 58.0 (23.9) SNF costs (2014$) 9,497 (5,294) 11,159 (6,494) 12,993 (6,279) SNF Medicare payments (2014$) 11,441 (5,722) 13,586 (7,077) 15,679 (6,606) Notes: Standard deviations are in parentheses. SNF costs and Medicare payments are average costs and payments per SNF per year, unweighted for the number of patients received per SNF per year, adjusted for inflation to 2014 dollars using market basket indices from the CMS. 17 Figure 1.1 Unadjusted and adjusted annual rates of SNF productivity growth for three conditions, 2006–2014 Note. All rates are significantly different from zero (p<0.05). -2.94 -3.57 -4.47 -3.48 -3.95 -4.42 1.07 2.19 1.99 -5.0 -4.0 -3.0 -2.0 -1.0 0.0 1.0 2.0 3.0 Lower Extremity Joint Replacement Stroke Hip Fracture Annual productivity growth rate (%) Unadjusted stays Stays adjusted with patient characteristics and severity of illness Adjusted survival with return to community 18 Figure 1.2 Annual rates of SNF productivity growth for three conditions, based on alternative definitions of quality, 2006-2014 Note. All rates are significantly different from zero (p<0.05). 1.25 2.59 2.58 1.07 2.19 1.99 1.44 2.06 2.26 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Lower Extremity Joint Replacement Stroke Hip Fracture annual productivity growth rate (%) Initial discharge to home or HHA Alive and residing in community in 90 days post admission No readmission in 90 days and continuously residing in community for at least 30 days 19 Figure 1.3 Annual rates of SNF productivity growth for three conditions based on costs vs. those based on payments, 2006–2014 Note. All rates are significantly different from zero (p<0.05). 1.07 2.19 1.99 -0.36 0.87 0.62 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 Lower Extremity Joint Replacement Stroke Hip Fracture annual productivity growth rate (%) Cost Payment 20 Figure 1.4 Annual rates of SNF productivity growth for three conditions, sensitivity analysis results, 2006–2014 Note. All rates are significantly different from zero (p<0.05). 1.07 2.19 1.99 0.4 0.95 1.37 1.19 2.52 1.81 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Lower Extremity Joint Replacement Stroke Hip Fracture Annual productivity growth rate (%) Primary analysis Add hospital length of stay and days between hospital discharge and SNF admission Add AHRQ risk-adjusted mortality rate 21 Figure 1.5 Cumulative productivity change for three conditions, 2006–2014 -25.0 -20.0 -15.0 -10.0 -5.0 0.0 5.0 10.0 15.0 20.0 2006 2007 2008 2009 2010 2011 2012 2013 2014 Productivity change rate compared to base year 2006 (%) Lower Extremity Joint Replacement Stroke Hip Fracture 22 Table 1.2 Estimated percent of productivity growth and interaction with vertical integration Lower Extremity Joint Replacement Stroke Hip Fracture Year 2.01** (0.09) 3.17** (0.17) 2.57** (0.10) Integrated 1.14** (0.01) 1.42** (0.02) 1.33** (0.02) Year × Integrated 0.82** (0.23) -0.59 (0.51) -0.43 (0.34) Note: Standard errors in parentheses. * p<0.05, ** p<0.01, *** p<0.001 23 Appendices Appendix 1. Annual rates of SNF productivity growth for three conditions, based on return to community rate in 30, 60, 90 (baseline) vs. 120 days, 2006-2014 Note. All rates are significantly different from zero (p<0.05). 0.81 0.76 0.78 1.03 2.08 1.87 1.07 2.19 1.99 0.92 1.62 1.52 0.0 0.5 1.0 1.5 2.0 2.5 Lower Extremity Joint Replacement Stroke Hip Fracture Annual productivity growth rate (%) 30 days 60 days 90 days (baseline) 120 days 24 Appendix 2. Vertical integration rate for three conditions, 2006-2014 0.02 0.03 0.04 0.05 0.06 0.07 0.08 2006 2007 2008 2009 2010 2011 2012 2013 2014 Low Extremity Joint Replacement Stroke Hip Fracture 25 Appendix 3. Algorithm to process the MDS data We used an algorithm to group assessment-level MDS data into stay-level data. First, we combined all of the years of the raw MDS data from 2006 to 2014. Then we sorted the dataset by Beneficiary ID and Assessment Date so that we could see the beneficiaries’ chronological progression through various nursing homes. We designated the first observation for each beneficiary as the starting point for the first stay. Then, we moved on to the next observation. If there was a discharge date, we flagged that as the last observation in the stay and designated the next observation for the same beneficiary as the first assessment of the second stay. If the discharge date was missing, we considered each observation after the first as an intermediate assessment within the same stay. We continued until either we found a discharge date or a new beneficiary appeared in the file. In the event that there were no discharge dates, we assumed that the beneficiary had a single stay and was still in the nursing home as of the last day of our data. The dataset changed from MDS2.0 to MDS3.0 in 2010. To smooth the MDS2.0 and MDS3.0 data together, we started by appending the two stay-level files together and sorted by beneficiary ID and entry date. Presumably, if the MDS2.0 stay flows into the MDS3.0 file, there should be one stay with an entry date before or on 9/30/2010 and a discharge date (which was imputed) on 9/30/2010. There should then be a stay in the next observation with an entry date occurring before 9/30/2010 and a discharge date after 9/30/2010. If those conditions were met, we kept the earliest Entry Date between the two adjacent stays and the latest Discharge Date between the two adjacent stays. We then had two identical stays by beneficiary ID, entry date, and discharge date, so we eliminated the duplicate stays on the basis of those variables. 26 Appendix 4. Changes in output and input over years 0.40 0.45 0.50 0.55 0.60 0.65 0.70 0.75 0.80 0.85 0.90 2006 2007 2008 2009 2010 2011 2012 2013 2014 Annual survival with return to community rate (%) Lower Extremity Joint Replacement Stroke Hip Fracture 6000 7000 8000 9000 10000 11000 12000 13000 14000 15000 2006 2007 2008 2009 2010 2011 2012 2013 2014 Average stay cost per SNF per year (2014$), unadjusted for number of patients Lower Extremity Joint Replacement Stroke Hip Fracture 27 References Afendulis, C. C., & Kessler, D. P. (2011). Vertical integration and optimal reimbursement policy. International journal of health care finance and economics, 11(3), 165-179. Agency for Healthcare Research and Quality. Inpatient Quality Indicators Overview. Retrieved from http://www.qualityindicators.ahrq.gov/Modules/iqi_resources.aspx Ashby, J., Guterman, S., & Greene, T. (2000). An analysis of hospital productivity and product change. Health Affairs, 19(5), 197-205. Baumol, W. J. (2012). The cost disease: Why computers get cheaper and health care doesn't: Yale university press. Birkmeyer, J. D., Siewers, A. E., Finlayson, E. V., Stukel, T. A., Lucas, F. L., Batista, I., . . . Wennberg, D. E. (2002). Hospital volume and surgical mortality in the United States. New England Journal of Medicine, 346(15), 1128-1137. Buntin, M. B., Colla, C. H., Deb, P., Sood, N., & Escarce, J. J. (2010). Medicare spending and outcomes after post-acute care for stroke and hip fracture. Medical care, 48(9), 776. Bureau of Labor Statistics. (2017). Multifactor Productivity. Retrieved from https://www.bls.gov/mfp/ Caves, D. W., Christensen, L. R., & Diewert, W. E. (1982). The economic theory of index numbers and the measurement of input, output, and productivity. Econometrica: Journal of the Econometric Society, 1393-1414. Centers for Medicare and Medicaid Services. (2012). Minimum Data Set 3.0 Public Reports. Retrieved from https://www.cms.gov/Research-Statistics-Data-and-Systems/Computer- Data-and-Systems/Minimum-Data-Set-3-0-Public-Reports/index.html 28 Centers for Medicare and Medicaid Services. (2016a). IMPACT Act of 2014 Data Standardization & Cross Setting Measures. Retrieved from https://www.cms.gov/Medicare/Quality-Initiatives-Patient-Assessment-Instruments/Post- Acute-Care-Quality-Initiatives/IMPACT-Act-of-2014/IMPACT-Act-of-2014-Data- Standardization-and-Cross-Setting-MeasuresMeasures.html Centers for Medicare and Medicaid Services. (2016b, Aug 8 2016). Medicare Provider Analysis and Review (MEDPAR) File. Retrieved from https://www.cms.gov/research-statistics- data-and-systems/statistics-trends-and-reports/medicarefeeforsvcpartsab/medpar.html Centers for Medicare and Medicaid Services. (2016c). National Health Expenditure Accounts: Methodology Paper, 2016, Definitions, Sources, and Methods. Retrieved from https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and- Reports/NationalHealthExpendData/downloads/dsm-16.pdf Centers for Medicare and Medicaid Services. (2017). Outcome Measures. Retrieved from https://www.cms.gov/medicare/quality-initiatives-patient-assessment- instruments/hospitalqualityinits/outcomemeasures.html Centers for Medicare and Medicaid Services. (2018a). Market Basket Data. Retrieved from https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and- Reports/MedicareProgramRatesStats/MarketBasketData.html Centers for Medicare and Medicaid Services. (2018b). Provider of Services Current Files. Retrieved from https://www.cms.gov/Research-Statistics-Data-and- Systems/Downloadable-Public-Use-Files/Provider-of-Services/ Centers for Medicare and Medicaid Services. (2018c). Skilled Nursing Facility (SNF) Quality Reporting Program (QRP) Data on Nursing Home Compare. Retrieved from 29 https://www.cms.gov/newsroom/fact-sheets/skilled-nursing-facility-snf-quality-reporting- program-qrp-data-nursing-home-compare Cutler, D. M., & McClellan, M. (2001). Is technological change in medicine worth it? Health Affairs, 20(5), 11-29. David, G., Rawley, E., & Polsky, D. (2013). Integration and task allocation: Evidence from patient care. Journal of economics & management strategy, 22(3), 617-639. DeJong, G., Horn, S. D., Smout, R. J., Tian, W., Putman, K., & Gassaway, J. (2009). Joint replacement rehabilitation outcomes on discharge from skilled nursing facilities and inpatient rehabilitation facilities. Archives of physical medicine and rehabilitation, 90(8), 1284-1296. Deutsch, A., Granger, C. V., Fiedler, R. C., DeJong, G., Kane, R. L., Ottenbacher, K. J., . . . Trevisan, M. (2005). Outcomes and reimbursement of inpatient rehabilitation facilities and subacute rehabilitation programs for Medicare beneficiaries with hip fracture. Medical care, 892-901. Einav, L., Finkelstein, A., & Mahoney, N. (2017). Provider incentives and healthcare costs: Evidence from long-term care hospitals. Retrieved from Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. (1998). Comorbidity measures for use with administrative data. Medical care, 36(1), 8-27. Evans, M. (2015). Hospitals select preferred SNFs to improve post-acute outcomes. Mod Healthc, 45, 14-15. Gu, W., & Li, J. (2015). Productivity in Residential Care Facilities in Canada, 1984-2009. International Productivity Monitor(29), 18. 30 Harper, M. J., Khandrika, B., Kinoshita, R., & Rosenthal, S. (2010). Nonmanufacturing industry contributions to multifactor productivity, 1987–2006. Monthly Labor Review, 133(6), 16- 31. Huckfeldt, P. J., Sood, N., Escarce, J. J., Grabowski, D. C., & Newhouse, J. P. (2014). Effects of Medicare payment reform: Evidence from the home health interim and prospective payment systems. Journal of Health Economics, 34, 1-18. Institute of Medicine, & Newhouse, J. P. (2013). Interim report of the committee on geographic variation in health care spending and promotion of high-value health care: Preliminary committee observations: National Academies Press. Jena, A. B., Sun, E. C., & Romley, J. A. (2013). Mortality among high-risk patients with acute myocardial infarction admitted to US teaching-intensive hospitals in July: a retrospective observational study. Circulation, 128(25), 2754-2763. Kibria, A., Mancher, M., McCoy, M. A., Graham, R. P., Garber, A. M., & Newhouse, J. P. (2013). Variation in health care spending: target decision making, not geography: National Academies Press. Kramer, A., Holthaus, D., Goodrish, G., & Epstein, A. (2006). A study of stroke post-acute care costs and outcomes: final report. In: US Department of Health and Human Services Washington. DC. Li, H., Morrow-Howell, N., & Proctor, E. K. (2004). Post-acute home care and hospital readmission of elderly patients with congestive heart failure. Health & social work, 29(4), 275-285. Lucarelli, C., & Nicholson, S. (2009). A quality-adjusted price index for colorectal cancer drugs. Retrieved from 31 Mechanic , R. (2014). Post-Acute Care — The Next Frontier for Controlling Medicare Spending. New England Journal of Medicine, 370(8), 692-694. doi:doi:10.1056/NEJMp1315607 Medicare Payment Advisory Commission. (2012). Care coordination in fee-for-service Medicare. Report to the Congress: Medicare and the Health Care Delivery System. Retrieved from Medicare Payment Advisory Commission. (2017). Chapter 8: Skilled nursing facilities services. Retrieved from http://www.medpac.gov/docs/default- source/reports/mar17_medpac_ch8.pdf Morley, M., Bogasky, S., Gage, B., Flood, S., & Ingber, M. J. (2014). Medicare post-acute care episodes and payment bundling. Medicare & medicaid research review, 4(1). Newhouse, J. P., & Garber, A. M. (2013a). Geographic variation in health care spending in the United States: insights from an Institute of Medicine report. Jama, 310(12), 1227-1228. Newhouse, J. P., & Garber, A. M. (2013b). Geographic variation in Medicare services. New England Journal of Medicine, 368(16), 1465-1468. Rahman, M., Foster, A. D., Grabowski, D. C., Zinn, J. S., & Mor, V. (2013). Effect of hospital– SNF referral linkages on rehospitalization. Health services research, 48(6pt1), 1898- 1919. Romley, J. A., Goldman, D. P., & Sood, N. (2015). US hospitals experienced substantial productivity growth during 2002-11. Health Affairs, 34(3), 511-518. doi:10.1377/hlthaff.2014.0587 Romley, J. A., Jena, A. B., & Goldman, D. P. (2011). Hospital spending and inpatient mortality: evidence from California: an observational study. Annals of internal medicine, 154(3), 160-167. 32 Romley, J. A., Jena, A. B., O’Leary, J. F., & Goldman, D. P. (2013). Spending and mortality in US acute care hospitals. The American journal of managed care, 19(2), e46. Sood, N., Buntin, M. B., & Escarce, J. J. (2008). Does how much and how you pay matter? Evidence from the inpatient rehabilitation care prospective payment system. Journal of Health Economics, 27(4), 1046-1059. doi:https://doi.org/10.1016/j.jhealeco.2008.01.003 Sood, N., & Higgins, A. (2012). Posing a framework to guide government’s role in payment and delivery system reform. Health Affairs, 31(9), 2043-2050. Spitalnic, P., Heffler, S., Dickensheets, B., & Knight, M. (2016). Hospital Multifactor Productivity: An Updated Presentation of Two Methodologies. Retrieved from https://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and- Reports/ReportsTrustFunds/Downloads/ProductivityMemo2016.pdf Zinn, J. S., & Mor, V. (1998). Organizational structure and the delivery of primary care to older Americans. Health services research, 33(2 Pt 2), 354. 33 Chapter 2 The effects of Accountable Care Organizations (ACOs) forming preferred Skilled Nursing Facility (SNF) network on patient volume, composition and outcomes Abstract Through participation in payment reforms such as bundled payment and accountable care organizations, hospitals are increasingly financially responsible for health care use and adverse health events occurring after hospital discharge. To improve the management and coordination of post-discharge care, hospitals are establishing a closer relationship with skilled nursing facilities (SNFs) through the formation of "preferred" SNF networks. We evaluated the effects of formation of preferred SNFs network on care patterns and outcomes. First, we investigated whether hospitals were able to effectively "steer" patients to preferred SNFs by examining patient volume and composition after network formation. Next, we investigated whether preferred network formation improved outcomes and lowered costs through care coordination. We found that preferred SNF network formation was not associated with higher patient volume or better outcomes for patients admitted to preferred SNFs. However, we found that preferred SNF network formation was associated with modest changes in patient composition at preferred SNFs, with evidence suggesting that hospitals sent more complex patients to their preferred SNFs. 34 Introduction Medicare expenditure on post-acute care rose rapidly in the first ten years of the 2000s (Medicare Payment Advisory Commission, 2018a). This growth has been often attributed to Medicare’s separate payment to hospitals and post-acute care providers, which provided little incentive for efficient use of post-acute care or coordinate care across settings (Jencks, Williams, & Coleman, 2009). In response to these concerns, the Affordable Care Act included several new payment approaches, such as bundled payment and accountable care organizations (ACOs), where some entity (e.g., a hospital or physician group) was responsible for post-discharge costs and patient outcomes (Sood & Higgins, 2012). A key challenge facing hospitals participating in ACOs was how to coordinate and avoid unnecessary post-acute care. In response, hospitals sought to develop closer relationships with post-acute care providers such as skilled nursing facilities (SNFs), either through formal vertical integration, defined by combined legal ownership, or informal integration, whereby hospitals selectively form strong ties with SNFs but remain under separate ownership. One example of informal integration between hospitals and SNFs is “preferred SNF networks”. Preferred SNFs are selected based on historical costs, quality of care, or historical volume (Evans, 2015; Livingston, 2017; McHugh et al., 2017; Medicare Payment Advisory Commission, 2015; Zhu, Patel, Shea, Neuman, & Werner, 2018). Preferred SNF networks play three key functions for ACOs. First, while hospitals are not allowed to compel patients to choose a particular SNF, they can provide patients with a suggested list of preferred providers (“soft steering”) (Huckfeldt, Weissblum, Escarce, Karaca‐Mandic, & Sood, 2018; Medicare Payment Advisory Commission, 2015). To the extent that these facilities provide lower cost and better quality of care, channeling patients to these SNFs may improve ACO performance on cost and quality. Second, preferred 35 SNF networks could promote care coordination between hospitals and preferred SNFs through electronic health record exchange, improving continuity of care by allowing physicians and nurses to monitor and treat patients across settings, and through shared quality improvement initiatives (Baicker & Levy, 2013; Konetzka, Stuart, & Werner, 2018; Zhu et al., 2018). Finally, in each Medicare ACO program, ACOs were allowed to designate a set of partner SNFs where attributed patients could be admitted without the typically required preceding 3-day hospital stay, which could facilitate greater coordination and efficiency in post-acute care (Centers for Medicare and Medicaid Services, 2019; L & M Policy Research, 2016). Several studies have examined the association between hospital-SNF collaboration and patient outcomes and found a strong negative correlation between the proportion of discharges received by a SNF from a hospital and the rate of hospital readmission (Rahman, Foster, Grabowski, Zinn, & Mor, 2013; Rahman, Gadbois, Tyler, & Mor, 2018; Schoenfeld et al., 2016). However, such positive effect of steering may not be present in all cases. For example, Rahman et al. found null effects of steering Medicare Advantage patients to SNFs in the MA contracts’ network (Rahman, Meyers, & Mor, 2018). Winblad et al. found that ACO-affiliated hospitals were more effective than other hospitals in reducing rehospitalizations from SNFs, but were not able to identify the precise mechanism (Winblad, Mor, McHugh, & Rahman, 2017). Furthermore, very few studies have investigated the effects of preferred SNF networks established by ACO hospitals. McHugh et al. found that hospitals developing SNF networks had reductions in readmission rates for patients discharged to SNFs, compared to hospitals without networks (McHugh et al., 2017). However, they were unable to infer a causal relationship due to a lack of information on when the preferred networks were formed. Huckfeldt et al. found preferred SNFs exhibited better performance prior to being selected to participate in preferred 36 SNF networks, indicating that health systems selected SNFs with lower resource use and better quality to participate in their networks (Huckfeldt et al., 2018). However, whether preferred network formation affects patient referral patterns and the effects of preferred network formation on patient outcomes and cost remains unknown. In this study, we investigated whether preferred SNF network formation improved outcomes and lowered costs for patients discharged from hospitals to preferred SNFs. We also investigated whether soft steering of patients to preferred SNFs led to change in patient volume and patient composition at preferred SNFs after network formation. We investigated these effects in 10 health systems that participated in Medicare ACO programs and established preferred SNF networks between 2014 and 2015. We identified SNFs that were included in an ACO’s preferred network and focused on patients discharged to these preferred SNFs. Within each SNF, we compared outcomes of patients who were admitted to the SNF from hospitals that participated in the ACO vs. patients who were discharged from other hospitals. We compared outcomes both prior to and after preferred SNF network formation to estimate how network formation changes outcomes. Methods Study population We selected 10 ACOs that formed preferred SNF networks between 2014 and early 2015. We determined which SNFs are included in an ACO’s preferred network and which hospitals participated in the ACOs based on public information on ACOs’ websites. We reported general information about the 10 ACOs, including the ACO types, the start dates of the ACOs and the preferred networks, Hospital Referral Regions (HRRs) in which they locate (Wennberg & 37 Cooper, 1996), whether they participate in the 3-day waiver program, whether they are physician-led ACOs, and the dates when the preferred SNF list that we used in the study was published. We searched archives of these webpages in order to find the version for the list that was as close to the initial network start date as possible. The study population included Medicare fee-for-service beneficiaries discharged from hospitals to freestanding SNFs that are included in the preferred SNF networks of the 10 ACOs between January 2012 and December 2013, when it was before the preferred network establishment, and between January 2015 and September 2016, when it was after network formation. The unit of analysis was an episode of care that included an initial hospital stay, a subsequent SNF stay and the 90 days following discharge from SNF. Patients who have been residing in SNFs within 30 days before hospital admission were excluded as the likelihood of going back to the same SNF is very high. We only included freestanding SNFs in our sample and excluded hospital-based SNFs for two reasons. First, hospital-based SNFs account for only about 5% of SNFs in the market and are structurally different from free-standing SNFs. Second, the sample size of hospital-based SNFs was not large enough to estimate a separate analysis of hospital-based SNFs. Data sources We obtained information on hospital characteristics from American Hospital Association hospital surveys (American Hospital Association, 2012). We identified SNF stays, preceding hospitalizations, hospital readmissions and inpatient rehabilitation facility (IRF) stays using Medicare Provider Analysis and Review (MEDPAR) files (Centers for Medicare & Medicaid Services, 2012-2016c). We identified home health episodes in Home Health Standard Analytic 38 Files (Centers for Medicare & Medicaid Services, 2012-2016a). We identified long‐term nursing home (or custodial nursing home) stays using the Minimum Data Set (MDS) (Centers for Medicare & Medicaid Services, 2012-2016d). We obtained patient characteristics, Medicare enrollment information, mortality, and patient comorbidities from the Master Beneficiary Summary and Chronic Condition segment files (Centers for Medicare & Medicaid Services, 2012-2016b). Study measures SNF episode outcomes The first set of study outcomes included health care use during the index SNF stay and preceding hospitalization, including the length of the preceding hospitalization, the initial SNF stay, and Medicare spending on the initial SNF stay. Next, we examined patient outcomes during the 90 days following the initial SNF discharge, including hospital readmissions, community residence on the last day of the episode, mortality, and total Medicare spending (initial SNF stay and all hospital and post-acute care spending in the subsequent 90 days). Finally, we examined subsequent post-acute care utilization in the 90-day episode, including whether a patient received any SNF, home health, IRF, long-term care or custodial nursing home care, and total post-acute spending. The final outcome was the change in functional status, measured by the difference between the last and the first non-missing Activity of Daily Living (ADL) score obtained from MDS files. Primary explanatory variables 39 Each patient in our sample first stayed in a hospital and then was discharged to a SNF that was part of an ACO's preferred network. We refer to these SNFs as “preferred SNFs”. Patients were admitted to preferred SNFs from two types of hospitals. First, patients were admitted from hospitals that participated in the same ACO as a preferred SNF. We designated the hospital as the SNF’s “ACO hospital”. Second, patients were admitted to preferred SNFs from hospitals that did not participate in any ACOs. We designated these hospitals as non-ACO hospitals. We categorized patients into the treatment group if a patient was from a SNF’s ACO hospital, and patients into the control group if a patient was admitted from a non-ACO hospital. We only included patients of preferred SNFs in our sample, as only these SNFs had variation in treatment status across patients. We excluded patients from hospitals that belong to some other ACOs besides the 10 ACOs in our sample, as we do not have information on their preferred SNF lists thus are not able to assign them to treatment or control group. Patient and hospital characteristics We controlled for patient characteristics including demographic measures (gender, age, and race and ethnicity), socio-economic status (Medicaid coverage and eligibility for the Part D low-income subsidy), Medicare Severity-Diagnosis Related Group (MS-DRG) for the preceding hospital stay, and the comorbidities listed on the preceding hospital claim (Elixhauser, Steiner, Harris, & Coffey, 1998). Hospital characteristics included ownership (non-profit, for-profit, or government), teaching status, hospital size indicated by the number of beds, and urban location. Empirical approach Patient volume 40 First, we evaluated the effects of network formation on patient volume. We hypothesized that hospitals would be more likely to discharge patients to their preferred SNFs after network formation. We used a sample of all patients from ACO hospitals, both sent to preferred and non- preferred SNFs, to calculate the percentage of patients sent to preferred SNFs in all patients from each hospital in pre and post period, and examined within-hospital changes in the probability of discharging patients to preferred SNFs after network formation (see Appendix Text 1, which provides details of regression). Patient composition Next, we evaluated the effects of network formation on the composition of patients of preferred SNFs. We hypothesized that ACO hospitals would be more likely to discharge more complex patients to preferred SNFs after network formation as more complex patients might benefit more from care coordination. We first evaluated the changes in characteristics of patients admitted to preferred SNFs from the SNF’s ACO hospital after network formation. Next, we compared the pre-post changes in characteristics of patients from ACO hospitals (treatment group) vs. patients from non-ACO hospitals (control group), by running a difference-in- difference (DID) regression model with SNF fixed effects (see Appendix Text 1, which provides details of regression). Patient outcomes Finally, we evaluated the effects of network formation on patient outcomes. We hypothesized that network formation would improve patient outcomes of patients from the preferred SNF’s ACO hospital relative to patients from non-ACO hospitals, due to improved information sharing and coordination between partnering hospitals and SNFs. We first evaluated 41 the pre-post changes in outcomes of patients admitted to preferred SNFs from the SNF’s ACO hospitals (treatment group), then compared the pre-post changes in outcomes for the treatment group relative to the control group (patients admitted to preferred SNFs from non-ACO hospitals) by estimating DID models with SNF fixed effects (see Appendix Text 1, which provides details of regression). We also conducted an event study to test the identifying assumption that time trends in outcomes for the treatment and control groups did not differ prior to network formation (see Appendix Text 1, which provides details of regression). It is important to note that the change in outcomes after network formation for patients from ACO hospitals could come from two sources. First, better care coordination after network formation could lead to improved outcomes for patients from ACO hospitals relative to patients from non-ACO hospitals. Second, the change in patient outcomes could be driven by unobservable changes in patient characteristics for treatment group relative to control group. For example, ACO hospitals might send more complex patients to their preferred SNFs after network formation and some of these changes of patient complexity might be unobservable to the researcher. To address this issue, we adopted an approach developed by Oster to explore the extent to which our results may be sensitive to selection based on unobservables (see Appendix Text 1, for details of Oster Bounds methods) (Oster, 2019). Results Table 2.1 summarized basic information of the 10 ACOs included in our sample. Most of the ACOs started in 2012 and they formed preferred SNF networks between 2014 and 2015. The 42 10 ACOs covered various parts of the US. The majority of these ACOs participated in the 3-day waiver program and only 2 ACOs are physician-led ACOs. Patient volume Table 2.2 displays the average percentage of patients who went to preferred SNFs in all SNF patients from each hospital before and after network formation and whether the within- hospital pre-post differences are statistically significant. Prior to network formation, hospitals affiliated with the 10 ACOs sent 30.9% patients to their future preferred SNFs; after network formation, they sent marginally higher 31.4% patients to preferred SNFs. The slight increase was not statistically significant. Hospitals that did not participate in the 10 ACOs sent 12.4% patients to preferred SNFs in the pre-period and the number remained the same after network formation. We also reported the fraction of all SNF patients within an HRR that went to preferred SNFs before and after network formation (see Appendix Table 1, for changes in market share). Consistent with Table 2.2 we did not find significant pre-post changes in the market shares of these preferred SNFs. Patient characteristics Table 2.3 first shows whether demographic, socio-economic, clinical and functional characteristics of preferred SNF patients from ACO hospitals changed after network formation. We find that patients in the treatment group are wealthier and sicker in the post-period compared to the pre-period. Table 2.3 then shows the results from the DID regressions which tested whether characteristics of patients from ACO hospital changed after network formation relative to patients from non-ACO hospitals. The DID regressions show that preferred network formation was generally not associated with significant changes in observed patient characteristics. The 43 exceptions were a small decrease in the fraction of Hispanic patients and a small increase in the average number of Elixhauser comorbidities for patients being admitted from a SNF’s ACO hospital after network formation. Patient outcomes Table 2.4 first shows results from regressions that evaluated changes in patient outcomes after network formation. We found that within each SNF, patients from ACO hospitals had shorter initial SNF stay, lower initial SNF payment, higher return to community rate after preferred SNF networks were formed, compared to the pre-period. They also had higher mortality rate and higher use of post-acute care. However, when we compared the pre-post differences of patients in treated group to patients in control group, we did not observe significant relative differences in any the outcomes. The DID models show that preferred SNF formation was not associated with significant changes in patient outcomes for patients admitted from the SNF’s ACO hospital relative to patients from the SNF’s non-ACO hospital. We conducted an event study which aims to test the identifying assumption that time trends in outcomes for the treatment and control groups did not differ prior to network formation (see Appendix Table 2, for event study results). In general, the results supported the parallel time trends assumption as for most time points in the pre-period the treatment-control difference was not significantly different and there was no systematic trend. We also constructed bounds for the estimated DID effects that considered bias due to selection on unobservables and found that estimates accounting for selection on unobservables were similar to estimates accounting for selection on observables only. In addition, both sets of estimates are consistent with the finding that network formation was not associated with 44 outcomes among patients admitted from a SNF’s ACO hospital (see Appendix Table 3, for Oster bounds of estimates). One potential explanation for our finding is that preferred SNF formation did change outcomes at preferred SNFs but the effects were not isolated to patients from ACO hospitals. That is, practice pattern spillover effects might have led to similar changes in outcomes for both patients admitted from ACO hospitals and non-ACO hospitals. To investigate this issue, we also compared the performance between preferred SNFs and non-preferred SNFs pre and post network formation (see Appendix Table 4, for comparison results). In both pre and post period, patients of preferred SNFs had better outcomes and lower costs compared to patients sent to out- of-network SNFs, and there were no significant pre-post differences. These results show that SNFs selected for ACOs’ preferred networks had better performance on an array of cost and quality measures even before network formation, however, network formation does not further improve performance of preferred SNFs. Discussion ACOs have an incentive to improve post-acute care outcomes and lower costs. Establishing preferred SNF networks is one approach ACOs and their affiliated hospitals are using to improve care coordination with SNFs. In our study, we hypothesized that after the formation of these preferred networks, hospitals would send more of their patients or more complex patients to their preferred SNFs, and would have improved patient outcomes. We used a DID analysis with SNF fixed effects to test this hypothesis. In particular we compared within SNF changes in outcomes after preferred network formation for preferred SNF patients from ACO hospital relative to patients from non-ACO hospitals. We found little change in patient 45 volume and outcomes and a modest increase in the number of comorbidities of patients from ACO hospitals after network formation. Similar to previous studies, we found that preferred SNFs themselves were better than non-preferred SNFs, even before preferred networks were actually established (Huckfeldt et al., 2018; Lage, Rusinak, Carr, Grabowski, & Ackerly, 2015). This result suggests that hospitals were able to accurately identify higher quality and lower cost SNFs to be included in their networks. However, when we investigated the effects of network establishment, we did not find hospitals sent more patients to preferred SNFs or that preferred SNF patients from ACO hospitals had improved outcomes. In addition, despite the fact that some of the SNFs participated in “3-day waiver” with ACOs where patients could be directly admitted to the SNF or have a hospital stay less than 3 days, we found no change in hospital length of stay. This result likely reflects that our sample conditioned on an index hospital stay occurring. In addition, prior work found that among 3-day waiver SNF stays in the Pioneer ACO program, 76% were direct admissions to SNFs and only 24% were initiated by hospital stays less than 3 days (L & M Policy Research, 2016). There are several possible explanations for our null results. First, even though hospitals may have accurate information on the quality of SNFs – and despite the fact that hospitals are financially responsible for post-discharge care and outcomes – they are not allowed to recommend specific SNFs under Medicare regulations guaranteeing freedom of choice of provider (Medicare Payment Advisory Commission, 2018b). Hospital discharge planners may not provide patients with quality information on potential SNFs for fear of violating these regulations (Tyler et al., 2017). For example, according to a study that interviewed staff members and patients of 16 hospitals and 25 SNFs, patients were not given quality-of-care information 46 about SNFs when discharged from hospitals (Tyler et al., 2017). Most patients choose SNFs based on location, which would not be altered by preferred SNF network formation. Second, prior research on the performance of ACOs models shows mixed results. Some studies document that ACOs are associated with modest reduction or slower increase in spending and improved performance on certain quality measures, while other studies find no evidence of savings or improved performance (Colla et al., 2016; Herrel et al., 2016; McWilliams, Hatfield, Chernew, Landon, & Schwartz, 2016; Nyweide et al., 2015). It remains an open question whether ACOs can achieve significant cost savings and improvements in quality of care. Our results also provided some evidence that ACO hospitals send more complex patients to preferred SNFs after network formation. Patients with multiple comorbid conditions usually have complex medical needs and are responsible for the greatest proportion of spending. Sending those patients to preferred SNFs may be a way for ACO hospitals to better manage them, in order to achieve the largest effects on both patient outcomes and financial rewards (Colla et al., 2016). Our finding suggests that ACOs developed preferred network of SNFs and targeted its care coordination and management efforts toward patients with more complex needs. Our study has several limitations. First, we examined 10 ACOs that publicly posted the list of their preferred SNFs, but we do not have information about other ACOs. Therefore, our results might not be representative of other preferred networks. Second, although we carefully controlled for hospital and patient characteristics, we acknowledge that there may be unobserved differences between patients from ACO hospitals vs. non-ACO hospitals within each SNF, which may confound the observed results. Although we calculated bounds for our DID estimation that accounted for potential omitted variables bias and showed robust results, these bounds are subject to assumptions and thus not conclusive. Finally, we identify patients 47 discharged from hospitals that were part of an ACO. However, it is possible that not all patients from these hospitals were attributed to the ACO. We do not have data on outpatient care and office visits to determine ACO attribution. In conclusion, after the formation of preferred SNF networks, ACO hospitals sent more complex patients to their preferred SNFs, but there was no change in the volume of patients received by these SNFs. Furthermore, preferred SNF network formation was not associated with improvement in patient outcomes. This situation might be ameliorated by giving ACOs greater control over patient choice of SNFs. 48 Tables and Figures Table 2.1 Background information for ACOs and their preferred SNF networks in the sample ACO name ACO type ACO start date Preferred SNF network start date HRR 3-day waiver Date list published physician- led? 1 Allina Pioneer & Next Generation 2012 2014 MN yes 1/1/16 no 2 Monarch Pioneer & Next Generation 2012 2014 CA yes 8/1/15 yes 3 Michigan Pioneer Pioneer & Next Generation 2012 2014 MI yes 2015 no 4 Banner Pioneer & MSSP 2012 2014 AZ yes 08/2014 no 5 OSF Healthcare Pioneer & MSSP 2012 2015 IL, MI, MO yes 3/17/16 no 6 Partners Pioneer & Next Generation 2012 2014 MA yes 01/2017 no 7 Atrius Pioneer & Next Generation 2012 2014 MA yes 3/31/16 no 8 Cleveland Clinic MSSP 2015 2015 OH no 5/18/17 no 9 BJC HealthCare ACO MSSP 2012 2015 MO no 1/8/16 no 10 Torrance Memorial Integrated Physicians Next Generation 2012 2015 CA no 8/15/16 yes 49 Table 2.2 Percent of patients sent from hospitals to preferred SNFs in pre and post period (%) Pre Post Pre-post difference Difference with SNF FE ACO-hospitals Overall (%) 30.9 31.4 0.5 0.4 Partners (%) 43.1 45.4 2.3 1.1 Cleveland (%) 27.1 25.8 -1.3 -0.9 Atrius (%) 46.1 45.2 -0.9 -0.4 Banner (%) 41.8 44.4 2.6 2.7* Monarch (%) 21.9 23.3 1.4 1.1 Allina (%) 41.2 37.6 -3.6 -0.4 Michigan (%) 30.0 29.3 -0.7 -0.8 OSF (%) 20.7 28 7.3 5.5** BJC (%) 5.7 5.1 -0.6 -1.5 Torrance (%) 83.8 79.1 -4.7 -4.6** Non-ACO hospitals (%) 12.4 12.4 0.0 0.3 50 Table 2.3 Change in patient composition after preferred SNF network was formed Patients from ACO hospitals Pre-post diff for patients from ACO hospitals with SNF FE Pre-post diff for patients from ACO hospitals relative to patients from non-ACO hospitals (DID) with SNF FE Pre Post Number of episodes 22267 17378 Demographic characteristics Age 82.2 81.8 -0.3** -0.4 Male (%) 35.0 37.2 2.0** 0.2 White (%) 89.8 89.6 -0.6 0.1 Black (%) 6.4 6.5 0.6 0.1 Asian (%) 1.6 1.6 0.0 0.2 Hispanic (%) 1.0 0.7 -0.4** -0.3* Other (%) 1.3 1.7 0.3** -0.1 Socioeconomic characteristics Dual eligible (%) 16.5 14.6 -1.9*** -0.7 Low income subsidy (%) 18.0 15.9 -2.0*** -1.0 Clinical status Elixhauser comorbidities counts 2.5 3.1 0.6*** 0.1* Selected comorbidities (%) Hypertension 56.7 64.6 7.8*** 0.9 Diabetes 22.8 27.6 4.8*** 1.1 Congestive heart failure 16.8 20.3 3.6*** 0.5 Renal failure 15.6 22.2 6.7*** 1.7 Metastatic cancer 2.5 2.6 0.0 -0.2 Top 5 MS-DRGs (%) 470: Major Joint Replacement or Reattachment of Lower Extremity w/o MCC 11.9 9.9 -2.2*** 1.2 871: Septicemia or severe sepsis w/o MV 96+ hours w MCC 3.7 6.0 2.3*** 0.5 481: Hip & femur procedures except major joint w CC 3.6 3.7 0.1 -0.2 51 690: Kidney & urinary tract infections w/o MCC 2.6 1.7 -0.8*** -0.3 291: Heart failure & shock w MCC 1.8 2.4 0.6*** 0.1 Functional status 14.5 14.4 -0.1 -0.1 52 Table 2.4 Change in patient outcomes after preferred SNF network was formed Patients from ACO hospitals Unadjusted models with SNF FE Adjusted models with SNF FE Pre Post Pre-post diff for patients from ACO hospitals DID Pre-post diff for patients from ACO hospitals DID Number of episodes 22267 17378 Initial SNF use and hospitalization Hospital length of stay 6.1 6.2 0.0 -0.1 -0.2*** -0.2 Initial SNF length of stay 25.0 22.6 -2.6*** 0.1 -2.5*** 0.5 Initial SNF payment ($) 11096.3 10704.0 -518.5* -82.8 -518.2** 75.2 Episode outcomes occurring within 90 days of initial SNF discharge Readmission rate (%) 35.8 35.7 0.1 -0.7 -1.9** -0.6 Return to community rate (%) 68.0 71.4 3.5*** 1.5 4.3*** 0.8 Mortality rate (%) 16.1 18.5 2.3*** 0.1 1.3** 0.5 Total Medicare payment ($) 21142.2 21516.5 224.5 226.6 -397.3 307.1 Post-acute care use within 90 days of initial SNF discharge PAC payment ($) 5145.5 5054.2 -126.4 146.7 -253.1** 148.3 Any use of PAC (%) 71.4 74.1 2.4*** 0.3 1.9** 0.0 Any use of SNF (%) 23.7 22.6 -1.0* -0.8 -2.0*** -0.6 Any use of HHA (%) 61.4 65.6 4.0*** 0.9 3.8*** 0.5 Any use of IRF (%) 0.9 1.1 0.2 -0.1 0.2 -0.1 Any use of custodial nursing home (%) 16.0 10.9 -5.0*** -1.7 -4.6*** -1.0 Any use of long-term care (%) 1.5 1.2 -0.3* -0.1 -0.5** -0.1 Change in functional status Difference in ADL score -2.2 -1.9 0.3*** 0.0 0.2*** 0.0 53 Appendices Appendix Text 1: Details of regression models and analysis methods Patient volume For patient volume analysis, we examined within-hospital changes in the probability of discharging patients to preferred SNFs after the formation of preferred SNF networks. In particular, we estimated the following regression: 𝑝𝑟𝑒𝑓𝑒𝑟𝑆𝑁𝐹 EFG = 𝛼+𝛽 8 𝑃𝑜𝑠𝑡 G +𝐻𝑜𝑠𝑝 F ′𝛽 < +𝑢 EFG (1) where preferSNF indicates whether the patient went to a SNF that is within the preferred SNF networks of the hospital. Post indicates that the period is after network formation. Hosp’ indicates hospital fixed effects. We also estimated changes in the share of patients sent to preferred SNFs over all patients within each market (HRR) before and after preferred network were formed. We tested whether the within-market pre-post difference is statistically significant by estimating the following regression: 𝑃𝑟𝑒𝑓𝑒𝑟𝑆𝑁𝐹 ELG = 𝛼+𝛽 8 𝑝𝑜𝑠𝑡 G +𝐻𝑅𝑅 L ′𝛽 < +𝑢 ELG (2) where 𝑃𝑟𝑒𝑓𝑒𝑟𝑆𝑁𝐹 indicates the whether the patient went to a preferred SNF. Post indicates that the period is after network formation. HRR’ indicates HRR fixed effects. While the previous estimate focused on patients from ACO hospitals to preferred SNFs, this estimate captured patients in preferred SNFs from both ACO and non-ACO hospitals. Patient composition We compared the pre-post changes in characteristics of patients from ACO hospitals (treatment group) vs. patients from non-ACO hospitals (control group), by running a difference- in-difference (DID) regression model with SNF fixed effects as shown below in equation (3): 54 𝑃 EFNG = 𝛼+𝛽 8 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN +𝛽 < 𝑝𝑜𝑠𝑡 G +𝛽 O 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN ×𝑝𝑜𝑠𝑡 G +𝑆𝑁𝐹 N Q +𝑢 EFNG (3) where 𝑃 EFNG is the value of a particular patient characteristic for patient i in time period t., 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN is a binary variable indicating that the patient was admitted from the SNF’s ACO hospital, 𝑝𝑜𝑠𝑡 G is a binary variable indicating that the period is after network formation and 𝑆𝑁𝐹 N Q are SNF fixed effects. 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN ×𝑝𝑜𝑠𝑡 G is an interaction term between 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN and 𝑝𝑜𝑠𝑡 G . The coefficient on this interaction term 𝛽 O is the coefficient of interest. It measures the average within-SNF change in patient characteristics after preferred network formation for patients from the SNF’s ACO hospitals relative to the within-SNF change in patient characteristics after network formation for patients from the SNF’s non-ACO hospitals. Patient outcomes We compared the pre-post changes in outcomes for preferred SNF patients from ACO hospitals (treatment group) relative to patients from non-ACO hospitals (control group) by estimating DID models with SNF fixed effects In particular, we estimated the model shown in equation (4): 𝑌 EFNG = 𝛼+𝜃 8 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN +𝜃 < 𝑝𝑜𝑠𝑡 G +𝜃 O 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN ×𝑝𝑜𝑠𝑡 G +𝑋 EFNG Q 𝛽 8 +𝐻 F Q 𝛽 < +𝑆𝑁𝐹 N Q 𝛽 O +𝑢 EFNG (4) where 𝑌 EFNG is an outcome for patient i admitted to hospital h and SNF s at time t. 𝑋 E Q are a set of patient characteristics, and 𝐻 F Q are a set of hospital characteristics. 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN , 𝑝𝑜𝑠𝑡 G 𝑆𝑁𝐹 N Q and 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN ×𝑝𝑜𝑠𝑡 G have the same definition as in equation (2). The main coefficient of interest is 𝜃 O . It measures the average within-SNF change in patient outcomes after preferred network formation for patients from the SNF’s ACO hospitals relative to the change in patient outcomes for patients from the SNF’s non-ACO hospitals. Event study 55 To test the identifying assumption that time trends in outcomes for the treatment and control groups did not differ prior to network formation, we conducted an event study where we estimated within-SNF differences between treatment and control group for each outcome during each quarter from 2012 to 2016 after adjusting for patient and hospital characteristics, and then tested whether the estimated difference at each time point was significantly different from the difference at a reference time point (the quarter prior to network formation). 𝑌 EFNG = 𝛼+𝜃 8 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN +𝜃 < 𝑞𝑢𝑎𝑟𝑡𝑒𝑟 G ′+𝜃 O 𝑡𝑟𝑒𝑎𝑡𝑒𝑑 FN ×𝑞𝑢𝑎𝑟𝑡𝑒𝑟 G ′+𝑋 EFNG Q 𝛽 8 +𝐻 F Q 𝛽 < +𝑆𝑁𝐹 N Q 𝛽 O +𝑢 EFNG (5) Oster bounds This method infers the potential importance of unobserved patient characteristics based on the sensitivity of coefficient estimates to the addition of observed patient characteristics relative to an uncontrolled regression with no patient characteristics. It infers the importance of selection on unobservables under the assumption that selection on unobservable patient characteristics is proportionally related to selection based on observed patient characteristics. This assumption can be used to create bounds for our main estimates to show how our results would change if we were able to control for both observed and unobserved patient characteristics. 56 Appendix Table 1. Share of patients sent to preferred SNFs in all patients within each market (HHR) in pre and post period Pre Post Pre-post difference Difference with SNF FE Partners (%) 22.4 23.7 1.3 0.6 Cleveland (%) 17.0 16.0 -1.0 -1.0** Atrius (%) 28.9 31.2 2.3 0.7 Banner (%) 39.9 41.8 1.9 1.2 Monarch (%) 13.4 13.3 -0.1 0.1 Allina (%) 15.9 17.5 1.6 0.9 Michigan (%) 9.9 6.6 -3.3 -2.1 OSF (%) 15.5 17.8 2.3 0.8 BJC (%) 1.6 1.3 -0.3 -0.4 Torrance (%) 5.4 5.6 0.2 0.2 Overall (%) 19.5 20.6 1.1 0.3 57 Appendix Table 2. Event Study Differe nce in ADL score -0.1 -0.1 0.1 -0.2 -0.2 -0.2 -0.1 -0.2 0.0 -0.2 -0.1 -0.2 Any custodial nursing home 1.1 -1.6 1.6 1.1 0.8 -0.6 1.1 1.9 0.0 1.2 3.0 2.8 Any use of long- term care -0.4 -0.4 -0.3 -0.9 -0.6 -0.5 -1.2** -1.5** 0.0 -0.6 -0.3 0.4 Any use of HHA 2.8 3.0 0.0 2.1 1.8 4.3 0.3 -0.8 0.0 1.7 -1.1 0.8 Any use of IRF -0.5 -0.3 0.0 0.1 -0.1 -0.1 -0.1 -0.4 0.0 -0.4 -0.1 -2.8* Any use of SNF -1.3 1.5 2.3 -0.4 1.6 3.0 1.9 0.9 0.0 3.0 3.8 0.7 Any use of PAC 1.1 2.6 1.6 0.7 1.3 3.3 1.0 -0.6 0.0 2.2 3.7 -0.4 PAC payment -510 379 -36 -668 256 246 49 -260 0 332 381 -360 Total Medicare payment -531 741 -167 -1,525* 593 578 -156 -704 0 962 1018 2132 Mortality rate -0.2 -3.0 -0.5 -1.6 -2.5 -1.0 -1.3 -3.3* 0.0 -0.8 -0.3 -1.6 Return to commun ity rate 0.8 4.2* -0.4 2.2 2.0 0.9 0.7 2.6 0.0 0.5 -1.3 0.9 Read missi on rate -4.2 1.3 0.5 -1.5 -0.6 1.4 3.0 -2.7 0.0 -0.5 2.3 1.6 Initial SNF paym ent 362 242 155 -364 182 -210 -605 209 0 5 -384 740 Initial SNF length of stay -0.8 -0.2 -1.1 -2.6* -0.8 -1.4 -2.2** -0.6 0.0 -0.3 -1.3 2.8 Hospi tal length of stay 0.2 0.3 -0.1 0.1 0.1 0.1 -0.1 0.2 0.0 0.1 -0.5 0.1 Quart er 1 2 3 4 5 6 7 8 1 2 3 4 Perio d pre (2012 - 2013) 2014 58 -0.1 -0.4* -0.1 -0.2 0.0 -0.1 -0.1 Note: quarter 8 is the reference quarter. * means the difference in outcomes between treatment and control group is significantly different from the difference at the reference quarter with 0.01<p-value <0.05, adjusted for patient and hospital characteristics with SNF fixed effects; ** means 0.001< p-value <0.01. -1.3 0.5 1.4 -0.1 -0.4 -0.6 -2.0 -0.2 -0.7 -0.6 -0.3 -1.0* -1.3** 0.1 3.2 2.2 0.2 0.4 2.5 1.7 4.6 -0.5 -0.6 -0.6 0.2 -0.1 -0.8 -0.7 -0.2 2.1 -0.4 3.9 2.4 -1.6 0.7 1.9 1.9 0.7 2.1 2.3 1.2 2.0 -66 154 -68 597 288 -459 132 -437 -54 75 1135 1012 -628 1035 -1.8 -1.8 0.5 -1.1 -0.3 -0.2 -0.1 2.2 2.9 0.5 0.7 1.3 0.8 1.6 -1.7 -0.1 -0.6 1.2 0.7 -0.3 0.4 -228 -278 -128 137 197 -173 -36 -0.7 -1.0 -0.6 0.7 -0.4 -0.7 -1.0 0.2 -0.3 -0.1 0.2 0.2 -0.3 -0.5 1 2 3 4 5 6 7 post (2015 - 2016) 59 Appendix Table 3. Oster bounds for DID estimates Uncontrolled model: only control for SNF fixed effects Controlled model: adds patient and hospital characteristics Estimates incorporating potential omitted variable bias DID R2 DID R2 DID Initial SNF use and hospitalization Hospital length of stay -0.14 0.03 -0.21 0.34 -0.23 Initial SNF length of stay 0.09 0.09 0.51 0.17 0.76 Initial SNF payment ($) -82.75 0.12 75.19 0.24 166.94 Episode outcomes occurring within 90 days of initial SNF discharge Readmission rate (%) -0.70 0.02 -0.60 0.09 -0.57 Return to community rate (%) 1.54 0.06 0.76 0.24 0.46 Mortality rate (%) 0.12 0.02 0.48 0.19 0.59 Total Medicare payment ($) 226.58 0.06 307.12 0.14 348.94 Post-acute care use within 90 days of initial SNF discharge PAC payment ($) 146.72 0.02 148.29 0.06 148.99 Any use of PAC (%) 0.29 0.05 -0.02 0.10 -0.19 Any use of SNF (%) -0.77 0.01 -0.60 0.06 -0.53 Any use of HHA (%) 0.86 0.07 0.45 0.15 0.22 Any use of IRF (%) -0.07 0.01 -0.09 0.03 -0.09 Any use of custodial nursing home (%) -1.69 0.08 -1.03 0.19 -0.68 Any use of long-term care (%) -0.06 0.01 -0.07 0.03 -0.08 Change in functional status Difference in ADL score -0.02 0.13 0.01 0.22 0.04 60 Appendix Table 4. Performance of preferred SNFs vs. non-preferred SNFs Pre Post Patients from non- preferred SNFs Patients from preferred SNFs Diff Patients from non- preferred SNFs Patients from preferred SNFs Diff Number of episodes 62182 36569 45250 27493 Initial SNF use and hospitalization Hospital length of stay 6.7 6.1 -0.5*** 6.7 6.2 -0.4*** Initial SNF length of stay 28.6 22.9 -5.7*** 27.1 21.5 -5.5*** Initial SNF payment ($) 11170 10216.9 -953.1*** 11107.9 10144.9 -963.1*** Episode outcomes occurring within 90 days of initial SNF discharge Readmission rate (%) 38.3 34.7 -3.7*** 36.8 34.5 -2.3*** Return to community rate (%) 53.8 67.7 13.9*** 63.6 71.2 7.6*** Mortality rate (%) 19.9 16.2 -3.7*** 21.1 18.6 -2.5*** Total Medicare payment ($) 21724.4 20813.5 -910.8*** 21867.4 21278.8 -588.6*** Post-acute care use within 90 days of initial SNF discharge PAC payment ($) 5237.6 5652.3 414.7*** 4783 5316.1 533.2*** Any use of PAC (%) 63 73.9 10.8*** 64 73.8 9.8*** Any use of SNF (%) 29.4 25.5 -3.9*** 27.1 23.8 -3.2*** Any use of HHA (%) 45.3 62.8 17.5*** 48.9 64.4 15.6*** Any use of IRF (%) 0.9 1.0 0.1 0.9 1.2 0.3*** Any use of custodial nursing home (%) 29.2 15.9 -13.3*** 18.8 11 -7.8*** Any use of long- term care (%) 1.9 1.7 -0.2* 1.3 1.2 -0.1 61 References American Hospital Association. (2012). American Hospital Association Annual Survey. Baicker, K., & Levy, H. (2013). Coordination versus Competition in Health Care Reform. The New England Journal of Medicine, 369(9), 789-791. Centers for Medicare & Medicaid Services. (2012-2016a). Home Health Agency (Fee-For- Service). Centers for Medicare & Medicaid Services. (2012-2016b). Medicare Beneficiary Summary Files. Centers for Medicare & Medicaid Services. (2012-2016c). Medicare Provider Analysis and Review. Centers for Medicare & Medicaid Services. (2012-2016d). Minimum Data Set 3.0. Centers for Medicare and Medicaid Services. (2019). Medicare Shared Savings Program Skilled Nursing Facility 3-Day Waiver. Retrieved from https://www.cms.gov/Medicare/Medicare-Fee-for-Service- Payment/sharedsavingsprogram/Downloads/SNF-Waiver-Guidance.pdf Colla, C. H., Lewis, V. A., Kao, L.-S., O’Malley, A. J., Chang, C.-H., & Fisher, E. S. (2016). Association Between Medicare Accountable Care Organization Implementation and Spending Among Clinically Vulnerable Beneficiaries. JAMA Internal Medicine, 176(8), 1167-1175. doi:10.1001/jamainternmed.2016.2827 Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. (1998). Comorbidity measures for use with administrative data. Medical care, 8-27. Evans, M. (2015). Hospitals select preferred SNFs to improve post-acute outcomes. Mod Healthc, 45, 14-15. 62 Herrel, L. A., Norton, E. C., Hawken, S. R., Ye, Z., Hollenbeck, B. K., & Miller, D. C. (2016). Early impact of Medicare accountable care organizations on cancer surgery outcomes. Cancer, 122(17), 2739-2746. doi:10.1002/cncr.30111 Huckfeldt, P. J., Weissblum, L., Escarce, J. J., Karaca ‐Mandic, P., & Sood, N. (2018). Do Skilled Nursing Facilities Selected to Participate in Preferred Provider Networks Have Higher Quality and Lower Costs? Health services research, 53(6), 4886-4905. Jencks, S. F., Williams, M. V., & Coleman, E. A. (2009). Rehospitalizations among patients in the Medicare fee-for-service program. New England Journal of Medicine, 360(14), 1418- 1428. Konetzka, R. T., Stuart, E. A., & Werner, R. M. (2018). The effect of integration of hospitals and post-acute care providers on Medicare payment and patient outcomes. Journal of health economics, 61, 244-258. L & M Policy Research. (2016). Evaluation of Skilled Nursing Facility 3-Day Pioneer ACO Waiver – Final Report. Retrieved from https://innovation.cms.gov/Files/reports/pioneeraco-snf-evalrpt.pdf Lage, D. E., Rusinak, D., Carr, D., Grabowski, D. C., & Ackerly, D. C. (2015). Creating a network of high ‐quality skilled nursing facilities: Preliminary data on the postacute care quality improvement experiences of an accountable care organization. Journal of the American Geriatrics Society, 63(4), 804-808. Livingston, S. (2017). New pay models mean hospitals need stellar post-acute networks to thrive. Modern healthcare, 47(1), 28. 63 McHugh, J. P., Foster, A., Mor, V., Shield, R. R., Trivedi, A. N., Wetle, T., . . . Tyler, D. A. (2017). Reducing hospital readmissions through preferred networks of skilled nursing facilities. Health Affairs, 36(9), 1591-1598. McWilliams, J. M., Hatfield, L. A., Chernew, M. E., Landon, B. E., & Schwartz, A. L. (2016). Early Performance of Accountable Care Organizations in Medicare. N Engl J Med, 374(24), 2357-2366. doi:10.1056/NEJMsa1600142 Medicare Payment Advisory Commission. (2015). Medicare’s post-acute care: Trends and ways to rationalize payments. In Report to the Congress: Medicare Payment Policy. Washington, DC. Medicare Payment Advisory Commission. (2018a). A Data Book: Health Care Spending and the Medicare Program. Medicare Payment Advisory Commission. (2018b). Encouraging Medicare beneficiaries to use higher quality post-acute care providers. In Report to the Congress : Medicare and the Health Care Delivery System Washington DC. Nyweide, D. J., Lee, W., Cuerdon, T. T., Pham, H. H., Cox, M., Rajkumar, R., & Conway, P. H. (2015). Association of Pioneer Accountable Care Organizations vs Traditional Medicare Fee for Service With Spending, Utilization, and Patient Experience. JAMA, 313(21), 2152-2161. doi:10.1001/jama.2015.4930 Oster, E. (2019). Unobservable selection and coefficient stability: Theory and evidence. Journal of Business & Economic Statistics, 37(2), 187-204. Rahman, M., Foster, A. D., Grabowski, D. C., Zinn, J. S., & Mor, V. (2013). Effect of hospital– SNF referral linkages on rehospitalization. Health services research, 48(6pt1), 1898- 1919. 64 Rahman, M., Gadbois, E. A., Tyler, D. A., & Mor, V. (2018). Hospital–Skilled Nursing Facility Collaboration: A Mixed ‐Methods Approach to Understanding the Effect of Linkage Strategies. Health services research, 53(6), 4808-4828. Rahman, M., Meyers, D. J., & Mor, V. (2018). The Effects of Medicare Advantage Contract Concentration on Patients’ Nursing Home Outcomes. Health services research, 53(6), 4087-4105. Schoenfeld, A. J., Zhang, X., Grabowski, D. C., Mor, V., Weissman, J. S., & Rahman, M. (2016). Hospital-skilled nursing facility referral linkage reduces readmission rates among Medicare patients receiving major surgery. Surgery, 159(5), 1461-1468. Sood, N., & Higgins, A. (2012). Posing a framework to guide government’s role in payment and delivery system reform. Health Affairs, 31(9), 2043-2050. Tyler, D. A., Gadbois, E. A., McHugh, J. P., Shield, R. R., Winblad, U., & Mor, V. (2017). Patients are not given quality-of-care data about skilled nursing facilities when discharged from hospitals. Health Affairs, 36(8), 1385-1391. Wennberg, J. E., & Cooper, M. M. (1996). The Dartmouth atlas of health care. The Center for the Evaluative Clinical Sciences, Dartmouth Medical School, American Hospital Publishing, 15-20. Winblad, U., Mor, V., McHugh, J. P., & Rahman, M. (2017). ACO-affiliated hospitals reduced rehospitalizations from skilled nursing facilities faster than other hospitals. Health Affairs, 36(1), 67-73. Zhu, J. M., Patel, V., Shea, J. A., Neuman, M. D., & Werner, R. M. (2018). Hospitals using bundled payment report reducing skilled nursing facility use and improving care integration. Health Affairs, 37(8), 1282-1289. 65 Chapter 3 Do larger vertically integrated Pharmacy Benefit Managers (PBMs) benefit health plans and consumers? Abstract In this paper, we analyzed whether larger PBMs benefit consumers and health plans and whether these effects are different for health plans vertically integrated with the PBM. We used the significant expansion in the size of OptumRx in 2012 – the in-house PBM of UnitedHealthcare – as a natural experiment to analyze these effects. In particular, we used data from Medicare Part D market and compared monthly premiums and costs for top 100 drugs for United plans before and after the expansion relative to other plans in the same market. We also examined the experience of Pennsylvania Public School Employees plans, who used OptumRx as its PBM both before and after the expansion, relative to other plans in the same market. We find that United plans and consumers benefited from the increased size of OptumRx in the form of decreased premiums and drug costs. In contrast, we find no evidence that the larger size of OptumRx benefited Pennsylvania Public School Employee plans. Our results indicate that it is possible that an in-house PBM tend to pass cost savings to their vertically integrated insurer but not other clients. 66 Introduction Pharmacy Benefit Managers (PBMs) are one of the most important intermediaries in the pharmaceutical supply chain. They buy drugs from manufacturers, distribute them to patients, and manage drug benefits for insurers and employers. PBMs create value for health insurers in three distinct ways (Frakt & Garthwaite, 2018). First, they provide an administrative service for insurers, processing and paying prescription drug claims. Second, they negotiate prices and discounts with pharmacies and drug manufacturers. By pooling purchasing power across many insurers, they are able to negotiate larger discounts and rebates compared to individual insurers. Finally, they also help design tiered formularies to encourage cost-effective drug use. One source of income for PBMs is the negotiated rebates with manufacturers, part of which is shared with the client organization. Another source of income is price spreads (markups), i.e., a PBM charges the client organization a higher price for a drug than what the PBM pays to the pharmacy on behalf of the client (Kouvelis, Xiao, & Yang, 2015). The PBM market becomes more and more concentrated over years. In 2016, the top three PBMs comprised 78% of the total market (Balto, 2015). On the one hand, PBMs with larger size have greater bargaining power and are able to negotiate bigger rebates from manufacturer and steeper discounts from pharmacies; on the other hand, the highly concentrated PBM market may discourage PBMs from sharing these savings with health plans and consumers (Garthwaite & Morton, 2017). Therefore, it remains an open question whether insurers and consumers would benefit when the size of their PBM increases. Besides the consolidation in the PBM market, in recent years, there is also a trend towards vertical integration between insurers and PBMs (Schulman & Richman, 2018). 67 Examples include the mergers in 2018 between CVS and Aetna and Cigna and Express Scripts (LaVito, 2018). Insurers who are acquiring PBMs believe that such vertical mergers better align incentives of PBMs with insurers, resulting in cost savings for the insurer. For example, the in- house PBM will be incentivized to design better formularies, negotiate with manufacturers harder for rebates, process claims in a more timely manner, etc. With an in-house PBM, an insurer could reduce administrative costs while benefiting from superior purchasing power (Padhiari, 2019). These cost savings might be passed on to consumers as lower premiums, which attracts more enrollees and leads to a greater market share for the insurer (Porter, 2018). However, others argue that harms of such integration will outweigh the potential efficiencies because the merger will exacerbate the lack of competition in the insurance and PBM markets (Bresnick, 2018). After the merger, the insurer and their in-house PBM will have incentives to disadvantage competing health plans who are using the service of the PBM by reducing pass-through of rebates, not optimizing formulary design, slowing down claim processing, not negotiating hard with manufacturers and pharmacies, etc. In this way, there will be an increase in drug costs and total health care costs for the competing plans, therefore increased premiums faced by their consumers and market loss for the competing plans (Abelson, 2018; Sood, 2018). The insurance market will become even more concentrated when the integrated insurer is gaining market power and its competing plans are losing market. Research has shown that on average, consolidation in health insurance markets does not benefit consumers. Although greater insurance market concentration tends to lower provider prices, there is no evidence the cost savings are passed through to consumers in the form of lower premiums. On the contrary, premiums tend to rise with increased insurer concentration (Dafny, 2015; Gaynor, 2018). 68 To our knowledge, no studies have used real-world data to empirically examine the association between the size of a PBM and outcomes of its client organization and consumers and analyze whether these effects differ for health plans vertically integrated with the PBM. We analyzed this issue by using the expansion of OptumRx in 2012, an in-house PBM of UnitedHealthcare, as a case study. Since 2007, UnitedHealthcare, the largest health insurer in the US, had an in-house PBM called Prescription Solutions. Prescription Solutions was managing United plans in the Medicare Part D market while Medco, a standalone PBM, was managing United plans in the commercial market. However, at the end of 2012, United terminated its contract with Medco and gave the business to its own PBM. United also rebranded Prescription Solutions, its now much larger PBM, as OptumRx. Thus, OptumRx experienced a major expansion in 2013. Theoretically, the much larger OptumRx should be able to negotiate larger rebates and get larger discounts from pharmacies and manufacturers, but it is unclear whether it will pass these discounts and rebates back to health plans it was serving. We hypothesized that OptumRx will have incentives to help Unitedhealthcare plans but not other plans who were using OptumRx as their PBM. Therefore, after the expansion, we expect to see reductions in plan costs and/or reduction in premiums and out-of-pocket costs for United plan consumers with little or no effects for other plans using OptumRx as their PBM. Methods Study sample We included Medicare Part D plans from 2007 to 2016. Medicare Advantage prescription drug plans, plans in the territories and plans that were terminated during the study period were 69 excluded. Our sample included both standalone prescription drug plans (PDPs) and employer- sponsored plans and we did analysis on the two markets separately. Data sources The sample is constructed based on restricted use data obtained from the Centers for Medicare and Medicaid Services (CMS). Cost was calculated using Part D Drug Event Files. Premiums, deductibles, out-of-pocket threshold and other and plan characteristics were obtained from premium files and plan characteristics files. Each plan’s PBM was identified using Managed Market Surveyor data by Decision Resources Group. Study measures Outcomes Outcomes investigated included monthly premiums, patient paid costs, third party paid costs and total cost for top 100 drugs per month. We used Part D claims data to identify the most common filled 100 drugs in each year and calculated the average patient paid, third party paid and total cost for these drugs per patient per month. Third party paid cost represents the amount paid by the plan, but may also include the amount paid Part D low-income subsidy and other third-party payment when applicable. Primary explanatory variables We did analysis separately for standalone PDP market and employer-sponsored market. For standalone PDP market, we compared United plans vs. other plans. For employer-sponsored market, we identified plans offered by Pennsylvania Public School Employees as having 70 OptumRx as their PBM. Thus we compared trends in outcomes pre-post 2013 for Pennsylvania Public School Employees plans with other plans available in the northeastern area of the US, including states of New York, New Jersey, Maine, Vermont, Massachusetts, Rhode Island, Connecticut, New Hampshire, and Pennsylvania. Controls Control variables include whether the plan is a benchmark plan, whether the plan is a basic plan, as well as PDP region fixed effects. Empirical approach We compared selected outcomes for United plans before and after the expansion of OptumRx relative to other plans in the standalone PDP market. We first did an event study to test whether the outcomes for United and other plans share parallel trends before OptumRx’s expansion. Specifically, we estimated differences in outcomes between United and other plans during each year after adjusting for plan characteristics and PDP regions, and tested whether the estimated difference at each year was significantly different from the difference at the reference year 2012, which is the year prior to the expansion. Based on the results from event study that the trends in outcomes for United plans are different from other plans prior to the expansion, we adopted an interrupted times series model to estimate the difference as shown in equation 1, controlling for different trends for United and other plans: 𝑂𝑢𝑡𝑐𝑜𝑚𝑒 Y = 𝛽 8 𝑈𝑛𝑖𝑡𝑒𝑑 Y +𝛽 < 𝑃𝑜𝑠𝑡 G +𝛽 O 𝑇𝑟𝑒𝑛𝑑 G +𝛽 \ 𝑈𝑛𝑖𝑡𝑒𝑑 Y ×𝑇𝑟𝑒𝑛𝑑 G +𝛽 ] 𝑈𝑛𝑖𝑡𝑒𝑑 Y ×𝑃𝑜𝑠𝑡 G +𝛽 ^ 𝑅𝑒𝑔𝑖𝑜𝑛 Q +𝛽 ` 𝑃𝑙𝑎𝑛 Q (1) 71 In this equation, 𝑈𝑛𝑖𝑡𝑒𝑑 Y indicates whether the plan is a United plan, 𝑃𝑜𝑠𝑡 G indicates whether the time is after the expansion of OptumRx, 𝑇𝑟𝑒𝑛𝑑 G represents the linear trend for outcome of the plan. 𝑃𝑙𝑎𝑛′ is a vector for plan characteristics as control variables, and 𝑅𝑒𝑔𝑖𝑜𝑛 Q is a vector for PDP region fixed effects. the coefficient of interest is 𝛽 ] , which represents the pre-post difference in outcome for United plans, relative to pre-post difference for other plans. Similarly, in the employer-sponsored market we compared plans of Pennsylvania Public School Employees before and after the expansion relative to other plans in the northeastern area of the US. The model is shown in equation 2: 𝑂𝑢𝑡𝑐𝑜𝑚𝑒 Y = 𝛽 8 𝑃𝑒𝑛𝑛 Y +𝛽 < 𝑃𝑜𝑠𝑡 G +𝛽 O 𝑇𝑟𝑒𝑛𝑑 G +𝛽 \ 𝑃𝑒𝑛𝑛 Y ×𝑇𝑟𝑒𝑛𝑑 G +𝛽 ] 𝑃𝑒𝑛𝑛 Y ×𝑃𝑜𝑠𝑡 G (2) As there’s no information for PDP for employer-sponsored plans, for control group we only included other plans offered in the northeastern area of the US, which have comparable geographic locations with plans of treated group. We also did similar regressions assuming a quadratic trend to test whether our results are robust. We have two additional sensitivity analyses to compare United plans vs. other plans in the PDP market. Firstly, we restricted the sample of other plans to plans that are offered in the northeastern areas only, including states of New York, New Jersey, Maine, Vermont, Massachusetts, Rhode Island, Connecticut, New Hampshire, and Pennsylvania, making the results more comparable to the analysis of the employer-sponsored market. Secondly, we restricted the sample to renewed plans from 2007 to 2016 and re-did the analysis with plan fixed effects, in order to control for both observed and unobserved plan characteristics. Results 72 Table 3.1 shows summary statistics for outcomes and control variables by treatment group and control group and by whether the time is before or after the expansion of OptumRx. Before the expansion, the average monthly premiums for United plans is $45.1, which is close to the average monthly premiums for other plans in the PDP market. After the expansion, the average premium increased slightly for United plans to $46.7, while the average premium for other plans had a greater increase, from $45.0 to $54.3. For United plans, patient paid cost for top 100 drugs per month decreased from $20.6 to $9.4 after the expansion, while the amount for other plans decreased from $15.3 to $7.9. Similarly, total monthly cost for top 100 drugs decreased from $77.7 to $49.6 for United plans, and from $64.6 to $50.1 for other plans. Third party paid cost for top 100 drugs also decreased after the expansion for both United plans and other plans, however, United plans had a larger decrease from $57.1 to 40.3, compared to a decrease from $49.3 to $42.3 for other plans. There is an increase in the percent of benchmark plans for both United and other plans after OptumRx’s expansion, but United plans had a greater increase by 11%. The percent of basic plans decreased from 57.1% to 42.4% for United plans, while this number remains stable for other plans. We also studied three indicators for plans’ generosity, including deductible, initial coverage limit and out-of-pocket threshold. The average numbers for these three variables by year and by United vs. other plans are shown in Appendix table 1. We find that there are no differences in initial coverage limit and out-of-pocket threshold between United and other plans in each year. Therefore, in further regression analysis we didn’t include these two variables as control or outcomes. In terms of deductible, we find that all United plans have zero deductible in 2011 and 2012, which might be a data issue. We didn’t include deductible in further analysis as data might be unreliable. 73 Table 3.2 shows the results for event study. We find that the differences in outcomes between United plans and other plans in almost every year are significantly different from the difference in the reference year, indicating that outcomes for United plans and other plans have different trends prior to the expansion. Table 3.3 reported results from main regression analysis comparing outcomes of United plans vs. other plans before and after the expansion in the standalone PDP market. We find that after the expansion of OptumRx, monthly premiums of United plans decreased by $14.4, relative to the pre-post change in premiums of other plans in the same market. Compared to other plans in the same market, total monthly cost of top 100 drugs for United plans decreased by $18.3. 36.1% of such decrease is passed through to consumers as the monthly cost of top 100 drugs paid by patients of United plans decreased by $6.6, relative to patients of other plans. The remaining savings are kept by the plan (which could be passed to consumers as lower premiums), as the monthly cost paid by third parties for United plans decreased by $11.7, compared to other plans. Table 3.4 shows results comparing plans of Pennsylvania Public School Employees with other plans offered in the northeastern area before and after the expansion in the employer- sponsored plans market. We didn’t find any significant differences in the pre-post changes in outcomes between Penn plans and other plans. The expansion of OptumRx didn’t have a significant impact on the cost paid by consumers or the plans. In Table 3.5 and Table 3.6, we did similar regressions with quadratic trends and found similar results. After the expansion of OptumRx, monthly premiums of United plans decreased by $10.3, relative to the pre-post change in premiums of other plans in the same market. Compared to other plans in the same market, total monthly cost of top 100 drugs for United plans 74 decreased by $13.7. 34.4% of such decrease is passed through to consumers. On the contrary, we didn’t find any significant differences for the pre-post changes in outcomes between Penn plans and other plans. The expansion of OptumRx didn’t have a significant impact on the cost paid by consumers or the plans. Two sensitivity analyses also showed robust results. In Appendix Table 2, we re-did the analysis comparing United plans vs. other plans by restricting the sample to renewed plans only from 2007 to 2016. We find that compared to other plans in the same market, monthly premiums of United plans decreased by $9.9 after the expansion. Total monthly cost of United plans decreased by $14.7 relative to other plans, and $8.1 of such savings were passed through to consumers. Another sensitivity analysis comparing United plans with other plans available in the northeastern area of the US also showed similar findings (Appendix Table 3). Discussion In recent years, there has been an increasing trend towards consolidation in the PBM market. This trend has been coincident with vertical integration of PBMs and health insurers. However, how increased horizontal and vertical consolidation in the PBM market affects consumers and health plans remains unknown. In this study, we hypothesized that the increase in the size of an insurer-owned PBM would benefit its own insurers and its consumers, in the form of decreased plan costs and/or decreased premiums and decreased out-of-pocket costs. We also hypothesized that there would not be a similar size effect on the PBM’s other client insurers. We compared United plans with other plans before and after the expansion of its PBM and also examined any differences in premiums and costs for OptumRx’s other client organization prior and post the expansion. We find that when the size of OptumRx increased, United plans and its 75 consumers benefited from the increasing bargaining power of OptumRx, as they experienced decreased premiums and drug prices. However, for the other insurer called Pennsylvania Public School Employees that OptumRx was also serving, we didn’t find similar effects on plans and its consumers. Our results are consistent with previous literature that evaluated the importance of size effects in healthcare. Sorensen finds that discounts extracted by insurers for hospital services increased in payer size (Sorensen, 2003). Ellison and Snyder also found that buyer size reduced the price of antibiotics, although supplier competition is a prerequisite for it (Ellison & Snyder, 2010). Lakdawalla and Yin showed that increase in insurer size because of Medicare Part D enrollment lowered negotiated drug prices, which had positive spillover effects on insured consumers external to Part D (Lakdawalla & Yin, 2015). Theoretically, when the size of PBMs increases, they would gain greater bargaining power to lower drug prices (Galbraith, 1993). However, it is unclear whether they would pass the savings to its client insurers or keep the profits to their own. Our results confirmed the size effect for the insurer who is vertically integrated with the PBM, but found null effect for the PBM’s other client insurer. These results suggest that an insurer-owned PBM may be treating clients differently – cost savings from increased bargaining power may be passed through to its integrated insurer instead of other clients. Under such circumstances, insurers would seek more opportunities to merge with PBMs and the insurer and PBM market would become even more concentrated. Although we didn’t find evidence that consumers will be harmed through such concentration, previous literature has warned that lack of competition and increasing concentration of market would make consumers worse off. 76 Our study has several limitations. First, we used the expansion of one single PBM to do a case study. Therefore, the results may not be generalizable to the whole market. Second, due to limitations in data availability, we only investigated a few outcomes and we do not have information about rebates PBMs negotiated and the list prices of each plan. Therefore, we used the costs for the most common 100 drugs as a proxy for drug prices, and these proxies may not accurately represent true drug prices sometimes. Third, although we find different effects of PBM expansion on the PBM’s integrated insurer vs. other clients, the two analyses were done separately in two different markets. Due to data limitation, we were not able to compare the effects in one single model and/or in the same market. Therefore, we are not able to conclude that such differences in size effect are due to vertical integration. Unobserved differences between standalone PDP market and employer-sponsored market is also possible to contribute to the differences in outcomes. Finally, we didn’t investigate whether size effect is associated with the level of market competition. It is possible that within a market where United plans have a large market share, United would have more incentives to utilize its own PBM to attract consumers, thus a greater size effect might be observed in a more concentrated market. This hypothesis serves an important direction for future studies. 77 Tables and Figures Table 3.1 Summary statistics for UnitedHealthcare plans and plans of Pennsylvania Public Employees vs. other plans before and after the expansion of OptumRx United plans Other plans Pre Post Pre Post Plan total premiums ($) 45.09 46.68 44.97 54.31 Patient paid cost per month for top 100 drugs ($) 20.60 9.36 15.33 7.86 Total cost per month for top 100 drugs ($) 77.72 49.63 64.64 50.13 Third party paid cost per month for top 100 drugs ($) 57.13 40.27 49.31 42.27 Benchmark plan (%) 24.51 35.61 26.66 27.59 Basic plan (%) 57.14 42.42 50.00 49.53 Penn plans Other plans Pre Post Pre Post Patient paid cost per month for top 100 drugs ($) 19.05 10.56 14.43 7.33 Total cost per month for top 100 drugs ($) 61.00 34.54 63.09 47.25 Third party paid cost per month for top 100 drugs ($) 41.92 23.98 48.66 39.93 78 Table 3.2 Event study Monthly premium Patient paid cost per month for top 100 drugs Total cost per month for top 100 drugs Third party paid cost for top 100 drugs 2007 -14.77 *** -5.89 *** -6.91 *** -1.02 2008 -13.30 *** -5.69 *** -5.55 *** 0.13 2009 -11.24 *** -3.55 *** -0.38 3.17 *** 2010 -7.59 *** -1.86 ** 5.56 *** 7.41 *** 2011 -3.53 *** -0.44 1.11 1.54 * 2012 0.00 0.00 0.00 0.00 2013 -15.70 *** -4.99 *** -14.04 *** -9.05 *** 2014 -9.64 *** -4.66 *** -9.99 *** -5.33 *** 2015 -26.10 *** -8.34 *** -16.17 *** -7.84 *** 2016 -25.47 *** -8.87 *** -17.66 *** -8.79 *** 79 Table 3.3 Regression results comparing United plans before and after the expansion (linear trend) (1) (2) (3) (4) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Monthly premium United 10.00*** 3.75*** 6.24*** 0.68* (0.79) (0.33) (0.64) -0.31 Trend -1.46*** -1.55*** 0.09 3.12*** (0.17) (0.05) (0.15) -0.12 Post -6.86*** 0.56** -7.42*** -6.99*** (1.12) (0.17) (1.06) (0.74) United × trend 1.19*** 0.64*** 0.55** 0.66*** (0.19) (0.07) (0.16) (0.12) United × post -18.33*** -6.63*** -11.70*** -14.42*** (1.02) (0.51) (0.87) (0.90) Mean value for United plans in pre period 77.72 20.60 57.13 45.09 Robust standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 80 Table 3.4 Regression results comparing Penn plans before and after the expansion (linear trend) (1) (2) (3) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Penn 6.13* 5.96*** 0.17 (2.51) -0.67 (2.23) Trend -2.08*** -1.51*** -0.58*** (0.17) (0.04) (0.15) Post -5.65*** 0.27 -5.92*** (0.94) (0.25) (0.84) Penn × trend -2.51*** -0.27 -2.24*** (0.73) (0.20) (0.65) Penn × post -0.07 -0.76 0.69 (3.82) (1.02) (3.39) Mean value for Penn plans in pre period 60.97 19.05 41.92 Standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 81 Table 3.5 Regression results comparing United plans before and after the expansion (quadratic trend) (1) (2) (3) (4) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Monthly premium United 7.18*** 2.86*** 4.32*** -2.40*** (0.76) (0.34) (0.66) (0.33) Trend -1.11** -2.14*** 1.03*** 5.04*** (0.32) (0.10) (0.28) (0.23) Trend 2 -0.05 0.091*** -0.15*** -0.30*** (0.03) (0.01) (0.03) (0.02) Post -6.04*** -0.83*** -5.21*** -2.41*** (0.93) (0.11) (0.89) (0.66) United × trend 3.93*** 1.52*** 2.41*** 3.62*** (0.27) (0.15) (0.31) (0.30) United × trend 2 -0.38*** -0.13*** -0.25*** -0.39*** (0.04) (0.01) (0.04) (0.04) United × post -13.66*** -4.70*** -8.95*** -10.33*** (1.24) (0.40) (1.04) (1.06) Mean value for United plans in pre period 77.72 20.60 57.13 45.09 Robust standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 82 Table 3.6 Regression results comparing Penn plans before and after the expansion (quadratic trend) (1) (2) (3) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Penn 9.39** 7.65*** 1.74 (3.59) (0.95) (3.19) Trend -3.82*** -2.50*** -1.33*** (0.33) (0.09) (0.29) Trend 2 0.22*** 0.13*** 0.10** (0.04) (0.01) (0.03) Post -7.75*** -0.93*** -6.82*** (1.00) (0.27) (0.89) Penn × trend -3.98* -1.01* -2.97* (1.64) (0.43) (1.46) Penn × trend 2 0.14 0.07 0.07 (0.17) (0.04) (0.15) Penn × post -0.21 -0.78 0.56 (3.95) (1.05) (3.51) Mean value for Penn plans in pre period 60.97 19.05 41.92 Standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 83 Appendices Appendix Table 1. Average initial coverage limit, out-of-pocket threshold and deductible by year Initial Coverage Limit Out-of-pocket threshold Deductible United Other United Other United Other 2007 2400 2385 3850 3850 53 94 2008 2510 2495 4050 4050 110 104 2009 2700 2691 4350 4350 74 118 2010 2830 2821 4550 4550 103 146 2011 2840 2840 4550 4550 0 160 2012 2930 2930 4700 4700 0 165 2013 2970 2970 4750 4750 102 167 2014 2850 2850 4550 4550 105 160 2015 2960 2960 4700 4700 160 163 2016 3310 3310 4850 4850 229 221 84 Appendix Table 2. Sensitivity analysis comparing United plans before and after expansion with plan fixed effects (1) (2) (3) (4) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Monthly premium Trend -1.32*** -1.32*** 0.002 4.37*** (0.16) (0.08) (0.13) (0.22) Post -3.81*** -0.87** -2.95** -8.92*** (1.06) (0.31) (0.99) (0.90) United × trend 2.16*** 1.44*** 0.72*** 2.44*** (0.25) (0.14) (0.19) (0.26) United × post -14.66*** -8.07*** -6.59*** -9.90*** (2.05) (0.91) (1.42) (1.39) Mean value for United plans in pre period 76.53 22.04 54.49 47.89 Robust standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 85 Appendix Table 3. Sensitivity analysis comparing United plans before and after expansion with control plans available in the northeastern area (1) (2) (3) (4) Total cost per month for top 100 drugs Patient paid cost per month for top 100 drugs Third party paid cost for top 100 drugs Monthly premium United 8.45*** 3.60*** 4.85*** 1.20 (1.02) (0.61) (0.63) (0.65) Trend -1.95*** -1.69*** -0.26 3.37*** (0.43) (0.17) (0.30) (0.30) Post -6.69** 0.76 -7.45*** -9.18*** (2.23) (0.45) (1.85) (2.12) United × trend 1.72*** 0.82*** 0.91** 0.48 (0.40) (0.17) (0.29) (0.28) United × post -19.19*** -7.25*** -11.94*** -13.08*** (2.17) (0.53) (1.84) (2.12) Mean value for United plans in pre period 77.72 20.60 57.13 45.09 Standard errors in parentheses *** p<0.001, ** p<0.01, * p<0.05 86 References Abelson, R. (2018). CVS Health and Aetna $69 Billion Merger Is Approved With Conditions. The New York Times. Retrieved from https://www.nytimes.com/2018/10/10/health/cvs- aetna-merger.html Balto, D. A. (2015). The State of Competition in the Pharmacy Benefits Manager and Pharmacy Marketplaces. Retrieved from https://judiciary.house.gov/wp- content/uploads/2016/02/Balto-Testimony-1.pdf Bresnick, J. (2018). AMA: CVS-Aetna Merger Would Reduce PBM Competition, Raise Prices. Retrieved from https://healthpayerintelligence.com/news/ama-cvs-aetna-merger-would- reduce-pbm-competition-raise-prices Dafny, L. (2015). Evaluating the Impact of Health Insurance Industry Consolidation: Learning from Experience. Retrieved from https://www.commonwealthfund.org/publications/issue-briefs/2015/nov/evaluating- impact-health-insurance-industry-consolidation Ellison, S. F., & Snyder, C. M. (2010). Countervailing power in wholesale pharmaceuticals. The Journal of Industrial Economics, 58(1), 32-53. Frakt, A. B., & Garthwaite, C. (2018). The CVS–Aetna Merger: Another Large Bet on the Changing US Health Care Landscape. Annals of Internal Medicine, 168(7), 511-512. Galbraith, J. K. (1993). American capitalism: The concept of countervailing power (Vol. 619): Transaction Publishers. Garthwaite, C., & Morton, F. S. (2017). Perverse Market Incentives Encourage High Prescription Drug Prices. ProMarket Blog Post. Online at: https://promarket. org/perversemarket- incentives-encourage-high-prescription-drug-prices. 87 Gaynor, M. (2018). Examining the Impact of Health Care Consolidation. Statement before the Energy and Commerce Oversight Committee, US House of Representatives. Kouvelis, P., Xiao, Y., & Yang, N. (2015). PBM competition in pharmaceutical supply chain: Formulary design and drug pricing. Manufacturing & Service Operations Management, 17(4), 511-526. Lakdawalla, D., & Yin, W. (2015). Insurers’ negotiating leverage and the external effects of Medicare part D. Review of Economics and Statistics, 97(2), 314-331. LaVito, A. (2018). Justice Department Reportedly Close to Approving CVS-Aetna, Cigna- Express Scripts Deals. CNBC. Retrieved from https://www.cnbc.com/2018/09/05/justice- department-reportedly-close-to-approving-cvs-aetna-cigna-express-scripts-deals.html Padhiari, S. (2019). Health insurers merging with PBMs – the good, the bad and the ugly. Retrieved from https://www.omnihealthdata.com/blog/health-insurers-merging-pbms- good-bad-and-ugly Porter, S. (2018). 3 reasons why health insurers and PBMs are merging. Retrieved from https://www.healthleadersmedia.com/strategy/3-reasons-why-health-insurers-and-pbms- are-merging Schulman, K. A., & Richman, B. D. (2018). The Evolving Pharmaceutical Benefits Market. JAMA, 319(22), 2269-2270. Sood, N. (2018). Potential Effects of the Proposed CVS Acquisition of Aetna on Competition and Consumer Welfare. Retrieved from California Department of Insurance: http://www.insurance.ca.gov/01-consumers/110-health/60-resources/upload/Sood-AMA- finalv3.pdf 88 Sorensen, A. T. (2003). Insurer‐hospital bargaining: negotiated discounts in post‐deregulation Connecticut. The Journal of Industrial Economics, 51(4), 469-490.
Abstract (if available)
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Essays in pharmaceutical and health economics
PDF
The impact of healthcare interventions using electronic health records: an evaluation within an integrated healthcare system
PDF
Advances and applications for economic evaluation methods in health technology assessment (HTA)
PDF
Crohn’s disease: health outcomes and resource utilization in the biologic era
PDF
Discriminating changes in health using patient-reported outcomes
PDF
Developing an agent-based simulation model to evaluate competition in private health care markets with an assessment of accountable care organizations
PDF
Assessing value defects in limb preservation care
PDF
Burden of illness in hemophilia A: taking the patient’s perspective
PDF
Delivering better care for children with special health care needs: analyses of patient-centered medical home and types of insurance
PDF
Economic aspects of obesity
PDF
Impact of pharmacy-based transitional care on healthcare utilization and costs
PDF
Implementation of peer providers in integrated health care settings
PDF
Evaluating treatment options for metastatic, castration-resistant prostate cancer: a comprehensive value assessment
PDF
Value in oncology care and opportunities for improvement
PDF
Investigation of health system performance: effects of integrated triple element method of high reliability, patient safety, and care coordination
PDF
Close the health gap: improving patient access to psychiatric treatment through primary care and telepsychiatry integration
PDF
Essays in health economics and provider behavior
PDF
Three essays on estimating the effects of government programs and policies on health care among disadvantaged population
PDF
Economic, clinical, and behavioral outcomes from medical and pharmaceutical treatments
PDF
Outcomes of antibiotic use among children with acute respiratory tract infections
Asset Metadata
Creator
Gu, Jing
(author)
Core Title
Value in health in the era of vertical integration
School
School of Pharmacy
Degree
Doctor of Philosophy
Degree Program
Health Economics
Publication Date
07/26/2020
Defense Date
06/01/2020
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
health insurance,health outcomes,healthcare cost,Medicare,OAI-PMH Harvest,pharmacy benefit manager,post-acute care,vertical integration
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Sood, Neeraj (
committee chair
), Romley, John (
committee member
), Seabury, Seth (
committee member
)
Creator Email
gujing@usc.edu,gujing0924@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-c89-347925
Unique identifier
UC11664202
Identifier
etd-GuJing-8767.pdf (filename),usctheses-c89-347925 (legacy record id)
Legacy Identifier
etd-GuJing-8767.pdf
Dmrecord
347925
Document Type
Dissertation
Rights
Gu, Jing
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Access Conditions
The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...
Repository Name
University of Southern California Digital Library
Repository Location
USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA
Tags
health insurance
health outcomes
healthcare cost
Medicare
pharmacy benefit manager
post-acute care
vertical integration